%Aigaion2 BibTeX export from Idiap Publications
%Tuesday 03 February 2026 05:12:59 PM
All publications for Sébastien Marcel
@ARTICLE{Akhtar_IEEEMM_2017,
author = {Akhtar, Zahid and Hadid, Abdenour and Nixon, Mark and Tistarelli, Massimo and Dugelay, Jean-Luc and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Biometrics: In Search of Identity and Security (Q & A)},
journal = {IEEE MultiMedia},
volume = {PP},
year = {2017},
doi = {http://ieeexplore.ieee.org/document/7948983/},
abstract = {To address the issues like identity theft and security threats, a continuously evolving technology known as biometrics is presently being deployed in a wide range of personal, government, and commercial applications. Despite the great progress in the field, several exigent problems have yet to be addressed to unleash biometrics full potential. This article aims to present an overview of biometric research and more importantly the significant progress that has been attained over the recent years. The paper is envisaged to further not only the understanding of general audiences and policy makers but also interdisciplinary research. Most importantly, this article is intended to complement earlier articles with updates on most recent topics and developments related to e.g. spoofing, evasion, obfuscation, face reconstruction from DNA, Big data issues in biometrics, etc.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Akhtar_IEEEMM_2017.pdf}
}
@INPROCEEDINGS{Anjos_Bob_ACMMM12,
author = {Anjos, Andr{\'{e}} and El Shafey, Laurent and Wallace, Roy and G{\"{u}}nther, Manuel and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {TABULA RASA, BEAT},
month = oct,
title = {Bob: a free signal processing and machine learning toolbox for researchers},
journal = {Association for Computing Machinery's Multimedia Conference 2012},
booktitle = {Proceedings of the ACM Multimedia Conference},
year = {2012},
url = {https://www.idiap.ch/software/bob/},
crossref = {Anjos_Idiap-RR-25-2012},
abstract = {Bob is a free signal processing and machine learning toolbox originally developed by the Biometrics group at Idiap Research Institute, Switzerland. The toolbox is designed to meet the needs of researchers by reducing development time and efficiently processing data. Firstly, Bob provides a researcher-friendly Python environment for rapid development. Secondly, efficient processing of large amounts of multimedia data is provided by fast C++ implementations of identified bottlenecks. The Python environment is integrated seamlessly with the C++ library, which ensures the library is easy to use and extensible. Thirdly, Bob supports reproducible research through its integrated experimental protocols for several databases. Finally, a strong emphasis is placed on code clarity, documentation, and thorough unit testing. Bob is thus an attractive resource for researchers due to this unique combination of ease of use, efficiency, extensibility and transparency. Bob is an open-source library and an ongoing community effort.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Anjos_Bob_ACMMM12.pdf}
}
@INPROCEEDINGS{Anjos_ICML2017-2_2017,
author = {Anjos, Andr{\'{e}} and G{\"{u}}nther, Manuel and de Freitas Pereira, Tiago and Korshunov, Pavel and Mohammadi, Amir and Marcel, S{\'{e}}bastien},
projects = {BEAT, SWAN},
month = aug,
title = {Continuously Reproducing Toolchains in Pattern Recognition and Machine Learning Experiments},
booktitle = {Thirty-fourth International Conference on Machine Learning},
year = {2017},
location = {Sidney, Australia},
note = {https://openreview.net/group?id=ICML.cc/2017/RML},
url = {https://www.idiap.ch/software/bob/},
abstract = {Pattern recognition and machine learning research work often contains experimental results on real-world data, which corroborates hypotheses and provides a canvas for the development and comparison of new ideas. Results, in this context, are typically summarized as a set of tables and figures, allowing the comparison of various methods, highlighting the advantages of the proposed ideas. Unfortunately, result reproducibility is often an overlooked feature of original research publications, competitions, or benchmark evaluations. The main reason for such a gap is the complexity on the development of software associated with these reports. Software frameworks are difficult to install, maintain, and distribute, while scientific experiments often consist of many steps and parameters that are difficult to report. The increasingly rising complexity of research challenges make it even more difficult to reproduce experiments and results. In this paper, we emphasize that a reproducible research work should be repeatable, shareable, extensible, and stable, and discuss important lessons we learned in creating, distributing, and maintaining software and data for reproducible research in pattern recognition and machine learning. We focus on a specific use-case of face recognition and describe in details how we can make the recognition experiments reproducible in practice.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Anjos_ICML2017-2_2017.pdf}
}
@INPROCEEDINGS{Anjos_ICML2017_2017,
author = {Anjos, Andr{\'{e}} and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {BEAT},
month = aug,
title = {BEAT: An Open-Science Web Platform},
booktitle = {Thirty-fourth International Conference on Machine Learning},
year = {2017},
location = {Sydney, Australia},
note = {https://openreview.net/group?id=ICML.cc/2017/RML},
url = {https://beat-eu.org/platform},
abstract = {With the increased interest in computational sciences, machine learning (ML), pattern recognition (PR) and big data, governmental agencies, academia and manufacturers are overwhelmed by the constant influx of new algorithms and techniques promising improved performance, generalization and robustness. Sadly, result reproducibility is often an overlooked feature accompanying original research publications, competitions and benchmark evaluations. The main reasons behind such a gap arise from natural complications in research and development in this area: the distribution of data may be a sensitive issue; software frameworks are difficult to install and maintain; Test protocols may involve a potentially large set of intricate steps which are difficult to handle.
To bridge this gap, we built an open platform for research in computational sciences related to pattern recognition and machine learning, to help on the development, reproducibility and certification of results obtained in the field. By making use of such a system, academic, governmental or industrial organizations enable users to easily and socially develop processing toolchains, re-use data, algorithms, workflows and compare results from distinct algorithms and/or parameterizations with minimal effort. This article presents such a platform and discusses some of its key features, uses and limitations. We overview a currently operational prototype and provide design insights.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Anjos_ICML2017_2017.pdf}
}
@TECHREPORT{Anjos_Idiap-Com-02-2012,
author = {Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {analysis biometrics python tool},
projects = {TABULA RASA},
month = {4},
title = {ScoreToolKit Documentation},
type = {Idiap-Com},
number = {Idiap-Com-02-2012},
year = {2012},
institution = {Idiap},
abstract = {The ScoreToolKit is a tool for analysing biometric system performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Anjos_Idiap-Com-02-2012.pdf}
}
@TECHREPORT{Anjos_Idiap-RR-14-2017,
author = {Anjos, Andr{\'{e}} and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, machine learning, machine-learning software, pattern recognition, software},
projects = {BEAT},
month = {4},
title = {BEAT: An Open-Source Web-Based Open-Science Platform},
type = {Idiap-RR},
number = {Idiap-RR-14-2017},
year = {2017},
institution = {Idiap},
abstract = {With the increased interest in computational sciences, machine learning (ML), pattern recognition (PR) and big data, governmental agencies, academia and manufacturers are overwhelmed by the constant influx of new algorithms and techniques promising improved performance, generalization and robustness. Sadly, result reproducibility is often an overlooked feature accompanying original research publications, competitions and benchmark evaluations. The main reasons behind such a gap arise from natural complications in research and development in this area: the distribution of data may be a sensitive issue; software frameworks are difficult to install and maintain; Test protocols may involve a potentially large set of intricate steps which are difficult to handle. Given the raising complexity of research challenges and the constant increase in data volume, the conditions for achieving reproducible research in the domain are also increasingly difficult to meet.
To bridge this gap, we built an open platform for research in computational sciences related to pattern recognition and machine learning, to help on the development, reproducibility and certification of results obtained in the field. By making use of such a system, academic, governmental or industrial organizations enable users to easily and socially develop processing toolchains, re-use data, algorithms, workflows and compare results from distinct algorithms and/or parameterizations with minimal effort. This article presents such a platform and discusses some of its key features, uses and limitations. We overview a currently operational prototype and provide design insights.},
pdf = {https://publications.idiap.ch/attachments/reports/2016/Anjos_Idiap-RR-14-2017.pdf}
}
@ARTICLE{Anjos_IETBIOMETRICS_2013,
author = {Anjos, Andr{\'{e}} and Chakka, Murali Mohan and Marcel, S{\'{e}}bastien},
projects = {TABULA RASA},
month = jul,
title = {Motion-Based Counter-Measures to Photo Attacks in Face Recognition},
journal = {Institution of Engineering and Technology Journal on Biometrics},
year = {2013},
url = {http://pypi.python.org/pypi/antispoofing.optflow},
abstract = {Identity spoofing is a contender for high-security face recognition applications. With the advent of social media and globalized search, our face images and videos are wide-spread on the internet and can be potentially used to attack biometric systems without previous user consent. Yet, research to counter these threats is just on its infancy – we lack public standard databases, protocols to measure spoofing vulnerability and baseline methods to detect these attacks. The contributions of this work to the area are three-fold: firstly we introduce a publicly available PHOTO-ATTACK database with associated protocols to measure the effectiveness of counter-measures. Based on the data available, we conduct a study on current state-of-the-art spoofing detection algorithms based on motion analysis, showing they fail under the light of these new dataset. By last, we propose a new technique of counter-measure solely based on foreground/background motion correlation using Optical Flow that outperforms all other algorithms achieving nearly perfect scoring with an equal-error rate of 1.52\% on the available test data. The source code leading to the reported results is made available for the replicability of findings in this article.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Anjos_IETBIOMETRICS_2013.pdf}
}
@INPROCEEDINGS{Anjos_IJCB_2011,
author = {Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {Attack, Counter-Measures, Counter-Spoofing, Disguise, Dishonest Acts, Face Recognition, Forgery, Liveness Detection, Replay, Spoofing, Trick},
projects = {TABULA RASA},
month = oct,
title = {Counter-Measures to Photo Attacks in Face Recognition: a public database and a baseline},
journal = {International Joint Conference on Biometrics 2011},
booktitle = {International Joint Conference on Biometrics 2011},
year = {2011},
url = {http://pypi.python.org/pypi/antispoofing.motion},
abstract = {A common technique to by-pass 2-D face recognition systems is to use photographs of spoofed identities. Unfortunately, research in counter-measures to this type of attack have not kept-up - even if such threats have been known for nearly a decade, there seems to exist no consensus on best practices, techniques or protocols for developing and testing spoofing-detectors for face recognition. We attribute the reason for this delay, partly, to the unavailability of public databases and protocols to study solutions and compare results. To this purpose we introduce the publicly available PRINT-ATTACK database and exemplify how to use its companion protocol with a motion-based algorithm that detects correlations between the person's head movements and the scene context. The results are to be used as basis for comparison to other counter-measure techniques. The PRINT-ATTACK database contains 200 videos of real-accesses and 200 videos of spoof attempts using printed photographs of 50 different identities.},
pdf = {https://publications.idiap.ch/attachments/papers/2011/Anjos_IJCB_2011.pdf}
}
@INCOLLECTION{Anjos_SPRINGER-2_2014,
author = {Anjos, Andr{\'{e}} and Chingovska, Ivana and Marcel, S{\'{e}}bastien},
editor = {Z.Li, Stan and Jain, Anil},
projects = {TABULA RASA},
title = {Anti-Spoofing: Face Databases},
booktitle = {Encyclopedia of Biometrics},
edition = {second edition},
year = {2014},
publisher = {Springer US},
isbn = {978-3-642-27733-7},
url = {http://link.springer.com/referenceworkentry/10.1007/978-3-642-27733-7_9067-2},
doi = {10.1007/978-3-642-27733-7_9212-2},
abstract = {Datasets for the evaluation of face verification system vulnerabilities to spoofing attacks and for the evaluation of face spoofing countermeasures.}
}
@INCOLLECTION{Anjos_SPRINGER_2014,
author = {Anjos, Andr{\'{e}} and Komulainen, Jukka and Marcel, S{\'{e}}bastien and Hadid, Abdenour and Pietikainen, Matti},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Z.Li, Stan},
projects = {TABULA RASA},
title = {Face Anti-spoofing: Visual Approach},
booktitle = {Handbook of Biometric Anti-Spoofing},
chapter = {4},
year = {2014},
pages = {65-82},
publisher = {Springer-Verlag},
isbn = {978-1-4471-6523-1},
doi = {10.1007/978-1-4471-6524-8},
abstract = {User authentication is an important step to protect information and in this regard face biometrics is advantageous. Face biometrics is natural, easy to use and less human-invasive. Unfortunately, recent work revealed that face biometrics is quite vulnerable to spoofing attacks. This chapter presents the different modalities of attacks to visual spectrum face recognition systems. We introduce public datasets for the evaluation of vulnerability of recognition systems and performance of counter-measures. Finally, we build a comprehensive view of anti-spoofing techniques for visual spectrum face recognition and provide an outlook of issues that remain unaddressed.}
}
@INCOLLECTION{Anjos_SPRINGER_2019,
author = {Anjos, Andr{\'{e}} and Tome, Pedro and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Fierrez, Julian and Evans, Nicholas},
projects = {Idiap, 3DFINGERVEIN},
title = {An Introduction to Vein Presentation Attacks and Detection},
booktitle = {Handbook of Biometric Anti-Spoofing},
edition = {2nd},
chapter = {18},
year = {2019},
publisher = {Springer International Publishing},
isbn = {978-3-319-92627-8},
url = {https://www.springer.com/us/book/9783319926261},
doi = {10.1007/978-3-319-92627-8},
abstract = {The domain of presentation attacks (PA), including vulnerability studies and detection (PAD) remains very much unexplored by available scientific literature in biometric vein recognition. Contrary to other modalities that use visual spectral sensors for capturing biometric samples, vein biometrics is typically implemented with near-infrared imaging. The use of invisible light spectra challenges the cre- ation PA instruments, but does not render it impossible. In this chapter, we provide an overview of current landscape for PA manufacturing in possible attack vectors for vein recognition, describe existing public databases and baseline techniques to counter such attacks. The reader will also find material to reproduce experiments and findings for fingervein recognition systems. We provide this material with the hope it will be extended to other vein recognition systems and improved in time.}
}
@TECHREPORT{Atanasoaei_Idiap-RR-02-2012,
author = {Atanasoaei, Cosmin and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {1},
title = {Face detection using boosted Jaccard distance-based regression},
type = {Idiap-RR},
number = {Idiap-RR-02-2012},
year = {2012},
institution = {Idiap},
note = {Submitted to CVPR 2011},
abstract = {This paper presents a new face detection method. We train a model that predicts the Jaccard distance between a sample sub-window and the ground truth face location. This model produces continuous outputs as opposite to the binary output produced by the widely used boosted cascade classifiers. To train this model we introduce a generalization of the binary classification boosting algorithms in which arbitrary smooth loss functions can be optimized. This way single output regression and binary classification models can be trained with the same procedure.
Our method presents several significant advantages. First, it circumvents the need for a specific discretization of the location and scale during testing. Second, it provides an approximation of the search direction (in location and scale) towards the nearest ground truth location. And finally, the training set consists of more diverse samples (e.g. samples covering portions of the faces) that cannot be used to train a classifier. We provide experimental results on BioID face dataset to compare our method with the sliding-windows approach.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Atanasoaei_Idiap-RR-02-2012.pdf}
}
@TECHREPORT{Atanasoaei_Idiap-RR-43-2010,
author = {Atanasoaei, Cosmin and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {12},
title = {On Improving Face Detection Performance by Modelling Contextual Information},
type = {Idiap-RR},
number = {Idiap-RR-43-2010},
year = {2010},
institution = {Idiap},
abstract = {In this paper we present a new method to enhance
object detection by removing false alarms and merging multiple
detections in a principled way with few parameters. The method
models the output of an object classi{\"{\i}}¬er which we consider as
the context. A hierarchical model is built using the detection
distribution around a target sub-window to discriminate between
false alarms and true detections. Next the context is used
to iteratively re{\"{\i}}¬ne the detections. Finally the detections are
clustered using the Adaptive Mean Shift algorithm.
The speci{\"{\i}}¬c case of face detection is chosen for this work as
it is a mature {\"{\i}}¬eld of research. We report results that are better
than baseline method on XM2VTS, BANCA and MIT+CMU
face databases. We signi{\"{\i}}¬cantly reduce the number of false
acceptances while keeping the detection rate at approximately
the same level and in certain conditions we recover miss-aligned
detections.},
pdf = {https://publications.idiap.ch/attachments/reports/2009/Atanasoaei_Idiap-RR-43-2010.pdf}
}
@TECHREPORT{bengio:2001:idiap-01-21,
author = {Bengio, Samy and Mari{\'{e}}thoz, Johnny and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Evaluation of Biometric Technology on {XM2VTS}},
type = {Idiap-RR},
number = {Idiap-RR-21-2001},
year = {2001},
institution = {IDIAP},
note = {also available as the deliverable D71 of the European Project BANCA},
pdf = {https://publications.idiap.ch/attachments/reports/2001/rr01-21.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2001/rr01-21.ps.gz},
ipdmembership={speech, learning},
}
@ARTICLE{bengio:2002:if,
author = {Bengio, Samy and Marcel, Christine and Marcel, S{\'{e}}bastien and Mari{\'{e}}thoz, Johnny},
projects = {Idiap},
title = {Confidence Measures for Multimodal Identity Verification},
journal = {Information Fusion},
volume = {3},
number = {04},
year = {2002},
crossref = {bengio:2001:idiap-01-38},
pdf = {https://publications.idiap.ch/attachments/reports/2002/bengio_2002_if.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/bengio_2002_if.ps.gz},
ipdmembership={speech, learning, vision},
}
@INPROCEEDINGS{Bhattacharjee_BIOSIG2017_2017,
author = {Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
keywords = {3D Masks, Face Presentation Attack Detection, LWIR, NIR},
projects = {Idiap, Tesla},
title = {What you can't see can help you -- extended-range imaging for 3D-mask presentation attack detection},
booktitle = {Proceedings of the 16th International Conference on Biometrics Special Interest Group.},
year = {2017},
publisher = {Gesellschaft fuer Informatik e.V. (GI)},
location = {Darmstadt (Germany)},
issn = {1617-5468},
isbn = {978-3-88579-664-0},
abstract = {We show how imagery in NIR and LWIR bandwidths can be used to detect 2D as well as mask based presentation attacks on face-recognition systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Bhattacharjee_BIOSIG2017_2017.pdf}
}
@INPROCEEDINGS{Bhattacharjee_BTAS2018_2018,
author = {Bhattacharjee, Sushil and Mohammadi, Amir and Marcel, S{\'{e}}bastien},
projects = {Idiap, SWAN, Tesla},
month = oct,
title = {Spoofing Deep Face Recognition With Custom Silicone Masks},
booktitle = {Proceedings of BTAS2018},
year = {2018},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Bhattacharjee_BTAS2018_2018.pdf}
}
@INPROCEEDINGS{Bhattacharjee_ICPR_2024,
author = {Bhattacharjee, Sushil and Geissbuhler, David and Clivaz, G. and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Idiap, Innosuisse CANDY},
month = dec,
title = {Vascular Biometrics Experiments on Candy -- A New Contactless Finger-Vein Dataset},
booktitle = {Proceedings of the International Conference on Pattern Recognition (ICPR)},
year = {2024},
location = {Calcutta (India)},
abstract = {The topic of finger-vein (FV) biometrics is an active and growing topic of research. Most FV systems available today rely on contact sensors that capture vein patterns of a single finger at a time. We have recently completed a project aimed at designing a contactless vein sensing platform, named sweet. In this paper we present a new FV dataset collected using sweet. The dataset includes multiple FV samples from 120 subjects, and 280 presentation attack instruments (PAI), captured in a contactless manner. Further, we present baseline FV authentication (FVA) results achieved for proposed dataset. The sweet platform is equipped to capture a sequence of images suitable for photometric-stereo (PS) reconstruction of 3D surfaces. We present a FV presentation attack detection (PAD) method based on PS reconstruction, and the corresponding baseline FV PAD results on the proposed dataset.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Bhattacharjee_ICPR_2024.pdf}
}
@INCOLLECTION{Bhattacharjee_SPRINGER_2019,
author = {Bhattacharjee, Sushil and Mohammadi, Amir and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Fierrez, Julian and Evans, Nicholas},
projects = {Idiap, Tesla, SWAN},
month = apr,
title = {Recent Advances in Face Presentation Attack Detection},
booktitle = {Handbook of Biometric Anti-Spoofing},
edition = {2nd},
series = {Advances in Computer Vision and Pattern Recognition},
chapter = {10},
year = {2019},
publisher = {Springer},
isbn = {978-3-319-92627-8},
url = {https://www.springer.com/us/book/9783319926261},
abstract = {The undeniable convenience of face-recognition (FR) based biometrics has made it an attractive tool for access control in various applications, from immigration-control to remote banking. Widespread adopti
on of face biometrics, however, depends on the how secure such systems are perceive
d to be. One particular vulnerability of FR systems comes from presentation attacks (PA), where a
subject A attempts to impersonate another subject B, by presenting, for example, a photograph of
B to the biometric sensor (i.e., the camera). PAs are the most likely forms of attacks on face biometric systems, as the camera is the only component of the biometric system that is exposed to the outside world. Robust presentation attack detection (PAD) methods are necessary to construct secure FR based access control
systems. The first edition of the Handbook of Biometric Anti-spoofing included two chapters on face-PAD. In this chapter we present the significant advances in face-PAD research since the publication of the first edition
of this book. In addition to PAD methods designed to work with color images, we also discuss advances in
face-PAD methods using other imaging modalities, namely, near-infrared (NIR) and thermal imaging. This chapter also presents a number of recently published datasets for face-PAD experiments.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Bhattacharjee_SPRINGER_2019.pdf}
}
@INPROCEEDINGS{Boulkenafet_IJCB-2017_2017,
author = {Boulkenafet, Z. and Komulainen, J. and Akhtar, Zahid and Benlamoudi, A. and Bekhouche, SE. and Dornaika, F. and Ouafi, A. and Mohammadi, Amir and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
projects = {SWAN, Tesla},
month = oct,
title = {A Competition on Generalized Software-based Face Presentation Attack Detection in Mobile Scenarios},
booktitle = {Proceedings of the International Joint Conference on Biometrics, 2017},
year = {2017},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Boulkenafet_IJCB-2017_2017.pdf}
}
@TECHREPORT{Bourlard_Idiap-RR-05-2023,
author = {Bourlard, Herv{\'{e}} and Gatica-Perez, Daniel and Odobez, Jean-Marc and Garner, Philip N. and Motlicek, Petr and Magimai-Doss, Mathew and Calinon, Sylvain and Marcel, S{\'{e}}bastien and K{\"{a}}mpf, J{\'{e}}r{\^{o}}me and Luisier, Raphaelle and Liebling, Michael and van der Plas, Lonneke and Teney, Damien and Kodrasi, Ina and Senft, Emmanuel and Henderson, James and Freitas, Andre and Anjos, Andr{\'{e}}},
projects = {Idiap},
month = {7},
title = {Idiap Scientific Report 2022},
type = {Idiap-RR},
number = {Idiap-RR-05-2023},
year = {2023},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2023/Bourlard_Idiap-RR-05-2023.pdf}
}
@INPROCEEDINGS{Bros_BIOSIG_2021,
author = {Bros, Victor and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Vein Enhancement with Deep Auto-Encoders to improve Finger Vein Recognition},
booktitle = {Biometrics Special Interest Group (BIOSIG 2021)},
year = {2021},
isbn = {978-3-88579-709-8},
abstract = {The field of Vascular Biometric Recognition has drawn a lot of attention recently with theemergence of new computer vision techniques. The different methods using Deep Learning involvea new understanding of deeper features from the vascular network. The specific architecture of theveins needs complex model capable of comprehending the vascular pattern. In this paper, we presentan image enhancement method using Deep Convolutional Neural Network. For this task, a residualconvolutional auto-encoder architecture has been trained in a supervised way to enhance the veinpatterns in near-infrared images. The method has been evaluated on several databases with promisingresults on the UTFVP database as a main result. In including the model as a preprocessing in thebiometric pipelines of recognition for finger vein patterns, the error rate has been reduced from 2.1\%to 1.0\%.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/Bros_BIOSIG_2021.pdf}
}
@INPROCEEDINGS{Cardinaux02NSI,
author = {Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Verification using {MLP} and {SVM}},
booktitle = {XI Journees NeuroSciences et Sciences pour l'Ingenieur ({NSI} 2002)},
number = {21},
year = {2002},
address = {La Londe Les Maures, France},
crossref = {cardinaux02rr},
abstract = {The performance of machine learning algorithms has steadily improved over the past few years, such as MLP or more recently SVM. In this paper, we compare two successful discriminant machine learning algorithms apply to the problem of face verification: MLP and SVM. These two algorithms are tested on a benchmark database, namely XM2VTS. Results show that a MLP is better than a SVM on this particular task.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr-02-21.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr-02-21.ps.gz},
ipdmembership={vision},
language={English},
}
@TECHREPORT{cardinaux03RR1,
author = {Cardinaux, Fabien and Sanderson, Conrad and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Comparison of {MLP} and {GMM} Classifiers for Face Verification on {XM2VTS}},
type = {Idiap-RR},
number = {Idiap-RR-10-2003},
year = {2003},
institution = {IDIAP},
abstract = {We compare two classifier approaches, namely classifiers based on Multi Layer Perceptrons (MLPs) and Gaussian Mixture Models (GMMs,',','),
for use in a face verification system. The comparison is carried out in terms of performance, robustness and practicability. Apart from structural differences, the two approaches use different training criteria; the MLP approach uses a discriminative criterion, while the GMM approach uses a combination of Maximum Likelihood (ML) and Maximum a Posteriori (MAP) criteria. Experiments on the XM2VTS database show that for low resolution faces the MLP approach has slightly lower error rates than the GMM approach; however, the GMM approach easily outperforms the MLP approach for high resolution faces and is significantly more robust to imperfectly located faces. The experiments also show that the computational requirements of the GMM approach can be significantly smaller than the MLP approach at a cost of small loss of performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-10.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-10.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{cardinaux03_avbpa,
author = {Cardinaux, Fabien and Sanderson, Conrad and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Comparison of {MLP} and {GMM} Classifiers for Face Verification on {XM2VTS}},
booktitle = {4th International Conference on AUDIO- and VIDEO-BASED BIOMETRIC PERSON AUTHENTICATION},
number = {10},
year = {2003},
address = {University of Surrey, Guildford, UK},
crossref = {cardinaux02rr},
abstract = {We compare two classifier approaches, namely classifiers based on Multi Layer Perceptrons (MLPs) and Gaussian Mixture Models (GMMs,',','),
for use in a face verification system. The comparison is carried out in terms of performance, robustness and practicability. Apart from structural differences, the two approaches use different training criteria; the MLP approach uses a discriminative criterion, while the GMM approach uses a combination of Maximum Likelihood (ML) and Maximum a Posteriori (MAP) criteria. Experiments on the XM2VTS database show that for low resolution faces the MLP approach has slightly lower error rates than the GMM approach; however, the GMM approach easily outperforms the MLP approach for high resolution faces and is significantly more robust to imperfectly located faces. The experiments also show that the computational requirements of the GMM approach can be significantly smaller than the MLP approach at a cost of small loss of performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-10.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-10.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{Cernak_INTERSPEECH_2017,
author = {Cernak, Milos and Komaty, Alain and Mohammadi, Amir and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {Bob toolbox, Kaldi toolkit, open science, Reproducible research, speaker verification},
projects = {Idiap, SWAN},
month = aug,
title = {Bob Speaks Kaldi},
booktitle = {Proc. of Interspeech},
year = {2017},
abstract = {This paper introduces and demonstrates Kaldi integration into Bob signal-processing and machine learning toolbox. The motivation for this integration is two-fold. Firstly, Bob benefits from using advanced speech processing tools developed in Kaldi. Secondly, Kaldi benefits from using complementary Bob modules, such as modulation-based VAD with an adaptive thresholding. In addition, Bob is designed as an open science tool, and this integration might offer to the Kaldi speech community a framework for better reproducibility of state-of-the-art research results.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Cernak_INTERSPEECH_2017.pdf}
}
@INPROCEEDINGS{Chakka_IJCB2011_2011,
author = {Chakka, Murali Mohan and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien and Tronci, Roberto and Muntoni, Daniele and Fadda, Gianluca and Pili, Maurizio and Sirena, Nicola and Murgia, Gabriele and Ristori, Marco and Roli, Fabio and Yan, Junjie and Yi, Dong and Lei, Zhen and Zhang, Zhiwei and Z.Li, Stan and Schwartz, William Robson and Rocha, Anderson and Pedrini, Helio and Lorenzo-Navarro, Javier and Castrill{\'{o}}n-Santana, Modesto and Maatta, Jukka and Hadid, Abdenour and Pietikainen, Matti},
keywords = {2-D Face, Anti-spoofing, Competition, Counter-Measures, Face Recognition, Replay Attacks, Spoofing Attacks},
projects = {Idiap, TABULA RASA},
month = oct,
title = {Competition on Counter Measures to 2-D Facial Spoofing Attacks},
booktitle = {Proceedings of IAPR IEEE International Joint Conference on Biometrics (IJCB), Washington DC, USA},
year = {2011},
abstract = {Spoofing identities using photographs is one of the most common techniques to attack 2-D face recognition systems. There seems to exist no comparative studies of different techniques using the same protocols and data. The motivation behind this competition is to compare the performance of different state-of-the-art algorithms on the same database using a unique evaluation method. Six different teams from universities around the world have participated in the contest. Use of one
or multiple techniques from motion, texture analysis and liveness detection appears to be the common trend in this competition. Most of the algorithms are able to clearly separate spoof attempts from real accesses. The results suggest the investigation of more complex attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2011/Chakka_IJCB2011_2011.pdf}
}
@INPROCEEDINGS{Chingovska_CVPRWORKSHOPONBIOMETRICS_2013,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {biometric recognition, Counter-Measures, Fusion, Spoofing, trustworthy, vulnerability},
projects = {Idiap, TABULA RASA, BEAT},
month = jun,
title = {Anti-spoofing in action: joint operation with a verification system},
booktitle = {Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Workshop on Biometrics},
year = {2013},
location = {Portland, Oregon},
crossref = {Chingovska_Idiap-RR-19-2013},
abstract = {Besides the recognition task, today's biometric systems need to cope with additional problem: spoofing attacks. Up to date, academic research considers spoofing as a binary classification problem: systems are trained to discriminate between real accesses and attacks. However, spoofing counter-measures are not designated to operate stand-alone, but as a part of a recognition system they will protect. In this paper, we study techniques for decision-level and score-level fusion to integrate a recognition and anti-spoofing systems, using an open-source framework that handles the ternary classification problem (clients, impostors and attacks) transparently. By doing so, we are able to report the impact of different spoofing counter-measures, fusion techniques and thresholding on the overall performance of the final recognition system. For a specific use-case covering face verification, experiments show to what extent simple fusion improves the trustworthiness of the system when exposed to spoofing attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Chingovska_CVPRWORKSHOPONBIOMETRICS_2013.pdf}
}
@INPROCEEDINGS{Chingovska_ICB2013_2013,
author = {Chingovska, Ivana and Yang, Jinwei and Lei, Zhen and Yi, Dong and Z.Li, Stan and K{\"{a}}hm, Olga and Damer, Naser and Glaser, Christian and Kuijper, Arjan and Nouak, Alexander and Komulainen, Jukka and de Freitas Pereira, Tiago and Gupta, Shubham and Bansal, Shubham and Khandelwal, Shubham and Rai, Ayush and Krishna, Tarun and Goyal, Dushyant and Waris, Muhammad-Adeel and Zhang, Honglei and Ahmad, Iftikhar and Kiranyaz, Serkan and Gabbouj, Moncef and Tronci, Roberto and Pili, Maurizio and Sirena, Nicola and Roli, Fabio and Galbally, Javier and Fierrez, Julian and Pinto, Allan and Pedrini, Helio and Schwartz, William Robson and Rocha, Anderson and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Competition, Counter-Measures, face spoofing, presentation attack},
projects = {Idiap, TABULA RASA},
month = jun,
title = {The 2nd competition on counter measures to 2D face spoofing attacks},
booktitle = {International Conference of Biometrics 2013},
year = {2013},
location = {Madrid, Spain},
crossref = {Chingovska_Idiap-RR-18-2013},
abstract = {As a crucial security problem, anti-spoofing in biometrics, and particularly for the face modality, has achieved great progress in the recent years. Still, new threats arrive in form of better, more realistic and more sophisticated spoofing attacks. The objective of the 2nd Competition on Counter Measures to 2D Face Spoofing Attacks is to challenge researchers to create counter measures effectively detecting a variety of attacks. The submitted propositions are evaluated on the Replay-Attack database and the achieved results are presented in this paper.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Chingovska_ICB2013_2013.pdf}
}
@INPROCEEDINGS{Chingovska_IEEEBIOSIG2012_2012,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA},
month = sep,
title = {On the Effectiveness of Local Binary Patterns in Face Anti-spoofing},
booktitle = {Proceedings of the 11th International Conference of the Biometrics Special Interes Group},
year = {2012},
crossref = {Chingovska_Idiap-RR-19-2012},
abstract = {Spoofing attacks are one of the security traits that
biometric recognition systems are proven to be vulnerable to.
When spoofed, a biometric recognition system is bypassed by
presenting a copy of the biometric evidence of a valid user. Among
all biometric modalities, spoofing a face recognition system is
particularly easy to perform: all that is needed is a simple
photograph of the user.
In this paper, we address the problem of detecting face spoofing
attacks. In particular, we inspect the potential of texture features
based on Local Binary Patterns (LBP) and their variations on
three types of attacks: printed photographs, and photos and
videos displayed on electronic screens of different sizes. For
this purpose, we introduce REPLAY-ATTACK, a novel publicly
available face spoofing database which contains all the mentioned
types of attacks. We conclude that LBP, with 15\% Half Total
Error Rate, show moderate discriminability when confronted
with a wide set of attack types.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Chingovska_IEEEBIOSIG2012_2012.pdf}
}
@ARTICLE{Chingovska_IEEETIFS_2014,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA, BEAT},
month = dec,
title = {Biometrics Evaluation Under Spoofing Attacks},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {9},
number = {12},
year = {2014},
pages = {2264-2276},
issn = {1556-6013},
doi = {10.1109/Tifs.2014.2349158},
crossref = {Chingovska_Idiap-RR-12-2014},
abstract = {Abstract—While more accurate and reliable than ever, the
trustworthiness of biometric verification systems is compromised
by the emergence of spoofing attacks. Responding to this threat,
numerous research publications address isolated spoofing detection,
resulting in efficient counter-measures for many biometric
modes. However, an important, but often overlooked issue
regards their engagement into a verification task and how to
measure their impact on the verification systems themselves.
A novel evaluation framework for verification systems under
spoofing attacks, called Expected Performance and Spoofability
(EPS) framework, is the major contribution of this paper. Its
purpose is to serve for an objective comparison of different verification
systems with regards to their verification performance
and vulnerability to spoofing, taking into account the system’s
application-dependent susceptibility to spoofing attacks and cost
of the errors. The convenience of the proposed open-source
framework is demonstrated for the face mode, by comparing
the security guarantee of four baseline face verification systems
before and after they are secured with anti-spoofing algorithms.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Chingovska_IEEETIFS_2014.pdf}
}
@INCOLLECTION{Chingovska_SPRINGER-2_2014,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
editor = {Z.Li, Stan and Jain, Anil},
projects = {TABULA RASA},
title = {Anti-spoofing: Evaluation Methodologies},
booktitle = {Encyclopedia of Biometrics},
edition = {2nd edition},
year = {2014},
publisher = {Springer US},
isbn = {978-3-642-27733-7},
doi = {10.1007/978-3-642-27733-7},
abstract = {Following the definition of the task of the anti-spoofing systems to discriminate between real accesses and spoofing attacks, anti-spoofing can be regarded as a binary classification problem. The spoofing databases and the evaluation methodologies for anti-spoofing systems most often comply to the standards for binary classification problems. However, the anti-spoofing systems are not destined to work stand-alone, and their main purpose is to protect a verification system from spoofing attacks. In the process of combining the decision of an anti-spoofing and a recognition system, effects on the recognition performance can be expected. Therefore, it is important to analyze the problem of anti-spoofing under the umbrella of biometric recognition systems. This brings certain requirements in the database design, as well as adapted concepts for evaluation of biometric recognition systems under spoofing attacks.}
}
@INCOLLECTION{Chingovska_SPRINGER_2014,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Z.Li, Stan},
projects = {Idiap, TABULA RASA},
title = {Evaluation Methodologies},
booktitle = {Handbook of Biometric Antispoofing},
year = {2014},
publisher = {Springer},
isbn = {978-1-4471-6523-1}
}
@INCOLLECTION{Chingovska_SPRINGER_2016,
author = {Chingovska, Ivana and Erdogmus, Nesli and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {TABULA RASA},
month = feb,
title = {Face Recognition Systems Under Spoofing Attacks},
booktitle = {Face Recognition Systems Under Spoofing Attacks},
edition = {1st},
chapter = {8},
year = {2016},
pages = {165-194},
publisher = {Springer International Publishing},
isbn = {978-3-319-28501-6},
url = {http://link.springer.com/chapter/10.1007%2F978-3-319-28501-6_8},
doi = {10.1007/978-3-319-28501-6_8},
crossref = {Chingovska_Idiap-RR-18-2020},
abstract = {In this chapter, we give an overview of spoofing attacks and spoofing countermeasures for face recognition systems , with a focus on visual spectrum systems (VIS) in 2D and 3D, as well as near-infrared (NIR) and multispectral systems . We cover the existing types of spoofing attacks and report on their success to bypass several state-of-the-art face recognition systems. The results on two different face spoofing databases in VIS and one newly developed face spoofing database in NIR show that spoofing attacks present a significant security risk for face recognition systems in any part of the spectrum. The risk is partially reduced when using multispectral systems. We also give a systematic overview of the existing anti-spoofing techniques, with an analysis of their advantages and limitations and prospective for future work.}
}
@INCOLLECTION{Chingovska_SPRINGER_2019,
author = {Chingovska, Ivana and Mohammadi, Amir and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Fierrez, Julian and Evans, Nicholas},
projects = {SWAN},
title = {Evaluation Methodologies for Biometric Presentation Attack Detection},
booktitle = {Handbook of Biometric Anti-Spoofing},
edition = {2nd},
chapter = {20},
year = {2019},
publisher = {Springer International Publishing},
isbn = {978-3-319-92627-8},
url = {https://www.springer.com/us/book/9783319926261},
doi = {10.1007/978-3-319-92627-8},
abstract = {Presentation attack detection (PAD, also known as anti-spoofing) systems, regardless of the technique, biometric mode or degree of independence of external equipment, are most commonly treated as binary classification systems. The two classes that they differentiate are bona-fide and presentation attack samples. From this perspective, their evaluation is equivalent to the established evaluation standards for the binary classification systems. However, PAD systems are designed to operate in conjunction with recognition systems and as such can affect their performance. From the point of view of a recognition system, the presentation attacks are a separate class that they need to be detected and rejected. As the problem of presentation attack detection grows to this pseudo-ternary status, the evaluation methodologies for the recognition systems need to be revised and updated. Consequentially, the database requirements for presentation attack databases become more specific. The focus of this chapter is the task of biometric verification and its scope is three-fold: firstly, it gives the definition of the presentation attack detection problem from the two perspectives. Secondly, it states the database requirements for a fair and unbiased evaluation. Finally, it gives an overview of the existing evaluation techniques for presentation attacks detection systems and verification systems under presentation attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Chingovska_SPRINGER_2019.pdf}
}
@INPROCEEDINGS{Colbois_BIOSIG_2022,
author = {Colbois, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = sep,
title = {On the detection of morphing attacks generated by GANs},
booktitle = {21st International Conference of the Biometrics Special Interest Group (BIOSIG 2022)},
year = {2022},
crossref = {Colbois_Idiap-RR-07-2022},
abstract = {Recent works have demonstrated the feasibility of GAN-based morphing attacks that reach similar success rates as more traditional landmark-based methods. This new type of "deep" morphs might require the development of new adequate detectors to protect face recognition systems. We explore simple deep morph detection baselines based on spectral features and LBP histograms features, as well as on CNN models, both in the intra-dataset and cross-dataset case. We observe that simple LBP-based systems are already quite accurate in the intra-dataset setting, but struggle with generalization, a phenomenon that is partially mitigated by fusing together several of those systems at score-level. We conclude that a pretrained ResNet effective for GAN image detection is the most effective overall, reaching close to perfect accuracy. We note however that LBP-based systems maintain a level of interest : additionally to their lower computational requirements and increased interpretability with respect to CNNs, LBP+ResNet fusions sometimes also showcase increased performance versus ResNet-only, hinting that LBP-based systems can focus on meaningful signal that is not necessarily picked up by the CNN detector.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Colbois_BIOSIG_2022.pdf}
}
@INPROCEEDINGS{Colbois_IJCB2024_2024,
author = {Colbois, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = sep,
title = {Evaluating the Effectiveness of Attack-Agnostic Features for Morphing Attack Detection},
booktitle = {International Joint Conference on Biometrics},
year = {2024},
abstract = {Morphing attacks have diversified significantly over the past years, with new methods based on generative adversarial networks (GANs) and diffusion models posing substantial threats to face recognition systems. Recent research has demonstrated the effectiveness of features extracted from large vision models pretrained on bonafide data only (attack-agnostic features) for detecting deep generative images. Building on this, we investigate the potential of these image representations for morphing attack detection (MAD). We develop supervised detectors by training a simple binary linear SVM on the extracted features and one-class detectors by modeling the distribution of bonafide features with a Gaussian Mixture Model (GMM). Our method is evaluated across a comprehensive set of attacks and various scenarios, including generalization to unseen attacks, different source datasets, and print-scan data. Our results indicate that attack-agnostic features can effectively detect morphing attacks, outperforming traditional supervised and one-class detectors from the literature in most scenarios. Additionally, we provide insights into the strengths and limitations of each considered representation and discuss potential future research directions to further enhance the robustness and generalizability of our approach.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Colbois_IJCB2024_2024.pdf}
}
@INPROCEEDINGS{Colbois_IJCB_2021,
author = {Colbois, Laurent and de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
title = {On the use of automatically generated synthetic image datasets for benchmarking face recognition},
booktitle = {International Joint Conference on Biometrics (IJCB 2021)},
year = {2021},
note = {Accepted for Publication in IJCB2021},
abstract = {The availability of large-scale face datasets has been key in the progress of face recognition. However, due to licensing issues or copyright infringement, some datasets are not available anymore (e.g. MS-Celeb-1M). Recent advances in Generative Adversarial Networks (GANs), to synthesize realistic face images, provide a pathway to replace real datasets by synthetic datasets, both to train and benchmark face recognition (FR) systems. The work presented in this paper provides a study on benchmarking FR systems using
a synthetic dataset. First, we introduce the proposed methodology to generate a synthetic dataset, without the need for human intervention, by exploiting the latent structure of a StyleGAN2 model with multiple controlled factors of variation. Then, we confirm that (i) the generated synthetic identities are not data subjects from the GAN's training dataset, which is verified on a synthetic dataset with 10K+ identities; (ii) benchmarking results on the synthetic dataset are a good substitution, often providing error rates and system ranking similar to the benchmarking on the real dataset.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/Colbois_IJCB_2021.pdf}
}
@INPROCEEDINGS{Colbois_IJCB_2023,
author = {Colbois, Laurent and Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center, TRESPASS-ETN},
month = sep,
title = {Approximating Optimal Morphing Attacks using Template Inversion},
booktitle = {IEEE International Joint Conference on Biometric},
year = {2023},
issn = {2474-9680},
isbn = {979-8-3503-3726-6},
doi = {https://doi.org/10.1109/IJCB57857.2023.10448752},
crossref = {Colbois_Idiap-RR-07-2023},
abstract = {Recent works have demonstrated the feasibility of inverting face recognition systems, enabling to recover convincing face images using only their embeddings. We leverage such template inversion models to develop a novel type of deep morphing attack based on inverting a theoretical optimal morph embedding, which is obtained as an average of the face embeddings of source images. We experiment with two variants of this approach : the first one exploits a fully self-contained embedding-to-image inversion model, while the second leverages the synthesis network of a pretrained StyleGAN for increased morph realism. We generate morphing attacks from several source datasets and study the effectiveness of those attacks against several face recognition networks. We showcase that our method can compete with and regularly beat the previous state of the art for deep-learning based morph generation in terms of effectiveness, both in white-box and black-box attack scenarios, and is additionally much faster to run. We hope this might facilitate the development of large scale deep morph datasets for training detection models.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Colbois_IJCB_2023.pdf}
}
@INPROCEEDINGS{Costa-Pazo_BIOSIG2016_2016,
author = {Costa-Pazo, Artur and Bhattacharjee, Sushil and Vazquez-Fernandez, Esteban and Marcel, S{\'{e}}bastien},
projects = {Idiap, Tesla, SWAN},
month = sep,
title = {The REPLAY-MOBILE Face Presentation-Attack Database},
booktitle = {Proceedings of the International Conference on Biometrics Special Interests Group},
year = {2016},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Costa-Pazo_BIOSIG2016_2016.pdf}
}
@INPROCEEDINGS{Das_IJCB2020_2020,
author = {Das, Priyanka and McGrath, Joseph and Fang, Zhaoyuan and Boyd, Aidan and Jang, Ganghee and Mohammadi, Amir and Purnapatra, Sandip and Yambay, David and Marcel, S{\'{e}}bastien and Trokielewicz, Mateusz and Maciejewicz, Piotr and Bowyer, Kevin and Czajka, Adam and Schuckers, Stephanie},
projects = {Idiap},
title = {Iris Liveness Detection Competition (LivDet-Iris) – The 2020 Edition},
booktitle = {INTERNATIONAL JOINT CONFERENCE ON BIOMETRICS (IJCB 2020)},
year = {2020},
url = {https://arxiv.org/abs/2009.00749},
abstract = {Launched in 2013, LivDet-Iris is an international competition series open to academia and industry with the aim to assess and report advances in iris Presentation Attack Detection (PAD). This paper presents results from the fourth competition of the series: LivDet-Iris 2020. This year's competition introduced several novel elements: (a) incorporated new types of attacks (samples displayed on a screen, cadaver eyes and prosthetic eyes), (b) initiated LivDet-Iris as an on-going effort, with a testing protocol available now to everyone via the Biometrics Evaluation and Testing (BEAT)open-source platform to facilitate reproducibility and benchmarking of new algorithms continuously, and (c) performance comparison of the submitted entries with three baseline methods (offered by the University of Notre Dame and Michigan State University), and three open-source iris PAD methods available in the public domain. The best performing entry to the competition reported a weighted average APCER of 59.10\% and a BPCER of 0.46\% over all five attack types. This paper serves as the latest evaluation of iris PAD on a large spectrum of presentation attack instruments.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Das_IJCB2020_2020.pdf}
}
@ARTICLE{deFreitasPereira_ARXIV_2020,
author = {de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = nov,
title = {Fairness in Biometrics: a figure of merit to assess biometric verification systems},
journal = {arXiv},
year = {2020},
abstract = {Machine learning-based (ML) systems are being largely deployed since the last decade in a myriad of scenarios impacting several instances in our daily lives. With this vast sort of applications, aspects of fairness start to rise in the spotlight due to the social impact that this can get in minorities. In this work aspects of fairness in biometrics are addressed. First, we introduce the first figure of merit that is able to evaluate and compare fairness aspects between multiple biometric verification systems, the so-called Fairness Discrepancy Rate (FDR). A use case with two synthetic biometric systems is introduced and demonstrates the potential of this figure of merit in extreme cases of fair and unfair behavior. Second, a use case using face biometrics is presented where several systems are
evaluated compared with this new figure of merit using three public datasets exploring gender and race demographics.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/deFreitasPereira_ARXIV_2020.pdf}
}
@INPROCEEDINGS{deFreitasPereira_BTAS2015_2015,
author = {de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
keywords = {authentication task, biometrics (access control), CPqD mobile biometric database, Face Recognition, mobile computing, mobile environment, MOBIO database, periocular biometrics, person recognition, security of data, session variability},
projects = {BEAT, HFACE},
month = sep,
title = {Periocular Biometrics in Mobile Environment},
booktitle = {IEEE Seventh International Conference on Biometrics: Theory, Applications and Systems},
year = {2015},
pages = {1-7},
publisher = {IEEE},
location = {Arlington, USA},
doi = {10.1109/BTAS.2015.7358785},
abstract = {In this work we study periocular biometrics in a challenging scenario: a mobile environment, where person recognition can take place on a mobile device.
The proposed technique, that models session variability, is evaluated for the authentication task on the MOBIO database, previously used in face recognition, and on a novel mobile biometric database named the CPqD Biometric Database, as well as compared to prior work.
We show that in this particular mobile environment the periocular region is complementary to face recognition, but not superior, unlike shown in a previous study on a more controlled environment.
We show also that a combination with face recognition brings a relative improvement of 5.84\% in terms of HTER.
Finally, the results of this paper will be reproducible using an open software and a novel Web platform.},
pdf = {https://publications.idiap.ch/attachments/papers/2015/deFreitasPereira_BTAS2015_2015.pdf}
}
@ARTICLE{deFreitasPereira_EURASIPJIVP_2014,
author = {de Freitas Pereira, Tiago and Komulainen, Jukka and Anjos, Andr{\'{e}} and De Martino, Jos{\'{e}} Mario and Hadid, Abdenour and Pietikainen, Matti and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Counter-Measures, Face Recognition, temporal pattern extraction, Texture Analysis},
projects = {TABULA RASA},
month = jan,
title = {Face liveness detection using dynamic texture},
journal = {EURASIP Journal on Image and Video Processing},
volume = {2},
year = {2014},
url = {https://pypi.python.org/pypi/antispoofing.lbptop},
doi = {10.1186/1687-5281-2014-2},
abstract = {User authentication is an important step to protect information, and in this context, face biometrics is potentially advantageous. Face biometrics is natural, intuitive, easy to use, and less human-invasive. Unfortunately, recent work has revealed that face biometrics is vulnerable to spoofing attacks using cheap low-tech equipment. This paper introduces a novel and appealing approach to detect face spoofing using the spatiotemporal (dynamic texture) extensions of the highly popular local binary pattern operator. The key idea of the approach is to learn and detect the structure and the dynamics of the facial micro-textures that characterise real faces but not fake ones. We evaluated the approach with two publicly available databases (Replay-Attack Database and CASIA Face Anti-Spoofing Database). The results show that our approach performs better than state-of-the-art techniques following the provided evaluation protocols of each database.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/deFreitasPereira_EURASIPJIVP_2014.pdf}
}
@INPROCEEDINGS{deFreitasPereira_ICB_2013,
author = {de Freitas Pereira, Tiago and Anjos, Andr{\'{e}} and De Martino, Jos{\'{e}} Mario and Marcel, S{\'{e}}bastien},
projects = {BEAT, TABULA RASA},
month = jun,
title = {Can face anti-spoofing countermeasures work in a real world scenario?},
booktitle = {International Conference on Biometrics},
year = {2013},
location = {Madrid, Spain},
url = {http://pypi.python.org/pypi/antispoofing.crossdatabase},
abstract = {User authentication is an important step to protect in- formation and in this field face biometrics is advantageous. Face biometrics is natural, easy to use and less human-invasive. Unfortunately, recent work has revealed that face biometrics is vulnerable to spoofing attacks using low-tech equipments. This article assesses how well existing face anti-spoofing countermeasures can work in a more realistic condition. Experiments carried out with two freely available video databases (Replay Attack Database and CASIA Face Anti-Spoofing Database) show low generalization and possible database bias in the evaluated countermeasures. To generalize and deal with the diversity of attacks in a real world scenario we introduce two strategies that show promising results.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/deFreitasPereira_ICB_2013.pdf}
}
@TECHREPORT{deFreitasPereira_Idiap-RR-09-2022,
author = {de Freitas Pereira, Tiago and Schmidli, Dominic and Linghu, Yu and Zhang, Xinyi and Marcel, S{\'{e}}bastien and G{\"{u}}nther, Manuel},
keywords = {Face Recognition, Reproducible research},
projects = {Biometrics Center, CITeR},
month = {8},
title = {Eight Years of Face Recognition Research: Reproducibility, Achievements and Open Issues},
type = {Idiap-RR},
number = {Idiap-RR-09-2022},
year = {2022},
institution = {Idiap},
url = {https://gitlab.idiap.ch/bob/bob.paper.8years},
abstract = {Automatic face recognition is a research area with high popularity. Many different face recognition algorithms have been proposed in the last thirty years of intensive research in the field. With the popularity of deep learning and its capability to solve a huge variety of different problems, face recognition researchers have concentrated effort on creating better models under this paradigm. From the year 2015, state-of-the-art face recognition has been rooted in deep learning models. Despite the availability of large-scale and diverse datasets for evaluating the performance of face recognition algorithms, many of the modern datasets just combine different factors that influence face recognition, such as face pose, occlusion, illumination, facial expression and image quality. When algorithms produce errors on these datasets, it is not clear which of the factors has caused this error and, hence, there is no guidance in which direction more research is required. This work is a followup from our previous works developed in 2014 and eventually published in 2016, showing the impact of various facial aspects on face recognition algorithms. By comparing the current state-of-the-art with the best systems from the past, we demonstrate that faces under strong occlusions, some types of illumination, and strong expressions are problems mastered by deep learning algorithms, whereas recognition with low-resolution images, extreme pose variations, and open-set recognition is still an open problem. To show this, we run a sequence of experiments using six different datasets and five different face recognition algorithms in an open-source and reproducible manner. We provide the source code to run all of our experiments, which is easily extensible so that utilizing your own deep network in our evaluation is just a few minutes away.},
pdf = {https://publications.idiap.ch/attachments/reports/2022/deFreitasPereira_Idiap-RR-09-2022.pdf}
}
@INPROCEEDINGS{deFreitasPereira_IEEECOMPUTERSOCIETYWORKSHOPONBIOMETRICS_2016,
author = {de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT, HFACE, SWAN},
month = jun,
title = {Heterogeneous Face Recognition using Inter-Session Variability Modelling},
booktitle = {IEEE Computer Society Workshop on Biometrics},
year = {2016},
publisher = {IEEE},
location = {Las Vegas - USA},
abstract = {The task of Heterogeneous Face Recognition consists in to match face images that were sensed in different modalities, such as sketches to photographs, thermal images to photographs or near infrared to photographs. In this preliminary work we introduce a novel and generic approach based on Inter-session Variability Modelling to handle this task. The experimental evaluation conducted with two dif-
ferent image modalities showed an average rank-1 identification rates of 96.93\% and 72.39\% for the CUHK-CUFS (Sketches) and CASIA NIR-VIS 2.0 (Near infra-red) respectively. This work is totally reproducible and all the source code for this approach is made publicly available.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/deFreitasPereira_IEEECOMPUTERSOCIETYWORKSHOPONBIOMETRICS_2016.pdf}
}
@ARTICLE{deFreitasPereira_IEEET-IFS_2019,
author = {de Freitas Pereira, Tiago and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = feb,
title = {Heterogeneous Face Recognition Using Domain Specific Units},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2019},
pages = {13},
doi = {10.1109/TIFS.2018.2885284},
abstract = {The task of Heterogeneous Face Recognition consists in matching face images that are sensed in different domains, such as sketches to photographs (visual spectra images), thermal images to photographs or near-infrared images to photographs.
In this work we suggest that high level features of Deep Convolutional Neural Networks trained on visual spectra images are potentially domain independent and can be used to encode faces sensed in different image domains.
A generic framework for Heterogeneous Face Recognition is proposed by adapting Deep Convolutional Neural Networks low level features in, so called, ``Domain Specific Units''.
The adaptation using Domain Specific Units allow the learning of shallow feature detectors specific for each new image domain.
Furthermore, it handles its transformation to a generic face space shared between all image domains.
Experiments carried out with four different face databases covering three different image domains show substantial improvements, in terms of recognition rate, surpassing the state-of-the-art for most of them.
This work is made reproducible: all the source code, scores and trained models of this approach are made publicly available.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/deFreitasPereira_IEEET-IFS_2019.pdf}
}
@INPROCEEDINGS{deFreitasPereira_LBP_2012,
author = {de Freitas Pereira, Tiago and Anjos, Andr{\'{e}} and De Martino, Jos{\'{e}} Mario and Marcel, S{\'{e}}bastien},
keywords = {Attack, Counter-Measures, Counter-Spoofing, Face Recognition, Liveness Detection, Replay, Spoofing},
projects = {TABULA RASA},
month = nov,
title = {LBP-TOP based countermeasure against face spoofing attacks},
booktitle = {International Workshop on Computer Vision With Local Binary Pattern Variants - ACCV},
year = {2012},
pages = {12},
abstract = {User authentication is an important step to protect information and in this eld face biometrics is advantageous. Face biometrics is natural, easy to use and less human-invasive. Unfortunately, recent work has revealed that face biometrics is vulnerable to spoofing attacks using low-tech cheap equipments. This article presents a countermeasure against such attacks based on the LBP-TOP operator combining both space and time information into a single multiresolution texture descriptor. Experiments carried out with the REPLAY ATTACK database show a Half Total Error Rate (HTER) improvement from 15:16\% to 7:60\%.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/deFreitasPereira_LBP_2012.pdf}
}
@ARTICLE{deFreitasPereira_TBIOM_2021,
author = {de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Fairness in Biometrics: a figure of merit to assess biometric verification systems},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2021},
doi = {10.1109/TBIOM.2021.3102862},
abstract = {Machine learning-based (ML) systems are being largely deployed since the last decade in a myriad of scenarios impacting several instances in our daily lives. With this vast sort of applications, aspects of fairness start to rise in the spotlight due to the social impact that this can get in some social groups. In this work aspects of fairness in biometrics are addressed. First, we introduce a figure of merit that is able to evaluate and compare fairness aspects between multiple biometric verification systems, the so-called Fairness Discrepancy Rate (FDR). A use case with two synthetic biometric systems is introduced and demonstrates the potential of this figure of merit in extreme cases of demographic differentials. Second, a use case using face biometrics is presented where several systems are evaluated compared with this new figure of merit using three public datasets exploring gender and race demographics.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/deFreitasPereira_TBIOM_2021.pdf}
}
@ARTICLE{Dutta_IETBIOMETRICS_2014,
author = {Dutta, Abhishek and G{\"{u}}nther, Manuel and El Shafey, Laurent and Marcel, S{\'{e}}bastien and Veldhuis, Raymond and Spreeuwers, Luuk},
keywords = {automatic eye detectors, commercial face recognition systems, error characteristics, eye detection, eye localization errors, face normalization, face recognition algorithms, face recognition performance, facial feature alignment, manual eye annotations, open source implementations, query phases},
projects = {BBfor2, Idiap},
month = jan,
title = {Impact of Eye Detection Error on Face Recognition Performance},
journal = {IET Biometrics},
year = {2015},
issn = {2047-4938},
url = {http://digital-library.theiet.org/content/journals/10.1049/iet-bmt.2014.0037},
abstract = {The location of the eyes is the most commonly used features to perform face normalization (i.e., alignment of facial features), which is an essential preprocessing stage of many face recognition systems. In this paper, we study the sensitivity of open source implementations of five face recognition algorithms to misalignment caused by eye localization errors. We investigate the ambiguity in location f the eyes by comparing the difference between two independent manual eye annotations. We also study the error characteristics of automatic eye detectors present in two commercial face recognition systems. Furthermore, we explore the impact of using different eye detectors for training/enrollment and query phases of a face recognition system. These experiments provide an insight into the influence of eye localization errors on the performance of face recognition systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Dutta_IETBIOMETRICS_2014.pdf}
}
@INPROCEEDINGS{ElShafey_BIOSIG_2012,
author = {El Shafey, Laurent and Wallace, Roy and Marcel, S{\'{e}}bastien},
keywords = {Face Recognition, Gabor, Gaussian Mixture Models (GMM)},
projects = {Idiap, BBfor2},
month = sep,
title = {Face Verification using Gabor Filtering and Adapted Gaussian Mixture Models},
booktitle = {Proceedings of the 11th International Conference of the Biometrics Special Interest Group},
year = {2012},
pages = {397-408},
publisher = {GI-Edition},
location = {Darmstadt, Germany},
issn = {1617-5468},
isbn = {978-3-88579-290-1},
crossref = {ElShafey_Idiap-RR-37-2011},
abstract = {The search for robust features for face recognition in uncontrolled environments is an important topic of research. In particular, there is a high interest in Gabor-based features which have invariance properties to simple geometrical transformations. In this paper, we first reinterpret Gabor filtering as a frequency decomposition into bands, and analyze the influence of each band separately for face recognition. Then, a new face verification scheme is proposed, combining the strengths of Gabor filtering with Gaussian Mixture Model (GMM) modelling. Finally, this new system is evaluated on the BANCA and MOBIO databases with respect to well known face recognition algorithms. The proposed system demonstrates up to 52\% relative improvement in verification error rate compared to a standard GMM approach, and outperforms the state-of-the-art Local Gabor Binary Pattern Histogram Sequence (LGBPHS) technique for several face verification protocols on two different databases.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/ElShafey_BIOSIG_2012.pdf}
}
@INPROCEEDINGS{ElShafey_EABRA_2014,
author = {El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {BBfor2, BEAT},
month = sep,
title = {Scalable Probabilistic Models: Applied to Face Identification in the Wild},
booktitle = {8th European Biometrics Research and Industry Awards},
year = {2014},
location = {Darmstadt, Germany},
organization = {European Association for Biometrics},
note = {Research paper supporting my application to the EAB Awards},
url = {http://www.eab.org/award/reports/report2014.html?ts=1410764719629},
pdf = {https://publications.idiap.ch/attachments/papers/2014/ElShafey_EABRA_2014.pdf}
}
@INPROCEEDINGS{ElShafey_IJCB_2014,
author = {El Shafey, Laurent and Khoury, Elie and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BEAT},
month = oct,
title = {Audio-Visual Gender Recognition in Uncontrolled Environment Using Variability Modeling Techniques},
booktitle = {International Joint Conference on Biometrics},
year = {2014},
pages = {1 - 8},
publisher = {IEEE},
location = {Clearwater, Florida, USA},
url = {http://pypi.python.org/pypi/xbob.gender.bimodal},
doi = {10.1109/BTAS.2014.6996271},
pdf = {https://publications.idiap.ch/attachments/papers/2014/ElShafey_IJCB_2014.pdf}
}
@ARTICLE{ElShafey_TPAMI_2013,
author = {El Shafey, Laurent and McCool, Chris and Wallace, Roy and Marcel, S{\'{e}}bastien},
keywords = {Expectation maximization, face verification, PLDA, Probablistic Model},
projects = {BBfor2, TABULA RASA},
month = jul,
title = {A Scalable Formulation of Probabilistic Linear Discriminant Analysis: Applied to Face Recognition},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
volume = {35},
number = {7},
year = {2013},
pages = {1788-1794},
url = {https://pypi.python.org/pypi/xbob.paper.tpami2013},
doi = {10.1109/TPAMI.2013.38},
crossref = {ElShafey_Idiap-RR-07-2013},
abstract = {In this paper we present a scalable and exact solution for probabilistic linear discriminant analysis (PLDA). PLDA is a probabilistic model that has been shown to provide state-of-the-art performance for both face and speaker recognition. However, it has one major drawback, at training time estimating the latent variables requires the inversion and storage of a matrix whose size grows quadratically with the number of samples for the identity (class). To date two approaches have been taken to deal with this problem, to: i) use an exact solution which calculates this large matrix and is obviously not scalable with the number of samples or ii) derive a variational approximation to the problem. We present a scalable derivation which is theoretically equivalent to the previous non-scalable solution and so obviates the need for a variational approximation. Experimentally, we demonstrate the efficacy of our approach in two ways. First, on Labelled Faces in the Wild we illustrate the equivalence of our scalable implementation with previously published work. Second, on the large Multi-PIE database, we illustrate the gain in performance when using more training samples per identity (class), which is made possible by the proposed scalable formulation of PLDA.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/ElShafey_TPAMI_2013.pdf}
}
@INPROCEEDINGS{Erdogmus_BIOSIG_2012,
author = {Erdogmus, Nesli and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA},
month = sep,
title = {Spoofing Attacks To 2D Face Recognition Systems With 3D Masks},
booktitle = {International Conference of the Biometrics Special Interes Group},
year = {2013},
location = {Darmstadt, Germany},
crossref = {Erdogmus_Idiap-RR-42-2013},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Erdogmus_BIOSIG_2012.pdf}
}
@INPROCEEDINGS{Erdogmus_BTAS_2013,
author = {Erdogmus, Nesli and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Face Recognition, Mask attack, Spoofing},
projects = {Idiap, TABULA RASA},
month = sep,
title = {Spoofing in 2D Face Recognition with 3D Masks and Anti-spoofing with Kinect},
booktitle = {Biometrics: Theory, Applications and Systems},
year = {2013},
location = {Washington DC, USA},
crossref = {Erdogmus_Idiap-RR-27-2013},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Erdogmus_BTAS_2013.pdf}
}
@ARTICLE{Erdogmus_IEEETIFS_2014,
author = {Erdogmus, Nesli and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA},
month = jul,
title = {Spoofing Face Recognition with 3D Masks},
journal = {IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY},
year = {2014},
pages = {1084-1097},
issn = {1556-6013},
doi = {10.1109/TIFS.2014.2322255},
abstract = {Spoofing is the act of masquerading as a valid user by falsifying data to gain an illegitimate access. Vulnerability of recognition systems to spoofing attacks (presentation attacks) is still an open security issue in biometrics domain and among all biometric traits, face is exposed to the most serious threat, since it is particularly easy to access and reproduce. In the literature, many different types of face spoofing attacks have been examined and various algorithms have been proposed to detect them. Mainly focusing on 2D attacks forged by displaying printed photos or replaying recorded videos on mobile devices, a significant portion of these studies ground their arguments on the flatness of the spoofing material in front of the sensor. However, with the advancements in 3D reconstruction and printing technologies, this assumption can no longer be maintained. In this paper, we aim to inspect the spoofing potential of subject-specific 3D facial masks for different recognition systems and address the detection problem of this more complex attack type. In order to assess the spoofing performance of 3D masks against 2D, 2.5D and 3D face recognition and to analyse various texture based countermeasures using both 2D and 2.5D data, a parallel study with comprehensive experiments is performed on two datasets: The Morpho database which is not publicly available and the newly distributed 3D mask attack database.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Erdogmus_IEEETIFS_2014.pdf}
}
@INPROCEEDINGS{Erdogmus_MMSP_2015,
author = {Erdogmus, Nesli and Vanoni, Matthias and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = sep,
title = {Within- and Cross- Database Evaluations for Gender Classification via BeFIT Protocols},
booktitle = {International Workshop on Multimedia Signal Processing},
year = {2014},
pages = {1-6},
url = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6958797},
doi = {10.1109/MMSP.2014.6958797},
abstract = {With its wide range of applicability, gender classification is an important task in face image analysis and it has drawn a great interest from the pattern recognition community. In this paper, we aim to deal with this problem using Local Binary Pattern Histogram Sequences as feature vectors in general. Differently from what has been done in similar studies, the algorithm parameters used in cropping and feature extraction steps are selected after an extensive grid search using BANCA and MOBIO databases. The final system which is evaluated on FERET, MORPH-II and LFW with gender balanced and imbalanced training sets is shown to achieve commensurate and better results compared to other state-of-the-art performances on those databases. The system is additionally tested for cross-database training in order to assess its accuracy in real world conditions. For LFW and MORPH-II, BeFIT protocols are used.},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Erdogmus_MMSP_2015.pdf}
}
@INPROCEEDINGS{Galbally_BIDS_2009,
author = {Galbally, Javier and McCool, Chris and Fierrez, Julian and Marcel, S{\'{e}}bastien and Ortega-Garcia, Javier},
projects = {Idiap, MOBIO},
title = {Hill-Climbing Attack to an Eigenface-Based Face Verification System},
booktitle = {Proceedings of the First IEEE International Conference on Biometrics, Identity and Security (BIdS)},
year = {2009},
abstract = {We use a general hill-climbing attack algorithm based on Bayesian adaption to test the vulnerability of an Eigenface-based approach for face recognition against indirect attacks. The attacking technique uses the scores provided by the matcher to adapt a global distribution, computed from a development set of users, to the local specificities of the client being attacked. The proposed attack is evaluated on an Eigenfacebased verification system using the XM2VTS database. The results show a very high efficiency of the hill-climbing algorithm, which successfully bypassed the system for over 85\% of the attacked accounts.},
pdf = {https://publications.idiap.ch/attachments/papers/2009/Galbally_BIDS_2009.pdf}
}
@ARTICLE{Galbally_PR_2009,
author = {Galbally, Javier and McCool, Chris and Fierrez, Julian and Marcel, S{\'{e}}bastien and Ortega-Garcia, Javier},
projects = {Idiap, MOBIO},
title = {On the vulnerability of face verification systems to hill-climbing attacks},
journal = {Pattern Recognition},
year = {2009},
abstract = {In this paper, we use a hill-climbing attack algorithm based on Bayesian adaption to test the vulnerability of two face recognition systems to indirect attacks. The attacking technique uses the scores provided by the matcher to adapt a global distribution computed from an independent set of users, to the local specificities of the client being attacked. The proposed attack is evaluated on an eigenface-based and a parts-based face verification system using the XM2VTS database. Experimental results demonstrate that the hill-climbing algorithm is very efficient and is able to bypass over 85\% of the attacked accounts (for both face recognition systems). The security flaws of the analyzed systems are pointed out and possible countermeasures to avoid them are also proposed.}
}
@ARTICLE{Geissbuhler_ARXIV_2024,
author = {Geissbuhler, David and Bhattacharjee, Sushil and Kotwal, Ketan and Clivaz, G. and Marcel, S{\'{e}}bastien},
projects = {Idiap, Innosuisse CANDY},
month = apr,
title = {SWEET - An Open Source Modular Platform for Contactless Hand Vascular Biometric Experiments},
journal = {arXiv},
year = {2024},
url = {https://arxiv.org/abs/2404.09376},
doi = {https://doi.org/10.48550/arXiv.2404.09376},
abstract = {Current finger-vein or palm-vein recognition systems usually require direct contact of the subject with the apparatus. This can be problematic in environments where hygiene is of primary importance. In this work we present a contactless vascular biometrics sensor platform named SWEET which can be used for hand vascular biometrics studies (wrist-, palm- and finger-vein) and surface features such as palmprint. It supports several acquisition modalities such as multi-spectral Near-Infrared (NIR), RGB-color, Stereo Vision (SV) and Photometric Stereo (PS). Using this platform we collect a dataset consisting of the fingers, palm and wrist vascular data of 120 subjects and develop a powerful 3D pipeline for the pre-processing of this data. We then present biometric experimental results, focusing on Finger-Vein Recognition (FVR).}
}
@INPROCEEDINGS{Geissbuhler_ICML_2025,
author = {Geissbuhler, David and Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER, CITeR},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
title = {Synthetic Face Datasets Generation via Latent Space Exploration from Brownian Identity Diffusion},
booktitle = {The Forty-second International Conference on Machine Learning (ICML)},
year = {2025},
url = {https://openreview.net/pdf?id=QxOgS8WwCr},
abstract = {Face recognition models are trained on large-scale datasets, which have privacy and ethical concerns. Lately, the use of synthetic data to complement or replace genuine data for the training of face recognition models has been proposed. While promising results have been obtained, it still remains unclear if generative models can yield diverse enough data for such tasks. In this work, we introduce a new method, inspired by the physical motion of soft particles subjected to stochastic Brownian forces, allowing us to sample identities distributions in a latent space under various constraints. We introduce three complementary algorithms, called Langevin, Dispersion, and DisCo, aimed at generating large synthetic face datasets. With this in hands, we generate several face datasets and benchmark them by training face recognition models, showing that data generated with our method exceeds the performance of previously GAN-based datasets and achieves competitive performance with state-of-the-art diffusion-based synthetic datasets. While diffusion models are shown to memorize training data, we prevent leakage in our new synthetic datasets, paving the way for more responsible synthetic datasets. Project page: https://www.idiap.ch/paper/synthetics-disco}
}
@ARTICLE{George_ACCESS_2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Attacking Face Recognition with T-shirts: Database, Vulnerability Assessment and Detection},
journal = {IEEE Access},
year = {2023},
crossref = {George_Idiap-RR-08-2023},
pdf = {https://publications.idiap.ch/attachments/papers/2023/George_ACCESS_2023.pdf}
}
@INPROCEEDINGS{George_CVPR_2021,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Cross Modal Focal Loss for RGBD Face Anti-Spoofing},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2021},
abstract = {Automatic methods for detecting presentation attacks are
essential to ensure the reliable use of facial recognition
technology. Most of the methods available in the litera-
ture for presentation attack detection (PAD) fails in gen-
eralizing to unseen attacks. In recent years, multi-channel
methods have been proposed to improve the robustness of
PAD systems. Often, only a limited amount of data is avail-
able for additional channels, which limits the effectiveness
of these methods. In this work, we present a new framework
for PAD that uses RGB and depth channels together with a
novel loss function. The new architecture uses complemen-
tary information from the two modalities while reducing the
impact of overfitting. Essentially, a cross-modal focal loss
function is proposed to modulate the loss contribution of
each channel as a function of the confidence of individual
channels. Extensive evaluations in two publicly available
datasets demonstrate the effectiveness of the proposed ap-
proach.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/George_CVPR_2021.pdf}
}
@INPROCEEDINGS{George_ICASSP2024_2024,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Heterogeneous Face Recognition Using Domain Invariant Units},
booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/George_ICASSP2024_2024.pdf}
}
@INPROCEEDINGS{George_ICB2019,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Deep Pixel-wise Binary Supervision for Face Presentation Attack Detection},
booktitle = {International Conference on Biometrics},
year = {2019},
pdf = {https://publications.idiap.ch/attachments/papers/2019/George_ICB2019.pdf}
}
@INPROCEEDINGS{George_ICCV-2_2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
additionalresearchprograms = {AI for Everyone},
title = {EdgeDoc: Hybrid CNN-Transformer Model for Accurate Forgery Detection and Localization in ID Documents},
booktitle = {ICCV},
year = {2025},
crossref = {George_Idiap-RR-08-2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/George_ICCV-2_2025.pdf}
}
@INPROCEEDINGS{George_ICCV_2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, CARMEN},
additionalresearchprograms = {AI for Everyone},
title = {Enhancing Domain Diversity in Synthetic Data Face Recognition with Dataset Fusion},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/George_ICCV_2025.pdf}
}
@TECHREPORT{George_Idiap-RR-01-2024,
author = {George, Anjith and Ecabert, Christophe and Otroshi Shahreza, Hatef and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Idiap, SAFER},
month = {1},
title = {EdgeFace: Efficient Face Recognition Model for Edge Devices},
type = {Idiap-RR},
number = {Idiap-RR-01-2024},
year = {2024},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2023/George_Idiap-RR-01-2024.pdf}
}
@TECHREPORT{George_Idiap-RR-02-2022,
author = {George, Anjith and Geissbuhler, David and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
month = {2},
title = {A Comprehensive Evaluation on Multi-channel Biometric Face Presentation Attack Detection},
type = {Idiap-RR},
number = {Idiap-RR-02-2022},
year = {2022},
institution = {Idiap},
abstract = {The vulnerability against presentation attacks is
a crucial problem undermining the wide-deployment of face
recognition systems. Though presentation attack detection (PAD)
systems try to address this problem, the lack of generalization
and robustness continues to be a major concern. Several works
have shown that using multi-channel PAD systems could alleviate
this vulnerability and result in more robust systems. However,
there is a wide selection of channels available for a PAD system
such as RGB, Near Infrared, Shortwave Infrared, Depth, and
Thermal sensors. Having a lot of sensors increases the cost of
the system, and therefore an understanding of the performance
of different sensors against a wide variety of attacks is necessary
while selecting the modalities. In this work, we perform a
comprehensive study to understand the effectiveness of various
imaging modalities for PAD. The studies are performed on a
multi-channel PAD dataset, collected with 14 different sensing
modalities considering a wide range of 2D, 3D, and partial
attacks. We used the multi-channel convolutional network-based
architecture, which uses pixel-wise binary supervision. The model
has been evaluated with different combinations of channels, and
different image qualities on a variety of challenging known and
unknown attack protocols. The results reveal interesting trends
and can act as pointers for sensor selection for safety-critical
presentation attack detection systems. The source codes and
protocols to reproduce the results are made available publicly
making it possible to extend this work to other architectures.},
pdf = {https://publications.idiap.ch/attachments/reports/2021/George_Idiap-RR-02-2022.pdf}
}
@TECHREPORT{George_Idiap-RR-09-2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, HARDENING},
month = {11},
title = {Bridging the Gap: Heterogeneous Face Recognition with Conditional Adaptive Instance Modulation},
type = {Idiap-RR},
number = {Idiap-RR-09-2023},
year = {2023},
institution = {Idiap},
abstract = {Heterogeneous Face Recognition (HFR) aims to match face images across different domains, such as thermal and visible spectra, expanding the applicability of Face Recognition (FR) systems to challenging scenarios. However, the domain gap and limited availability of large-scale datasets in the target domain make training robust and invariant \hfr models from scratch difficult. In this work, we treat different modalities as distinct styles and propose a framework to adapt feature maps, bridging the domain gap. We introduce a novel \fullcaim (\caim) module that can be integrated into pre-trained FR networks, transforming them into \hfr networks. The \caim block modulates intermediate feature maps, to adapt the style of the target modality effectively bridging the domain gap. Our proposed method allows for end-to-end training with a minimal number of paired samples. We extensively evaluate our approach on multiple challenging benchmarks, demonstrating superior performance compared to state-of-the-art methods. The source code and protocols for reproducing the findings will be made publicly available.},
pdf = {https://publications.idiap.ch/attachments/reports/2023/George_Idiap-RR-09-2023.pdf}
}
@TECHREPORT{George_Idiap-RR-12-2020,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
month = {6},
title = {Can Your Face Detector Do Anti-spoofing? Face Presentation Attack Detection with a Multi-Channel Face Detector},
type = {Idiap-RR},
number = {Idiap-RR-12-2020},
year = {2020},
institution = {Idiap},
abstract = {In a typical face recognition pipeline, the task of
the face detector is to localize the face region. However, the face
detector localizes regions that look like a face, irrespective of the
liveliness of the face, which makes the entire system susceptible
to presentation attacks. In this work, we try to reformulate the
task of the face detector to detect real faces, thus eliminating
the threat of presentation attacks. While this task could be
challenging with visible spectrum images alone, we leverage the
multi-channel information available from off the shelf devices
(such as color, depth, and infrared channels) to design a multi-
channel face detector. The proposed system can be used as a
live-face detector obviating the need for a separate presentation
attack detection module, making the system reliable in practice
without any additional computational overhead. The main idea
is to leverage a single-stage object detection framework, with
a joint representation obtained from different channels for the
PAD task. We have evaluated our approach in the multi-channel
WMCA dataset containing a wide variety of attacks to show the
effectiveness of the proposed framework.},
pdf = {https://publications.idiap.ch/attachments/reports/2019/George_Idiap-RR-12-2020.pdf}
}
@TECHREPORT{George_Idiap-RR-15-2020,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Convolutional neural network, Face Recognition, Presentation Attack Detection, Reproducible research, Unseen Attack Detection.},
projects = {Idiap, ODIN/BATL},
month = {7},
title = {Learning One Class Representations for Presentation Attack Detection using Multi-channel Convolutional Neural Networks},
type = {Idiap-RR},
number = {Idiap-RR-15-2020},
year = {2020},
institution = {Idiap},
abstract = {Face recognition has evolved as a widely used biometric modality. However, its vulnerability against presentation attacks poses a significant security threat. Though presentation attack detection (PAD) methods try to address this issue, they often fail in generalizing to unseen attacks. In this work, we propose a new framework for PAD using a one-class classifier, where the representation used is learned with a Multi-Channel Convolutional Neural Network (MCCNN). A novel loss function is introduced, which forces the network to learn a compact embedding for bonafide} class while being far from the representation of attacks. A one-class Gaussian Mixture Model is used on top of these embeddings for the PAD task. The proposed framework introduces a novel approach to learn a robust PAD system from bonafide and available (known) attack classes. This is particularly important as collecting bonafide data and simpler attacks are much easier than collecting a wide variety of expensive attacks. The proposed system is evaluated on the publicly available WMCA multi-channel face PAD database, which contains a wide variety of 2D and 3D attacks. Further, we have performed experiments with MLFP and SiW-M datasets using RGB channels only. Superior performance in unseen attack protocols shows the effectiveness of the proposed approach. Software, data, and protocols to reproduce the results are made available publicly.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/George_Idiap-RR-15-2020.pdf}
}
@ARTICLE{George_IEEETBIOM-2_2024,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
title = {From Modalities to Styles: Rethinking the Domain Gap in Heterogeneous Face Recognition},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/George_IEEETBIOM-2_2024.pdf}
}
@ARTICLE{George_IEEETBIOM_2024,
author = {George, Anjith and Ecabert, Christophe and Otroshi Shahreza, Hatef and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Idiap, IARPA GRAIL, SAFER, TRESPASS-ETN},
title = {EdgeFace : Efficient Face Recognition Model for Edge Devices},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/George_IEEETBIOM_2024.pdf}
}
@ARTICLE{George_IEEETIFS_2022,
author = {George, Anjith and Mohammadi, Amir and Marcel, S{\'{e}}bastien},
projects = {Idiap, HARDENING},
title = {Prepended Domain Transformer: Heterogeneous Face Recognition without Bells and Whistles},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2022},
abstract = {Heterogeneous Face Recognition (HFR) refers to matching face images captured in different domains, such as thermal to visible images (VIS), sketches to visible images, near-infrared to visible, and so on. This is particu- larly useful in matching visible spectrum images to images captured from other modalities. Though highly useful, HFR is challenging because of the domain gap between the source and target domain. Often, large-scale paired heterogeneous face image datasets are absent, preventing training models specifically for the heterogeneous task. In this work, we propose a surprisingly simple, yet, very effective method for matching face images across different sensing modalities. The core idea of the proposed approach is to add a novel neural network block called Prepended Domain Transformer (PDT) in front of a pre-trained face recognition (FR) model to address the domain gap. Retraining this new block with few paired samples in a contrastive learning setup was enough to achieve state- of-the-art performance in many HFR benchmarks. The PDT blocks can be retrained for several source-target combinations using the proposed general framework. The proposed approach is architecture agnostic, meaning they can be added to any pre-trained FR models. Further, the approach is modular and the new block can be trained with a minimal set of paired samples, making it much easier for practical deployment. The source code and protocols will be made available publicly.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/George_IEEETIFS_2022.pdf}
}
@INPROCEEDINGS{George_IJCB2023-2_2023,
author = {George, Anjith and Ecabert, Christophe and Otroshi Shahreza, Hatef and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {EFaR 2023: Efficient Face Recognition Competition},
booktitle = {IEEE International Joint Conference on Biometrics (IJCB 2023)},
year = {2023},
issn = {2474-9680},
isbn = {979-8-3503-3726-6},
abstract = {This paper presents the summary of the Efficient Face Recognition Competition (EFaR) held within the 2023 In- ternational Joint Conference on Biometrics (IJCB 2023). The competition received 17 submissions from 6 different teams. To drive further development of efficient face recog- nition models, the submitted solutions are ranked based on a weighted score of the achieved verification accura- cies on a diverse set of benchmarks, as well as the de- ployability given by the number of floating-point operations and model size. The evaluation of submissions is extended to bias, cross-quality, and large-scale recognition bench- marks. Overall, the paper gives an overview of the achieved performance values of the submitted solutions as well as a diverse set of baselines, the methodologies used to achieve lightweight and efficient face recognition solutions, and out- looks on possible techniques that are underrepresented in current solutions.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/George_IJCB2023-2_2023.pdf}
}
@INPROCEEDINGS{George_IJCB2023-3_2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {The Unconstrained Ear Recognition Challenge 2023: Maximizing Performance and Minimizing Bias},
booktitle = {IEEE International Joint Conference on Biometrics (IJCB 2023)},
year = {2023},
abstract = {The paper provides a summary of the 2023 Uncon- strained Ear Recognition Challenge (UERC), a benchmark- ing effort focused on ear recognition from images acquired in uncontrolled environments. The objective of the chal- lenge was to evaluate the effectiveness of current ear recog- nition techniques on a challenging ear dataset while ana- lyzing the techniques from two distinct aspects, i.e., veri- fication performance and bias with respect to specific de- mographic factors, i.e., gender and ethnicity. Seven re- search groups participated in the challenge and submitted a seven distinct recognition approaches that ranged from descriptor-based methods and deep-learning models to en- semble techniques that relied on multiple data representa- tions to maximize performance and minimize bias. A com- prehensive investigation into the performance of the submit- ted models is presented, as well as an in-depth analysis of bias and associated performance differentials due to differ- ences in gender and ethnicity. The results of the challenge suggest that a wide variety of models (e.g., transformers, convolutional neural networks, ensemble models) is capa- ble of achieving competitive recognition results, but also that all of the models still exhibit considerable performance differentials with respect to both gender and ethnicity. To promote further development of unbiased and effective ear recognition models, the starter kit of UERC 2023 together with the baseline model, and training and test data is made available from: http://ears.fri.uni-lj.si/.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/George_IJCB2023-3_2023.pdf}
}
@INPROCEEDINGS{George_IJCB2023_2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
title = {Bridging the Gap: Heterogeneous Face Recognition with Conditional Adaptive Instance Modulation},
booktitle = {IJCB},
year = {2023},
abstract = {Heterogeneous Face Recognition (HFR) aims to match face images acrossdifferent domains, such as thermal and visible spectra, expanding theapplicability of Face Recognition (FR) systems to challenging scenarios.However, the domain gap and limited availability of large-scale datasets in thetarget domain make training robust and invariant HFR models from scratchdifficult. In this work, we treat different modalities as distinct styles andpropose a framework to adapt feature maps, bridging the domain gap. Weintroduce a novel Conditional Adaptive Instance Modulation (CAIM) module thatcan be integrated into pre-trained FR networks, transforming them into HFRnetworks. The CAIM block modulates intermediate feature maps, to adapt thestyle of the target modality effectively bridging the domain gap. Our proposedmethod allows for end-to-end training with a minimal number of paired samples.We extensively evaluate our approach on multiple challenging benchmarks,demonstrating superior performance compared to state-of-the-art methods. Thesource code and protocols for reproducing the findings will be made publiclyavailable.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/George_IJCB2023_2023.pdf}
}
@INPROCEEDINGS{George_IJCB2025-2_2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, CARMEN},
additionalresearchprograms = {AI for Everyone},
title = {xEdgeFace: Efficient Cross-Spectral Face Recognition for Edge Devices},
booktitle = {International Joint Conference on Biometrics (IJCB 2025), IEEE.},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/George_IJCB2025-2_2025.pdf}
}
@INPROCEEDINGS{George_IJCB2025_2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, CARMEN},
additionalresearchprograms = {AI for Everyone},
title = {The Invisible Threat: Evaluating the Vulnerability of Cross-Spectral Face Recognition to Presentation Attacks},
booktitle = {International Joint Conference on Biometrics (IJCB 2025), IEEE.},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/George_IJCB2025_2025.pdf}
}
@INPROCEEDINGS{George_IJCB_2021,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {On the Effectiveness of Vision Transformers for Zero-shot Face Anti-Spoofing},
booktitle = {International Joint Conference on Biometrics (IJCB 2021)},
year = {2021},
crossref = {George_Idiap-RR-30-2020},
abstract = {The vulnerability of face recognition systems to presentation attacks has limited their application in security-critical scenarios. Automatic methods of detecting such malicious attempts are essential for the safe use of facial recognition technology. Although various methods have been suggested for detecting such attacks, most of them over-fit the training set and fail in generalizing to unseen attacks and environments. In this work, we use transfer learning from the vision transformer model for the zero-shot anti-spoofing task. The effectiveness of the proposed approach is demonstrated through experiments in publicly available datasets. The proposed approach outperforms the state-of-the-art methods in the zero-shot protocols in the HQ-WMCA and SiW-M datasets by a large margin. Besides, the model achieves a significant boost in cross-database performance as well.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/George_IJCB_2021.pdf}
}
@INPROCEEDINGS{George_IJCB_2024,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, IARPA GRAIL},
title = {Modality Agnostic Heterogeneous Face Recognition with Switch Style Modulators},
booktitle = {IEEE International Joint Conference on Biometrics},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/George_IJCB_2024.pdf}
}
@INCOLLECTION{George_SPRINGERNATURE_2024,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, HARDENING},
title = {Heterogeneous Face Recognition with Prepended Domain Transformers},
booktitle = {Face Recognition Across the Imaging Spectrum},
year = {2024},
publisher = {Springer},
pdf = {https://publications.idiap.ch/attachments/papers/2024/George_SPRINGERNATURE_2024.pdf}
}
@INCOLLECTION{George_SPRINGER_2021,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Multi-channel Face Presentation Attack Detection Using Deep Learning},
booktitle = {Deep Learning-Based Face Analytics},
year = {2021},
publisher = {Springer International Publishing},
pdf = {https://publications.idiap.ch/attachments/papers/2021/George_SPRINGER_2021.pdf}
}
@INCOLLECTION{George_SPRINGER_2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Robust Face Presentation Attack Detection with Multi-channel Neural Networks},
booktitle = {Handbook of Biometric Anti-Spoofing},
year = {2023},
publisher = {Springer},
crossref = {George_Idiap-RR-03-2022},
pdf = {https://publications.idiap.ch/attachments/papers/2023/George_SPRINGER_2023.pdf}
}
@ARTICLE{George_TIFS_2019,
author = {George, Anjith and Mostaani, Zohreh and Geissbuhler, David and Nikisins, Olegs and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Biometric Face Presentation Attack Detection with Multi-Channel Convolutional Neural Network},
journal = {IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY},
year = {2019},
abstract = {Face recognition is a mainstream biometric authentication method. However, vulnerability to presentation attacks (a.k.a spoofing) limits its usability in unsupervised applications. Even though there are many methods available for tackling presentation attacks (PA), most of them fail to detect sophisticated attacks such as silicone masks.
As the quality of presentation attack instruments improves over time, achieving reliable PA detection with visual spectra alone remains very challenging. We argue that analysis in multiple channels might help to address this issue. In this context, we propose a multi-channel Convolutional Neural Network based approach for presentation attack detection (PAD).
We also introduce the new Wide Multi-Channel presentation Attack (WMCA) database for face PAD which contains a wide variety of 2D and 3D presentation attacks for both impersonation and obfuscation attacks. Data from different channels such as color, depth, near-infrared and thermal are available to advance the research in face PAD. The proposed method was compared with feature-based approaches and found to outperform the baselines achieving an ACER of 0.3\% on the introduced dataset. The database and the software to reproduce the results are made available publicly.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/George_TIFS_2019.pdf}
}
@ARTICLE{George_TIFS_2020,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Learning One Class Representations for Face Presentation Attack Detection using Multi-channel Convolutional Neural Networks},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2020},
abstract = {Face recognition has evolved as a widely used biometric modality. However, its vulnerability against presentation attacks poses a significant security threat. Though presentation attack detection (PAD) methods try to address this issue, they often fail in generalizing to unseen attacks. In this work, we propose a new framework for PAD using a one-class classifier, where the representation used is learned with a Multi-Channel Convolutional Neural Network (MCCNN). A novel loss function is introduced, which forces the network to learn a compact embedding for bonafide class while being far from the representation of attacks. A one-class Gaussian Mixture Model is used on top of these embeddings for the PAD task. The proposed framework introduces a novel approach to learn a robust PAD system from bonafide and available (known) attack classes. This is particularly important as collecting bonafide data and simpler attacks are much easier than collecting a wide variety of expensive attacks. The proposed system is evaluated on the publicly available WMCA multi-channel face PAD database, which contains a wide variety of 2D and 3D attacks. Further, we have performed experiments with MLFP and SiW-M datasets using RGB channels only. Superior performance in unseen attack protocols shows the effectiveness of the proposed approach. Software, data, and protocols to reproduce the results are made available publicly.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/George_TIFS_2020.pdf}
}
@INPROCEEDINGS{George_WACV2025_2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, SAFER},
title = {Digi2Real: Bridging the Realism Gap in Synthetic Data Face Recognition via Foundation Models},
booktitle = {2025 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/George_WACV2025_2025.pdf}
}
@INPROCEEDINGS{Goncalves_BIOSIG_2017,
author = {Goncalves, Andre{\'{e}} R. and Korshunov, Pavel and Violato, Ricardo P. V. and Sim{\~{o}}es, Fl{\'{a}}vio O. and Marcel, S{\'{e}}bastien},
editor = {Br{\"{o}}mme, A. and Busch, Christoph and Dantcheva, A. and Rathgeb, C. and Uhl, A.},
projects = {Idiap, SWAN, Tesla},
month = sep,
title = {On the Generalization of Fused Systems in Voice Presentation Attack Detection},
booktitle = {16th International Conference of the Biometrics Special Interest Group},
year = {2017},
address = {Darmstadt, Germany},
abstract = {This paper describes presentation attack detection systems developed for the Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof 2017). The submitted systems, using calibration and score fusion techniques, combine different sub-systems (up to 18), which are based on eight state of the art features and rely on Gaussian mixture models and feed-forward neural network classifiers. The systems achieved the top five performances in the competition. We present the proposed systems and analyze the calibration and fusion strategies employed. To assess the systems' generalization capacity, we evaluated it on an unrelated larger database recorded in Portuguese language, which is different from the English language used in the competition. These extended evaluation results show that the fusion-based system, although successful in the scope of the evaluation, lacks the ability to accurately discriminate genuine data from attacks in unknown conditions, which raises the question on how to assess the generalization ability of attack detection systems in practical application scenarios.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Goncalves_BIOSIG_2017.pdf}
}
@INPROCEEDINGS{Gunther_BEFIT2012_2012,
author = {G{\"{u}}nther, Manuel and Wallace, Roy and Marcel, S{\'{e}}bastien},
editor = {Fusiello, Andrea and Murino, Vittorio and Cucchiara, Rita},
keywords = {Biometrics, Face Recognition, Open Source, Reproducible research},
projects = {Idiap, FP 7},
month = oct,
title = {An Open Source Framework for Standardized Comparisons of Face Recognition Algorithms},
booktitle = {Computer Vision - ECCV 2012. Workshops and Demonstrations},
series = {Lecture Notes in Computer Science},
volume = {7585},
year = {2012},
pages = {547-556},
publisher = {Springer Berlin},
location = {Heidelberg},
organization = {Idiap Research Institute},
address = {Rue Marconi 19, CH - 1920 Martigny, Switzerland},
note = {The source code to re-generate the results of this paper can be downloaded from the URL below},
isbn = {978-3-642-33884-7},
url = {http://pypi.python.org/pypi/xfacereclib.paper.BeFIT2012},
doi = {10.1007/978-3-642-33885-4_55},
crossref = {Gunther_Idiap-RR-29-2012},
abstract = {In this paper we introduce the facereclib, the first software library that allows to compare a variety of face recognition algorithms on most of the known facial image databases and that permits rapid prototyping of novel ideas and testing of meta-parameters of face recognition algorithms. The facereclib is built on the open source signal processing and machine learning library Bob. It uses well-specified face recognition protocols to ensure that results are comparable and reproducible. We show that the face recognition algorithms implemented in Bob as well as third party face recognition libraries can be used to run face recognition experiments within the framework of the facereclib. As a proof of concept, we execute four different state-of-the-art face recognition algorithms: local Gabor binary pattern histogram sequences (LGBPHS), Gabor graph comparisons with a Gabor phase based similarity measure, inter-session variability modeling (ISV) of DCT block features, and the linear discriminant analysis on two different color channels (LDA-IR) on two different databases: The Good, The Bad, & The Ugly, and the BANCA database, in all cases using their fixed protocols. The results show that there is not one face recognition algorithm that outperforms all others, but rather that the results are strongly dependent on the employed database.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Gunther_BEFIT2012_2012.pdf}
}
@INPROCEEDINGS{Gunther_ICB2013_2013,
author = {G{\"{u}}nther, Manuel and Costa-Pazo, Artur and Ding, Changxing and Boutellaa, Elhocine and Chiachia, Giovani and Zhang, Honglei and de Assis Angeloni, Marcus and Struc, Vitomir and Khoury, Elie and Vazquez-Fernandez, Esteban and Tao, Dacheng and Bengherabi, Messaoud and Cox, David and Kiranyaz, Serkan and de Freitas Pereira, Tiago and Zganec-Gros, Jerneja and Argones-R{\'{u}}a, Enrique and Pinto, Nicolas and Gabbouj, Moncef and Sim{\~{o}}es, Fl{\'{a}}vio and Dobrisek, Simon and Gonz{\'{a}}lez-Jim{\'{e}}nez, Daniel and Rocha, Anderson and Uliani Neto, M{\'{a}}rio and Pavesic, Nikola and Falc{\~{a}}o, Alexandre and Violato, Ricardo and Marcel, S{\'{e}}bastien},
projects = {Idiap, BBfor2, BEAT},
month = jun,
title = {The 2013 Face Recognition Evaluation in Mobile Environment},
booktitle = {The 6th IAPR International Conference on Biometrics},
year = {2013},
crossref = {Gunther_Idiap-RR-36-2013},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Gunther_ICB2013_2013.pdf}
}
@TECHREPORT{Gunther_Idiap-RR-13-2017,
author = {G{\"{u}}nther, Manuel and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {BBfor2},
month = {4},
title = {2D Face Recognition: An Experimental and Reproducible Research Survey},
type = {Idiap-RR},
number = {Idiap-RR-13-2017},
year = {2017},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Gunther_Idiap-RR-13-2017.pdf}
}
@INCOLLECTION{Gunther_SPRINGER_2016,
author = {G{\"{u}}nther, Manuel and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
editor = {Bourlai, Thirimachos},
keywords = {Face Recognition, Reproducible research},
projects = {BBfor2, BEAT},
month = feb,
title = {Face Recognition in Challenging Environments: An Experimental and Reproducible Research Survey},
booktitle = {Face Recognition Across the Imaging Spectrum},
edition = {1},
year = {2016},
publisher = {Springer},
pdf = {https://publications.idiap.ch/attachments/papers/2016/Gunther_SPRINGER_2016.pdf}
}
@ARTICLE{Hadid_SIGPRO_2015,
author = {Hadid, Abdenour and Evans, Nicholas and Marcel, S{\'{e}}bastien and Fierrez, Julian},
keywords = {Anti-spoofing, Biometrics, Spoofing},
projects = {TABULA RASA, BEAT},
title = {Biometrics systems under spoofing attack: an evaluation methodology and lessons learned},
journal = {IEEE Signal Processing Magazine},
year = {2015},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Hadid_SIGPRO_2015.pdf}
}
@TECHREPORT{heusch-com-05-03,
author = {Heusch, Guillaume and Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Lighting Normalization Algorithms for Face Verification},
type = {Idiap-Com},
number = {Idiap-Com-03-2005},
year = {2005},
institution = {IDIAP},
abstract = {In this report, we address the problem of face verification across illumination, since it has been identified as one of the major factor degrading the performance of face recognition systems. First, a brief overview of face recognition together with its main challenges is made, before reviewing state-of-the-art approaches to cope with illumination variations. We then present investigated approaches, which consists in applying a pre-processing step to the face images, and we also present the underlying theory. Namely, we will study the effect of various photometric normalization algorithms on the performance of a system based on local feature extraction and generative models (Gaussian Mixture Models). Studied algorithms include the Multiscale Retinex, as well as two state-of-the-art approaches: the Self Quotient Image and an anisotropic diffusion based normalization. This last involves the resolution of large sparse system of equations, and hence different approaches to solve such problems are described, including the efficient multigrid framework. Performances of the normalization algorithms are assessed with the challenging BANCA database and its realistic protocols. Conducted experiments showed significant improvements in terms of verification error rates and are comparable to other state-of-the-art face verification systems on the same database.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/heusch-2005.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/heusch-2005.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{Heusch-rr-05-45,
author = {Heusch, Guillaume and Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Efficient Diffusion-based Illumination Normalization for Face Verification},
type = {Idiap-RR},
number = {Idiap-RR-46-2005},
year = {2005},
institution = {IDIAP},
abstract = {In this paper, the problem of face verification across illumination is addressed. In order to cope with different lighting conditions, a preprocessing step is applied to the face image so as to make it independent on the illumination conditions. The illuminant invariant representation of the image is obtained by first applying an anisotropic diffusion process to the original image. Hence, it implies the numerical resolution of an elliptic partial differential equation on a large grid: the image. So, a comparison is performed on two methods to resolve such diffusion problems, namely the Gauss-Seidel relaxation and the Multigrid V-cycle. The preprocessing algorithm with its different resolution schemes is applied prior to the task of face verification. Experiments conducted on the challenging BANCA database showed a significant improvement in terms of face verification error rate, while staying computationally efficient.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/rr05-46.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/rr05-46.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{heusch:AFGR:2006,
author = {Heusch, Guillaume and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Local Binary Patterns as an Image Preprocessing for Face Authentication},
booktitle = {{IEEE} Int. Conf. on Automatic Face and Gesture Recognition ({AFGR})},
year = {2006},
note = {IDIAP-RR 05-76},
crossref = {heusch:rr05-76},
abstract = {One of the major problem in face authentication systems is to deal with variations in illumination. In a \mbox{realistic} scenario, it is very likely that the lighting conditions of the probe image does not correspond to those of the gallery image, hence there is a need to handle such variations. In this work, we present a new preprocessing algorithm based on Local Binary Patterns (LBP): a texture representation is derived from the input face image before being forwarded to the classifier. The efficiency of the proposed approach is empirically demonstrated using both an appearance-based (LDA) and a feature-based (HMM) face authentication systems on two databases: BANCA and XM2VTS (with its darkened set). Conducted experiments show a significant improvement in terms of verification error rates and compare to results obtained with state-of-the-art preprocessing techniques.},
pdf = {https://publications.idiap.ch/attachments/papers/2006/heusch-AFGR-2006.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2006/heusch-AFGR-2006.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{heusch:ICB:2007,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Authentication with {S}alient {L}ocal {F}eatures and {S}tatic {B}ayesian Network},
booktitle = {I{EEE} / {IAPR} Intl. {C}onf. On {B}iometrics ({ICB})},
year = {2007},
note = {IDIAP-RR 07-04},
crossref = {heusch:rr07-04},
abstract = {In this paper, the problem of face authentication using salient facial features together with statistical generative models is adressed. Actually, classical generative models, and Gaussian Mixture Models in particular make strong assumptions on the way observations derived from face images are generated. Indeed, systems proposed so far consider that local observations are independent, which is obviously not the case in a face. Hence, we propose a new generative model based on Bayesian Networks using only salient facial features. We compare it to Gaussian Mixture Models using the same set of observations. Conducted experiments on the BANCA database show that our model is suitable for the face authentication task, since it outperforms not only Gaussian Mixture Models, but also classical appearance-based methods, such as Eigenfaces and Fisherfaces.},
pdf = {https://publications.idiap.ch/attachments/papers/2007/heusch-ICB-2007.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2007/heusch-ICB-2007.ps.gz},
ipdmembership={vision},
}
@ARTICLE{Heusch_ARXIV_2017,
author = {Heusch, Guillaume and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap, COHFACE},
month = sep,
title = {A reproducible study on remote heart rate measurement},
journal = {arXiv},
year = {2017},
url = {https://arxiv.org/abs/1709.00962},
abstract = {This paper studies the problem of reproducible research in remote photoplethysmography (rPPG).
Most of the work published in this domain is assessed on privately-owned databases, making
it difficult to evaluate proposed algorithms in a standard and principled manner.
As a consequence, we present a new, publicly available database containing a
relatively large number of subjects recorded under two different lighting conditions.
Also, three state-of-the-art rPPG algorithms from the literature were selected, implemented and
released as open source free software. After a thorough, unbiased experimental evaluation in various
settings, it is shown that none of the selected algorithms is precise enough to be used in a real-world scenario.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Heusch_ARXIV_2017.pdf}
}
@INPROCEEDINGS{Heusch_BTAS_2018,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {ODIN/BATL},
title = {Pulse-based Features for Face Presentation Attack Detection},
booktitle = {Proceedings of BTAS 2018, special session on Image and Video Forensics in Biometrics},
year = {2018},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Heusch_BTAS_2018.pdf}
}
@INPROCEEDINGS{Heusch_ICB_2009,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap, GMFace},
title = {Bayesian Networks to Combine Intensity and Color Information in Face Recognition},
booktitle = {International Conference on Biometrics},
series = {Lectures Notes on Computer Science},
volume = {5558},
year = {2009},
publisher = {Springer},
crossref = {Heusch_Idiap-RR-27-2009},
pdf = {https://publications.idiap.ch/attachments/papers/2009/Heusch_ICB_2009.pdf}
}
@ARTICLE{Heusch_IEEETBIOM_2019,
author = {Heusch, Guillaume and de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {A Comprehensive Experimental and Reproducible Study on Selfie Biometrics in Multistream and Heterogeneous Settings},
journal = {IEEE Transactions on Biometrics, Behavior and Identity Science},
year = {2019},
url = {https://ieeexplore.ieee.org/document/8759986},
doi = {10.1109/TBIOM.2019.2927692},
crossref = {Heusch_Idiap-RR-09-2019},
abstract = {This contribution presents a new database to address current challenges in face recognition. It contains face video sequences of 75 individuals acquired either through a laptop webcam or when mimicking the front-facing camera of a smartphone. Sequences have been acquired with a device allowing to record visual, near-infrared and depth data at the same time. Recordings have been made across three sessions with different, challenging illumination conditions and variations in pose. Together with the database, several experimental protocols are provided and correspond to real world scenarios, when a mismatch in conditions between enrollment and probe images occurs. A comprehensive set of baseline experiments using publicly available baseline algorithms show that extreme illumination conditions and pose variations are remaining issues. However, the usage of different data domains -and their fusion -allows to mitigate such variation. Finally, experiments on heterogeneous face recognition are also presented using a state-of-the-art model based on deep neural networks, and showed better performance. When applied to other tasks, this model proved to surpass all existing baselines as well. The data, as well as the code to reproduce all experiments are made publicly available to help foster research in selfie biometrics using latest imaging devices.}
}
@ARTICLE{Heusch_IVC_2009,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap, GMFace},
title = {A novel statistical generative model dedicated to face recognition},
journal = {Image & Vision Computing},
year = {2009},
note = {in press},
crossref = {heusch:rr07-39},
pdf = {https://publications.idiap.ch/attachments/papers/2009/Heusch_IVC_2009.pdf}
}
@INCOLLECTION{Heusch_SPRINGER_2019,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Fierrez, Julian and Evans, Nicholas},
projects = {ODIN/BATL},
month = apr,
title = {Remote Blood Pulse Analysis for Face Presentation Attack Detection},
booktitle = {Handbook of Biometric Anti-Spoofing},
edition = {2nd},
series = {Advances in Computer Vision and Pattern Recognition},
chapter = {10},
year = {2019},
publisher = {Springer},
isbn = {978-3-319-92627-8},
url = {https://www.springer.com/us/book/9783319926261},
abstract = {In this chapter, the usage of Remote Photoplethysmography (rPPG\index{Remote Photoplethysmography (rPPG)}) as a mean for
face presentation attack detection is investigated. Remote
photoplethysmography consists in retrieving the heart-rate of a subject from a
video sequence containing some skin, and recorded at a distance. To get a
pulse signal, such methods take advantage of subtle color variation on skin
pixels due to the blood flowing through vessels. Since the inferred pulse
signal gives information on the liveness of the recorded subject, it can be
used for biometric presentation attack detection (PAD\index{Presentation Attack Detection (PAD)}). Inspired by work made
for speaker presentation attack detection, we propose to use long-term
spectral statistical features of the pulse signal to discriminate real
accesses from attack attempts. A thorough experimental evaluation, with
different rPPG and classification algorithms is carried on four publicly
available datasets containing a wide range of face presentation attacks.
Obtained results suggest that the proposed features are effective for this
task, and we empirically show that our approach performs better than
state-of-the-art rPPG-based presentation attack detection algorithms.}
}
@ARTICLE{Heusch_TBIOM_2020,
author = {Heusch, Guillaume and George, Anjith and Geissbuhler, David and Mostaani, Zohreh and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Deep Models and Shortwave Infrared Information to Detect Face Presentation Attacks},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2020},
abstract = {This paper addresses the problem of face presentation attack detection using different image modalities. In particular, the usage
of short wave infrared (SWIR) imaging is considered.
Face presentation attack detection is performed using recent models based on Convolutional Neural Networks
using only carefully selected SWIR image differences as input. Conducted experiments show superior performance over similar models
acting on either color images or on a combination of different modalities (visible, NIR, thermal and depth), as well as on a SVM-based classifier
acting on SWIR image differences. Experiments have been carried on a new public and freely available database,
containing a wide variety of attacks.
Video sequences have been recorded thanks to several
sensors resulting in 14 different streams in the visible, NIR, SWIR and thermal spectra, as well as depth data.
The best proposed approach is able to almost perfectly detect all impersonation attacks while ensuring low \bona classification
errors. On the other hand, obtained results show that obfuscation attacks are more difficult to detect. We hope that the proposed
database will foster research on this challenging problem.
Finally, all the code and instructions to reproduce presented
experiments is made available to the research community.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Heusch_TBIOM_2020.pdf}
}
@ARTICLE{Huang_IEEE-TIFS_2025,
author = {Huang, Junduan and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien and Kang, Wenxiong},
keywords = {authentication, Biometrics, Cameras, Convolutional Neural Networks, dynamic Transformer, Feature extraction, Finger vein, Fingers, full-view, Imaging, redundancy, Taxonomy, transformers, Vein recognition},
projects = {Idiap, Biometrics Center},
mainresearchprogram = {AI for Everyone},
title = {Study of Full-View Finger Vein Biometrics on Redundancy Analysis and Dynamic Feature Extraction},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2025},
issn = {1556-6021},
url = {https://ieeexplore.ieee.org/document/11236466},
doi = {10.1109/TIFS.2025.3630891},
abstract = {As a biometric trait drawing increasing attention, finger vein (FV) has been studied from many perspectives. One promising new direction in FV biometrics research is full-view FV biometrics, where multiple images, covering the entire surface of the presented finger, are captured. Full-view FV biometrics presents two main problems: increased computational load, and low performance-to-cost ratio for some views/regions. Both problems are related to the inherent redundancy in vascular information available in full-view FV images. In this work, we address this redundancy issue in full-view FV biometrics. Firstly, we propose a straightforward FV redundancy analysis (FVRA) method for quantifying the information redundancy in FV images. Our analysis shows that the redundancy ratio of full-view FV images is up to 83\%-87\%. Then, we propose a novel feature extraction model, named FV dynamic Transformer (FDT), whose architecture is configured based on the redundancy analysis results. The FDT focuses on both local (single-view) information as well as global (full view) information at different processing stages. Both stages provide the advantage of de-redundancy and noise avoidance. Additionally, the end-to-end architecture simplifies the full-view FV biometrics pipeline by enabling the direct, simultaneous processing of multiple input images, thus consolidating multiple steps into one. A series of rigorous experiments is conducted to evaluate the effectiveness of the proposed methods. Experimental results show that the proposed FDT achieves state of the art authentication performance on the MFFV-N dataset, yielding an EER of 0.97\% on the development set and an HTER of 1.84\% on the test set under the balanced protocol and EER criterion. The cross-domain generalization capability of FDT is also demonstrated on the LFMB-3DFB dataset, where it achieves an EER of 7.24\% and an HTER of 7.34\% under the same protocol and criterion. Code for the proposed methods can be access via: https://github.com/SCUT-BIP-Lab/FDT.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Huang_IEEE-TIFS_2025.pdf}
}
@ARTICLE{Huang_IEEETCSVT_2024,
author = {Huang, Junduan and Li, Zifeng and Bhattacharjee, Sushil and Kang, Wenxiong and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Finger vein, Full-view Authentication, Miura-Match, Multi-view, Vein recognition},
projects = {Idiap, Biometrics Center},
title = {Mirror-based Full-View Finger Vein Authentication with Illumination Adaptation},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2024},
doi = {DOI: 10.1109/TCSVT.2024.3490581},
abstract = {Full-view finger vein (FV) biometrics systems capture multiple FV images of the presented finger ensuring that the entire surface of the finger is covered. Existing full-view FV systems suffer from three common problems: large device size, high cost for multi-camera system, and sub-optimal illumination in the recorded FV images. To address the problem of device size, we propose a novel Mirror-based Full-view FV (MFFV) capture device. The MFFV device has a compact size by using mirror-reflection approach. We reduce the cost of the device by using low-cost components, in particular, consumer-grade cameras. To address the problems of lower-quality images captured by such cameras and obtain optimally illuminated FV images, we propose a two-step approach. The first step is a Multi-illumination Intensities FV (MIFV) capture strategy, which capture the FV image set with varying illumination intensities. In the second step, a FV illumination adaptation (FVIA) algorithm is proposed to select the optimally illuminated FV image from the MIFV image set. Using the proposed MFFV device, we collect a comprehensive dataset, namely MFFV dataset, along with reproducible baseline FV authentication results for both single-view and full-view FV. Our experimental results demonstrate that the MIFV capture strategy as well as the FVIA algorithm can effectively improve the authentication performance, and that the full-view FV authentication is significantly superior than the single-view FV authentication. The source-code and dataset for reproducing our experimental results are publicly available (https://github.com/SCUT-BIP-Lab/MFFV).}
}
@ARTICLE{I.Mantasari_IETBMT_2014,
author = {Mandasari, Miranti I. and G{\"{u}}nther, Manuel and Wallace, Roy and Saedi, Rahim and Marcel, S{\'{e}}bastien and Van Leeuwen, David},
keywords = {calibration performance evaluation, calibration performance metric, categorical calibration, face recognition system, Inter-session Variability Modelling, likelihood ratio interpretation, linear score transformation, linearly calibrated face recognition scores, mobile biometrics, speaker recognition field, surveillance camera face databases},
projects = {BBfor2},
month = feb,
title = {Score Calibration in Face Recognition},
journal = {IET Biometrics},
year = {2014},
pages = {1-11},
issn = {2047-4938},
url = {http://digital-library.theiet.org/content/journals/10.1049/iet-bmt.2013.0066},
doi = {10.1049/iet-bmt.2013.0066},
crossref = {I.Mantasari_Idiap-RR-01-2014},
abstract = {This paper presents an evaluation of verification and calibration performance of a face recognition system based on inter-session variability modeling. As an extension to the calibration through linear transformation of scores, categorical calibration is introduced as a way to include additional information of images to calibration. The cost of likelihood ratio, which is a well-known measure in the speaker recognition field, is used as a calibration performance metric. Evaluated on the challenging MOBIO and SCface databases, the results indicate that through linear calibration the scores produced by the face recognition system can be less misleading in its likelihood ratio interpretation. In addition, it is shown through the categorical calibration experiments that calibration can be used not only to assure likelihood ratio interpretation of scores, but also improving the verification performance of face recognition system.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/I.Mantasari_IETBMT_2014.pdf}
}
@INPROCEEDINGS{Igene_IJCB_2024,
author = {Igene, Lambert and Hossain, Afzal and Chowdhury, Mohammad Zahir Uddin and Rezaie, Humaira and Rollins, Ayden and Dykes, Jesse and Vijaykumar, Rahul and Komaty, Alain and Marcel, S{\'{e}}bastien and Schuckers, Stephanie and Tapia, Juan E. and Aravena, Carlos and Schulz, Daniel and Adami, Banafsheh and Karimian, Nima and Nunes, Diogo and Marcos, Jo{\~{a}}o and Gon{\c c}alves, Nuno and Sikosek, Lovro and Batagelj, Borut and Alenin, Aleksandr and Alkhaddour, Alhasan and Pimenov, Anton and Tregubov, Artem and Avdonin, Igor and Kazantsev, Maxim and Pozigun, Mikhail and Pryadchenko, Vasiliy and Schei, Nima and Pabon, David and Tiedemann, Manuela},
projects = {Idiap, SOTERIA},
title = {Face Liveness Detection Competition (LivDet-Face) - 2024},
booktitle = {IEEE International Joint Conference on Biometrics},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Igene_IJCB_2024.pdf}
}
@INPROCEEDINGS{Ivanova_TEA2018_2018,
author = {Ivanova, Malinka and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien and Rozeva, Anna and Durcheva, Mariana},
projects = {Idiap, Tesla},
month = dec,
title = {Enhancing Trust in eAssessment - the TeSLA System Solution},
booktitle = {Technology Enhanced Assessment Conference.},
year = {2018},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Ivanova_TEA2018_2018.pdf}
}
@INPROCEEDINGS{Jain_MMSP_2021,
author = {Jain, Anubhav and Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = oct,
title = {Improving Generalization of Deepfake Detection by Training for Attribution},
booktitle = {International Workshop on Multimedia Signal Processing},
year = {2021},
abstract = {Recent advances in automated video and audio editing tools, generative adversarial networks (GANs), and social media allow the creation and fast dissemination of high-quality tampered videos, which are commonly called deepfakes. Typically, in these videos, a face is automatically swapped with the face of another person. The simplicity and accessibility of tools for generating deepfakes pose a significant technical challenge for their detection and filtering. In response to the threat, several large datasets of deepfake videos and various methods to detect them were proposed recently. However, the proposed methods suffer from the problem of over-fitting on the training data and the lack of generalization across different databases and generative approaches. In this paper, we approach deepfake detection by solving the related problem of attribution, where the goal is to distinguish each separate type of a deepfake attack. Using publicly available datasets from Google and Jigsaw, FaceForensics++, Celeb-DF, DeepfakeTIMIT, and our own large database DF-Mobio, we demonstrate that an XceptionNet and EfficientNet based models trained for an attribution task generalize better to unseen deepfakes and different datasets, compared to the same models trained for a typical binary classification task. We also demonstrate that by training for attribution with a triplet-loss, the generalization in cross-database scenario improves even more, compared to the binary system, while the performance on the same database degrades only marginally.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/Jain_MMSP_2021.pdf}
}
@TECHREPORT{Johansson_Idiap-RR-07-2011,
author = {Johansson, Niklas and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {3},
title = {On-line unsupervised adaptation for face verification using Gaussian Mixture Models with multiple user models},
type = {Idiap-RR},
number = {Idiap-RR-07-2011},
year = {2011},
institution = {Idiap},
abstract = {In this paper, we present an initial study of on-line
unsupervised adaptation for face verification. To the authors{\^{a}}€™
knowledge this is the first study of this type. The key contributions
consist of four test scenarios for the BANCA database as
well as two novel adaptation strategies that use multiple user
models. We show that by using multiple user models for each
user, we can perform on-line unsupervised adaptation with a
consistent increase in verification performance. Finally, we show
that one of the proposed strategies performs better, or as well as
the baseline, in four test scenarios at all adaptation thresholds
evaluated. This suggests that this strategy is robust against both
changing conditions and inexact thresholds.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Johansson_Idiap-RR-07-2011.pdf}
}
@TECHREPORT{just-rr03-79,
author = {Just, Agn{\`{e}}s and Marcel, S{\'{e}}bastien and Bernier, O. and Viallet, J. E.},
projects = {Idiap},
title = {Reconnaissance de gestes 3D bi-manuels},
type = {Idiap-RR},
number = {Idiap-RR-79-2003},
year = {2003},
institution = {IDIAP},
abstract = {Cet article pr{\'{e}}sente une base de seize gestes dynamiques obtenus par suivi de diff{\'{e}}rentes parties color{\'{e}}es du corps, suivi r{\'{e}}alis{\'{e}} en temps r{\'{e}}el par l'algorithme EM. Ces gestes r{\'{e}}alis{\'{e}}s avec une ou deux mains sont appris et reconnus avec des HMM. Les pr{\'{e}}traitements effectu{\'{e}}s sur cette base de gestes ainsi que les r{\'{e}}sultats de classification sont pr{\'{e}}sent{\'{e}}s.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-79.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-79.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{just-rr05-24,
author = {Just, Agn{\`{e}}s and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Two-Handed Gesture Recognition},
type = {Idiap-RR},
number = {Idiap-RR-24-2005},
year = {2005},
institution = {IDIAP},
abstract = {Nowadays, computer interaction is mostly done using dedicated devices. But gestures are an easy mean of expression between humans that could be used to communicate with computers in a more natural manner. Most of the current research on hand gesture recognition for Human-Computer Interaction deals with one-handed gestures. But two-handed gestures can provide more efficient and easy to interact with user interfaces. It is particularly the case with two-handed gestures we do in the physical world, such as gestures to manipulate objects. It would be very valuable to permit to the user to interact with virtual objects in the same way that he/she interacts with physical ones. This paper presents a two-handed gesture database to manipulate virtual objects on the screen (mostly rotations) and some recognition experiment using Hidden Markov Models (HMMs). The results obtained with this state-of-the-art algorithm are really encouraging. These gestures would improve the interaction performance between the user and virtual reality applications.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/rr05-24.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/rr05-24.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{just2004,
author = {Just, Agn{\`{e}}s and Bernier, O. and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Recognition of Isolated Complex Mono- and Bi-Manual 3D Hand Gestures},
booktitle = {Proc. of the sixth International Conference on Automatic Face and Gesture Recognition},
year = {2004},
note = {IDIAP-RR 03-63},
crossref = {just2003},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-63.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-63.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{just2004:rr-04-39,
author = {Just, Agn{\`{e}}s and Bernier, O. and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {HMM and IOHMM for the Recognition of Mono- and Bi-Manual 3D Hand Gestures},
type = {Idiap-RR},
number = {Idiap-RR-39-2004},
year = {2004},
institution = {IDIAP},
abstract = {In this paper, we address the problem of the recognition of isolated complex mono- and bi-manual hand gestures. In the proposed system, hand gestures are represented by the 3D trajectories of blobs obtained by tracking colored body parts. In this paper, we study the results obtained on a complex database of mono- and bi-manual gestures. These results are obtained by using Input/Output Hidden Markov Model (IOHMM,',','),
implemented within the framework of an open source machine learning library, and are compared to Hidden Markov Model (HMM).},
pdf = {https://publications.idiap.ch/attachments/reports/2004/rr04-39.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2004/rr04-39.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{just:afgr:2006,
author = {Just, Agn{\`{e}}s and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Hand Posture Classification and Recognition using the Modified Census Transform},
booktitle = {{IEEE} Int. Conf. on Automatic Face and Gesture Recognition ({AFGR})},
year = {2006},
note = {IDIAP-RR 06-02},
crossref = {just:rr06-02},
abstract = {Developing new techniques for human-computer interaction is very challenging. Vision-based techniques have the advantage of being unobtrusive and hands are a natural device that can be used for more intuitive interfaces. But in order to use hands for interaction, it is necessary to be able to recognize them in images. In this paper, we propose to apply to the hand posture classification and recognition tasks an approach that has been successfully used for face detection~\cite{Froba04}. The features are based on the Modified Census Transform and are illumination invariant. For the classification and recognition processes, a simple linear classifier is trained, using a set of feature lookup-tables. The database used for the experiments is a benchmark database in the field of posture recognition. Two protocols have been defined. We provide results following these two protocols for both the classification and recognition tasks. Results are very encouraging.},
pdf = {https://publications.idiap.ch/attachments/papers/2006/just-afgr-2006.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2006/just-afgr-2006.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{keomany:rr06-07,
author = {Keomany, Jean and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Active Shape Models Using Local Binary Patterns},
type = {Idiap-RR},
number = {Idiap-RR-07-2006},
year = {2006},
institution = {IDIAP},
abstract = {This report addresses the problem of locating facial features in images of frontal faces taken under different lighting conditions. The well-known Active Shape Model method proposed by Cootes {\it et al.} is extended in order to improve its robustness to illumination changes. For that purpose, we introduce the use of Local Binary Patterns (LBP). Three different approaches combining ASM with LBP are presented: profile-based LBP-ASM, square-based LBP-ASM and divided-square-based LBP-ASM. Experiments performed on the standard and darkened image sets of the XM2VTS database demonstrate that the divided-square-based LBP-ASM gives superior performance compared to the state-of-the-art ASM. It achieves more accurate results and fails less frequently.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/keomany-idiap-rr-06-07.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/keomany-idiap-rr-06-07.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{Khoury_BTFS_2013,
author = {Khoury, Elie and G{\"{u}}nther, Manuel and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, BBfor2, SNSF-LOBI},
month = oct,
title = {On the Improvements of Uni-modal and Bi-modal Fusions of Speaker and Face Recognition for Mobile Biometrics},
booktitle = {Biometric Technologies in Forensic Science},
year = {2013},
location = {Nijmegen, The Netherlands},
crossref = {Khoury_Idiap-RR-35-2013},
abstract = {The MOBIO database provides a challenging test-bed for speaker and face recognition systems because it includes voice and face samples as they would appear in forensic scenarios.
In this paper, we investigate uni-modal and bi-modal multi-algorithm fusion using logistic regression.
The source speaker and face recognition systems were taken from the 2013 speaker and face recognition evaluations that were held in the context of the last International Conference on Biometrics (ICB-2013).
Using the unbiased MOBIO protocols, the employed evaluation measures are the equal error rate (EER), the half-total error rate (HTER) and the detection error trade-off (DET).
The results show that by uni-modal algorithm fusion, the HTER's of the speaker recognition system are reduced by around 35\%, and of the face recognition system by between 15\% and 20\%.
Bi-modal fusion drastically boosts recognition by a relative gain of 65\% - 70\% of performance compared to the best uni-modal system.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Khoury_BTFS_2013.pdf}
}
@INPROCEEDINGS{Khoury_ICASSP_2014,
author = {Khoury, Elie and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
keywords = {bob, Open Source, speaker recognition, toolbox},
projects = {SNSF-LOBI, BEAT},
month = may,
title = {SPEAR: An open source toolbox for speaker recognition based on Bob},
booktitle = {Proceedings of the 39th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
year = {2014},
pages = {1655 - 1659},
issn = {1520-6149},
url = {https://pypi.python.org/pypi/bob.spear},
doi = {10.1109/ICASSP.2014.6853879},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Khoury_ICASSP_2014.pdf}
}
@TECHREPORT{Khoury_Idiap-Com-04-2012,
author = {Khoury, Elie and Marcel, S{\'{e}}bastien and G{\"{u}}nther, Manuel},
projects = {Idiap, SNSF-LOBI},
month = {12},
title = {ICB 2013 - Competition on speaker recognition in mobile environment using the MOBIO database: The Evaluation Plan},
type = {Idiap-Com},
number = {Idiap-Com-04-2012},
year = {2012},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Khoury_Idiap-Com-04-2012.pdf}
}
@ARTICLE{Khoury_IMAVIS_2014,
author = {Khoury, Elie and El Shafey, Laurent and McCool, Chris and G{\"{u}}nther, Manuel and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BBfor2, BEAT},
title = {Bi-Modal Biometric Authentication on Mobile Phones in Challenging Conditions},
journal = {Image and Vision Computing},
year = {2014},
pages = {1147-1160},
url = {http://www.sciencedirect.com/science/article/pii/S0262885613001492},
doi = {http://dx.doi.org/10.1016/j.imavis.2013.10.001},
crossref = {Khoury_Idiap-RR-30-2013},
abstract = {This paper examines the issue of face, speaker and bi-modal authentication in mobile environments when there is significant condition mismatch. We introduce this mismatch by enrolling client models on high quality biometric samples obtained on a laptop computer and authenticating them on lower quality biometric samples acquired with a mobile phone. To perform these experiments we develop three novel authentication protocols for the large publicly available MOBIO database. We evaluate state-of-the-art face, speaker and bi-modal authentication techniques and show that inter-session variability modelling using Gaussian mixture models provides a consistently robust system for face, speaker and bi-modal authentication. It is also shown that multi-algorithm fusion provides a consistent performance improvement for face, speaker and bi-modal authentication. Using this bi-modal multi-algorithm system we derive a state-of-the-art authentication system that obtains a half total error rate of 6.3\% and 1.9\% for Female and Male trials, respectively.}
}
@INPROCEEDINGS{Khoury_INTERSPEECH_2014,
author = {Khoury, Elie and Kinnunen, Tomi and Sizov, Aleksandr and Wu, Zhizheng and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI},
title = {Introducing I-Vectors for Joint Anti-spoofing and Speaker Verification},
booktitle = {The 15th Annual Conference of the International Speech Communication Association},
year = {2014},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Khoury_INTERSPEECH_2014.pdf}
}
@INPROCEEDINGS{Khoury_NISTSRE_2012,
author = {Khoury, Elie and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI},
month = dec,
title = {The Idiap Speaker Recognition Evaluation System at NIST SRE 2012},
booktitle = {NIST Speaker Recognition Conference},
year = {2012},
location = {Orlando, USA},
organization = {NIST},
abstract = {In this paper, we present the Idiap Research Institute submission to the 2012 NIST Speaker Recognition Evaluation. Our system is based on the Inter-Session Variability
(ISV) modelling technique. The implementation of the system relies on Bob, a free signal processing and machine learning toolbox developed at Idiap. The NIST official results show the effectiveness of the proposed approach, especially on added noise
data.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Khoury_NISTSRE_2012.pdf}
}
@INPROCEEDINGS{Khoury_ODYSSEY_2014,
author = {Khoury, Elie and El Shafey, Laurent and Ferras, Marc and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BEAT, InEvent},
title = {Hierarchical speaker clustering methods for the NIST i-vector Challenge},
booktitle = {Odyssey: The Speaker and Language Recognition Workshop},
year = {2014},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Khoury_ODYSSEY_2014.pdf}
}
@INPROCEEDINGS{Komaty_IJCB2023_2023,
author = {Komaty, Alain and Krivokuca, Vedrana and Ecabert, Christophe and Marcel, S{\'{e}}bastien},
projects = {SOTERIA},
title = {Can personalised hygienic masks be used to attack face recognition systems?},
booktitle = {Proceedings of IEEE International Joint Conference on Biometrics (IJCB2023)},
year = {2023},
abstract = {The proliferation of automated face recognition (FR) necessitates increasingly accurate person identification. The COVID-19 pandemic has exposed the limitations of FR systems when presented with faces occluded by hygienic masks. However, the security risks of personalised hygienic mask attacks, whereby an attacker wears the mask on which the bottom part of an enrolled user's face is printed, have not yet been studied. To address this research gap, we introduce a novel face dataset consisting of smartphone-recorded videos of real (bona-fide) faces and personalised hygienic mask attacks. We also analyse the vulnerability of two state-of-the-art FR systems to this type of attack, using our dataset. Our results indicate that personalised hygienic mask attacks have the potential to compromise system security, particularly for FR systems that are tuned towards optimising user convenience. These findings underscore the importance of developing suitable Presentation Attack Detection (PAD) algorithms. Our dataset will help researchers and practitioners work towards this goal, thereby enhancing the security and reliability of FR systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Komaty_IJCB2023_2023.pdf}
}
@INPROCEEDINGS{Komaty_WACV_2025,
author = {Komaty, Alain and Otroshi Shahreza, Hatef and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, SOTERIA, CARMEN, Biometrics Center},
title = {Exploring ChatGPT for Face Presentation Attack Detection in Zero and Few-Shot in-Context Learning},
booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
year = {2025},
abstract = {This study highlights the potential of ChatGPT (specifically GPT-4o) as a competitive alternative for Face Presentation Attack Detection (PAD), outperforming several PAD models, including commercial solutions, in specific scenarios. Our results\footnote{\href{https://gitlab.idiap.ch/bob/bob.paper.wacv2025\_chatgpt_face_pad}{https://gitlab.idiap.ch/bob/bob.paper.wacv2025\_chatgpt\_face\_pad}} show that GPT-4o demonstrates high consistency, particularly in few-shot in-context learning, where its performance improves as more examples are provided (reference data). We also observe that detailed prompts enable the model to provide scores reliably, a behavior not observed with concise prompts. Additionally, explanation-seeking prompts slightly enhance the model's performance by improving its interpretability. Remarkably, the model exhibits emergent reasoning capabilities, correctly predicting the attack type (print or replay) with high accuracy in few-shot scenarios, despite not being explicitly instructed to classify attack types. Despite these strengths, GPT-4o faces challenges in zero-shot tasks, where its performance is limited compared to specialized PAD systems. Experiments were conducted on a subset of the SOTERIA dataset, ensuring compliance with data privacy regulations by using only data from consenting individuals. These findings underscore GPT-4o's promise in PAD applications, laying the groundwork for future research to address broader data privacy concerns and improve cross-dataset generalization.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Komaty_WACV_2025.pdf}
}
@INPROCEEDINGS{Komulainen_ICB_2013,
author = {Komulainen, Jukka and Hadid, Abdenour and Pietikainen, Matti and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {TABULA RASA, BEAT},
month = jun,
title = {Complementary Countermeasures for Detecting Scenic Face Spoofing Attacks},
booktitle = {International Conference on Biometrics},
year = {2013},
location = {Madrid, Spain},
url = {https://pypi.python.org/pypi/antispoofing.fusion},
abstract = {The face recognition community has finally started paying more attention to the long-neglected problem of spoofing attacks. The number of countermeasures is gradually increasing and fairly good results have been reported on the publicly available databases. There exists no superior anti-spoofing technique due to the varying nature of attack scenarios and acquisition conditions. Therefore, it is important to find out complementary countermeasures and study how they should be combined in order to construct an easily extensible anti-spoofing framework. In this paper, we address this issue by studying fusion of motion and texture based countermeasures under several types of scenic face attacks. We provide an intuitive way to explore the fusion potential of different visual cues and show that the performance of the individual methods can be vastly improved by performing fusion at score level. The Half-Total Error Rate (HTER) of the best individual countermeasure was decreased from 11.2\% to 5.1\% on the Replay Attack Database. More importantly, we question the idea of using complex classification schemes in individual countermeasures, since nearly same fusion performance is obtained by replacing them with a simple linear one. In this manner, the computational efficiency and also probably the generalization ability of the resulting anti-spoofing framework are increased.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Komulainen_ICB_2013.pdf}
}
@INPROCEEDINGS{Korshunov_ACMMM_2022,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, MARGIN},
month = oct,
title = {Face Anthropometry Aware Audio-visual Age Verification},
booktitle = {ACM Multimedia},
year = {2022},
abstract = {Protection of minors against destructive content or illegal advertising is an important problem, which is now under increasing societal and legislative pressure. The latest advancements in an automated age verification is a possible solution to this problem. There are however limitations of the current state of the art age verification methods, specifically, the lack of approaches focusing on video-based or even solely audio-based approaches, since the image domain is the one with the majority of publicly available datasets. In this paper, we consider the problem of age verification as a multimodal problem by proposing and evaluating several audio- and image-based models and their combinations. To that end, we annotated a set of publicly available videos with age labels, with a special focus on the children age labels. We also propose a new training strategy based on the adaptive label distribution learning (ALDL), which is driven by facial anthropometry and age-based skin degradation. This adaptive approach demonstrates the best accuracy when evaluated across several test databases.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Korshunov_ACMMM_2022.pdf}
}
@INPROCEEDINGS{Korshunov_AVFAKES_ICML_2019,
author = {Korshunov, Pavel and Halstead, Michael and Castan, Diego and Graciarena, Martin and McLaren, Mitchell and Burns, Brian and Lawson, Aaron and Marcel, S{\'{e}}bastien},
keywords = {inconsistencies detection, lip-syncing, Video tampering},
projects = {Idiap, SAVI},
month = jul,
title = {Tampered Speaker Inconsistency Detection with Phonetically Aware Audio-visual Features},
booktitle = {International Conference on Machine Learning},
series = {Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes},
year = {2019},
note = {Best paper award in ICML workshop "Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes"},
abstract = {The recent increase in social media based propaganda, i.e., ‘fake news’, calls for automated methods to detect tampered content. In this paper, we focus on detecting tampering in a video with a person speaking to a camera. This form of manipulation is easy to perform, since one can just replace a part of the audio, dramatically chang- ing the meaning of the video. We consider several detection approaches based on phonetic features and recurrent networks. We demonstrate that by replacing standard MFCC features with embeddings from a DNN trained for automatic speech recognition, combined with mouth landmarks (visual features), we can achieve a significant performance improvement on several challenging publicly available databases of speakers (VidTIMIT, AMI, and GRID), for which we generated sets of tampered data. The evaluations demonstrate a relative equal error rate reduction of 55\% (to 4.5\% from 10.0\%) on the large GRID corpus based dataset and a satisfying generalization of the model on other datasets.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Korshunov_AVFAKESICML_2019.pdf}
}
@INPROCEEDINGS{Korshunov_BTAS-2_2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT, SWAN},
month = sep,
title = {Joint Operation of Voice Biometrics and Presentation Attack Detection},
booktitle = {IEEE International Conference on Biometrics: Theory, Applications and Systems},
year = {2016},
address = {Niagara Falls, Buffalo, New York, USA},
note = {Open source software for the paper: https://pypi.python.org/pypi/bob.paper.btas_j2016},
url = {https://pypi.python.org/pypi/bob.paper.btas_j2016},
crossref = {Korshunov_Idiap-RR-25-2016},
abstract = {Research in the area of automatic speaker verification (ASV) has advanced enough for the industry to start using ASV systems in practical applications. However, as it was also shown for fingerprints, face, and other verification systems, ASV systems are highly vulnerable to spoofing or presentation attacks, limiting their wide practical deployment. Therefore, to protect against such attacks, effective anti-spoofing detection techniques, more formally known as presentation attack detection (PAD) systems, need to be developed. These techniques should be then seamlessly integrated into existing ASV systems for practical all-in-one solutions. In this paper, we focus on the integration of PAD and ASV systems. We consider the state of the art i-vector and ISV-based ASV systems and demonstrate the effect of score-based integration with a PAD system on the verification and attack detection accuracies. In our experiments, we rely on AVspoof database that contains realistic presentation attacks, which are considered by the industry to be the threat to practical ASV systems. Experimental results show a significantly increased resistance of the joint ASV-PAD system to the attacks at the expense of slightly degraded performance for scenarios without spoofing attacks. Also, an important contribution of the paper is an open source and an online-based implementations of the separate and joint ASV-PAD systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Korshunov_BTAS-2_2016.pdf}
}
@INPROCEEDINGS{Korshunov_BTAS_2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien and Muckenhirn, Hannah and Gon{\c c}alves, A. R. and Mello, A. G. Souza and Violato, R. P. Velloso and Sim{\~{o}}es, F. O. and Neto, M. U. and de Assis Angeloni, M. and Stuchi, J. A. and Dinkel, H. and Chen, N. and Qian, Y. and Paul, D. and Saha, G. and Sahidullah, Md},
projects = {Idiap, BEAT, SWAN},
month = sep,
title = {Overview of BTAS 2016 Speaker Anti-spoofing Competition},
booktitle = {IEEE International Conference on Biometrics: Theory, Applications and Systems},
year = {2016},
address = {Niagara Falls, NY, USA},
url = {https://pypi.python.org/pypi/bob.paper.btas_c2016},
crossref = {Korshunov_Idiap-RR-24-2016},
abstract = {This paper provides an overview of the Speaker Anti-spoofing Competition organized by Biometric group at Idiap Research Institute for the IEEE International Conference on Biometrics: Theory, Applications, and Systems (BTAS 2016). The competition used AVspoof database, which contains a comprehensive set of presentation attacks, including, (i) direct replay attacks when a genuine data is played back using a laptop and two phones (Samsung Galaxy S4 and iPhone 3G), (ii) synthesized speech replayed with a laptop, and (iii) speech created with a voice conversion algorithm, also replayed with a laptop.
The paper states competition goals, describes the database and the evaluation protocol, discusses solutions for spoofing or presentation attack detection submitted by the participants, and presents the results of the evaluation.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Korshunov_BTAS_2016.pdf}
}
@INPROCEEDINGS{Korshunov_EUSIPCO_2018,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
keywords = {Benchmarking, lip-syncing, LSTM, Video tampering},
projects = {Idiap, SAVI},
month = sep,
title = {Speaker Inconsistency Detection in Tampered Video},
booktitle = {European Signal Processing Conference},
year = {2018},
abstract = {With the increasing amount of video being consumed by people daily, there is a danger of the rise in maliciously modified video content (i.e., 'fake news') that could be used to damage innocent people or to impose a certain agenda, e.g., meddle in elections. In this paper, we consider audio manipulations in video of a person speaking to the camera. Such manipulation is easy to perform, for instance, one can just replace a part of audio, while it can dramatically change the message and the meaning of the video. With the goal to develop an automated system that can detect these audio-visual speaker inconsistencies, we consider several approaches proposed for lip-syncing and dubbing detection, based on convolutional and recurrent networks and compare them with systems that are based on more traditional classifiers. We evaluated these methods on publicly available databases VidTIMIT, AMI, and GRID, for which we generated sets of tampered data.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Korshunov_EUSIPCO_2018.pdf}
}
@INPROCEEDINGS{Korshunov_ICASSP_2014,
author = {Korshunov, Pavel and George, Anjith and {\"{O}}zbulak, G{\"{o}}khan and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = apr,
title = {Vulnerability of Face Age Verification to Replay Attacks},
booktitle = {IEEE International Conference on Acoustics, Speech, and Signal Processing},
year = {2024},
abstract = {Presentation attacks on biometric systems have long created significant security risks. The increase in the adoption of age verification systems, which ensure that only age-appropriate content is consumed online, raises the question of vulnerability of such systems to replay presentation attacks. In this paper, we analyze the vulnerability of face age verification to simple replay attacks and assess whether presentation attack detection (PAD) systems created for biometrics can be effective at detecting similar attacks on age verification. We used three types of attacks captured with iPhone 12, Galaxy S9, and Huawei Mate 30 phones from iPad Pro, which replayed the images from a commonly used UTKFace dataset of faces with true age labels. We evaluated four state of the art face age verification algorithms, including simple classification, distribution-based, regression via classification, and adaptive distribution approaches. We show that these algorithms are vulnerable to the attacks, since the accuracy of age verification on replayed images is only a couple of percentage points different compared to when the original images are used, which means an age verification system cannot distinguish attacks from bona fide images. Using two state of the art presentation attack detection systems, DeepPixBiS and CDCN, trained to detect similar attacks on biometrics, we demonstrate that they struggle to detect both: the types of attacks that are possible in age verification scenario and the type of bona fide images that are commonly used. These results highlight the need for the development of age verification specific attack detection systems for age verification to become practical.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Korshunov_ICASSP_2014.pdf}
}
@INPROCEEDINGS{Korshunov_ICASSP_2021,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, Verifake, Biometrics Center},
month = jun,
title = {Subjective and objective evaluation of deepfake videos},
booktitle = {The international Conference on Acoustics, Speech, and Signal Processing},
year = {2021},
abstract = {Practically anyone can now generate a realistic looking deepfake video. It is clear that the online prevalence of such fake videos will erode the societal trust in video evidence even further. To counter the looming threat, many methods to detect deepfakes were recently proposed by the research community. However, it is still unclear how realistic deepfake videos are for an average person and whether the algorithms are significantly better than humans at detecting them. Therefore, this paper, presents a subjective study, which, using 60 naive subjects, evaluates how hard it is for humans to see if a video is a deepfake or not. For the study, 120 videos (60 deepfakes and 60 originals) were manually selected from the Facebook database used in Kaggle's Deepfake Detection Challenge 2020.
The results of the subjective evaluation were compared with two state of the art deepfake detection methods, based on Xception and EfficientNet (B4 variant) neural network models pre-trained on two other public databases: Google and Jiqsaw subset from FaceForensics++ and Celeb-DF v2 dataset. The experiments demonstrate that while the human perception is very different from the perception of a machine, both successfully but in different ways are fooled by deepfakes. Specifically, algorithms struggle to detect the deepfake videos that humans find to be very easy to spot.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/Korshunov_ICASSP_2021.pdf}
}
@INPROCEEDINGS{Korshunov_ICASSP_2022,
author = {Korshunov, Pavel and Jain, Anubhav and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = may,
title = {Custom attribution loss for improving generalization and interpretability of deepfake detection},
booktitle = {International Conference on Acoustics, Speech, and Signal Processing},
year = {2022},
abstract = {The simplicity and accessibility of tools for generating deepfakes pose a significant technical challenge for their detection and filtering. Many of the recently proposed methods for deeptake detection focus on a 'blackbox' approach and therefore suffer from the lack of any additional information about the nature of fake videos beyond the fake or not fake labels. In this paper, we approach deepfake detection by solving the related problem of attribution, where the goal is to distinguish each separate type of a deepfake attack. We design a training approach with customized Triplet and ArcFace losses that allow to improve the accuracy of deepfake detection on several publicly available datasets, including Google and Jigsaw, FaceForensics++, HifiFace, DeeperForensics, Celeb-DF, DeepfakeTIMIT, and DF-Mobio. Using an example of Xception net as an underlying architecture, we also demonstrate that when trained for attribution, the model can be used as a tool to analyze the deepfake space and to compare it with the space of original videos.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Korshunov_ICASSP_2022.pdf}
}
@INPROCEEDINGS{Korshunov_ICBB_2019,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, SAVI},
month = oct,
title = {Vulnerability of Face Recognition to Deep Morphing},
booktitle = {International Conference on Biometrics for Borders},
year = {2019},
abstract = {It is increasingly easy to automatically swap faces in images and video or morph two faces into one using generative adversarial networks (GANs). The high quality of the resulted deep-morph raises the question of how vulnerable the current face recognition systems are to such fake images and videos. It also calls for automated ways to detect these GAN-generated faces. In this paper, we present the publicly available dataset of the Deepfake videos with faces morphed with a GAN-based algorithm. To generate these videos, we used open source software based on GANs, and we emphasize that training and blending parameters can significantly impact the quality of the resulted videos. We show that the state of the art face recognition systems based on VGG and Facenet neural networks are vulnerable to the deep morph videos, with 85.62 and 95.00 false acceptance rates, respectively, which means methods for detecting these videos are necessary. We consider several baseline approaches for detecting deep morphs and find that the method based on visual quality metrics (often used in presentation attack detection domain) leads to the best performance with 8.97 equal error rate. Our experiments demonstrate that GAN-generated deep morph videos are challenging for both face recognition systems and existing detection methods, and the further development of deep morphing technologies will make it even more so.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Korshunov_ICBB_2019.pdf}
}
@INPROCEEDINGS{Korshunov_ICB_2019,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
keywords = {Deepfakes, detection, Face Recognition, vulnerability},
projects = {Idiap, SAVI},
month = jun,
title = {Vulnerability assessment and detection of Deepfake videos},
booktitle = {IAPR International Conference on Biometrics},
year = {2019},
crossref = {Korshunov_Idiap-RR-18-2018},
abstract = {It is becoming increasingly easy to automatically replace a face of one person in a video with the face of another person by using a pre-trained generative adversarial network (GAN). Recent public scandals, e.g., the faces of celebrities being swapped onto pornographic videos, call for automated ways to detect these Deepfake videos. To help developing such methods, in this paper, we present the first publicly available set of Deepfake videos generated from videos of VidTIMIT database. We used open source software based on GANs to create the Deepfakes, and we emphasize that training and blending parameters can significantly impact the quality of the resulted videos. To demonstrate this impact, we generated videos with low and high visual quality (320 videos each) using differently tuned parameter sets. We showed that the state of the art face recognition systems based on VGG and Facenet neural networks are vulnerable to Deepfake videos, with 85.62\% and 95.00\% false acceptance rates (on high quality versions) respectively, which means methods for detecting Deepfake videos are necessary. By considering several baseline approaches, we found the best performing method based on visual quality metrics, which is often used in presentation attack detection domain, to lead to 8.97\% equal error rate on high quality Deepfakes. Our experiments demonstrate that GAN-generated Deepfake videos are challenging for both face recognition systems and existing detection methods, and the further development of face swapping technology will make it even more so.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Korshunov_ICB_2019.pdf}
}
@INPROCEEDINGS{Korshunov_ICCV_2025,
author = {Korshunov, Pavel and Vidit, Vidit and Mohammadi, Amir and Ecabert, Christophe and Shamoska, Nevena and Marcel, S{\'{e}}bastien and Yu, Zeqin and Tian, Ye and Ni, Jiangqun and Lazarevic, Lazar and Khizbullin, Renat and Evteeva, Anastasiia and Tochin, Alexey and Grishin, Aleksei and George, Anjith and DeAlcala, Daniel and Endrei, Tamas and Munoz-Haro, Javier and Tolosana, Ruben and Vera-Rodriguez, Ruben and Morales, Aythami and Fierrez, Julian and Cserey, Gyorgy and Sharma, Hardik and Chaudhary, Sachin and Dudhane, Akshay and Hambarde, Praful and Shukla, Amit and Shaily, Prateek and Kumar, Jayant and Hase, Ajinkya and Maurya, Satish and Sharma, Mridul and Dwivedi, Pallav},
projects = {Idiap, ROSALIND},
mainresearchprogram = {AI for Everyone},
additionalresearchprograms = {AI for Life},
month = oct,
title = {DeepID Challenge of Detecting Synthetic Manipulations in ID Documents},
booktitle = {International Conference on Computer Vision (ICCV)},
year = {2025},
abstract = {An increase in AI based manipulations of ID document images threatens KYC systems widely used in online banking and other digital authentication services. DeepID challenge aimed to advance the research in the methods for detecting synthetic manipulations in ID documents. For that purpose a FantasyID dataset of both bona fide and manipulated fantasy ID cards was provided to the participants for training and tuning of their systems. Participating submissions were evaluated on a test set of FantasyID card created with both seen and unseen attacks, and on an out-of-domain private dataset of 20K real ID documents containing both genuine bona fide and manipulated samples. The challenge included two tracks 1) a binary detection track to detect whether an ID document is manipulated or not and 2) a localization track, where the goal was to identify the manipulated regions of an ID document. The evaluations were based on the F1-score metric for both detection and localization track and the submissions were ranked based on the weighted average F1-score of FantasyID (with weight 0.3) and private (with weight 0.7) test sets. With more than 100 registrations in the challenge, 26 teams have participated and 6 of them managed to beat the provided TruFor baseline method in detection track and 4 teams in the localization track. Sunlight team from Sun Yat-sen University has won both tracks of the challenge and UAM-Biometrics has ranked best in the private dataset.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Korshunov_ICCV_2025.pdf}
}
@TECHREPORT{Korshunov_Idiap-RR-36-2020,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, Verifake},
month = {12},
title = {Deepfake detection: humans vs. machines},
type = {Idiap-RR},
number = {Idiap-RR-36-2020},
year = {2020},
institution = {Idiap},
abstract = {Deepfake videos, where a person’s face is automatically swapped with a face of someone else, are becoming easier to generate with more realistic results. In response to the threat such manipulations can pose to our trust in video evidence, several large datasets of deepfake videos and many methods to detect them were proposed recently. However, it is still unclear how realistic deepfake videos are for an average person and whether the algorithms are significantly better than humans at detecting them. In this paper, we present a subjective study conducted in a crowdsourcing-like scenario, which systematically evaluates how hard it is for humans to see if the video is deepfake or not. For the evaluation, we used 120 different videos (60 deepfakes and 60 originals) manually pre-selected from the Facebook deepfake database, which was provided in the Kaggle’s Deepfake Detection Challenge 2020. For each video, a simple question: "Is face of the person in the video real of fake?" was answered on average by 19 na ̈{\i}ve subjects. The results of the subjective evaluation were compared with the performance of two different state of the art deepfake detection methods, based on Xception and EfficientNets (B4 variant) neural networks, which were pre- trained on two other large public databases: the Google’s subset from FaceForensics++ and the recent Celeb-DF dataset. The evaluation demonstrates that while the human perception is very different from the perception of a machine, both successfully but in different ways are fooled by deepfakes. Specifically, algorithms struggle to detect those deepfake videos, which human subjects found to be very easy to spot.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/Korshunov_Idiap-RR-36-2020.pdf}
}
@INCOLLECTION{Korshunov_IET_2017,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
editor = {Vielhauer, Claus},
projects = {Idiap, SWAN, Tesla},
title = {Presentation attack detection in voice biometrics},
booktitle = {User-Centric Privacy and Security in Biometrics},
chapter = {7},
year = {2017},
publisher = {The Institution of Engineering and Technology},
address = {Savoy Place, London WC2R 0BL, UK},
abstract = {Recent years have shown an increase in both the accuracy of biometric systems and their practical use. The application of biometrics is becoming widespread with fingerprint sensors in smartphones, automatic face recognition in social networks and video-based applications, and speaker recognition in phone banking and other phone-based services. The popularization of the biometric systems, however, exposed their major flaw --- high vulnerability to spoofing attacks. A fingerprint sensor can be easily tricked with a simple glue-made mold, a face recognition system can be accessed using a printed photo, and a speaker recognition system can be spoofed with a replay of pre-recorded voice. The ease with which a biometric system can be spoofed demonstrates the importance of developing efficient anti-spoofing systems that can detect both known (conceivable now) and unknown (possible in the future) spoofing attacks.
Therefore, it is important to develop mechanisms that can detect such attacks, and it is equally important for these mechanisms to be seamlessly integrated into existing biometric systems for practical and attack-resistant solutions. To be practical, however, an attack detection should have (i) high accuracy, (ii) be well-generalized for different attacks, and (iii) be simple and efficient.
One reason for the increasing demand for effective presentation attack detection (PAD) systems is the ease of access to people's biometric data. So often, a potential attacker can almost effortlessly obtain necessary biometric samples from social networks, including facial images, audio and video recordings, and even extract fingerprints from high resolution images. Therefore, various privacy protection solutions, such as legal privacy requirements and algorithms for obfuscating personal information, e.g., visual privacy filters, as well as, social awareness of threats to privacy can also increase security of personal information and potentially reduce the vulnerability of biometric systems.
In this chapter, however, we focus on presentation attacks detection in voice biometrics, i.e., automatic speaker verification (ASV) systems. We discuss vulnerabilities of these systems to presentation attacks (PAs), present different state of the art PAD systems, give the insights into their performances, and discuss the integration of PAD and ASV systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Korshunov_IET_2017.pdf}
}
@INPROCEEDINGS{Korshunov_IJCB-2_2025,
author = {Korshunov, Pavel and Mohammadi, Amir and Vidit, Vidit and Ecabert, Christophe and Marcel, S{\'{e}}bastien},
projects = {ROSALIND},
mainresearchprogram = {AI for Everyone},
title = {FantasyID: A dataset for detecting digital manipulations of ID-documents},
booktitle = {Proceedings of IEEE International Joint Conference on Biometrics},
year = {2025},
abstract = {Advancements in image generation led to the availability of easy-to-use tools for malicious actors to create forged images. These tools pose a serious threat to the widespread Know Your Customer (KYC) applications, requiring robust systems for detection of the forged Identity Documents (IDs). To facilitate the development of the detection algorithms, in this paper, we propose a novel publicly available (including commercial use) dataset, FantasyID, which mimics real-world IDs but without tampering with legal documents and, compared to previous public datasets, it does not contain generated faces or specimen watermarks. FantasyID contains ID cards with diverse design styles, languages, and faces of real people. To simulate a realistic KYC scenario, the cards from FantasyID were printed and captured with three different devices, constituting the bonafide class. We have emulated digital forgery/injection attacks that could be performed by a malicious actor to tamper the IDs using the existing generative tools. The current state-of-the-art forgery detection algorithms, such as TruFor, MMFusion, UniFD, and FatFormer, are challenged by FantasyID dataset. It especially evident, in the evaluation conditions close to practical, with the operational threshold set on validation set so that false positive rate is at 10\%, leading to false negative rates close to 50\% across the board on the test set. The evaluation experiments demonstrate that FantasyID dataset is complex enough to be used as an evaluation benchmark for detection algorithms.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Korshunov_IJCB-2_2025.pdf}
}
@INPROCEEDINGS{Korshunov_IJCB_2023,
author = {Korshunov, Pavel and Chen, Haolin and Garner, Philip N. and Marcel, S{\'{e}}bastien},
projects = {Idiap, NAST, Biometrics Center},
month = sep,
title = {Vulnerability of Automatic Identity Recognition to Audio-Visual Deepfakes},
booktitle = {IEEE International Joint Conference on Biometrics},
year = {2023},
abstract = {The task of deepfakes detection is far from being solved by speech or vision researchers. Several publicly available databases of fake synthetic video and speech were built to aid the development of detection methods. However, existing databases either focus on visual or voice modalities. Also, the databases rarely contain or any evidence is shown that the resulted deepfakes realistically impersonate any real person. In this paper, we present the first realistic audio-visual database of deepfakes SWAN-DF, where lips and speech are well synchronized and video have high visual and audio qualities. We took the publicly available SWAN dataset of real videos with different identities to create audio-visual deepfakes using several models from DeepFaceLab and blending techniques for face swapping and HiFiVC, DiffVC, YourTTS, and FreeVC models for voice conversion. From the publicly available speech dataset LibriTTS, we also created a separate database of only audio deepfakes LibriTTS-DF using several latest text to speech methods: YourTTS, Adaspeech, and TorToiSe. We demonstrate the vulnerability of a state of the art speaker recognition system, such as ECAPA-TDNN-based model from SpeechBrain, to the synthetic voices. Similarly, we tested face recognition system based on the MobileFaceNet architecture to several variants of our visual deepfakes. The vulnerability assessment show that by tuning the existing pretrained deepfake models to specific identities, one can successfully spoof the face and speaker recognition systems in more than 90\% of the time and achieve a very realistic looking and sounding fake video of a given person.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Korshunov_IJCB_2023.pdf}
}
@INPROCEEDINGS{Korshunov_IJCB_2025,
author = {Korshunov, Pavel and Kotwal, Ketan and Ecabert, Christophe and Vidit, Vidit and Mohammadi, Amir and Marcel, S{\'{e}}bastien},
keywords = {Demographic bias, Fairness, Synthetic Dataset},
projects = {Idiap, ROSALIND},
mainresearchprogram = {AI for Everyone},
month = sep,
title = {Investigation of accuracy and bias in face recognition trained with synthetic data},
booktitle = {Proceedings of IEEE International Joint Conference on Biometrics},
year = {2025},
abstract = {Synthetic data has emerged as a promising alternative for training face recognition (FR) models, offering advantages in scalability, privacy compliance, and potential for bias mitigation. However, critical questions remain on whether both high accuracy and fairness can be achieved with synthetic data. In this work, we evaluate the impact of synthetic data on bias and performance of FR systems. We generate balanced face dataset, FairFaceGen, using two state of the art text-to-image generators, Flux.1-dev and Stable Diffusion v3.5 (SD35), and combine them with several identity augmentation methods, including Arc2Face and four IP-Adapters. By maintaining equal identity count across synthetic and real datasets, we ensure fair comparisons when evaluating FR performance on standard (LFW, AgeDB-30, etc.) and challenging IJB-B/C benchmarks and FR bias on Racial Faces in-the-Wild (RFW) dataset. Our results demonstrate that although synthetic data still lags behind the real datasets in the generalization on IJB-B/C, demographically balanced synthetic datasets, especially those generated with SD35, show potential for bias mitigation. We also observe that the number and quality of intra-class augmentations significantly affect FR accuracy and fairness. These findings provide practical guidelines for constructing fairer FR systems using synthetic data.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Korshunov_IJCB_2025.pdf}
}
@INPROCEEDINGS{Korshunov_INTERSPEECH_2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
keywords = {cross-database testing, Open Source, presentation attack, speaker anti-spoofing},
projects = {Idiap, BEAT, SWAN},
month = sep,
title = {Cross-database evaluation of audio-based spoofing detection systems},
booktitle = {Interspeech},
year = {2016},
location = {San Francisco, USA},
url = {https://pypi.python.org/pypi/bob.paper.interspeech_2016},
crossref = {Korshunov_Idiap-RR-23-2016},
abstract = {Since automatic speaker verification (ASV) systems are highly vulnerable to spoofing attacks, it is important to develop mechanisms that can detect such attacks. To be practical, however, a spoofing attack detection approach should have (i) high accuracy, (ii) be well-generalized for practical attacks, and (iii) be simple and efficient. Several audio-based spoofing detection methods have been proposed recently but their evaluation is limited to less realistic databases containing homogeneous data. In this paper, we consider eight existing presentation attack detection (PAD) methods and evaluate their performance using two major publicly available speaker databases with spoofing attacks: AVspoof and ASVspoof. We first show that realistic presentation attacks (speech is replayed to PAD system) are significantly more challenging for the considered PAD methods compared to the so called `logical access' attacks (speech is presented to PAD system directly). Then, via a cross-database evaluation, we demonstrate that the existing methods generalize poorly when different databases or different types of attacks are used for training and testing. The results question the efficiency and practicality of the existing PAD systems, as well as, call for creation of databases with larger variety of realistic speech presentation attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2016/Korshunov_INTERSPEECH_2016.pdf}
}
@INPROCEEDINGS{Korshunov_ISBA_2018,
author = {Korshunov, Pavel and Goncalves, Andre{\'{e}} R. and Violato, Ricardo P. V. and Sim{\~{o}}es, Fl{\'{a}}vio O. and Marcel, S{\'{e}}bastien},
projects = {Idiap, Tesla, SWAN},
month = jan,
title = {On the Use of Convolutional Neural Networks for Speech Presentation Attack Detection},
booktitle = {International Conference on Identity, Security and Behavior Analysis},
year = {2018},
abstract = {Research in the area of automatic speaker verification (ASV) has advanced enough for the industry to start using ASV systems in practical applications. However, these systems are highly vulnerable to spoofing or presentation attacks (PAs), limiting their wide deployment. Several speech-based presentation attack detection (PAD) methods have been proposed recently but most of them are based on hand crafted frequency or phase-based features. Although convolutional neural networks (CNN) have already shown breakthrough results in face recognition, little is understood whether CNNs are as effective in detecting presentation attacks in speech. In this paper, to investigate the applicability of CNNs for PAD, we consider shallow and deep examples of CNN architectures implemented using Tensorflow and compare their performances with the state of the art MFCC with GMM-based system on two large databases with presentation attacks: publicly available voicePA and proprietary BioCPqD-PA. We study the impact of increasing the depth of CNNs on the performance, and note how they perform on unknown attacks, by using one database to train and another to evaluate. The results demonstrate that CNNs are able to learn a database significantly better (increasing depth also improves the performance), compared to hand crafted features. However, CNN-based PADs still lack the ability to generalize across databases and are unable to detect unknown attacks well.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Korshunov_ISBA_2018.pdf}
}
@INCOLLECTION{Korshunov_SPRINGER_2018,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Fierrez, Julian and Evans, Nicholas},
projects = {Idiap, SWAN, Tesla},
month = nov,
title = {A Cross-database Study of Voice Presentation Attack Detection},
booktitle = {Handbook of Biometric Anti-Spoofing: Presentation Attack Detection, 2nd Edition},
chapter = {19},
year = {2018},
publisher = {Springer},
abstract = {Despite an increasing interest in speaker recognition technologies, a significant obstacle still hinders their wide deployment --- their high vulnerability to spoofing or presentation attacks. These attacks can be easy to perform. For instance, if an attacker has access to a speech sample from a target user, he/she can replay it using a loudspeaker or a smartphone to the recognition system during the authentication process. The ease of executing presentation attacks and the fact that no technical knowledge of the biometric system is required makes these attacks especially threatening in practical application. Therefore, late research focuses on collecting data databases with such attacks and on development of presentation attack detection (PAD) systems. In this chapter, we present an overview of the latest databases and the techniques to detect presentation attacks. We consider several prominent databases that contain bona fide and attack data, including: ASVspoof 2015, ASVspoof 2017, AVspoof, voicePA, and BioCPqD-PA (the only proprietary database). Using these databases, we focus on the performance of PAD systems in the cross-database scenario or in the presence of 'unknown' (not available during training) attacks, as these scenarios are closer to practice, when pre-trained systems need to detect attacks in unforeseen conditions. We first present and discuss the performance of PAD systems based on handcrafted features and traditional Gaussian mixture model (GMM) classifiers. We then demonstrate whether the score fusion techniques can improve the performance of PADs. We also present some of the latest results of using neural networks for presentation attack detection. The experiments show that PAD systems struggle to generalize across databases and mostly unable to detect unknown attacks, with systems based on neural networks demonstrating better performance compared to the systems based on handcraft features.}
}
@ARTICLE{Korshunov_STSP_2017,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
keywords = {Presentation Attack Detection, score fusion, speaker database, speaker recognition, voice biometrics},
projects = {Idiap, SWAN},
month = jun,
title = {Impact of score fusion on voice biometrics and presentation attack detection in cross-database evaluations},
journal = {IEEE Journal of Selected Topics in Signal Processing},
volume = {11},
number = {4},
year = {2017},
pages = {695 - 705},
doi = {10.1109/JSTSP.2017.2692389},
abstract = {Research in the area of automatic speaker verification (ASV) has been advanced enough for the industry to start using ASV systems in practical applications. However, these systems are highly vulnerable to spoofing or presentation attacks, limiting their wide deployment. Therefore, it is important to develop mechanisms that can detect such attacks, and it is equally important for these mechanisms to be seamlessly integrated into existing ASV systems for practical and attack-resistant solutions. To be practical, however, an attack detection should (i) have high accuracy, (ii) be well-generalized for different attacks, and (iii) be simple and efficient. Several audio-based presentation attack detection (PAD) methods have been proposed recently but their evaluation was usually done on a single, often obscure, database with limited number of attacks. Therefore, in this paper, we conduct an extensive study of eight state-of-the-art PAD methods and evaluate their ability to detect known and unknown attacks (e.g., in a cross-database scenario) using two major publicly available speaker databases with spoofing attacks: AVspoof and ASVspoof. We investigate whether combining several PAD systems via score fusion can improve attack detection accuracy. We also study the impact of fusing PAD systems (via parallel and cascading schemes) with two i-vector and inter-session variability based ASV systems on the overall performance in both bona fide (no attacks) and spoof scenarios. The evaluation results question the efficiency and practicality of the existing PAD systems, especially when comparing results for individual databases and cross-database data. Fusing several PAD systems can lead to a slightly improved performance; however, how to select which systems to fuse remains an open question. Joint ASV-PAD systems show a significantly increased resistance to the attacks at the expense of slightly degraded performance for bona fide scenarios.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Korshunov_STSP_2017.pdf}
}
@ARTICLE{Korshunov_TBIOM_2021,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, Verifake, Biometrics Center},
month = dec,
title = {Improving Generalization of Deepfake Detection with Data Farming and Few-Shot Learning},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2021},
abstract = {Recent advances in automated video and audio editing tools, generative adversarial networks (GANs), and social media allow creation and fast dissemination of high quality tampered videos, which are generally called deepfakes. Typically, in these videos, a face is swapped with someone else's using GANs. Accessible open source software and apps for the face swapping led to a wide and rapid dissemination of the generated deepfakes, posing a significant technical challenge for their detection and filtering. In response to the threat, which deepfake videos can pose to our trust in video evidence, several large datasets of deepfake videos and several methods to detect them were proposed recently. However, the proposed methods suffer from a problem of overfitting on the training data and the lack of the generalization across different databases and the generative models. Therefore, in this paper, we investigate the techniques for improving the generalization of deepfake detection methods that can be employed in practical settings. We have selected two popular state of the art deepfake detectors: based on Xception and EfficientNet models, and we use five databases: from Google and Jigsaw, FaceForensics++, DeeperForensics, Celeb-DF, and our own publicly available large dataset DF-Mobio. To improve generalization, we apply different augmentation strategies used during training, including a proposed aggressive `data farming' technique based on random patches. We also tested two few-shot tuning methods, when either a first convolutional layer or a last layer of a pre-trained model is tuned on 100 seconds from a training set of the test database. The experimental results clearly expose the generalization problem of deepfake detection methods, since the accuracy drops significantly when a model is trained on one dataset and evaluated on another. However, the silver lining is that an aggressive augmentation during training and a few-shot tuning on the test database can improve the accuracy of the detection methods in a cross-database scenario. As a side observation, we show the importance of database selection for training and evaluation, as FaceForensics++ is found to be better to use for training, while DeeperForensics is found to be significantly more challenging as a test database.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/Korshunov_TBIOM_2021.pdf}
}
@INPROCEEDINGS{Kotwal_CVPR-W_2022,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
keywords = {Finger vein, residual CNN, vein enhancement},
projects = {Biometrics Center, Innosuisse CANDY},
month = jun,
title = {Residual Feature Pyramid Network for Enhancement of Vascular Patterns},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2022},
abstract = {The accuracy of finger vein recognition systems gets degraded due to low and uneven contrast between veins and
surroundings, often resulting in poor detection of vein patterns. We propose a finger-vein enhancement technique,
ResFPN (Residual Feature Pyramid Network), as a generic preprocessing method agnostic to the recognition pipeline.
A bottom-up pyramidal architecture using the novel Structure Detection block (SDBlock) facilitates extraction of
veins of varied widths. Using a feature aggregation module (FAM), we combine these vein-structures, and train the
proposed ResFPN for detection of veins across scales. With enhanced presentations, our experiments indicate a reduction upto 5\% in the average recognition errors for commonly used recognition pipeline over two publicly available datasets. These improvements are persistent even in cross-dataset scenario where the dataset used to train the ResFPN is different from the one used for recognition.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Kotwal_CVPR-W_2022.pdf}
}
@INPROCEEDINGS{Kotwal_ICPR_FAIRBIO_2022,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Demographic, Fairness, Fairness Evaluation.},
projects = {Idiap},
month = aug,
title = {Fairness Index Measures to Evaluate Bias in Biometric Recognition},
booktitle = {International Conference on Pattern Recognition Workshops},
year = {2022},
abstract = {The demographic disparity of biometric systems has led to serious concerns regarding their societal impact as well as applicability of such systems in private and public domains. A quantitative evaluation of demographic fairness is an important step towards understanding, assessment, and mitigation of demographic bias in biometric applications. While few, existing fairness measures are based on post-decision data (such as verification accuracy) of biometric systems, we discuss how pre-decision data (score distributions) provide useful insights towards demographic fairness. In this paper, we introduce multiple measures, based on the statistical characteristics of score distributions, for the evaluation of demographic fairness of a generic biometric verification system. We also propose different variants for each fairness measure depending on how the contribution from constituent demographic groups needs to be combined towards the final measure. In each case, the behavior of the measure has been illustrated numerically and graphically on synthetic data. The demographic imbalance in benchmarking datasets is often overlooked during fairness assessment. We provide a novel weighing strategy to reduce the effect of such imbalance through a non-linear function of sample sizes of demographic groups. The proposed measures are independent of the biometric modality, and thus, applicable across commonly used biometric modalities (e.g., face, fingerprint, etc.).},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Kotwal_ICPR_FAIRBIO_2022.pdf}
}
@TECHREPORT{Kotwal_Idiap-RR-03-2024,
author = {Kotwal, Ketan and {\"{O}}zbulak, G{\"{o}}khan and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Iris Recognition, Periocular PAD, Periocular Recognition, semantic segmentation, Virtual Reality Dataset},
projects = {Biometrics Center},
month = {7},
title = {VRBiom: A New Periocular Dataset for Biometric Applications of HMD},
type = {Idiap-RR},
number = {Idiap-RR-03-2024},
year = {2024},
institution = {Idiap},
abstract = {With advancements in hardware, high-quality head-mounted display (HMD) devices are being developed by numerous companies, driving increased consumer interest in AR, VR, and MR applications. This proliferation of HMD devices opens up possibilities of wide range of applications beyond entertainment. Most commercially available HMD devices are equipped with internal inward-facing cameras to record the periocular areas. Given the nature of these devices and captured data, many applications such as biometric authentication and gaze analysis become feasible. To effectively explore the potential of HMDs for these diverse use-cases and to enhance the corresponding techniques, it is essential to have an HMD dataset that captures realistic scenarios.
In this work, we present a new dataset, called VRBiom, of periocular videos acquired using a Virtual Reality headset. The VRBiom, targeted at biometric applications, consists of 900 short videos acquired from 25 individuals recorded in the NIR spectrum. These 10s long videos have been captured using the internal tracking cameras of Meta Quest Pro at 72 FPS. To encompass real-world variations, the dataset includes recordings under three gaze conditions: steady, moving, and partially closed eyes. We have also ensured an equal split of recordings without and with glasses to facilitate the analysis of eye-wear. These videos, characterized by non-frontal views of the eye and relatively low spatial resolutions (400 × 400), can be instrumental in advancing state-of-the-art research across various biometric applications. The VRBiom dataset can be utilized to evaluate, train, or adapt models for biometric use-cases such as iris and/or periocular recognition and associated sub-tasks such as detection and semantic segmentation.
In addition to data from real individuals, we have included around 1100 presentation attacks constructed from 92 PA instruments. These PAIs fall into six categories constructed through combinations of print attacks (real and synthetic identities), fake 3D eyeballs, plastic eyes, and various types of masks and mannequins. These PA videos, combined with genuine (bona-fide) data, can be utilized to address concerns related to spoofing, which is a significant threat if these devices are to be used for authentication.
The VRBiom dataset is publicly available for research purposes related to biometric applications only.},
pdf = {https://publications.idiap.ch/attachments/reports/2024/Kotwal_Idiap-RR-03-2024.pdf}
}
@INPROCEEDINGS{Kotwal_IEEEICIP_2020,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Tesla, Idiap},
month = oct,
title = {CNN Patch Pooling for Detecting 3D Mask Presentation Attacks in NIR},
booktitle = {IEEE International Conference on Image Processing},
year = {2020},
crossref = {Kotwal_Idiap-RR-10-2020},
abstract = {Presentation attacks using 3D masks pose a serious threat to face recognition systems. Automatic detection of these attacks is challenging due to hyper-realistic nature of masks. In this work, we consider presentations acquired in near infrared (NIR) imaging channel for detection of mask-based attacks. We propose a patch pooling mechanism to learn complex textural features from lower layers of a convolutional neural network CNN). The proposed patch pooling layer can be used in conjunction with a pretrained face recognition CNN without fine-tuning or adaptation. The pretrained CNN, in fact, can also be trained from visual spectrum data. We demonstrate efficacy of the proposed method on mask attacks in NIR channel from WMCA and MLFP datasets. It achieves near perfect results on WMCA data, and outperforms existing benchmark on MLFP dataset by a large margin.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Kotwal_IEEEICIP_2020.pdf}
}
@ARTICLE{Kotwal_IEEETBIOM_2022,
author = {Kotwal, Ketan and Bhattacharjee, Sushil and Abbet, Philip and Mostaani, Zohreh and Wei, Huang and Wenkang, Xu and Yaxi, Zhao and Marcel, S{\'{e}}bastien},
keywords = {CNN, domain adaptation, Face presentation attack detection (PAD), near-infrared},
projects = {Idiap},
title = {Domain-Specific Adaptation of CNN for Detecting Face Presentation Attacks in NIR},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2022},
abstract = {For the automotive industry moving towards personalized applications and experiences, the identification of the person inside vehicle is necessary; and it must be carried out in a secure manner. In this paper, we propose a unique face presentation attack detection (PAD) system for operation inside a passenger vehicle. A typical in-vehicular face PAD system is required to function with several constraints such as bounded sensing (imaging) capabilities, limited computing resources on embedded devices, real-time inference, and essentially, very high accuracy. In this work, we develop a face PAD system for automotive domain, relying on a single NIR camera, to continually verify whether the driver’s face is bona-fide or not. Our work has two main contributions: first, a lightweight face PAD framework has been developed using a 9-layer convolutional neural network (CNN). With its compact size and limited set of operators, it can be deployed in a resource constrained embedded device to achieve a near real-time inference. To alleviate the problem of limited training data (face PAD in NIR) for a given system, we develop an efficient mechanism to obtain this CNN through the combination of adaptation of domain-specific layers and task-specific fine-tuning of a base CNN. As the second contribution, we collect a large face PAD dataset with 5800+ videos, acquired in NIR (940 nm) illumination, for in-vehicular use-cases. This dataset, named VFPAD, captures several real-world variations in terms of environmental settings, illumination, subject’s pose, and appearances. Based on the VFPAD dataset, we demonstrate that the proposed face PAD method achieves very high performance (overall accuracy ≈ 98.0\%), and also outperforms several baseline face PAD methods. The dataset will be shared with the wider scientific community for research purposes.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Kotwal_IEEETBIOM_2022.pdf}
}
@ARTICLE{Kotwal_IEEETRANS.BIOM_2019,
author = {Kotwal, Ketan and Mostaani, Zohreh and Marcel, S{\'{e}}bastien},
keywords = {AIM, deep learning, Face Presentation Attack Detection, Makeup Attack Detection, Makeup Attacks, Old-Age Makeups, score fusion, Shape descriptor, Texture Descriptor},
projects = {ODIN/BATL},
title = {Detection of Age-Induced Makeup Attacks on Face Recognition Systems Using Multi-Layer Deep Features},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2019},
pages = {11},
abstract = {Makeup is a simple and easy instrument that can alter the appearance of a person’s face, and hence, create a presentation attack on face recognition (FR) systems. These attacks, especially the ones mimicking ageing, are difficult to detect due to their close resemblance with genuine (non-makeup) appearances. Makeups can also degrade the performance of recognition systems and of various algorithms that use human face as an input. The detection of facial makeups is an effective prohibitory measure to minimize these problems.
This work proposes a deep learning-based presentation attack detection (PAD) method to identify facial makeups. We propose the use of a convolutional neural network (CNN) to extract features that can distinguish between presentations with age-induced facial makeups (attacks), and those without makeup (bona-fide). These feature descriptors, based on shape and texture cues, are constructed from multiple intermediate layers of a CNN. We introduce a new dataset AIM (Age Induced Makeups) consisting of 200+ video presentations of old-age makeups and bona-fide, each. Our experiments indicate makeups in AIM result in 14\% decrease in the median matching scores of a recent CNN-based FR system. We demonstrate accuracy of the proposed PAD method where 93\% presentations in the AIM dataset are correctly classified. In additional testing, it also outperforms existing methods of detection of generic makeups. A simple score-level fusion, performed on the classification scores of shape- and texture-based features, can further improve the accuracy of the proposed makeup detector.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Kotwal_IEEETRANS.BIOM-2_2019.pdf}
}
@INPROCEEDINGS{Kotwal_IJCB-2_2024,
author = {Kotwal, Ketan and {\"{O}}zbulak, G{\"{o}}khan and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = sep,
title = {Assessing the Reliability of Biometric Authentication on Virtual Reality Devices},
booktitle = {Proceedings of IEEE International Joint Conference on Biometrics},
year = {2024},
crossref = {Kotwal_Idiap-RR-04-2024},
abstract = {Recent developments in Virtual Reality (VR) headsets have unlocked a plethora of innovative use-cases, many of which were previously unimaginable. However, as these use-cases, such as personalized immersive experiences, necessitate user authentication, ensuring robustness and resistance to spoofing attacks becomes imperative. The absence of appropriate dataset has constrained our understanding and assessment of VR devices’ susceptibility to presentation attacks. To address this research gap, we introduce VRBiom: a new periocular video dataset acquired from a VR headset (Meta Quest Pro), comprising 900 genuine and 1104 presentation attack videos, each spanning 10 seconds. The bona-fide videos consist of variations in terms of gaze and glasses; while the attacks are constructed with 6 different types of instruments. Additionally, we evaluate the performance of two prominent CNN architectures trained using various configurations for detecting presentation attacks in the newly created VRBiom dataset. Our benchmarking on VRBiom reveals the presence of spoofing threats in VR headsets. While baseline models exhibit considerable efficacy in attack detection, substantial scope exists for improvement in detecting attacks on periocular videos. Our dataset will be a useful resource for researchers aiming to enhance the security and reliability of VR-based authentication systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Kotwal_IJCB-2_2024.pdf}
}
@INPROCEEDINGS{Kotwal_IJCB_2024,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {SAFER, Biometrics Center},
month = sep,
title = {Demographic Fairness Transformer for Bias Mitigation in Face Recognition},
booktitle = {Proceedings of IEEE International Joint Conference on Biometrics (IJCB2024)},
year = {2024},
abstract = {Demographic bias in deep learning-based face recognition systems has led to serious concerns. Often, the biased nature of models is attributed to severely imbalanced datasets used for training. However, several studies have shown that biased models can emerge even when trained on balanced data due to factors in the data acquisition process. Considering the impact of input data on demographic bias, we propose an image to image transformer for demographic fairness (DeFT). This transformer can be applied before the pretrained recognition CNN to selectively enhance the image representation with the goal of reducing the bias through overall recognition pipeline. The multi-head encoders of DeFT provide multiple transformation paths to the input which are then combined based on its demographic information implicitly inferred through soft-attention mechanism applied to intermittent layers of DeFT. We compute probabilistic weights for demographic information, as opposed to conventional hard labels, simplifying the learning process and enhancing the robustness of the DeFT. Our experiments demonstrate that in a cross-dataset testing (pretrained as well as locally trained models), integrating the DeFT leads to fairer models, reducing the variation in accuracies while often slightly improving average recognition accuracy over baselines.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Kotwal_IJCB_2024.pdf}
}
@ARTICLE{Kotwal_TBIOM_2019,
author = {Kotwal, Ketan and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
projects = {Idiap, Tesla},
title = {Multispectral Deep Embeddings As a Countermeasure To Custom Silicone Mask Presentation Attacks},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2019},
abstract = {This work focuses on detecting presentation attacks (PA) mounted using custom silicone masks. Face recognition (FR) systems have been shown to be highly vulnerable to PAs based on such masks [1, 2]. Here
we explore the use of multispectral data (color imagery, near infrared (NIR) imagery and thermal imagery) for face presentation attack detection (PAD), specifically against the custom silicone mask attacks. Using a
new dataset (XCSMAD) representing 21 custom made masks, we establish the baseline performance of several commonly used face-PAD methods, on the different imaging channels. Considering thermal imagery in particular, our experiments show that low-cost thermal imaging devices are as effective in face-PAD as more expensive thermal cameras, for mask-based attacks. This result reinforces the case for the use of thermal data in face-PAD.
We also demonstrate that fusing information from multiple channels leads to significant improvement in face-PAD performance. Finally, we propose a new approach to face-PAD of custom silicone masks using a convolutional neural network (CNN). On individual spectral channels, the proposed approach achieves state-of-the-art results. Using multispectral-fusion, the proposed CNN-based method significantly outperforms the
baseline methods. The new dataset and source-code for our experiments is freely available for research purposes.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Kotwal_TBIOM_2019.pdf}
}
@ARTICLE{Kotwal_TBIOM_2025,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
keywords = {bias, Biometrics, demographic fairness, Differential Performance, Face Recognition, Trustworthy AI},
projects = {SAFER, Biometrics Center},
mainresearchprogram = {AI for Everyone},
title = {Review of Demographic Fairness in Face Recognition},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2025},
crossref = {Kotwal_Idiap-RR-01-2025},
abstract = {The issue of difference in face recognition (FR) performance across demographic groups has emerged as a critical area of research, given its impact on fairness, equity, and reliability across diverse applications. As FR technologies are increasingly deployed globally, disparities in performance across demographic groups– such as race, ethnicity, and gender– have garnered significant attention. These differences or biases not only compromise the credibility of FR systems but also raise ethical concerns, especially when these technologies are employed in sensitive domains. This review consolidates extensive research efforts providing a comprehensive overview of the multifaceted aspects of demographic fairness in FR.
We systematically examine the primary causes, datasets, assessment metrics, and mitigation approaches associated with performance differences in FR across demographic groups. By categorizing key contributions in these areas, this work provides a structured approach to understanding and addressing the complexity of this issue. Finally, we highlight current advancements and identify emerging challenges that need further investigation. This article aims to provide researchers with a unified perspective on the state-of-the-art while emphasizing the critical need for equitable and trustworthy FR systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Kotwal_TBIOM_2025.pdf}
}
@INPROCEEDINGS{Kotwal_WACV-W_2024,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
keywords = {bias mitigation, Demographic bias, Fairness, regularization},
projects = {Idiap},
month = jan,
title = {Mitigating Demographic Bias in Face Recognition via Regularized Score Calibration},
booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision Workshops},
year = {2024},
publisher = {IEEE/CVF},
abstract = {Demographic bias in deep learning-based face recognition systems has led to serious concerns. Several existing works attempt to mitigate bias by incorporating demographic-specific processing during inference, which requires knowledge or learning of demographic attribute with an additional cost. We propose to regularize training of the face recognition CNN, for demographic fairness, by imposing constraints on the distributions of matching scores. Our regularization term enforces the score distributions from different demographic groups to respect a predefined probability distribution, as well as it penalizes misalignment of distributions across demographic groups. The proposed method improves fairness of face recognition models without compromising the recognition accuracy, and does not require extra resources during inference. Our experiments indicate that in a cross-dataset testing, the regularized CNN can reduce the variation in accuracies (i.e., more fairness) of different demographic groups up to 25\% while slightly improving recognition accuracy over baselines.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Kotwal_WACV-W_2024.pdf}
}
@ARTICLE{Krivokuca_ARXIV_2024,
author = {Krivokuca, Vedrana and Maceiras, J{\'{e}}r{\'{e}}my and Komaty, Alain and Abbet, Philip and Marcel, S{\'{e}}bastien},
projects = {MBP4DR},
title = {in-Car Biometrics (iCarB) Datasets for Driver Recognition: Face, Fingerprint, and Voice},
journal = {arXiv},
year = {2024},
url = {https://arxiv.org/abs/2411.17305},
doi = {https://doi.org/10.48550/arXiv.2411.17305},
abstract = {We present three biometric datasets (iCarB-Face, iCarB-Fingerprint, iCarB-Voice) containing face videos, fingerprint images, and voice samples, collected inside a car from 200 consenting volunteers. The data was acquired using a near-infrared camera, two fingerprint scanners, and two microphones, while the volunteers were seated in the driver's seat of the car. The data collection took place while the car was parked both indoors and outdoors, and different "noises" were added to simulate non-ideal biometric data capture that may be encountered in real-life driver recognition. Although the datasets are specifically tailored to in-vehicle biometric recognition, their utility is not limited to the automotive environment. The iCarB datasets, which are available to the research community, can be used to: (i) evaluate and benchmark face, fingerprint, and voice recognition systems (we provide several evaluation protocols); (ii) create multimodal pseudo-identities, to train/test multimodal fusion algorithms; (iii) create Presentation Attacks from the biometric data, to evaluate Presentation Attack Detection algorithms; (iv) investigate demographic and environmental biases in biometric systems, using the provided metadata. To the best of our knowledge, ours are the largest and most diverse publicly available in-vehicle biometric datasets. Most other datasets contain only one biometric modality (usually face), while our datasets consist of three modalities, all acquired in the same automotive environment. Moreover, iCarB-Fingerprint seems to be the first publicly available in-vehicle fingerprint dataset. Finally, the iCarB datasets boast a rare level of demographic diversity among the 200 data subjects, including a 50/50 gender split, skin colours across the whole Fitzpatrick-scale spectrum, and a wide age range (18-60+). So, these datasets will be valuable for advancing biometrics research.}
}
@ARTICLE{Krivokuca_arxiv_PolyProtect_v1,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Towards Protecting Face Embeddings in Mobile Face Verification Scenarios},
journal = {arXiv},
year = {2021},
note = {Version 1 -- Submitted to IEEE T-BIOM},
url = {https://arxiv.org/abs/2110.00434v1},
crossref = {Krivokuca_arxiv_PolyProtect_v2},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_arxiv_PolyProtect_v1.pdf}
}
@ARTICLE{Krivokuca_arxiv_survey_2021,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques},
journal = {arXiv},
year = {2021},
note = {Version 1 -- Submitted to IEEE TIFS},
url = {https://arxiv.org/abs/2110.05044v1},
crossref = {Krivokuca_arxiv_survey_2022},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_arxiv_survey_2021.pdf}
}
@INPROCEEDINGS{Krivokuca_ISBA2018_2018,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Towards Quantifying the Entropy of Fingervein Patterns across Different Feature Extractors},
booktitle = {2018 IEEE 4th International Conference on Identity, Security, and Behavior Analysis (ISBA)},
year = {2018},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Krivokuca_ISBA2018_2018.pdf}
}
@INCOLLECTION{Krivokuca_SPRINGEROPEN-2_2019,
author = {Krivokuca, Vedrana and Gomez-Barrero, Marta and Marcel, S{\'{e}}bastien and Rathgeb, Christian and Busch, Christoph},
editor = {Uhl, Andreas and Busch, Christoph and Marcel, S{\'{e}}bastien and Veldhuis, Raymond},
projects = {SWAN},
title = {Towards Measuring the Amount of Discriminatory Information in Finger Vein Biometric Characteristics Using a Relative Entropy Estimator},
booktitle = {Handbook of Vascular Biometrics},
chapter = {17},
year = {2019},
pages = {507-525},
publisher = {Springer Open},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Krivokuca_SPRINGEROPEN-2_2019.pdf}
}
@INCOLLECTION{Krivokuca_SPRINGEROPEN_2019,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
editor = {Uhl, Andreas and Busch, Christoph and Marcel, S{\'{e}}bastien and Veldhuis, Raymond},
projects = {SWAN},
title = {On the Recognition Performance of BioHash-Protected Finger Vein Templates},
booktitle = {Handbook of Vascular Biometrics},
chapter = {15},
year = {2019},
pages = {465-480},
publisher = {Springer Open},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Krivokuca_SPRINGEROPEN_2019.pdf}
}
@INPROCEEDINGS{KucurErgunay_IEEEBTAS_2015,
author = {Kucur Ergunay, Serife and Khoury, Elie and Lazaridis, Alexandros and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BEAT},
month = sep,
title = {On the Vulnerability of Speaker Verification to Realistic Voice Spoofing},
booktitle = {IEEE International Conference on Biometrics: Theory, Applications and Systems},
year = {2015},
pages = {1-8},
publisher = {IEEE},
url = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7358783},
doi = {10.1109/BTAS.2015.7358783},
pdf = {https://publications.idiap.ch/attachments/papers/2015/KucurErgunay_IEEEBTAS_2015.pdf}
}
@INPROCEEDINGS{Lazaridis_ODYSSEY_2014,
author = {Lazaridis, Alexandros and Khoury, Elie and Goldman, Jean-Philippe and Avanzi, Mathieu and Marcel, S{\'{e}}bastien and Garner, Philip N.},
keywords = {Accent Identification, French Regional Accents, GMM Modelling, i-vectors, SVM},
projects = {Idiap, SIWIS, SNSF-LOBI},
title = {SWISS FRENCH REGIONAL ACCENT IDENTIFICATION},
booktitle = {Odyssey: The Speaker and Language Recognition Workshop},
year = {2014},
abstract = {In this paper an attempt is made to automatically recognize the speaker’s accent among regional Swiss French accents from four different regions of Switzerland, i.e. Geneva (GE), Martigny (MA), Neuchˆatel (NE) and Nyon (NY). To achieve this goal, we rely on a generative probabilistic framework for classification based on Gaussian mixture modelling (GMM). Two different GMM-based algorithms are investigated: (1) the baseline technique of universal background modelling (UBM) followed by maximum-a-posteriori (MAP) adaptation, and (2) total variability (i-vector) modelling. Both systems perform well, with the i-vector-based system outperforming the baseline system, achieving a relative improvement of 17.1\% in the overall regional accent identification accuracy.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Lazaridis_ODYSSEY_2014.pdf}
}
@INPROCEEDINGS{Lee_NISTSRE2012_2012,
author = {Lee, Kong Aik and Saedi, Rahim and Hasan, Tawfik and Kinnunen, Tomi and Fauve, Benoit and Bousquet, Pierre-Michel and Khoury, Elie and Martinez, Pablo Luis Sordo and Thiruvaran, Tharmarajah and You, Changhuai and Rajan, Padmanabhan and Van Leeuwen, David and Sadjadi, Seyed Omid and Matrouf, Driss and El Shafey, Laurent and Mason, John and Ambikairajah, Eliathamby and Sun, Hanwu and Larcher, Anthony and Ma, Bin and Hautam{\"{a}}ki, Ville and Hanilci, Cemal and Braithwaite, Billy and Rosa, Gonzalez-Hautam{\"{a}}ki and Liu, Gang and Boril, Hynek and Shokouhi, Navid and Hansen, John and Bonastre, Jean-Fran{\c c}ois and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI},
month = dec,
title = {The I4U Submission to the 2012 NIST Speaker Recognition Evaluation},
booktitle = {NIST Speaker Recognition Conference},
year = {2012}
}
@INPROCEEDINGS{Linghu_IJCB_2024,
author = {Linghu, Yu and de Freitas Pereira, Tiago and Ecabert, Christophe and Marcel, S{\'{e}}bastien and G{\"{u}}nther, Manuel},
projects = {SAFER},
title = {Score Normalization for Demographic Fairness in Face Recognition},
booktitle = {IEEE International Joint Conference on Biometrics (IJCB 2024)},
year = {2024},
abstract = {Fair biometric algorithms have similar verification performance across different demographic groups given a single decision threshold. Unfortunately, for state-of-the-art face recognition networks, score distributions differ between demographics. Contrary to work that tries to align those distributions by extra training or fine-tuning, we solely focus on score post-processing methods. As proved, well-known sample-centered score normalization techniques, Z-norm and T-norm, do not improve fairness for high-security operating points. Thus, we extend the standard Z/T-norm to integrate demographic information in normalization. Additionally, we investigate several possibilities to incorporate cohort similarities for both genuine and impostor pairs per demographic to improve fairness across different operating points. We run experiments on two datasets with different demographics (gender and ethnicity) and show that our techniques generally improve the overall fairness of five state-of-the-art pre-trained face recognition networks, without downgrading verification performance. We also indicate that an equal contribution of False Match Rate (FMR) and False Non-Match Rate (FNMR) in fairness evaluation is required for the highest gains. Code and protocols are available.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Linghu_IJCB_2024.pdf}
}
@TECHREPORT{Madikeri_Idiap-RR-17-2019,
author = {Madikeri, Srikanth and Sarfjoo, Seyyed Saeed and Motlicek, Petr and Marcel, S{\'{e}}bastien},
projects = {CTI-Shaped, Tesla},
month = {11},
title = {Idiap submission to the NIST SRE 2018 Speaker Recognition Evaluation},
type = {Idiap-RR},
number = {Idiap-RR-17-2019},
year = {2019},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2018/Madikeri_Idiap-RR-17-2019.pdf}
}
@TECHREPORT{Marcel00IRR,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Approches g{\'{e}}n{\'{e}}ratives pour le traitement de s{\'{e}}quences d'images: application {\`{a}} la reconnaissance dynamique des gestes de la main},
type = {Idiap-RR},
number = {Idiap-RR-45-2000},
year = {2000},
institution = {IDIAP},
note = {Submitted: VALGO 2001, France, 2001},
abstract = {Cet article propose deux approches g{\'{e}}n{\'{e}}ratives pour le traitement de s{\'{e}}quences d'images appliqu{\'{e}}es {\`{a}} la reconnaissance dynamique des gestes de la main. Dans un premier temps, un mod{\`{e}}le probabiliste gaussien de r{\'{e}}gions homog{\`{e}}nes de teinte chair (blob) est pr{\'{e}}sent{\'{e}}. Les param{\`{e}}tres des blobs sont calcul{\'{e}}s par un algorithme EM (Expectation-Maximisation). Les blobs obtenus sont moins nombreux et plus stables que les zones de pixels de teinte chair connexes dont les blobs sont issus. Dans un second temps, l'article d{\'{e}}crit un mod{\`{e}}le hybride {\`{a}} base de r{\'{e}}seaux de neurones et de mod{\`{e}}les de Markov cach{\'{e}}s pour traiter des s{\'{e}}quences de donn{\'{e}}es et ainsi reconnaitre les trajectoires form{\'{e}}es par les blobs comme des gestes. Les param{\`{e}}tres du mod{\`{e}}le sont calcul{\'{e}}s {\'{e}}galement par un algorithme EM. Le mod{\`{e}}le obtenu est capable d'effectuer des t{\^{a}}ches de pr{\'{e}}diction et de classification. Une extension g{\'{e}}n{\'{e}}rative de ce mod{\`{e}}le est propos{\'{e}}e pour prendre en compte la probabilit{\'{e}} d'observation des entr{\'{e}}es. Ainsi, le nouveau mod{\`{e}}le g{\'{e}}n{\'{e}}ratif est capable de rejeter une s{\'{e}}quence d'entr{\'{e}}e qui n'a jamais {\'{e}}t{\'{e}} apprise.},
pdf = {https://publications.idiap.ch/attachments/reports/2000/rr00-45.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2000/rr00-45.ps.gz},
ipdinar={2000},
ipdmembership={vision},
language={French},
}
@TECHREPORT{Marcel02-34IRR,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Gestures for Multi-Modal Interfaces: A Review},
type = {Idiap-RR},
number = {Idiap-RR-34-2002},
year = {2002},
institution = {IDIAP},
abstract = {This document presents a review on gestures for multi-modal interfaces and focus on hand gestures. It first introduces the role that the gesture modality plays in human communication. It then describes different types of gestures. Finally, it gives an overview of many techniques for the recognition of hand gestures.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr02-34.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr02-34.ps.gz},
ipdinar={2002},
ipdmembership={vision},
language={English},
}
@TECHREPORT{Marcel02-49IRR,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Robust Face Verification using Skin Color and Neural Networks},
type = {Idiap-RR},
number = {Idiap-RR-49-2002},
year = {2002},
institution = {IDIAP},
abstract = {The performance of face verification systems has steadily improved over the last few years. State-of-the-art methods often use the gray-scale face image as input. In this paper, we use an additional feature to the face image: the skin color. The feature set is tested on a benchmark database, namely XM2VTS, using a simple discriminant artificial neural network. Results show that the proposed model achieves robust state-of-the-art results.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr02-49.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr02-49.ps.gz},
ipdinar={2002},
ipdmembership={vision},
language={English},
}
@TECHREPORT{Marcel02-50IRR,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Evaluation Protocols and Comparative Results for the {T}riesch Hand Posture Database},
type = {Idiap-RR},
number = {Idiap-RR-50-2002},
year = {2002},
institution = {IDIAP},
abstract = {Research efforts in the design of image-based gestural interfaces have steadily increased over the last few years. Numerous approaches have been investigated to recognize gestures such as facial expressions, hand gestures or hand postures. Nevertheless, there exist no reference databases and no standards for the evaluation and the comparison of developed algorithms in gesture recognition and especially in hand posture recognition. This document proposes an evaluation protocol for a benchmark hand posture database, namely the Triesch hand posture database. This document provides also comparative results on this database.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr02-50.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr02-50.ps.gz},
ipdinar={2002},
ipdmembership={vision},
language={English},
}
@INPROCEEDINGS{marcel:2002:cost,
author = {Marcel, S{\'{e}}bastien and Marcel, Christine and Bengio, Samy},
projects = {Idiap},
title = {A State-of-the-art Neural Network for Robust Face Verification},
booktitle = {Proceedings of the COST275 Workshop on The Advent of Biometrics on the Internet},
year = {2002},
address = {Rome, Italy},
crossref = {marcel02-36irr},
pdf = {https://publications.idiap.ch/attachments/reports/2002/marcel_2002_cost.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/marcel_2002_cost.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{marcel:2002:icpr,
author = {Marcel, S{\'{e}}bastien and Bengio, Samy},
projects = {Idiap},
title = {Improving Face Verification using Skin Color Information},
booktitle = {Proceedings of the 16th International Conference on Pattern Recognition},
year = {2002},
publisher = {IEEE Computer Society Press},
crossref = {marcel01-44irr},
pdf = {https://publications.idiap.ch/attachments/reports/2002/marcel_2002_icpr.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/marcel_2002_icpr.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{marcel:2004:afgr,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {A Symmetric Transformation for LDA-based Face Verification},
booktitle = {Proceedings of the 6th International Conference on Automatic Face and Gesture Recognition},
year = {2004},
publisher = {IEEE Computer Society Press},
crossref = {marcel:symfacelda:2003},
pdf = {https://publications.idiap.ch/attachments/reports/2004/marcel_2004_afgr.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2004/marcel_2004_afgr.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:2004:com04-02,
author = {Tiphaigne, Julien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {A video package for Torch},
type = {Idiap-Com},
number = {Idiap-Com-02-2004},
year = {2004},
institution = {IDIAP},
pdf = {https://publications.idiap.ch/attachments/reports/2004/com04-02.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2004/com04-02.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel::2004,
author = {Marcel, S{\'{e}}bastien and Jost, P. and Vandergheynst, P. and Thiran, Jean-Philippe},
projects = {Idiap},
title = {Face Authentication using Client-specific Matching Pursuit},
type = {Idiap-RR},
number = {Idiap-RR-78-2004},
year = {2004},
institution = {IDIAP},
abstract = {In this paper, we address the problem of finding image decompositions that allow good compression performance, and that are also efficient for face authentication. We propose to decompose the face image using Matching Pursuit and to perform the face authentication in the compressed domain using a MLP (Multi-Layer Perceptron) classifier. We provide experimental results and comparisons with PCA and LDA systems on the multi-modal benchmark database BANCA using its associated protocol.},
pdf = {https://publications.idiap.ch/attachments/reports/2004/rr04-78.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2004/rr04-78.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:bancaface:2003,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Verification using LDA and MLP on the BANCA database},
type = {Idiap-RR},
number = {Idiap-RR-66-2003},
year = {2003},
institution = {IDIAP},
abstract = {In this paper, we propose a system for face verification. It describes in detail each stage of the system: the modeling of the face, the extraction of relevant features and the classification of the input face as a client or an impostor. This system is based on LDA feature extraction, successfully used in previous st and MLP for classification. Experiments were carried out on a difficult multi-modal database, namely BANCA. Results show that our approach perform better than the state-of-the-art on the same database. Experiments show also contradictory results in the state-of-the-art literature.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-66.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-66.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:com06-06,
author = {Marcel, S{\'{e}}bastien and Rodriguez, Yann and Guillemot, Ma{\"{e}}l and Popescu-Belis, Andrei},
projects = {Idiap},
title = {Annotation of face detection: description of XML format and files},
type = {Idiap-Com},
number = {Idiap-Com-06-2006},
year = {2006},
institution = {IDIAP},
pdf = {https://publications.idiap.ch/attachments/reports/2006/marcel-idiap-com-06-06.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/marcel-idiap-com-06-06.ps.gz},
ipdmembership={vision},
}
@ARTICLE{marcel:ieee-tpamisi:2007,
author = {Marcel, S{\'{e}}bastien and Mill{\'{a}}n, Jos{\'{e}} del R.},
projects = {Idiap},
title = {Person Authentication using Brainwaves (EEG) and Maximum A Posteriori Model Adaptation},
journal = {{IEEE} TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE Special Issue on Biometrics},
year = {2007},
note = {IDIAP-RR 05-81},
crossref = {marcel:rr05-81},
abstract = {In this paper, we investigate the use of brain activity for person authentication. It has been shown in previous studies that the brain-wave pattern of every individual is unique and that the electroencephalogram (EEG) can be used for biometric identification. EEG-based biometry is an emerging research topic and we believe that it may open new research directions and applications in the future. However, very little work has been done in this area and was focusing mainly on person identification but not on person authentication. Person authentication aims to accept or to reject a person claiming an identity, i.e comparing a biometric data to one template, while the goal of person identification is to match the biometric data against all the records in a database. We propose the use of a statistical framework based on Gaussian Mixture Models and Maximum A Posteriori model adaptation, successfully applied to speaker and face authentication, which can deal with only one training session. We perform intensive experimental simulations using several strict train/test protocols to show the potential of our method. We also show that there are some mental tasks that are more appropriate for person authentication than others.},
pdf = {https://publications.idiap.ch/attachments/papers/2007/marcel-ieee-tpamisi-2007.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2007/marcel-ieee-tpamisi-2007.ps.gz},
ipdmembership={vision},
}
@ARTICLE{marcel:ijivp:2007,
author = {Marcel, S{\'{e}}bastien and Rodriguez, Yann and Heusch, Guillaume},
projects = {Idiap},
title = {On the Recent Use of Local Binary Patterns for Face Authentication},
journal = {International Journal on Image and Video Processing Special Issue on Facial Image Processing},
year = {2007},
note = {IDIAP-RR 06-34, accepted for publication but withdrawn because of author charges.},
crossref = {marcel:rr06-34},
abstract = {This paper presents a survey on the recent use of Local Binary Patterns (LBPs) for face recognition. LBP is becoming a popular technique for face representation. It is a non-parametric kernel which summarizes the local spacial structure of an image and it is invariant to monotonic gray-scale transformations. This is a very interesting property in face recognition. This probably explains the recent success of Local Binary Patterns in face recognition. In this paper, we describe the LBP technique and different approaches proposed in the literature to represent and to recognize faces. The most representatives are considered for experimental comparison on a common face authentication task. For that purpose, the XM2VTS and BANCA databases are used according to their respective experimental protocols.},
pdf = {https://publications.idiap.ch/attachments/papers/2007/marcel-ijivp-2007.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2007/marcel-ijivp-2007.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:improvefacesymface:2003,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Improving Face Verification using Symmetric Transformation},
type = {Idiap-RR},
number = {Idiap-RR-68-2003},
year = {2003},
institution = {IDIAP},
abstract = {One of the major problem in face verification is to deal with a few number of images per person to train the system. A solution to that problem is to generate virtual samples from an unique image by doing simple geometric transformations such as translation, scale, rotation and vertical mirroring. In this paper, we propose to use a symmetric transformation to generate a new virtual sample. This symmetric virtual sample is obtained by computing the average between the original image and the vertical mirrored image. The face verification system is based on LDA feature extraction, successfully used in previous studies, and MLP for classification. Experiments were carried out on a difficult multi-modal data\-base, namely BANCA. Results on this database show that our face verification system performs better that the state-of-the-art and also that the addition of the symmetric virtual sample improves the performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-68.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-68.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{marcel:mmua:2006,
author = {Marcel, S{\'{e}}bastien and Mari{\'{e}}thoz, Johnny and Rodriguez, Yann and Cardinaux, Fabien},
projects = {Idiap},
title = {Bi-Modal Face and Speech Authentication: a BioLogin Demonstration System},
booktitle = {Workshop on Multimodal User Authentication ({MMUA})},
year = {2006},
note = {IDIAP-RR 06-18},
crossref = {marcel:rr06-18},
abstract = {This paper presents a bi-modal (face and speech) authentication demonstration system that simulates the login of a user using its face and its voice. This demonstration is called BioLogin. It runs both on Linux and Windows and the Windows version is freely available for download. Bio\-Login is implemented using an open source machine learning library and its machine vision package.},
pdf = {https://publications.idiap.ch/attachments/papers/2006/marcel-mmua-2006.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2006/marcel-mmua-2006.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr06-47,
author = {Marcel, S{\'{e}}bastien and Keomany, Jean and Rodriguez, Yann},
projects = {Idiap},
title = {Robust-to-Illumination Face Localisation using Active Shape Models and Local Binary Patterns},
type = {Idiap-RR},
number = {Idiap-RR-47-2006},
year = {2006},
institution = {IDIAP},
note = {Submitted for publication},
abstract = {This paper addresses the problem of locating facial features in images of frontal faces taken under different lighting conditions. The well-known Active Shape Model method proposed by Cootes {\it et al.} is extended to improve its robustness to illumination changes. For that purpose, we introduce the use of Local Binary Patterns (LBP). Three different incremental approaches combining ASM with LBP are presented: profile-based LBP-ASM, square-based LBP-ASM and divided-square-based LBP-ASM. Experiments performed on the standard and darkened image sets of the XM2VTS database demonstrate that the divided-square-based LBP-ASM gives superior performance compared to the state-of-the-art ASM. It achieves more accurate results and fails less frequently.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/marcel-idiap-rr-06-47.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/marcel-idiap-rr-06-47.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr07-07,
author = {Marcel, S{\'{e}}bastien and Abbet, Philip and Guillemot, Ma{\"{e}}l},
projects = {Idiap},
title = {Google Portrait},
type = {Idiap-Com},
number = {Idiap-Com-07-2007},
year = {2007},
institution = {IDIAP},
abstract = {This paper presents a system to retrieve and browse images from the Internet containing only one particular object of interest: the human face. This system, called Google Portrait, uses Google Image search engine to retrieve images matching a text query and filters images containing faces using a face detector. Results and ranked by portraits and a tagging module is provided to change manually the label attached to faces.},
pdf = {https://publications.idiap.ch/attachments/reports/2007/marcel-idiap-com-07-07.pdf},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr07-14,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Joint Bi-Modal Face and Speaker Authentication using Explicit Polynomial Expansion},
type = {Idiap-RR},
number = {Idiap-RR-14-2007},
year = {2007},
institution = {IDIAP},
note = {Submitted for publication},
pdf = {https://publications.idiap.ch/attachments/reports/2007/marcel-idiap-rr-07-14.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2007/marcel-idiap-rr-07-14.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{Marcel_Idiap-RR-09-2010,
author = {Marcel, S{\'{e}}bastien and McCool, Chris and Matejka, Pavel and Ahonen, Timo and Cernocky, Jan},
projects = {MOBIO},
month = {5},
title = {Mobile Biometry (MOBIO) Face and Speaker Verification Evaluation},
type = {Idiap-RR},
number = {Idiap-RR-09-2010},
year = {2010},
institution = {Idiap},
address = {rue Marconi 19},
abstract = {This paper evaluates the performance of face and speaker verification techniques in
the context of a mobile environment. The mobile environment was chosen as it provides a realistic
and challenging test-bed for biometric person verification techniques to operate. For instance the
audio environment is quite noisy and there is limited control over the illumination conditions and
the pose of the subject for the video. To conduct this evaluation, a part of a database captured during
the ``Mobile Biometry'' (MOBIO) European Project was used. In total there were nine participants
to the evaluation who submitted a face verification system and five participants who submitted
speaker verification systems.
The nine face verification systems all varied significantly in terms of both verification algorithms
and face detection algorithms. Several systems used the OpenCV face detector while the better systems
used proprietary software for the task of face detection. This ended up making the evaluation of
verification algorithms challenging.
The five speaker verification systems were based on one of two paradigms: a Gaussian Mixture Model (GMM)
or Support Vector Machine (SVM) paradigm. In general the systems based on the SVM paradigm performed
better than those based on the GMM paradigm.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Marcel_Idiap-RR-09-2010.pdf}
}
@TECHREPORT{Marcel_Idiap-RR-30-2010,
author = {Marcel, S{\'{e}}bastien and McCool, Chris and Matejka, Pavel and Ahonen, Timo and Cernocky, Jan and al},
projects = {MOBIO},
month = {8},
title = {On the Results of the First Mobile Biometry (MOBIO) Face and Speaker Verification Evaluation},
type = {Idiap-RR},
number = {Idiap-RR-30-2010},
year = {2010},
institution = {Idiap},
abstract = {This paper evaluates the performance of face and speaker verification
techniques in the context of a mobile environment. The mobile environment was
chosen as it provides a realistic and challenging test-bed for biometric person
verification techniques to operate. For instance the audio environment is quite
noisy and there is limited control over the illumination conditions and the pose of
the subject for the video. To conduct this evaluation, a part of a database captured
during the {\^{a}}€{\oe}Mobile Biometry{\^{a}}€ (MOBIO) European Project was used. In total
there were nine participants to the evaluation who submitted a face verification
system and five participants who submitted speaker verification systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Marcel_Idiap-RR-30-2010.pdf}
}
@TECHREPORT{Marcel_Idiap-RR-31-2010,
author = {Marcel, S{\'{e}}bastien and McCool, Chris and Atanasoaei, Cosmin and Tarsetti, Flavio and Pesan, Jan and Matejka, Pavel and Cernocky, Jan and Helistekangas, Mika and Turtinen, Markus},
projects = {MOBIO},
month = {8},
title = {MOBIO: Mobile Biometric Face and Speaker Authentication},
type = {Idiap-RR},
number = {Idiap-RR-31-2010},
year = {2010},
institution = {Idiap},
address = {rue Marconi 19},
abstract = {This paper presents a mobile biometric person authentication demonstration system. It consists of verifying a user's claimed identity by biometric means and more particularly using their face and their voice simultaneously on a Nokia N900 mobile device with its built-in sensors (frontal video camera and microphone).},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Marcel_Idiap-RR-31-2010.pdf}
}
@INPROCEEDINGS{McCool_ICB2009_2009,
author = {McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Parts-Based Face Verification using Local Frequency Bands},
booktitle = {in Proceedings of IEEE/IAPR International Conference on Biometrics},
year = {2009},
crossref = {McCool_Idiap-RR-03-2009},
pdf = {https://publications.idiap.ch/attachments/papers/2009/McCool_ICB2009_2009.pdf}
}
@TECHREPORT{McCool_Idiap-Com-02-2009,
author = {McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {11},
title = {MOBIO Database for the ICPR 2010 Face and Speech Competition},
type = {Idiap-Com},
number = {Idiap-Com-02-2009},
year = {2009},
institution = {Idiap},
abstract = {This document presents an overview of the mobile biometry (MOBIO) database. This document is written expressly for the face and speech organised for the 2010 International Conference on Pattern Recognition.},
pdf = {https://publications.idiap.ch/attachments/reports/2009/McCool_Idiap-Com-02-2009.pdf}
}
@TECHREPORT{McCool_Idiap-RR-06-2011,
author = {McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {3},
title = {Parts-Based Face Verification using Local Frequency Bands},
type = {Idiap-RR},
number = {Idiap-RR-06-2011},
year = {2011},
institution = {Idiap},
abstract = {In this paper we extend the Parts-Based approach of face verification by performing
a frequency-based decomposition. The Parts-Based approach divides
the face into a set of blocks which are then considered to be separate observations,
this is a spatial decomposition of the face. This paper extends the
Parts-Based approach by also dividing the face in the frequency domain and
treating each frequency response from an observation separately. This can be
expressed as forming a set of sub-images where each sub-image represents the
response to a different frequency of, for instance, the Discrete Cosine Transform.
Each of these sub-images is treated separately by a Gaussian Mixture Model
(GMM) based classifier. The classifiers from each sub-image are then combined
using weighted summation with the weights being derived using linear logistic
regression. It is shown on the BANCA database that this method improves the
performance of the system from an Average Half Total Error Rate of 26.59\% for
a baseline GMM Parts-Based system to 14.85\% for a column-based approach
on the frequency sub-images, for Protocol P.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/McCool_Idiap-RR-06-2011.pdf}
}
@INPROCEEDINGS{McCool_IEEEICMEWORKSHOPONHOTTOPICSINMOBILEMULTIMEDIA_2012,
author = {McCool, Chris and Marcel, S{\'{e}}bastien and Hadid, Abdenour and Pietikainen, Matti and Matejka, Pavel and Cernocky, Jan and Poh, Norman and Kittler, J. and Larcher, Anthony and Levy, Christophe and Matrouf, Driss and Bonastre, Jean-Fran{\c c}ois and Tresadern, Phil and Cootes, Timothy},
projects = {Idiap, MOBIO, TABULA RASA},
month = jul,
title = {Bi-Modal Person Recognition on a Mobile Phone: using mobile phone data},
booktitle = {IEEE ICME Workshop on Hot Topics in Mobile Multimedia},
year = {2012},
crossref = {McCool_Idiap-RR-13-2012},
abstract = {This paper presents a novel fully automatic bi-modal, face and speaker, recognition system which runs in real-time on a mobile phone. The implemented system runs in real-time on a Nokia N900 and demonstrates the feasibility of performing both automatic face and speaker recognition on a mobile phone. We evaluate this recognition system on a novel publicly-available mobile phone database and provide a well defined evaluation protocol. This database was captured almost exclusively using mobile phones and aims to improve research into deploying biometric techniques to mobile devices. We show, on this mobile phone database, that face and speaker recognition can be performed in a mobile environment and using score fusion can improve the performance by more than 25\% in terms of error rates.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/McCool_IEEEICMEWORKSHOPONHOTTOPICSINMOBILEMULTIMEDIA_2012.pdf}
}
@ARTICLE{McCool_IET_BMT_2013,
author = {McCool, Chris and Wallace, Roy and McLaren, Mitchell and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {BBfor2, TABULA RASA},
month = sep,
title = {Session variability modelling for face authentication},
journal = {IET Biometrics},
volume = {2},
number = {3},
year = {2013},
pages = {117-129},
issn = {2047-4938},
doi = {10.1049/iet-bmt.2012.0059},
crossref = {McCool_Idiap-RR-17-2013},
abstract = {This study examines session variability modelling for face authentication using Gaussian mixture models. Session variability modelling aims to explicitly model and suppress detrimental within-class (inter-session) variation. The authors examine two techniques to do this, inter-session variability modelling (ISV) and joint factor analysis (JFA), which were initially developed for speaker authentication. We present a self-contained description of these two techniques and demonstrate that they can be successfully applied to face authentication. In particular, they show that using ISV leads to significant error rate reductions of, on average, 26\% on the challenging and publicly available databases SCface, BANCA, MOBIO and multi-PIE. Finally, the authors show that a limitation of both ISV and JFA for face authentication is that the session variability model captures and suppresses a significant portion of between-class variation.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/McCool_IET_BMT_2013.pdf}
}
@ARTICLE{McCool_PRL_2010,
author = {McCool, Chris and Sanchez-Riera, Jordi and Marcel, S{\'{e}}bastien},
keywords = {3D face recognition, Face, Face Recognition, Feature Distribution Modelling, GMM, HMM},
projects = {Idiap, MOBIO},
title = {Feature distribution modelling techniques for 3D face recognition},
journal = {Pattern Recognition Letters},
volume = {31},
year = {2010},
pages = {1324-1330},
abstract = {This paper shows that Hidden Markov Models (HMMs) can be effectively ap-
plied to 3D face data. The examined HMM techniques are shown to be superior
to a previously examined Gaussian Mixture Model (GMM) technique. Experi-
ments conducted on the Face Recognition Grand Challenge database show that
the Equal Error Rate can be reduced from 0.88\% for the GMM technique to
0.36\% for the best HMM approach.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/McCool_PRL_2010.pdf}
}
@ARTICLE{Melzi_ACCESS_2025,
author = {Melzi, Pietro and Otroshi Shahreza, Hatef and Rathgeb, Christian and Tolosana, Ruben and Vera-Rodriguez, Ruben and Fierrez, Julian and Marcel, S{\'{e}}bastien and Busch, Christoph},
projects = {TRESPASS-ETN},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
title = {Cancelable Face Biometrics With Soft-Biometric Privacy Enhancement},
journal = {IEEE Access},
year = {2025},
url = {https://ieeexplore.ieee.org/abstract/document/11086598},
doi = {10.1109/ACCESS.2025.3590989},
abstract = {The storage of biometric data has raised significant privacy concerns, necessitating robust measures for secure storage. While traditional Privacy-Enhancing Technologies (PETs), like Cancelable Biometric (CB) schemes, excel at creating protected templates that fulfill criteria such as irreversibility and unlinkability, they often fail to preserve the privacy of soft-biometric information. To address this issue, we propose a hybrid technology that combines PETs, leveraging their different properties to comprehensively address multiple privacy requirements and enhance overall protection for biometric templates. In our approach, we integrate Multi Incremental Variable Elimination (Multi-IVE), a recent technology designed to remove soft-biometric information from biometric templates, with conventional CB schemes. We apply our hybrid technology to facial templates and assess the properties of the resulting protected templates. In the event of stolen secrets, the combination of Multi-IVE with CB schemes helps decrease the accuracy of estimating soft-biometric attributes without affecting recognition performance, compared to CB schemes alone.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Melzi_ACCESS_2025.pdf}
}
@INPROCEEDINGS{Melzi_WACVW_2023,
author = {Melzi, Pietro and Otroshi Shahreza, Hatef and Rathgeb, Christian and Tolosana, Ruben and Vera-Rodriguez, Ruben and Fierrez, Julian and Marcel, S{\'{e}}bastien and Busch, Christoph},
projects = {TRESPASS-ETN},
month = jan,
title = {Multi-IVE: Privacy Enhancement of Multiple Soft-Biometrics in Face Embeddings},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
year = {2023},
url = {https://openaccess.thecvf.com/content/WACV2023W/DVPBA/html/Melzi_Multi-IVE_Privacy_Enhancement_of_Multiple_Soft-Biometrics_in_Face_Embeddings_WACVW_2023_paper.html},
abstract = {This study focuses on the protection of soft-biometric attributes related to the demographic information of individuals that can be extracted from compact representations of face images, called embeddings. We consider a state-of-the-art technology for soft-biometric privacy enhancement, Incremental Variable Elimination (IVE), and propose Multi-IVE, a new method based on IVE to secure multiple soft-biometric attributes simultaneously. Several aspects of this technology are investigated, proposing different approaches to effectively identify and discard multiple soft-biometric attributes contained in face embeddings. In particular, we consider a domain transformation using Principle Component Analysis (PCA), and apply IVE in the PCA domain. A complete analysis of the proposed Multi-IVE algorithm is carried out studying the embeddings generated by state-of-the-art face feature extractors, predicting soft-biometric attributes contained within them with multiple machine learning classifiers, and providing a cross-database evaluation. The results obtained show the possibility to simultaneously secure multiple soft-biometric attributes and support the application of embedding domain transformations before addressing the enhancement of soft-biometric privacy.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Melzi_WACVW_2023.pdf}
}
@INPROCEEDINGS{Metha_SCIA_2015,
author = {Metha, Rakesh and G{\"{u}}nther, Manuel and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, BEAT},
title = {Gender Classification by LUT based boosting of Overlapping Block Patterns},
booktitle = {Scandinavian Conference on Image Analysis},
volume = {9127},
year = {2015},
pages = {530-542},
publisher = {Springer International Publishing},
url = {http://link.springer.com/chapter/10.1007%2F978-3-319-19665-7_45},
doi = {10.1007/978-3-319-19665-7_45},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Metha_SCIA_2015.pdf}
}
@ARTICLE{Mohammadi_IETBIOMETRICS_2017,
author = {Mohammadi, Amir and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
projects = {Idiap, Tesla, SWAN},
title = {Deeply Vulnerable -- a study of the robustness of face recognition to presentation attacks},
journal = {IET (The Institution of Engineering and Technology) -- Biometrics},
year = {2017},
pages = {1--13},
note = {Accepted on 29-Sept-2017},
issn = {2047-4938},
doi = {10.1049/iet-bmt.2017.0079},
abstract = {The vulnerability of deep-learning-based face-recognition (FR) methods, to presentation attacks (PA), is studied in this study. Recently, proposed FR methods based on deep neural networks (DNN) have been shown to outperform most other methods by a significant margin. In a trustworthy face-verification system, however, maximising recognition-performance alone is not sufficient – the system should also be capable of resisting various kinds of attacks, including PA. Previous experience has shown that the PA vulnerability of FR systems tends to increase with face-verification accuracy. Using several publicly available PA datasets, the authors show that DNN-based FR systems compensate for variability between bona fide and PA samples, and tend to score them similarly, which makes such FR systems extremely vulnerable to PAs. Experiments show the vulnerability of the studied DNN-based FR systems to be consistently higher than 90\%, and often higher than 98\%.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Mohammadi_IETBIOMETRICS_2017.pdf}
}
@INPROCEEDINGS{Mohammadi_InfoVAE_ICASSP_2020,
author = {Mohammadi, Amir and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
keywords = {cross-dataset evaluation, domain generalization, mobile biometrics, Presentation Attack Detection},
projects = {SWAN},
title = {IMPROVING CROSS-DATASET PERFORMANCE OF FACE PRESENTATION ATTACK DETECTION SYSTEMS USING FACE RECOGNITION DATASETS},
booktitle = {45th International Conference on Acoustics, Speech, and Signal Processing},
year = {2020},
publisher = {IEEE},
url = {https://gitlab.idiap.ch/bob/bob.paper.icassp2020_facepad_generalization_infovae},
abstract = {Presentation attack detection (PAD) is now considered critically important for any face-recognition (FR) based access-control system. Current deep-learning based PAD systems show excellent performance when they are tested in intra-dataset scenarios. Under cross-dataset evaluation the performance of these PAD systems drops significantly. This lack of generalization is attributed to domain-shift. Here, we propose a novel PAD method that leverages the large variability present in FR datasets to induce invariance to factors that cause domain-shift. Evaluation of the proposed method on several datasets, including datasets collected using mobile devices, shows performance improvements in cross-dataset evaluations.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Mohammadi_InfoVAE_ICASSP_2020.pdf}
}
@INPROCEEDINGS{Mohammadi_Pruning_ICASSP_2020,
author = {Mohammadi, Amir and Bhattacharjee, Sushil and Marcel, S{\'{e}}bastien},
keywords = {domain adaptation, domain generalization, feature selection, Presentation Attack Detection, pruning},
projects = {SWAN},
title = {DOMAIN ADAPTATION FOR GENERALIZATION OF FACE PRESENTATION ATTACK DETECTION IN MOBILE SETTINGS WITH MINIMAL INFORMATION},
booktitle = {45th International Conference on Acoustics, Speech, and Signal Processing},
year = {2020},
publisher = {IEEE},
location = {Barcelona, Spain},
url = {https://gitlab.idiap.ch/bob/bob.paper.icassp2020_domain_guided_pruning},
abstract = {With face-recognition (FR) increasingly replacing fingerprint sensors for user-authentication on mobile devices, presentation attacks (PA) have emerged as the single most significant hurdle for manufacturers of FR systems. Current machine-learning based presentation attack detection (PAD) systems, trained in a data-driven fashion, show excellent performance when evaluated in intra-dataset scenarios. Their performance typically degrades significantly in cross-dataset evaluations. This lack of generalization in current PAD systems makes them unsuitable for deployment in real-world scenarios. Considering each dataset as representing a different domain, domain adaptation techniques have been proposed as a solution to this generalization problem. Here, we propose a novel one class domain adaptation method which uses domain guided pruning to adapt a pre-trained PAD network to the target dataset. The proposed method works without the need of collecting PAs in the target domain (i.e., with minimal information in the target domain). Experimental results on several datasets show promising performance improvements in cross-dataset evaluations.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Mohammadi_ICASSP2020_2020.pdf}
}
@ARTICLE{Morales_IEEEACCESS_2016,
author = {Morales, Aythami and Fierrez, Julian and Tolosana, Ruben and Ortega-Garcia, Javier and Galbally, Javier and Gomez-Barrero, Marta and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {BEAT},
month = nov,
title = {Keystroke Biometrics Ongoing Competition},
journal = {IEEE Access},
volume = {4},
year = {2016},
pages = {7736-7746},
issn = {2169-3536},
url = {http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7738412},
doi = {10.1109/ACCESS.2016.2626718},
abstract = {This paper presents the first Keystroke Biometrics Ongoing Competition (KBOC) organized to establish a reproducible baseline in person authentication using keystroke biometrics. The competition has been developed using the BEAT platform and includes one of the largest keystroke databases publicly available based on a fixed text scenario. The database includes genuine and attacker keystroke sequences from 300 users acquired in 4 different sessions distributed in a four month time span. The sequences correspond to the user's name and surname and therefore each user comprises an individual and personal sequence. As baseline for KBOC we report the results of 31 different algorithms evaluated according to performance and robustness. The systems have achieved EERs as low as 5.32\% and high robustness against multisession variability with drop of performances lower than 1\% for probes separated by months. The entire database is publicly available at the competition website.},
pdf = {https://publications.idiap.ch/attachments/papers/2016/Morales_IEEEACCESS_2016.pdf}
}
@TECHREPORT{Mostaani_Idiap-RR-22-2020,
author = {Mostaani, Zohreh and George, Anjith and Heusch, Guillaume and Geissbuhler, David and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
month = {9},
title = {The High-Quality Wide Multi-Channel Attack (HQ-WMCA) database},
type = {Idiap-RR},
number = {Idiap-RR-22-2020},
year = {2020},
institution = {Idiap},
abstract = {The High-Quality Wide Multi-Channel Attack database (HQ-WMCA) database extends the previous Wide Multi-Channel Attack database(WMCA) \cite{george_mccnn_tifs2019}, with more channels
including color, depth, thermal, infrared (spectra), and short-wave infrared (spectra), and also a wide variety of attacks.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/Mostaani_Idiap-RR-22-2020.pdf}
}
@INPROCEEDINGS{Motlicek_ICPR_2012,
author = {Motlicek, Petr and El Shafey, Laurent and Wallace, Roy and McCool, Chris and Marcel, S{\'{e}}bastien},
keywords = {face verification, Speaker identification},
projects = {Idiap, BBfor2, TA2, TABULA RASA},
month = nov,
title = {Bi-Modal Authentication in Mobile Environments Using Session Variability Modelling},
booktitle = {Proceedings of the 21st International Conference on Pattern Recognition},
year = {2012},
crossref = {Motlicek_Idiap-RR-18-2012},
abstract = {We present a state-of-the-art bi-modal authentication system for mobile environments, using session variability modelling. We examine inter-session variability modelling (ISV) and joint factor analysis (JFA) for both face and speaker authentication and evaluate our system on the largest bi-modal mobile authentication database available, the MOBIO database, with over 61 hours of audio-visual data captured by 150 people in uncontrolled environments on a mobile phone. Our system achieves 2.6\% and 9.7\% half total error rate for male and female trials respectively – relative improvements of 78\% and 27\% compared to previous results.},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Motlicek_ICPR_2012.pdf}
}
@INPROCEEDINGS{Muckenhirn_BIOSIG_2016,
author = {Muckenhirn, Hannah and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {long-term statistics, presentation attack, speaker verification, Spoofing},
projects = {Idiap, UNITS, SWAN},
month = sep,
title = {Presentation Attack Detection Using Long-Term Spectral Statistics for Trustworthy Speaker Verification},
booktitle = {International Conference of the Biometrics Special Interest Group (BIOSIG)},
year = {2016},
abstract = {In recent years, there has been a growing interest in developing countermeasures against non zero-effort attacks for speaker verification systems. Until now, the focus has been on logical access attacks, where the spoofed samples are injected into the system through a software-based process. This paper investigates a more realistic type of attack, referred to as physical access or presentation attacks, where the spoofed samples are presented as input to the microphone. To detect such attacks, we propose a binary classifier based approach that uses long-term spectral statistics as feature input. Experimental studies on the AVspoof database, which contains presentation attacks based on replay, speech synthesis and voice conversion, shows that the proposed approach can yield significantly low detection error rate with a linear classifier (half total error rate of 0.038\%). Furthermore, an investigation on Interspeech 2015 ASVspoof challenge dataset shows that it is equally capable of detecting logical access attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Muckenhirn_BIOSIG_2016.pdf}
}
@INPROCEEDINGS{Muckenhirn_ICASSP_2018,
author = {Muckenhirn, Hannah and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {Convolutional neural network, End-to-end learning, Fundamental frequency, recognition, speaker verification},
projects = {Idiap, UNITS},
month = apr,
title = {Towards directly modeling raw speech signal for speaker verification using CNNs},
booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2018},
pages = {4884-4888},
location = {Calgary, CANADA},
isbn = {978-1-5386-4658-8},
crossref = {Muckenhirn_Idiap-RR-30-2017},
abstract = {Speaker verification systems traditionally extract and model cepstral features or filter bank energies from the speech signal. In this paper, inspired by the success of neural network-based approaches to model directly raw speech signal for applications such as speech recognition, emotion recognition and anti-spoofing, we propose a speaker verification approach where speaker discriminative information is directly learned from the speech signal by: (a) first training a CNN-based speaker identification system that takes as input raw speech signal and learns to classify on speakers (unknown to the speaker verification system); and then (b) building a speaker detector for each speaker in the speaker verification system by replacing the output layer of the speaker identification system by two outputs (genuine, impostor), and adapting the system in a discriminative manner with enrollment speech of the speaker and impostor speech data. Our investigations on the Voxforge database shows that this approach can yield systems competitive to state-of-the-art systems. An analysis of the filters in the first convolution layer shows that the filters give emphasis to information in low frequency regions (below 1000 Hz) and implicitly learn to model fundamental frequency information in the speech signal for speaker discrimination.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Muckenhirn_ICASSP_2018.pdf}
}
@INPROCEEDINGS{Muckenhirn_IJCB_2017,
author = {Muckenhirn, Hannah and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, UNITS, SWAN},
title = {End-to-End Convolutional Neural Network-based Voice Presentation Attack Detection},
booktitle = {International Joint Conference on Biometrics},
year = {2017},
location = {Denver, Colorado, USA},
abstract = {Development of countermeasures to detect attacks performed on speaker verification systems through presentation of forged or altered speech samples is a challenging and open research problem. Typically, this problem is approached by extracting features through conventional short-term speech processing and feeding them to a binary classifier. In this article, we develop a convolutional neural network-based approach that learns in an end-to-end manner both the features and the binary classifier from the raw signal. Through investigations on two publicly available databases, namely, ASVspoof and AVspoof, we show that it yields systems comparable to or better than the state-of-the-art approaches for both physical access attacks and logical access attacks. Furthermore, the approach is shown to be complementary to a spectral statistics-based approach, which, similarly to the proposed approach, does not use prior assumptions related to speech signals.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Muckenhirn_IJCB_2017.pdf}
}
@INPROCEEDINGS{Muckenhirn_INTERSPEECH_2018,
author = {Muckenhirn, Hannah and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {Convolutional neural network, End-to-end learning, Formants, Fundamental frequency, speaker verification},
projects = {Idiap},
month = sep,
title = {On Learning Vocal Tract System Related Speaker Discriminative Information from Raw Signal Using CNNs},
booktitle = {Proceedings of Interspeech},
year = {2018},
pages = {1116-1120},
location = {Hyderabad, INDIA},
issn = {2308-457X},
isbn = {978-1-5108-7221-9},
abstract = {In a recent work, we have shown that speaker verification systems can be built where both features and classifiers are directly learned from the raw speech signal with convolutional neural networks (CNNs). In this framework, the training phase also decides the block processing through cross validation. It was found that the first convolution layer, which processes about 20 ms speech, learns to model fundamental frequency information. In the present paper, inspired from speech recognition studies, we build further on that framework to design a CNN-based system, which models sub-segmental speech (about 2ms speech) in the first convolution layer, with an hypothesis that such a system should learn vocal tract system related speaker discriminative information. Through experimental studies on Voxforge corpus and analysis on American vowel dataset, we show that the proposed system (a) indeed focuses on formant regions, (b) yields competitive speaker verification system and (c) is complementary to the CNN-based system that models fundamental frequency information.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Muckenhirn_INTERSPEECH_2018.pdf}
}
@INPROCEEDINGS{Muckenhirn_INTERSPEECH_2019,
author = {Muckenhirn, Hannah and Abrol, Vinayak and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {CNN visualization, deep learning, raw waveforms},
projects = {Idiap, UNITS},
title = {Understanding and Visualizing Raw Waveform-based CNNs},
booktitle = {Proceedings of Interspeech},
year = {2019},
crossref = {Muckenhirn_Idiap-RR-11-2018},
abstract = {Modeling directly raw waveforms through neural networks for speech processing is gaining more and more attention. Despite its varied success, a question that remains is: what kind of information are such neural networks capturing or learning for different tasks from the speech signal? Such an insight is not only interesting for advancing those techniques but also for understanding better speech signal characteristics. This paper takes a step in that direction, where we develop a gradient based approach to estimate the relevance of each speech sample input on the output score. We show that analysis of the resulting ``relevance signal" through conventional speech signal processing techniques can reveal the information modeled by the whole network. We demonstrate the potential of the proposed approach by analyzing raw waveform CNN-based phone recognition and speaker identification systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Muckenhirn_INTERSPEECH_2019.pdf}
}
@ARTICLE{Muckenhirn_TASLP_2017,
author = {Muckenhirn, Hannah and Korshunov, Pavel and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, cross-database, Presentation Attack Detection, spectral statistics},
projects = {Idiap, UNITS, SWAN, Tesla},
month = nov,
title = {Long-Term Spectral Statistics for Voice Presentation Attack Detection},
journal = {IEEE/ACM Transactions on Audio, Speech and Language Processing},
volume = {25},
number = {11},
year = {2017},
pages = {2098-2111},
crossref = {Muckenhirn_Idiap-RR-11-2017},
abstract = {Automatic speaker verification systems can be spoofed through recorded, synthetic or voice converted speech of target speakers. To make these systems practically viable, the detection of such attacks, referred to as presentation attacks, is of paramount interest. In that direction, this paper investigates two aspects: (a) a novel approach to detect presentation attacks where, unlike conventional approaches, no speech signal modeling related assumptions are made, rather the attacks are detected by computing first order and second order spectral statistics and feeding them to a classifier, and (b) generalization of the presentation attack detection systems across databases. Our investigations on ASVspoof 2015 challenge database and AVspoof database show that, when compared to the approaches based on conventional short-term spectral features, the proposed approach with a linear discriminative classifier yields a better system, irrespective of whether the spoofed signal is replayed to the microphone or is directly injected into the system software process. Cross-database investigations show that neither the short-term spectral processing based approaches nor the proposed approach yield systems which are able to generalize across databases or methods of attack. Thus, revealing the difficulty of the problem and the need for further resources and research.},
pdf = {https://publications.idiap.ch/attachments/papers/2017/Muckenhirn_TASLP_2017.pdf}
}
@INPROCEEDINGS{Nguyen_IJCB_2020,
author = {Nguyen, Huy H. and Yamagishi, Junichi and Echizen, Isao and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, deep learning, Deepfakes, Face Presentation Attack Detection, Face Recognition, GANs},
projects = {Idiap},
month = sep,
title = {Generating Master Faces for Use in PerformingWolf Attacks on Face Recognition Systems},
booktitle = {International Join Conference on Biometrics},
year = {2020},
abstract = {Due to its convenience, biometric authentication, especial face authentication, has become increasingly mainstream and thus is now a prime target for attackers. Presentation attacks and face morphing are typical types of attack. Previous research has shown that finger-vein- and fingerprint-based authentication methods are susceptible to wolf attacks, in which a wolf sample matches many enrolled user templates. In this work, we demonstrated that wolf (generic) faces, which we call “master faces,” can also compromise face recognition systems and that the master face concept can be generalized in some cases. Motivated by recent similar work in the fingerprint domain, we generated high-quality master faces by using the state-of-the-art face generator StyleGAN in a process called latent variable evolution. Experiments demonstrated that even attackers with limited resources using only pre-trained models available on the Internet can initiate master face attacks. The results, in addition to demonstrating performance from the attacker’s point of view, can also be used to clarify and improve the performance of face recognition systems and harden face authentication systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Nguyen_IJCB_2020.pdf}
}
@INPROCEEDINGS{Nikisins_ICB2018_2018,
author = {Nikisins, Olegs and Mohammadi, Amir and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {ODIN/BATL, SWAN},
month = feb,
title = {On Effectiveness of Anomaly Detection Approaches against Unseen Presentation Attacks in Face Anti-Spoofing},
booktitle = {The 11th IAPR International Conference on Biometrics (ICB 2018)},
year = {2018},
abstract = {While face recognition systems got a significant boost in terms of recognition performance in recent years, they are known to be vulnerable to presentation attacks. Up to date, most of the research in the field of face anti-spoofing or presentation attack detection was considered as a two-class classification task: features of bona-fide samples versus features coming from spoofing attempts. The main focus has been on boosting the anti-spoofing performance for databases with identical types of attacks across both training and evaluation subsets. However, in realistic applications the types of attacks are likely to be unknown, potentially occupying a broad space in the feature domain. Therefore, a failure to generalize on unseen types of attacks is one of the main potential challenges in existing anti-spoofing approaches. First, to demonstrate the generalization issues of two-class anti-spoofing systems we establish new evaluation protocols for existing publicly available databases. Second, to unite the data collection efforts of various institutions we introduce a challenging Aggregated database composed of 3 publicly available datasets: Replay-Attack, Replay-Mobile and MSU MFSD, reporting the performance on it. Third, considering existing limitations we propose a number of systems approaching a task of presentation attack detection as an anomaly detection, or a one-class classification problem, using only bona-fide features in the training stage. Using less training data, hence requiring less effort in the data collection, the introduced approach demonstrates a better generalization properties against previously unseen types of attacks on the proposed Aggregated database.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Nikisins_ICB2018_2018.pdf}
}
@INPROCEEDINGS{Nikisins_ICB_2019,
author = {Nikisins, Olegs and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
title = {Domain Adaptation in Multi-Channel Autoencoder based Features for Robust Face Anti-Spoofing},
booktitle = {International Conference on Biometrics 2019, IEEE},
year = {2019},
abstract = {While the performance of face recognition systems has improved significantly in the last decade, they are proved to be highly vulnerable to presentation attacks (spoofing). Most of the research in the field of face presentation attack detection (PAD), was focused on boosting the performance of the systems within a single database. Face PAD datasets are usually captured with RGB cameras, and have very limited number of both bona-fide samples and presentation attack instruments. Training face PAD systems on such data leads to poor performance, even in the closed-set scenario, especially when sophisticated attacks are involved.
We explore two paths to boost the performance of the face PAD system against challenging attacks. First, by using multi-channel (RGB, Depth and NIR) data, which is still easily accessible in a number of mass production devices. Second, we develop a novel Autoencoders + MLP based face PAD algorithm. Moreover, instead of collecting more data for training of the proposed deep architecture, the domain adaptation technique is proposed, transferring the knowledge of facial appearance from RGB to multi-channel domain. We also demonstrate, that learning the features of individual facial regions, is more discriminative than the features learned from an entire face. The proposed system is tested on a very recent publicly available multi-channel PAD database with a wide variety of presentation attacks.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Nikisins_ICB_2019.pdf}
}
@INPROCEEDINGS{Nikisins_IWBF2018_2018,
author = {Nikisins, Olegs and Eglitis, Teodors and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {BIOWAVE, 3DFINGERVEIN},
month = jun,
title = {Fast cross-correlation based wrist vein recognition algorithm with rotation and translation compensation},
booktitle = {Sixth International Workshop on Biometrics and Forensics},
year = {2018},
abstract = {Most of the research on vein biometrics addresses the problems of either palm or finger vein recognition with a considerably smaller emphasis on wrist vein modality. This paper paves the way to a better understanding of capabilities and challenges in the field of wrist vein verification.
This is achieved by introducing and discussing a fully automatic cross-correlation based wrist vein verification technique.
Overcoming the limitations of ordinary cross-correlation, the proposed system is capable of compensating for scale, translation and rotation between vein patterns in a computationally efficient way. Introduced comparison algorithm requires only two cross-correlation operations to compensate for both translation and rotation, moreover the well known property of log-polar transformation of Fourier magnitudes is not involved in any form.
To emphasize the veins, a two-layer Hessian-based vein enhancement approach with adaptive brightness normalization is introduced, improving the connectivity and the stability of extracted vein patterns.
The experiments on the publicly available PUT Vein wrist database give promising results with FNMR of 3.75\% for FMR of 0.1\%. In addition we make this research reproducible providing the source code and instructions to replicate all findings in this work.},
pdf = {https://publications.idiap.ch/attachments/papers/2018/Nikisins_IWBF2018_2018.pdf}
}
@INPROCEEDINGS{Osorio-Roig_IJCB_2022,
author = {Osorio-Roig, Dail{\'{e}} and Rathgeb, Christian and Otroshi Shahreza, Hatef and Busch, Christoph and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Indexing Protected Deep Face Templates by Frequent Binary Patterns},
booktitle = {Proceedings of the 2022 International Joint Conference on Biometrics (IJCB)},
year = {2022},
publisher = {IEEE},
location = {Abu Dhabi, United Arab Emirates (UAE)},
url = {https://ieeexplore.ieee.org/abstract/document/10007939},
doi = {10.1109/IJCB54206.2022.10007939},
abstract = {In this work, we present a simple biometric indexing scheme which is binning and retrieving cancelable deep face templates based on frequent binary patterns. The simplicity of the proposed approach makes it applicable to unprotected as well as protected, i.e. cancelable, deep face templates. As such, this approach represents to the best of the authors' knowledge the first generic indexing scheme that can be applied to arbitrary cancelable face templates (o binary representation). In experiments, deep face templates are obtained from the Labelled Faces in the Wild (LFW) dataset using the ArcFace face recognition system for feature extraction. Protected templates are then generated by employing different cancelable biometric schemes, i.e. BioHashing and two variants of Index-of-Maximum Hashing. The proposed indexing scheme is evaluated on closed- and open-set identification scenarios. It is shown to maintain the recognition accuracy of the baseline system while reducing the penetration rate and hence the workload of identifications to approximately 40\%.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Osorio-Roig_IJCB_2022.pdf}
}
@ARTICLE{OtroshiShahreza_ACCESS-2_2024,
author = {Otroshi Shahreza, Hatef and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER},
title = {Knowledge Distillation for Face Recognition using Synthetic Data with Dynamic Latent Sampling},
journal = {IEEE Access},
year = {2024},
url = {https://ieeexplore.ieee.org/abstract/document/10766575},
doi = {10.1109/ACCESS.2024.3505621}
}
@ARTICLE{OtroshiShahreza_ACCESS_2024,
author = {Otroshi Shahreza, Hatef and Shkel, Yanina Y. and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {On Measuring Linkability of Multiple Protected Biometric Templates using Maximal Leakage},
journal = {IEEE Access},
year = {2024},
url = {https://ieeexplore.ieee.org/abstract/document/10609363},
doi = {10.1109/ACCESS.2024.3433536},
abstract = {With the rapid development of biometric recognition systems, users can be simultaneously enrolled in multiple biometric recognition systems, either with a single or multiple biometric characteristics (e.g., face, voice, etc.). With such a growth of biometric systems, it is important to secure the sensitive information used within these systems. In particular, considering the privacy issues in such systems, several biometric template protection schemes are proposed in the literature. According to the ISO/IEC 24745 standard, each template protection scheme should satisfy the unlinkability property. While previous measures to evaluate unlinkability were based on two protected templates, the adversary may have access to more information. Such information can correspond to multiple templates from different biometric systems, a single multi-modal biometric system, or even a single unimodal biometric system. In this paper, we focus on measuring the linkability of multiple protected biometric templates, and define maximal linkability in the presence of multiple similarity scores. We define different scenarios where the adversary gains access to multiple similarity scores and evaluate the linkability of protected templates in each scenario. We investigate the theoretical properties of the maximal linkability measure, and compare the theoretical prediction with the calculated linkability of the compositive systems in our experiments. To our knowledge, this is the first work on measuring the linkability of multiple protected biometric templates. The source codes of our measure and all experiments are publicly available.}
}
@INPROCEEDINGS{OtroshiShahreza_BIOSIG_2023,
author = {Otroshi Shahreza, Hatef and Bassit, Amina and Marcel, S{\'{e}}bastien and Veldhuis, Raymond},
projects = {TRESPASS-ETN},
title = {Remote Cancelable Biometric System for Verification and Identification Applications},
booktitle = {Proceedings of the International Conference of the Biometrics Special Interest Group (BIOSIG)},
year = {2023},
url = {https://ieeexplore.ieee.org/abstract/document/10345984},
doi = {10.1109/BIOSIG58226.2023.10345984},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_BIOSIG_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_CVPRW_2024,
author = {Otroshi Shahreza, Hatef and George, Anjith and Unnervik, Alexander and Rahimi, Parsa and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER},
title = {Second Edition FRCSyn Challenge at CVPR 2024: Face Recognition Challenge in the Era of Synthetic Data},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
year = {2024},
pages = {3173-3183},
url = {https://openaccess.thecvf.com/content/CVPR2024W/FRCSyn/html/Deandres-Tame_Second_Edition_FRCSyn_Challenge_at_CVPR_2024_Face_Recognition_Challenge_CVPRW_2024_paper.html}
}
@INPROCEEDINGS{OtroshiShahreza_CVPRW_2025,
author = {Otroshi Shahreza, Hatef and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER},
month = jun,
title = {Face Reconstruction from Face Embeddings using Adapter to a Face Foundation Model},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2025},
abstract = {Face recognition systems extract embedding vectors from face images and use these embeddings to verify or identify individuals. Face reconstruction attack (also known as template inversion) refers to reconstructing face images from embeddings and using the reconstructed image to enter a face recognition system. In this paper, we propose to use a face foundation model to reconstruct face images from the embeddings of a blackbox face recognition model. The foundation model is trained with 42M images to generate face images from the facial embeddings of a fixed face recognition model. We propose to use an adapter (called Face Adapter) to translate target embeddings into the embedding space of the foundation model. The generated images are evaluated on different face recognition models and different datasets, demonstrating the effectiveness of our method to translate embeddings of different face recognition models. We also evaluate the transferability of reconstructed face images when attacking different face recognition models. Our experimental results show that our reconstructed face images outperform previous reconstruction attacks against face recognition models. Project page: https://www.idiap.ch/paper/face_adapter},
pdf = {https://publications.idiap.ch/attachments/papers/2025/OtroshiShahreza_CVPRW_2025.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_EUSIPCO_2023,
author = {Otroshi Shahreza, Hatef and Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Face Recognition, Hashing, MultiLayer Perceptron (MLP), template protection},
projects = {TRESPASS-ETN},
title = {MLP-Hash: Protecting Face Templates via Hashing of Randomized Multi-Layer Perceptron},
booktitle = {Proceedings of the 31st European Signal Processing Conference},
year = {2023},
location = {Helsinki, Finland},
url = {https://ieeexplore.ieee.org/document/10289780},
doi = {10.23919/EUSIPCO58844.2023.10289780},
abstract = {Applications of face recognition systems for authentication purposes are growing rapidly. Although state-of-the-art (SOTA) face recognition systems have high recognition accuracy, the features which are extracted for each user and are stored in the system's database contain privacy-sensitive information. Accordingly, compromising this data would jeopardize users' privacy. In this paper, we propose a new cancelable template protection method, dubbed MLP-hash, which generates protected templates by passing the extracted features through a user-specific randomly-weighted multi-layer perceptron (MLP) and binarizing the MLP output. We evaluated the unlinkability, irreversibility, and recognition accuracy of our proposed biometric template protection method to fulfill the ISO/IEC 30136 standard requirements. Our experiments with SOTA face recognition systems on the MOBIO and LFW datasets show that our method has competitive performance with the BioHashing and IoM Hashing (IoM-GRP and IoM-URP) template protection algorithms. We provide an open-source implementation of all the experiments presented in this paper so that other researchers can verify our findings and build upon our work.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/OtroshiShahreza_EUSIPCO_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_FG-2_2024,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Breaking Template Protection: Reconstruction of Face Images from Protected Facial Templates},
booktitle = {18th International Conference on Automatic Face and Gesture Recognition (FG)},
year = {2024},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_FG-2_2024.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_FG2024_2024,
author = {Otroshi Shahreza, Hatef and Ecabert, Christophe and George, Anjith and Unnervik, Alexander and Marcel, S{\'{e}}bastien},
projects = {Idiap, SAFER},
title = {SDFR: Synthetic Data for Face Recognition Competition},
booktitle = {IEEE FG 2024 : 18th IEEE International Conference on Automatic Face and Gesture Recognition},
year = {2024}
}
@INPROCEEDINGS{OtroshiShahreza_FG_2024,
author = {Otroshi Shahreza, Hatef and Ecabert, Christophe and George, Anjith and Unnervik, Alexander and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER},
title = {SDFR: Synthetic Data for Face Recognition Competition},
booktitle = {2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG)},
year = {2024},
publisher = {IEEE},
url = {https://ieeexplore.ieee.org/abstract/document/10581946},
doi = {10.1109/FG59268.2024.10581946},
abstract = {Large-scale face recognition datasets are collected by crawling the Internet and without individuals' consent, raising legal, ethical, and privacy concerns. With the recent advances in generative models, recently several works proposed generating synthetic face recognition datasets to mitigate concerns in web-crawled face recognition datasets. This paper presents the summary of the Synthetic Data for Face Recognition (SDFR) Competition held in conjunction with the 18th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2024) and established to investigate the use of synthetic data for training face recognition models. The SDFR competition was split into two tasks, allowing participants to train face recognition systems using new synthetic datasets and/or existing ones. In the first task, the face recognition backbone was fixed and the dataset size was limited, while the second task provided almost complete freedom on the model backbone, the dataset, and the training pipeline. The submitted models were trained on existing and also new synthetic datasets and used clever methods to improve training with synthetic data. The submissions were evaluated and ranked on a diverse set of seven benchmarking datasets. The paper gives an overview of the submitted face recognition models and reports achieved performance compared to baseline models trained on real and synthetic datasets. Furthermore, the evaluation of submissions is extended to bias assessment across different demography groups. Lastly, an outlook on the current state of the research in training face recognition models using synthetic data is presented, and existing problems as well as potential future directions are also discussed.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_FG_2024.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICASSP-2_2024,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Face Reconstruction from Partially Leaked Facial Embeddings},
booktitle = {Proceedings of the 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing},
year = {2024},
url = {https://ieeexplore.ieee.org/abstract/document/10445870},
doi = {10.1109/ICASSP48485.2024.10445870},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_ICASSP-2_2024.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICASSP_2021,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Auto-encoder, Biohashing, Convolutional neural network, deep learning, finger vein recognition, template protection},
projects = {TRESPASS-ETN},
title = {Deep Auto-Encoding and Biohashing for Secure Finger Vein Recognition},
booktitle = {Proceedings of the 2021 IEEE International Conference on Acoustics, Speech, and Signal Processing},
year = {2021},
location = {Toronto, Canada},
organization = {IEEE},
url = {https://ieeexplore.ieee.org/abstract/document/9414498},
doi = {10.1109/ICASSP39728.2021.9414498},
abstract = {Biometric recognition systems relying on finger vein have gained a lot of attention in recent years. Besides security, the privacy of finger vein recognition systems is always a crucial concern. To address the privacy concerns, several biometric template protection (BTP) schemes are introduced in the literature. However, despite providing privacy, BTP algorithms often affect the recognition performance. In this paper, we propose a deep-learning-based approach for secure finger vein recognition. We use a convolutional auto-encoder neural network with a multi-term loss function. In addition to the auto-encoder loss function, we deploy triplet loss for the embedding features. Next, we apply Biohashing to our deep features to generate protected templates. The experimental results indicate that the proposed method achieves superior performance to previous finger vein recognition methods protected with Biohashing. Besides, our proposed method has less execution time and requires less memory.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/OtroshiShahreza_ICASSP_2021.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICASSP_2024,
author = {Otroshi Shahreza, Hatef and Veuthey, Alexandre and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Face Recognition Using Lensless Camera},
booktitle = {Proceedings of the 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing},
year = {2024},
url = {https://ieeexplore.ieee.org/abstract/document/10446710},
doi = {10.1109/ICASSP48485.2024.10446710},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_ICASSP_2024.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICCVW_2025,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Face, FaceLLM, Large Language Models, LLM, MLLM},
mainresearchprogram = {AI for Everyone},
title = {FaceLLM: A Multimodal Large Language Model for Face Understanding},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
year = {2025},
url = {https://arxiv.org/pdf/2507.10300}
}
@INPROCEEDINGS{OtroshiShahreza_ICCV_2023,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
month = oct,
title = {Template Inversion Attack against Face Recognition Systems using 3D Face Reconstruction},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
year = {2023},
pages = {19662-19672},
issn = {1550-5499},
isbn = {979-8-3503-0718-4},
url = {https://openaccess.thecvf.com/content/ICCV2023/html/Shahreza_Template_Inversion_Attack_against_Face_Recognition_Systems_using_3D_Face_ICCV_2023_paper.html},
doi = {https://doi.org/10.1109/ICCV51070.2023.01801},
abstract = {Face recognition systems are increasingly being used in different applications. In such systems, some features (also known as embeddings or templates) are extracted from each face image. Then, the extracted templates are stored in the system's database during the enrollment stage and are later used for recognition. In this paper, we focus on template inversion attacks against face recognition systems and introduce a novel method (dubbed GaFaR) to reconstruct 3D face from facial templates. To this end, we use a geometry-aware generator network based on generative neural radiance fields (GNeRF), and learn a mapping from facial templates to the intermediate latent space of the generator network. We train our network with a semi-supervised learning approach using real and synthetic images simultaneously. For the real training data, we use a Generative Adversarial Network (GAN) based framework to learn the distribution of the latent space. For the synthetic training data, where we have the true latent code, we directly train in the latent space of the generator network. In addition, during the inference stage, we also propose optimization on the camera parameters to generate face images to improve the success attack rate (up to 17.14\% in our experiments). We evaluate the performance of our method in the whitebox and blackbox attacks against state-of-the-art face recognition models on the LFW and MOBIO datasets. To our knowledge, this paper is the first work on 3D face reconstruction from facial templates. The project page is available at: https://www.idiap.ch/paper/gafar}
}
@INPROCEEDINGS{OtroshiShahreza_ICIP_2022,
author = {Otroshi Shahreza, Hatef and Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
keywords = {embedding, Face Recognition, face reconstruction, template inversion},
projects = {TRESPASS-ETN},
title = {Face Reconstruction from Deep Facial Embeddings using a Convolutional Neural Network},
booktitle = {Proceedings of the 2022 IEEE International Conference on Image Processing (ICIP)},
year = {2022},
publisher = {IEEE},
location = {Bordeaux, France},
url = {https://ieeexplore.ieee.org/abstract/document/9897535},
doi = {10.1109/ICIP46576.2022.9897535},
abstract = {State-of-the-art (SOTA) face recognition systems generally use deep convolutional neural networks (CNNs) to extract deep features, called embeddings, from face images. The face embeddings are stored in the system’s database and are used for recognition of the enrolled system users. Hence, these features convey important information about the user’s identity, and therefore any attack using the face embeddings jeopardizes the user’s security and privacy. In this paper, we propose a CNN-based structure to reconstruct face images from face embeddings and we train our network with a multi-term loss function. In our experiments, our network is trained to reconstruct face images from SOTA face recognition models (ArcFace and ElasticFace) and we evaluate our face reconstruction network on the MOBIO and LFW datasets. The source code of all the experiments presented in this paper is publicly available so our work can be fully reproduced.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/OtroshiShahreza_ICIP_2022.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICIP_2023,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Blackbox, embedding, Face Recognition, face reconstruction, template inversion},
projects = {TRESPASS-ETN},
month = oct,
title = {Blackbox Face Reconstruction from Deep Facial Embeddings Using A Different Face Recognition Model},
booktitle = {Proceedings of the IEEE International Conference on Image Processing (ICIP)},
year = {2023},
pages = {2435-2439},
location = {Kuala Lumpur, Malaysia},
isbn = {978-1-7281-9835-4},
url = {https://ieeexplore.ieee.org/abstract/document/10222312},
doi = {10.1109/ICIP49359.2023.10222312},
abstract = {Face recognition systems generally store features (called embeddings) extracted from each face image during the enrollment stage, and then compare the extracted embeddings with the stored embeddings during the recognition stage. In this paper, we focus on the blackbox face reconstruction from facial embeddings stored in the face recognition database. We use a convolutional neural network (CNN) to reconstruct face images and train our network with a multi-term loss function. In particular, we use a different feature extractor trained for face recognition (which the adversary has the whitebox knowledge of it) to minimize the distance of embeddings extracted from the original and reconstructed face images. We evaluate our method in blackbox attacks against five state-of-the-art face recognition models on the MOBIO and LFW datasets. Our experimental results show that our proposed method outperforms previous face reconstruction methods in the literature. The source code of our experiments is publicly available to facilitate the reproducibility of our work.},
pdf = {https://publications.idiap.ch/attachments/papers/2023/OtroshiShahreza_ICIP_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICIP_2024,
author = {Hassanpour, Ahmad and Kowsari, Yasamin and Otroshi Shahreza, Hatef and Yang, Bian and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {ChatGPT and biometrics: an assessment of face recognition, gender detection, and age estimation capabilities},
booktitle = {2024 IEEE International Conference on Image Processing (ICIP)},
year = {2024},
url = {https://ieeexplore.ieee.org/document/10647924},
doi = {10.1109/ICIP51287.2024.10647924},
abstract = {This paper explores the application of large language models (LLMs), like ChatGPT, for biometric tasks. We specifically examine the capabilities of ChatGPT in performing biometric-related tasks, with an emphasis on face recognition, gender detection, and age estimation. Since biometrics are considered as sensitive information, ChatGPT avoids answering direct prompts, and thus we crafted a prompting strategy to bypass its safeguard and evaluate the capabilities for biometrics tasks. Our study reveals that ChatGPT recognizes facial identities and differentiates between two facial images with considerable accuracy. Additionally, experimental results demonstrate remarkable performance in gender detection and reasonable accuracy for the age estimation tasks. Our findings shed light on the promising potentials in the application of LLMs and foundation models for biometrics.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_ICIP_2024.pdf}
}
@ARTICLE{OtroshiShahreza_IEEE-TIFS_2023,
author = {Otroshi Shahreza, Hatef and Shkel, Yanina Y. and Marcel, S{\'{e}}bastien},
keywords = {biometric template protection, Biometrics, Differential Privacy, linkability, maximal leakage, statistical hypothesis testing, template},
title = {Measuring Linkability of Protected Biometric Templates Using Maximal Leakage},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {18},
year = {2023},
pages = {2262 - 2275},
url = {https://ieeexplore.ieee.org/abstract/document/10098649},
doi = {10.1109/TIFS.2023.3266170},
abstract = {As the applications of biometric recognition systems are increasing rapidly, there is a growing need to secure the sensitive data used within these systems. Considering privacy challenges in such systems, different biometric template protection (BTP) schemes were proposed in the literature, and the ISO/IEC 24745 standard defined a number of requirements for protecting biometric templates. While there are several studies on evaluating different requirements of the ISO/IEC 24745 standard, there have been few studies on how to measure the linkability of biometric templates. In this paper, we propose a new method for measuring linkability of protected biometric templates. The proposed method is based on maximal leakage, which is a well-studied measure in information-theoretic literature. We show that the resulting linkability measure has a number of important theoretical properties and an operational interpretation in terms of statistical hypothesis testing. We compare the proposed measure to two other linkability measures: one previously introduced in the literature, and a similar measure based on differential privacy. In our experiments, we use the proposed measure to evaluate the linkability of biometric templates from different biometric characteristics (face, voice, and finger vein), which are protected with different BTP schemes. The source codes of our proposed measure and all experiments are publicly available.}
}
@ARTICLE{OtroshiShahreza_IEEE-TIFS_2024,
author = {Otroshi Shahreza, Hatef and Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, embedding, Face Recognition, face reconstruction, template inversion, vulnerability evaluation},
projects = {TRESPASS-ETN},
title = {Vulnerability of State-of-the-Art Face Recognition Models to Template Inversion Attack},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {19},
year = {2024},
pages = {4585-4600},
issn = {1556-6013 1556-6021},
url = {https://ieeexplore.ieee.org/document/10478940},
doi = {10.1109/TIFS.2024.3381820},
abstract = {Face recognition systems use the templates (extracted from users’ face images) stored in the system’s database for recognition. In a template inversion attack, the adversary gains access to the stored templates and tries to enter the system using images reconstructed from those templates. In this paper, we propose a framework to evaluate the vulnerability of face recognition systems to template inversion attacks. We build our framework upon a real-world scenario and measure the vulnerability of the system in terms of the adversary’s success attack rate in entering the system using the reconstructed face images. We propose a face reconstruction network based on a new block called “enhanced deconvolution using cascaded convolution and skip connections” (shortly, DSCasConv), and train it with a multi-term loss function. We use our framework to evaluate the vulnerability of state-of-the-art face recognition models, with different network structures and loss functions (in total 31 models), on the MOBIO, LFW, and AgeDB face datasets. Our experiments show that the reconstructed face images can be used to enter the system, which threatens the system’s security. Additionally, the reconstructed face images may reveal important information about each user’s identity, such as race, gender, and age, and hence jeopardize the users’ privacy.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_IEEE-TIFS_2024.pdf}
}
@ARTICLE{OtroshiShahreza_IEEE-TIFS_2025,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {ALM, Audio Language Models, Biometrics, Foundation Models, Large Language Models, Large Multi-modal Models, LLM, privacy, security, Vision language models, VLM},
mainresearchprogram = {Sustainable & Resilient Societies},
title = {Foundation Models and Biometrics: A Survey and Outlook},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2025},
url = {https://ieeexplore.ieee.org/document/11137396},
doi = {10.1109/TIFS.2025.3602233},
abstract = {This paper provides an overview of the recent advancements in foundation models and discusses potential applications of these models in the field of biometrics. Foundation models (such as large language models, vision language models, audio-language models, and large multi-modal models) are based on large neural networks which are trained with massive amounts of data and enable robust feature extraction for transfer learning. These models allow efficient zero-shot and few-shot learning, achieving state-of-the-art performance in downstream tasks. Foundation models have been studied and used in different domains, including natural language processing, computer vision, audio processing, and multi-modal processing. Biometrics is also an active field of research, which involves various research problems, ranging from robust recognition to security and privacy in biometric systems. In this paper, we present an in-depth analysis of state-of-the-art methodologies regarding foundation multi-modal models, their advancements, and their applicability to biometrics tasks. We also highlight current limitations and provide insights into potential future research directions in the applications of foundation models in biometrics. To our knowledge, this paper is the first survey which investigates the applications of foundation models in biometrics.}
}
@ARTICLE{OtroshiShahreza_IEEE-TPAMI_2023,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Comprehensive Vulnerability Evaluation of Face Recognition Systems to Template Inversion Attacks Via 3D Face Reconstruction},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
volume = {45},
number = {12},
year = {2023},
pages = {14248-14265},
issn = {0162-8828 1939-3539},
url = {https://ieeexplore.ieee.org/document/10239446},
doi = {10.1109/TPAMI.2023.3312123},
abstract = {In this paper, we comprehensively evaluate the vulnerability of state-of-the-art face recognition systems to template inversion attacks using 3D face reconstruction. We propose a new method (called GaFaR) to reconstruct 3D faces from facial templates using a pretrained geometry-aware face generation network, and train a mapping from facial templates to the intermediate latent space of the face generator network. We train our mapping with a semi-supervised approach using real and synthetic face images. For real face images, we use a generative adversarial network (GAN)-based framework to learn the distribution of generator intermediate latent space. For synthetic face images, we directly learn the mapping from facial templates to the generator intermediate latent code. Furthermore, to improve the success attack rate, we use two optimization methods on the camera parameters of the GNeRF model. We propose our method in the whitebox and blackbox attacks against face recognition systems and compare the transferability of our attack with state-of-the-art methods across other face recognition systems on the MOBIO and LFW datasets. We also perform practical presentation attacks on face recognition systems using the digital screen replay and printed photographs, and evaluate the vulnerability of face recognition systems to different template inversion attacks. The project page is available at https://www.idiap.ch/paper/gafar .}
}
@ARTICLE{OtroshiShahreza_IF_2025,
author = {Otroshi Shahreza, Hatef and George, Anjith and Parsa, Rahimi and Unnervik, Alexander and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
additionalresearchprograms = {AI for Everyone},
title = {Second FRCSyn-onGoing: Winning Solutions and Post-Challenge Analysis to Improve Face Recognition with Synthetic Data},
journal = {Information Fusion},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/OtroshiShahreza_IF_2025.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_IJCB-2_2023,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Inversion of Deep Facial Templates using Synthetic Data},
booktitle = {Proceedings of the IEEE International Joint Conference on Biometric},
year = {2023},
issn = {2474-9680},
isbn = {979-8-3503-3726-6},
url = {https://ieeexplore.ieee.org/abstract/document/10449033},
doi = {10.1109/IJCB57857.2023.10449033},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_IJCB-2_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_IJCB2023_2023,
author = {Otroshi Shahreza, Hatef and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = sep,
title = {SynthDistill: Face Recognition with Knowledge Distillation from Synthetic Data},
booktitle = {IEEE International Joint Conference on Biometrics (IJCB 2023)},
year = {2023},
issn = {2474-9680},
isbn = {979-8-3503-3726-6},
doi = {https://doi.org/10.1109/IJCB57857.2023.10448642},
abstract = {State-of-the-art face recognition networks are often com- putationally expensive and cannot be used for mobile appli- cations. Training lightweight face recognition models also requires large identity-labeled datasets. Meanwhile, there are privacy and ethical concerns with collecting and using large face recognition datasets. While generating synthetic datasets for training face recognition models is an alter- native option, it is challenging to generate synthetic data with sufficient intra-class variations. In addition, there is still a considerable gap between the performance of models trained on real and synthetic data. In this paper, we propose a new framework (named SynthDistill) to train lightweight face recognition models by distilling the knowledge of a pre- trained teacher face recognition model using synthetic data. We use a pretrained face generator network to generate syn- thetic face images and use the synthesized images to learn a lightweight student network. We use synthetic face im- ages without identity labels, mitigating the problems in the intra-class variation generation of synthetic datasets. In- stead, we propose a novel dynamic sampling strategy from the intermediate latent space of the face generator network to include new variations of the challenging images while further exploring new face images in the training batch. The results on five different face recognition datasets demon- strate the superiority of our lightweight model compared to models trained on previous synthetic datasets, achiev- ing a verification accuracy of 99.52\% on the LFW dataset with a lightweight network. The results also show that our proposed framework significantly reduces the gap between training with real and synthetic data. The source code for replicating the experiments will be publicly released},
pdf = {https://publications.idiap.ch/attachments/papers/2023/OtroshiShahreza_IJCB2023_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_IJCB_2022,
author = {Otroshi Shahreza, Hatef and Rathgeb, Christian and Osorio-Roig, Dail{\'{e}} and Krivokuca, Vedrana and Marcel, S{\'{e}}bastien and Busch, Christoph},
projects = {TRESPASS-ETN},
title = {Hybrid Protection of Biometric Templates by Combining Homomorphic Encryption and Cancelable Biometrics},
booktitle = {Proceedings of the 2022 International Joint Conference on Biometrics (IJCB)},
year = {2022},
publisher = {IEEE},
location = {Abu Dhabi, United Arab Emirates (UAE)},
url = {https://ieeexplore.ieee.org/abstract/document/10007960},
doi = {10.1109/IJCB54206.2022.10007960},
abstract = {Homomorphic Encryption (HE) has become a well-known tool for privacy-preserving recognition in biometric systems. Despite some important advantages of HE (such as preservation of recognition accuracy), there are two main drawbacks in the application of HE to biometric recognition systems: first, the security of the system solely depends on the secrecy of the private (decryption) key; second, the computational costs of the operations on the ciphertexts are expensive. To address these challenges, in this paper we propose a hybrid scheme for the protection of biometric templates, which combines cancelable biometrics (CB) methods and HE. Applying CB prior to HE enhances both the security and privacy of the overall system, since the protected templates remain irreversible even if the secret keys are leaked (commonly referred to as the full disclosure scenario). In addition, we can reduce the dimensionality of templates using CB before applying HE, which speeds up the computation over the ciphertexts. We use BioHashing, Multi-Layer Perceptron (MLP) hashing, and Index-of-Maximum (IoM) hashing as different CB methods, and for each of these schemes, we propose a method for computing scores between hybrid-protected templates in the encrypted domain. We evaluate our proposed hybrid scheme using different state-of-the-art face recognition models (ArcFace, ElasticFace, and FaceNet) on the MOBIO and LFW datasets. The source code of our experiments is publicly available, so our work can be fully reproduced.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/OtroshiShahreza_IJCB_2022.pdf}
}
@ARTICLE{OtroshiShahreza_LSENS_2023,
author = {Otroshi Shahreza, Hatef and Veuthey, Alexandre and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Toward High-Resolution Face Image Generation From Coded Aperture Camera},
journal = {IEEE Sensors Letters},
year = {2023},
url = {https://ieeexplore.ieee.org/document/10251561},
doi = {10.1109/LSENS.2023.3315248},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_LSENS_2023.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_MLSP-2_2025,
author = {Otroshi Shahreza, Hatef and Colbois, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, TRESPASS-ETN},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
title = {3D Face Morph Generation Using Geometry-Aware Template Inversion},
booktitle = {2025 IEEE 35th International Workshop on Machine Learning for Signal Processing (MLSP)},
year = {2025},
url = {https://ieeexplore.ieee.org/abstract/document/11204286},
doi = {10.1109/MLSP62443.2025.11204286},
abstract = {While face recognition systems have become a popular solution for applications which require automatic authentication, their vulnerability to morphing attacks has become a major concern in sensitive scenarios. This work proposes a novel method to generate 3D face morphs. Given two source images, we use their face embeddings to derive an optimal morph embedding, and then use a geometry-aware template inversion method based on Generative Neural Radiance Fields (GNeRF) to construct a 3D face morph from this optimal embedding. Leveraging from the GNeRF structure, we can generate morph images with any arbitrary view-point. Our experiments show that our method achieve comparable performance with previous morph generation methods from the literature, and has an additional advantage of generating 3D results. To our knowledge, this is the first work on generating 3D face morphs based on GNeRF models, and it can potentially be used for sophisticated morphing attacks. The source code of our experiments is publicly released.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/OtroshiShahreza_MLSP-2_2025.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_MLSP_2025,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Face Recognition, Foundation Model, Synthetic Data},
projects = {SAFER, TRESPASS-ETN, Idiap},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
title = {Generating Synthetic Face Recognition Datasets Using Brownian Identity Diffusion and a Foundation Model},
booktitle = {2025 IEEE 35th International Workshop on Machine Learning for Signal Processing (MLSP)},
year = {2025},
url = {https://ieeexplore.ieee.org/abstract/document/11204248},
doi = {10.1109/MLSP62443.2025.11204248},
abstract = {Training face recognition models requires a large amount of identity-labeled face images, which are often collected by crawling the web, and therefore have ethical and privacy concerns. Recently, generating synthetic face datasets and training face recognition models using synthetic datasets has emerged to be a viable solution. This paper presents BIF-Face, a new framework to generate synthetic face recognition datasets. We use the Brownian identity diffusion to generate synthetic identities, and then build synthetic face recognition datasets by generating different samples per each identity using a foundation model. In our experiments, we use the generated face datasets to train face recognition models and evaluate them on several real benchmarking dataset. Our experimental results show that face recognition models trained with BIF-Face achieve competitive performance with face recognition models trained on state-of-the-art synthetic face recognition datasets.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/OtroshiShahreza_MLSP_2025.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_NEURIPS_2023,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Face Reconstruction from Facial Templates by Learning Latent Space of a Generator Network},
booktitle = {Thirty-seventh Conference on Neural Information Processing Systems},
year = {2023},
url = {https://openreview.net/pdf?id=hI6EPhq70A},
abstract = {In this paper, we focus on the template inversion attack against face recognition systems and propose a new method to reconstruct face images from facial templates. Within a generative adversarial network (GAN)-based framework, we learn a mapping from facial templates to the intermediate latent space of a pre-trained face generation network, from which we can generate high-resolution realistic reconstructed face images. We show that our proposed method can be applied in whitebox and blackbox attacks against face recognition systems. Furthermore, we evaluate the transferability of our attack when the adversary uses the reconstructed face image to impersonate the underlying subject in an attack against another face recognition system. Considering the adversary’s knowledge and the target face recognition system, we define five different attacks and evaluate the vulnerability of state-of-the-art face recognition systems. Our experiments show that our proposed method achieves high success attack rates in whitebox and blackbox scenarios. Furthermore, the reconstructed face images are transferable and can be used to enter target face recognition systems with a different feature extractor model. We also explore important areas in the reconstructed face images that can fool the target face recognition system.}
}
@INPROCEEDINGS{OtroshiShahreza_NEURIPS_2024,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Face Recognition, Identity, information leakage, Membership Inference (MI) Attack, Memorisation, Synthetic Data},
projects = {TRESPASS-ETN, SAFER},
title = {Unveiling Synthetic Faces: How Synthetic Datasets Can Expose Real Identities},
booktitle = {NeurIPS Workshop on New Frontiers in Adversarial Machine Learning},
year = {2024},
note = {Project page: https://www.idiap.ch/paper/unveiling_synthetic_faces/},
abstract = {Synthetic data generation is gaining increasing popularity in different computer vision applications. Existing state-of-the-art face recognition models are trained using large-scale face datasets, which are crawled from the Internet and raise privacy and ethical concerns. To address such concerns, several works have proposed generating synthetic face datasets to train face recognition models. However, these methods depend on generative models, which are trained on real face images. In this work, we design a simple yet effective membership inference attack to systematically study if any of the existing synthetic face recognition datasets leak any information from the real data used to train the generator model. We provide an extensive study on 6 state-of-the-art synthetic face recognition datasets, and show that in all these synthetic datasets, several samples from the original real dataset are leaked. To our knowledge, this paper is the first work which shows the leakage from training data of generator models into the generated synthetic face recognition datasets. Our study demonstrates privacy pitfalls in synthetic face recognition datasets and paves the way for future studies on generating responsible synthetic face datasets. Project page: https://www.idiap.ch/paper/unveiling_synthetic_faces/},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_NEURIPS_2024.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_NEURIPS_2024-2_2024,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {SAFER},
title = {HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere},
booktitle = {NeurIPS Safe Generative AI Workshop 2024},
year = {2024},
note = {The complete version of this paper is accepted in ICLR 2025: https://openreview.net/pdf?id=4YzVF9isgD},
url = {https://openreview.net/pdf?id=m0hUhbpP7G},
crossref = {OtroshiShahreza_ICLR_2025}
}
@ARTICLE{OtroshiShahreza_TBIOM_2021,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
keywords = {Auto-encoder, Biohashing, Biometrics, Deep neural network, Finger vein, template protection, vascular biometrics},
projects = {TRESPASS-ETN},
title = {Towards Protecting and Enhancing Vascular Biometric Recognition methods via Biohashing and Deep Neural Networks},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2021},
url = {https://ieeexplore.ieee.org/document/9419051},
doi = {10.1109/TBIOM.2021.3076444},
abstract = {Biometric template protection has been a crucial concern in biometric recognition systems. This is because biometric characteristics are irreplaceable, so the compromised templates can also be used in other applications. On the other side, deploying template protection algorithms often affects the performance of biometric systems. In this paper, we consider both raw and pre-processed finger vein images and propose a novel deep-learning-based framework to protect biometric templates and enhance recognition performance. We use a deep convolutional auto-encoder structure to reduce the dimension of the feature space, and then secure templates by applying the Biohashing algorithm on the features extracted at the bottleneck layer of our auto-encoder. The experimental results indicate that the protected templates through our framework achieve superior performance than Biohash protected templates of the raw features in the normal scenario. In the stolen scenario, where the Biohashing key is stolen, our model yields far better performance than Biohashing of raw features extracted by previous recognition methods.
We also evaluate the generalization of our proposed framework on other vascular biometric modalities. It is worth mentioning that we provide an open-source implementation of our framework so that other researchers can verify our findings and build upon our work.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/OtroshiShahreza_TBIOM_2021.pdf}
}
@ARTICLE{OtroshiShahreza_TBIOM_2024,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Template Inversion Attack Using Synthetic Face Images Against Real Face Recognition Systems},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
year = {2024},
url = {https://ieeexplore.ieee.org/document/10506232},
doi = {10.1109/TBIOM.2024.3391759},
abstract = {In this paper, we use synthetic data and propose a new method for template inversion attacks against face recognition systems. We use synthetic data to train a face reconstruction model to generate high-resolution (i.e., 1024×1024) face images from facial templates. To this end, we use a face generator network to generate synthetic face images and extract their facial templates using the face recognition model as our training set. Then, we use the synthesized dataset to learn a mapping from facial templates to the intermediate latent space of the same face generator network. We propose our method for both whitebox and blackbox TI attacks. Our experiments show that the trained model with synthetic data can be used to reconstruct face images from templates extracted from real face images. In our experiments, we compare our method with previous methods in the literature in attacks against different state-of-the-art face recognition models on four different face datasets, including the MOBIO, LFW, AgeDB, and IJB-C datasets, demonstrating the effectiveness of our proposed method on real face recognition datasets. Experimental results show our method outperforms previous methods on high-resolution 2D face reconstruction from facial templates and achieve competitive results with SOTA face reconstruction methods. Furthermore, we conduct practical presentation attacks using the generated face images in digital replay attacks against real face recognition systems, showing the vulnerability of face recognition systems to presentation attacks based on our TI attack (with synthetic train data) on real face datasets.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/OtroshiShahreza_TBIOM_2024.pdf}
}
@ARTICLE{OtroshiShahreza_TIFS_2025,
author = {Otroshi Shahreza, Hatef and Colbois, Laurent and Marcel, S{\'{e}}bastien},
keywords = {embedding, Face Recognition, generation, morph attack, optimal morph, template inversion},
projects = {TRESPASS-ETN, Biometrics Center},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
title = {On the Generation of Face Morphs by Inversion of Optimal Morph Embeddings},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2025},
url = {https://ieeexplore.ieee.org/document/11299120},
doi = {10.1109/TIFS.2025.3643785},
abstract = {Automatic face recognition systems are widely used in different applications which require authentication. Among various types of attacks against face recognition systems, morphing attacks have become a major concern, where face images of two subjects are combined into a face morph image which is submitted for enrolment. In a successful attack, both contributing subjects can then authenticate against the morph reference. In this work, we propose a new method to generate face morphs based on inversion of the optimal morph embeddings. To this end, we first find the optimal morph embeddings using the face embeddings of two source face images and then use state-of-the-art template inversion techniques to generate the morph. We use three different template inversion methods: the first one exploits a fully self-contained embedding-to-image inversion model, while the second and third leverage the realistic image generation of a pretrained StyleGAN network and a foundation model based on diffusion models, respectively. Furthermore, we use optimization methods to improve the performance of template inversion methods in the generation of face morph images from optimal morph embeddings. In our experiments, we evaluate the performance of generated face morph images and compare them with state-of-the-art morph generation methods, showing the superiority of our method. We showcase that our method can outperform state-of-the-art deep-learning-based morph generation methods, both in white-box and black-box attack scenarios, and compete with state-of-the-art landmark-based morph generation methods. Moreover, we perform a practical print-scan attack to simulate a real-world scenario and compare our method with previous methods in the literature, demonstrating the effectiveness and superiority of our method. The source code of our proposed method and all experiments are publicly available.},
pdf = {https://publications.idiap.ch/attachments/papers/2026/OtroshiShahreza_TIFS_2025.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_WIFS_2021,
author = {Otroshi Shahreza, Hatef and Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
keywords = {Biohashing, Biometrics, deep features, Face Recognition, template protection},
projects = {TRESPASS-ETN},
month = dec,
title = {On the Recognition Performance of BioHashing on state-of-the-art Face Recognition models},
booktitle = {Proceedings of the 13th IEEE International Workshop on Information Forensics and Security (WIFS)},
year = {2021},
publisher = {IEEE},
location = {Montpellier, France},
url = {https://ieeexplore.ieee.org/document/9648382},
doi = {10.1109/WIFS53200.2021.9648382},
abstract = {Face recognition has become a popular authentication tool in recent years. Modern state-of-the-art (SOTA) face recognition methods rely on deep neural networks, which extract discriminative features from face images. Although these methods have high recognition performance, the extracted features contain privacy-sensitive information. Hence, the users' privacy would be jeopardized if the features stored in the face recognition system were compromised. Accordingly, protecting the extracted face features (templates) is an essential task in face recognition systems. In this paper, we use BioHashing for face template protection and aim to establish the minimum BioHash length that would be required in order to maintain the recognition accuracy achieved by the corresponding unprotected system. We consider two hypotheses and experimentally show that the performance depends on the value of the BioHash length (as opposed to the ratio of the BioHash length to the dimension of the original features). To eliminate bias in our experiments, we use several SOTA face recognition models with different network structures, loss functions, and training datasets, and we evaluate these models on two different datasets (LFW and MOBIO). We provide an open-source implementation of all the experiments presented in this paper so that other researchers can verify our findings and build upon our work.},
pdf = {https://publications.idiap.ch/attachments/papers/2021/OtroshiShahreza_WIFS_2021.pdf}
}
@INPROCEEDINGS{Parsa_NEURIPS_2025,
author = {Parsa, Rahimi and Teney, Damien and Marcel, S{\'{e}}bastien},
projects = {SAFER},
mainresearchprogram = {AI for Everyone},
additionalresearchprograms = {AI for Everyone},
title = {AugGen: Synthetic Augmentation using Diffusion Models Can Improve Recognition},
booktitle = {The Thirty-ninth Annual Conference on Neural Information Processing Systems},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Parsa_NEURIPS_2025.pdf}
}
@INPROCEEDINGS{Patino_ODYSSEY_2018,
author = {Patino, Jose and Yin, Ruiqing and Delgado, Hector and Bredin, Herve and Komaty, Alain and Wisniewski, Guillaume and Barras, Claude and Evans, Nicholas and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODESSA},
month = jun,
title = {Low-latency speaker spotting with online diarization and detection},
booktitle = {The Speaker and Language Recognition Workshop (Odyssey)},
year = {2018},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Patino_ODYSSEY_2018.pdf}
}
@INPROCEEDINGS{Poh:2003:icassp,
author = {Poh, Norman and Marcel, S{\'{e}}bastien and Bengio, Samy},
projects = {Idiap},
title = {Improving Face Authetication Using Virtual Samples},
booktitle = {{IEEE} International Conference on Acoustics, Speech, and Signal Processing},
number = {40},
year = {2003},
note = {IDIAP-RR},
crossref = {poh_03_faceverif},
abstract = {In this paper, we present a simple yet effective way to improve a face verification system by generating multiple virtual samples from the unique image corresponding to an access request. These images are generated using simple geometric transformations. This method is often used during training to improve accuracy of a neural network model by making it robust against minor translation, scale and orientation change. The main contribution of this paper is to introduce such method during testing. By generating $N$ images from one single image and propagating them to a trained network model, one obtains $N$ scores. By merging these scores using a simple mean operator, we show that the variance of merged scores is decreased by a factor between 1 and $N$. An experiment is carried out on the XM2VTS database which achieves new state-of-the-art performances.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/norman_2003_icassp.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/norman_2003_icassp.ps.gz},
ipdmembership={learning},
}
@INPROCEEDINGS{Poh_ARTMETRICSATICCV_2025,
author = {Poh, Francois and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {interart},
mainresearchprogram = {AI for Everyone},
title = {ArtFace: Towards Historical Portrait Face Identification via Model Adaptation},
booktitle = {(Non-Archival)},
year = {2025},
pages = {4},
url = {https://www.idiap.ch/paper/artface/},
abstract = {Identifying sitters in historical paintings is a key task for art historians, offering insight into their lives and how they chose to be seen. However, the process is often subjective and limited by the lack of data and stylistic variations. Automated facial recognition is capable of handling challenging conditions and can assist, but while traditional facial recognition models perform well on photographs, they struggle with paintings due to domain shift and high intra-class variation. Artistic factors such as style, skill, intent, and influence from other works further complicate recognition. In this work, we investigate the potential of foundation models to improve facial recognition in artworks. By fine-tuning foundation models and integrating their embeddings with those from conventional facial recognition networks, we demonstrate notable improvements over current state-of-the-art methods. Our results show that foundation models can bridge the gap where traditional methods are ineffective.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Poh_ARTMETRICSATICCV_2025.pdf}
}
@TECHREPORT{Prasad_Idiap-RR-10-2025,
author = {Prasad, Amrutha and Otroshi Shahreza, Hatef and Carofilis, Andr{\'{e}}s and Farhadipour, Aref and Liu, Shiran and Madikeri, Srikanth and George, Anjith and Motlicek, Petr and Marcel, S{\'{e}}bastien and Chapariniya, Masoumeh and Perepelytsia, Valeriia and Vukovic, Teodora and Dellwo, Volker},
projects = {Idiap, armasuisse, Biometrics Center},
mainresearchprogram = {Sustainable & Resilient Societies},
month = {10},
title = {TEAM SWITZERLAND SUBMISSION TO NIST SRE24 SPEAKER RECOGNITION EVALUATION},
type = {Idiap-RR},
number = {Idiap-RR-10-2025},
year = {2025},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2024/Prasad_Idiap-RR-10-2025.pdf}
}
@INPROCEEDINGS{Purnapatra_IJCB2021_2021,
author = {Purnapatra, Sandip and Smalt, Nic and Bahmani, Keivan and Das, Priyanka and Yambay, David and Mohammadi, Amir and George, Anjith and Bourlai, Thirimachos and Marcel, S{\'{e}}bastien and Schuckers, Stephanie},
projects = {Idiap, BEAT, ODIN/BATL},
title = {Face Liveness Detection Competition (LivDet-Face) - 2021},
booktitle = {International Joint Conference on Biometrics},
year = {2021},
abstract = {Liveness Detection (LivDet)-Face is an international competition series open to academia and industry. The competition’s objective is to assess and report state-of-the-art in liveness / Presentation Attack Detection (PAD) for face recognition. Impersonation and presentation of false samples to the sensors can be classified as presentation attacks and the ability for the sensors to detect such attempts is known as PAD. LivDet-Face 2021 \footnote{\url{https://face2021.livdet.org/}} will be the first edition of the face liveness competition. This competition serves as an important benchmark in face presentation attack detection, offering (a) an independent assessment of the current state of the art in face PAD, and (b) a common evaluation protocol, availability of Presentation Attack Instruments (PAI) and live face image dataset through the Biometric Evaluation and Testing (BEAT) platform. The competition can be easily followed by researchers after it is closed, in a platform in which participants can compare their solutions against the LivDet-Face winners.}
}
@INPROCEEDINGS{Raghavendra_BTAS_2015,
author = {Raghavendra, Ramachandra and Avinas, Manasa and Busch, Christoph and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Biometrics, Finger vein, Spoofing},
projects = {Idiap},
month = sep,
title = {Finger vein Liveness Detection Using Motion Magnification},
booktitle = {IEEE International Conference on Biometrics: Theory, Applications and Systems},
year = {2015},
pages = {1-7},
doi = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7358762},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Raghavendra_BTAS_2015.pdf}
}
@TECHREPORT{Raghavendra_Idiap-RR-17-2020,
author = {Raghavendra, Ramachandra and Stokkenes, Martin and Mohammadi, Amir and Venkatesh, Sushma and Raja, Kiran B. and Wasnik, Pankaj and Poiret, Eric and Marcel, S{\'{e}}bastien and Busch, Christoph},
keywords = {Biometrics, database, presentation attacks, Smartphones, Spoofing},
projects = {Idiap},
month = {7},
title = {Smartphone Multi-modal Biometric Authentication: Database and Evaluation},
type = {Idiap-RR},
number = {Idiap-RR-17-2020},
year = {2020},
institution = {Idiap},
url = {https://arxiv.org/abs/1912.02487},
abstract = {Biometric-based verification is widely employed on the smartphones for various applications, including financial transactions. In this work, we present a new multimodal biometric dataset (face, voice, and periocular) acquired using a smartphone. The new dataset is comprised of 150 subjects that are captured in six different sessions reflecting real-life scenarios of smartphone assisted authentication. One of the unique features of this dataset is that it is collected in four different geographic locations representing a diverse population and ethnicity. Additionally, we also present a multimodal Presentation Attack (PA) or spoofing dataset using a low-cost Presentation Attack Instrument (PAI) such as print and electronic display attacks. The novel acquisition protocols and the diversity of the data subjects collected from different geographic locations will allow developing a novel algorithm for either unimodal or multimodal biometrics. Further, we also report the performance evaluation of the baseline biometric verification and Presentation Attack Detection (PAD) on the newly collected dataset.},
pdf = {https://publications.idiap.ch/attachments/reports/2019/Raghavendra_Idiap-RR-17-2020.pdf}
}
@INPROCEEDINGS{Raghavendra_IWBF2019_2019,
author = {Raghavendra, Ramachandra and Venkatesh, Sushma and Raja, Kiran B. and Bhattacharjee, Sushil and Wasnik, Pankaj and Marcel, S{\'{e}}bastien and Busch, Christoph},
projects = {Idiap, SWAN, Tesla},
month = may,
title = {Custom Silicone Face Masks - Vulnerability of Commercial Face Recognition Systems & Presentation Attack Detection},
booktitle = {Proceedings of 7th IAPR/IEEE International Workshop on Biometrics and Forensics},
year = {2019},
abstract = {The paper estimates the vulnerability of two commercial FR systems (Verilook and FaceVACS) to presentation attacks based on 3d custom silicone face-masks. The study is based on a new dataset, CSMAD-Mobile, introduced in the paper. Baseline performance values of several standard face-PAD methods ((LBP,LPQ,BSIF,IDA,color-textures)+SVM), computed using the new dataset, are also presented in the paper.},
pdf = {https://publications.idiap.ch/attachments/papers/2019/Raghavendra_IWBF2019_2019.pdf}
}
@INPROCEEDINGS{Rahimi_ECCV_2024,
author = {Rahimi, Parsa and Razeghi, Behrooz and Marcel, S{\'{e}}bastien},
keywords = {3D-Rendered Datasets, face recognition systems, Image-to-Image Translation, Photorealism in Synthetic Data, Realism Transfer},
projects = {SAFER},
title = {Synthetic to Authentic: Transferring Realism to 3D Face Renderings for Boosting Face Recognition},
journal = {European Conference on Computer Vision Workshops},
booktitle = {European Conference on Computer Vision Workshops},
year = {2024},
abstract = {In this paper, we investigate the potential of image-to-image translation (I2I) techniques for transferring realism to 3D-rendered facial images in the context of Face Recognition (FR) systems. The primary motivation for using 3D-rendered facial images lies in their ability to circumvent the challenges associated with collecting large real face datasets for training FR systems. These images are generated entirely by 3D rendering engines, facilitating the generation of synthetic identities. However, it has been observed that FR systems trained on such synthetic datasets underperform when compared to those trained on real datasets, on various FR benchmarks. In this work, we demonstrate that by transferring the realism to 3D-rendered images (i.e., making the 3D-rendered images look more real), we can boost the performance of FR systems trained on these more photorealistic images. This improvement is evident when these systems are evaluated against FR benchmarks like IJB-C, LFW which utilize real-world data by 2\% to \%5, thereby paving new pathways for employing synthetic data in real-world applications. The project page is
available at: \url{https://idiap.ch/paper/syn2auth}.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Rahimi_ECCV_2024.pdf}
}
@INPROCEEDINGS{Rahimi_IJCB_2023,
author = {Rahimi, Parsa and Ecabert, Christophe and Marcel, S{\'{e}}bastien},
keywords = {bias, Controlled Synthesis, Fairness, generative models, Synthetic Dataset},
projects = {Idiap},
title = {Toward responsible face datasets: modeling the distribution of a disentangled latent space for sampling face images from demographic groups},
booktitle = {IEEE International Joint Conference on Biometrics},
year = {2023},
abstract = {Recently, it has been exposed that some modern facial recognition systems could discriminate specific demographic groups and may lead to unfair attention with respect to various facial attributes such as gender and origin. The main reason are the biases inside datasets, unbalanced demographics, used to train theses models. Unfortunately, collecting a large-scale balanced dataset with respect to various demographics is impracticable.
In this paper, we investigate as an alternative the generation of a balanced and possibly bias-free synthetic dataset that could be used to train, to regularize or to evaluate deep learning-based facial recognition models. We propose to use a simple method for modeling and sampling a disentangled projection of a StyleGAN latent space to generate any combination of demographic groups (e.g. $hispanic-female$). Our experiments show that we can synthesis any combination of demographic groups effectively and the identities are different from the original training dataset. We also released the source code \footnote{\url{https://gitlab.idiap.ch/biometric/sg_latent_modeling}}},
pdf = {https://publications.idiap.ch/attachments/papers/2023/Rahimi_IJCB_2023.pdf}
}
@INPROCEEDINGS{Ramoly_IJCB_2024,
author = {Ramoly, Nathan and Komaty, Alain and Krivokuca, Vedrana and Younes, Lara and Awal, Ahmad-Montaser and Marcel, S{\'{e}}bastien},
projects = {Idiap, SOTERIA},
title = {A Novel and Responsible Dataset for Face Presentation Attack Detection on Mobile Devices},
booktitle = {The IEEE International Joint Conference on Biometrics},
year = {2024},
pages = {8},
location = {Buffalo, New York},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Ramoly_IJCB_2024.pdf}
}
@INPROCEEDINGS{Razeghi_ICASSP2024_2024,
author = {Razeghi, Behrooz and Rahimi, Parsa and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center, SAFER},
month = apr,
title = {Deep Variational Privacy Funnel: General Modeling with Applications in Face Recognition},
booktitle = {49th IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2024},
publisher = {IEEE},
url = {https://ieeexplore.ieee.org/document/10446646},
doi = {https://doi.org/10.1109/ICASSP48485.2024.10446646},
abstract = {In this study, we harness the information-theoretic Privacy Funnel (PF) model to develop a method for privacy-preserving representation learning using an end-to-end training framework. We rigorously address the trade-off between obfuscation and utility. Both are quantified through the logarithmic loss, a measure also recognized as self-information loss. This exploration deepens the interplay between information-theoretic privacy and representation learning, offering substantive insights into data protection mechanisms for both discriminative and generative models. Importantly, we apply our model to state-of-the-art face recognition systems. The model demonstrates adaptability across diverse inputs, from raw facial images to both derived or refined embeddings, and is competent in tasks such as classification, reconstruction, and generation.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Razeghi_ICASSP2024_2024.pdf}
}
@TECHREPORT{Razeghi_Idiap-RR-02-2024,
author = {Razeghi, Behrooz and Rahimi, Parsa and Marcel, S{\'{e}}bastien},
month = {3},
title = {Deep Privacy Funnel Model: From a Discriminative to a Generative Approach with an Application to Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-02-2024},
year = {2024},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2024/Razeghi_Idiap-RR-02-2024.pdf}
}
@ARTICLE{Razeghi_IEEEACCESS_2024,
author = {Razeghi, Behrooz and Gheisari, Marzieh and Atashin, Amir and Kostadinov, Dimche and Marcel, S{\'{e}}bastien and Gunduz, Deniz and Voloshynovskiy, Slava},
projects = {Idiap, Biometrics Center},
title = {Group Membership Verification via Nonlinear Sparsifying Transform Learning},
journal = {IEEE Access},
volume = {12},
year = {2024},
pages = {86739-86751},
url = {https://ieeexplore.ieee.org/document/10565897},
doi = {10.1109/ACCESS.2024.3417301},
abstract = {In today’s digitally interconnected landscape, confirming the genuine associations between entities—whether they are items, devices, or individuals—and specific groups is critical. This paper introduces a new group membership verification method while ensuring minimal information loss, coupled with privacy-preservation and discrimination priors. Instead of verifying based on a similarity score in the original data space, we use a min-max functional measure in a transformed space. This method comprises two stages: (i) generating candidate nonlinear transform representations, and (ii) evaluating the min-max measure over these representations for both group assignment and transform selection. We simultaneously determine group membership and pick the appropriate representation from the candidate set based on the evaluation score. To solve within this framework, we employ an iterative alternating algorithm that both learns the parameters of candidate transforms and assigns group membership. Our method’s efficacy is assessed on public datasets across various verification and identification scenarios and further tested on real-world image databases, CFP and LFW.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Razeghi_IEEEACCESS_2024.pdf}
}
@INPROCEEDINGS{rodrig:2004:bioaw,
author = {Marcel, S{\'{e}}bastien and Rodriguez, Yann},
projects = {Idiap},
title = {Boosting Pixel-based Classifiers for Face Verification},
booktitle = {Biometric Authentication Workshop of the 8th European Conference on Computer Vision, BIOAW2004},
year = {2004},
publisher = {Springer-Verlag},
institution = {IDIAP},
address = {Prague, Czech Republic},
crossref = {rodrig2003},
abstract = {The performance of face authentication systems has steadily improved over the last few years. State-of-the-art methods use the projection of the gray-scale face image into a Linear Discriminant subspace as input of a classifier such as Support Vector Machines or Multi-layer Perceptrons. Unfortunately, these classifiers involve thousands of parameters that are difficult to store on a smart-card for instance. Recently, boosting algorithms has emerged to boost the performance of simple (weak) classifiers by combining them iteratively. The famous AdaBoost algorithm have been proposed for object detection and applied successfully to face detection. In this paper, we investigate the use of AdaBoost for face authentication to boost weak classifiers based simply on pixel values. The proposed approach is tested on a benchmark database, namely XM2VTS. Results show that boosting only hundreds of classifiers achieved near state-of-the-art results. Furthermore, the proposed approach outperforms similar work on face authentication using boosting algorithms on the same database.},
pdf = {https://publications.idiap.ch/attachments/papers/2004/rodrig-bioaw-2004.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2004/rodrig-bioaw-2004.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{rodrig:2004:icpr,
author = {Popovici, Vlad and Thiran, Jean-Philippe and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {On Performance Evaluation of Face Detection and Localization Algorithms},
booktitle = {17th International Conference on Pattern Recognition ({ICPR})},
year = {2004},
note = {IDIAP-RR 03-80},
crossref = {rodrig2003-2},
abstract = {When comparing different methods for face detection or localization, one realizes that just simply comparing the reported results is misleading as, even if the results are reported on the same dataset, different authors have different views of what a correct detection/localization means. This paper addresses exactly this problem, proposing an objective measure for the goodness of a detection/localization for the case of frontal faces. The usage of the proposed technique insures a fair and unbiased way of reporting the results, making the experiment repeatable, measurable, and comparable by anybody else.},
pdf = {https://publications.idiap.ch/attachments/papers/2004/rodrig-icpr-2004.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2004/rodrig-icpr-2004.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{rodriguez:eccv:2006,
author = {Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Authentication Using Adapted Local Binary Pattern Histograms},
booktitle = {9th European Conference on Computer Vision ({ECCV})},
year = {2006},
note = {IDIAP-RR 06-06},
crossref = {rodriguez:rr06-06},
abstract = {In this paper, we propose a novel generative approach for face authentication, based on a Local Binary Pattern (LBP) description of the face. A generic face model is considered as a collection of LBP-histograms. Then, a client-specific model is obtained by an adaptation technique from this generic model under a probabilistic framework. We compare the proposed approach to standard state-of-the-art face authentication methods on two benchmark databases, namely XM2VTS and BANCA, associated to their experimental protocol. We also compare our approach to two state-of-the-art LBP-based face recognition techniques, that we have adapted to the verification task.},
pdf = {https://publications.idiap.ch/attachments/papers/2006/rodrig-eccv-2006.pdf},
postscript = {ftp://ftp.idiap.ch/pub/papers/2006/rodrig-eccv-2006.ps.gz},
ipdmembership={vision},
}
@INPROCEEDINGS{Roy_ACMSAC2010_2010,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI},
month = {3},
title = {Visual processing-inspired Fern-Audio features for Noise-Robust Speaker Verification},
booktitle = {ACM 25th Symposium on Applied Computing, 2010, Sierre, Switzerland},
year = {2010},
organization = {Association for Computing Machinery},
crossref = {Roy_Idiap-RR-29-2009},
abstract = {In this paper, we consider the problem of speaker verification as a two-class object detection problem in computer vision, where the object instances are 1-D short-time spectral vectors obtained from the speech signal. More precisely, we investigate the general problem of speaker verification in the presence of additive white Gaussian noise, which we consider as analogous to visual object detection under varying illumination conditions. Inspired by their recent success in illumination-robust object detection, we apply a certain class of binary-valued pixel-pair based features called Ferns for noise-robust speaker verification. Intensive experiments on a benchmark database according to a standard evaluation protocol have shown the advantage of the proposed features in the presence of moderate to extremely high amounts of additive noise.},
pdf = {https://publications.idiap.ch/attachments/papers/2009/Roy_ACMSAC2010_2010.pdf}
}
@INPROCEEDINGS{Roy_BMVC2009_2009,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI},
month = {9},
title = {Haar Local Binary Pattern Feature for Fast Illumination Invariant Face Detection},
booktitle = {British Machine Vision Conference 2009},
year = {2009},
crossref = {Roy_Idiap-RR-28-2009},
abstract = {Face detection is the first step in many visual processing systems like face recognition, emotion recognition and lip reading. In this paper, we propose a novel feature called Haar Local Binary Pattern (HLBP) feature for fast and reliable face detection, particularly in adverse imaging conditions. This binary feature compares bin values of Local Binary Pattern histograms calculated over two adjacent image subregions. These subregions are similar to those in the Haar masks, hence the name of the feature. They capture the region-specific variations of local texture patterns and are boosted using AdaBoost in a framework similar to that proposed by Viola and Jones. Preliminary results obtained on several standard databases show that it competes well with other face detection systems, especially in adverse illumination conditions.},
pdf = {https://publications.idiap.ch/attachments/papers/2009/Roy_BMVC2009_2009.pdf}
}
@INPROCEEDINGS{Roy_BTAS2010_2010,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
keywords = {audio and video classification, audio-visual speaker recognition, crossmodal matching, Multimodal biometrics},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
title = {Introducing Crossmodal Biometrics:Person Identification from Distinct Audio & Visual Streams},
booktitle = {IEEE Fourth International Conference on Biometrics: Theory, Applications and Systems},
number = {4},
year = {2010},
abstract = {Person identification using audio or visual biometrics is a well-studied problem in pattern recognition. In this scenario, both training and testing are done on the same modalities. However, there can be situations where this condition is not valid, i.e. training and testing has to be done on different modalities. This could arise, for example, in covert surveillance. Is there any person specific information common to both the audio and visual (video-only) modalities which could be exploited to identify a person in such a constrained situation? In this work, we investigate this question in a principled way and propose a framework which can perform this task consistently better than chance, suggesting that such crossmodal biometric information exists.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/Roy_BTAS2010_2010.pdf}
}
@INPROCEEDINGS{Roy_ICASSP11_2011,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
title = {Phoneme Recognition using Boosted Binary Features},
booktitle = {IEEE Intl. Conference on Acoustics, Speech and Signal Processing 2011},
year = {2011},
abstract = {In this paper, we propose a novel parts-based binary-valued feature for ASR. This
feature is extracted using boosted
ensembles of simple threshold-based classifiers. Each such classifier
looks at a specific pair of
time-frequency bins located on the spectro-temporal plane.
These features termed as Boosted Binary Features (BBF) are integrated
into standard HMM-based system by using
multilayer perceptron (MLP) and single layer perceptron (SLP).
Preliminary studies on TIMIT phoneme recognition task show that
BBF yields similar or better performance compared to MFCC
(67.8\% accuracy for BBF vs. 66.3\% accuracy for MFCC) using MLP,
while it yields significantly better performance than MFCC (62.8\%
accuracy for BBF vs. 45.9\% for MFCC) using SLP. This demonstrates the
potential of the proposed feature for speech recognition.},
pdf = {https://publications.idiap.ch/attachments/papers/2011/Roy_ICASSP11_2011.pdf}
}
@INPROCEEDINGS{Roy_ICASSP2010_2010,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
month = {3},
title = {BOOSTED BINARY FEATURES FOR NOISE-ROBUST SPEAKER VERIFICATION},
booktitle = {2010 IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2010},
location = {Dallas, Texas},
abstract = {The standard approach to speaker verification is to extract cepstral features from the speech spectrum and model them by generative or discriminative techniques. We propose a novel approach where a set of client-specific binary features carrying maximal discriminative information specific to the individual client are estimated from an ensemble of pair-wise comparisons of frequency components in magnitude spectra, using Adaboost algorithm. The final classifier is a simple linear combination of these selected features. Experiments on the XM2VTS database strictly according to a standard evaluation protocol have shown that although the proposed framework yields comparatively lower performance on clean speech, it significantly outperforms the state-of-the-art MFCC-GMM system in mismatched conditions with training on clean speech and testing on speech corrupted by four types of additive noise from the standard Noisex-92 database.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/Roy_ICASSP2010_2010.pdf}
}
@INPROCEEDINGS{Roy_ICPR2010_2010,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI},
month = {4},
title = {Crossmodal Matching of Speakers using Lip and Voice Features in Temporally Non-overlapping Audio and Video Streams},
booktitle = {20th International Conference on Pattern Recognition, Istanbul, Turkey},
year = {2010},
location = {Istanbul, Turkey},
organization = {International Association for Pattern Recognition (IAPR)},
crossref = {Roy_Idiap-RR-13-2010},
abstract = {Person identification using audio (speech) and visual (facial appearance, static or dynamic) modalities, either independently or jointly, is a thoroughly investigated problem in pattern recognition. In this work, we explore a novel task : person identification in a cross-modal scenario, i.e., matching the speaker in an audio recording to the same speaker in a video recording, where the two recordings have been made during different sessions, using speaker specific information which is common to both the audio and video modalities. Several recent psychological studies have shown how humans can indeed perform this task with an accuracy significantly higher than chance. Here we propose two systems which can solve this task comparably well, using purely pattern recognition techniques. We hypothesize that such systems could be put to practical use in multimodal biometric and surveillance systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/Roy_ICPR2010_2010.pdf}
}
@TECHREPORT{Roy_Idiap-RR-29-2010,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI, IM2},
month = {8},
title = {Introducing Crossmodal Biometrics: Person Identification from Distinct Audio & Visual Streams},
type = {Idiap-RR},
number = {Idiap-RR-29-2010},
year = {2010},
institution = {Idiap},
abstract = {Person identification using audio or visual biometrics is a well-studied problem in pattern recognition. In this scenario, both training and testing are done on the same modalities. However, there can be situations where this condition is not valid, i.e. training and testing has to be done on different modalities. This could arise, for example, in covert surveillance. Is there any person specific information common to both the audio and visual (video-only) modalities which could be exploited to identify a person in such a constrained situation? In this work, we investigate this question in a principled way and propose a framework which can perform this task consistently better than chance, suggesting that such crossmodal biometric information exists.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Roy_Idiap-RR-29-2010.pdf}
}
@TECHREPORT{Roy_Idiap-RR-35-2011,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
keywords = {continuous speech recognition boosted binary features resource management},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI, TABULA RASA},
month = {10},
title = {Continuous Speech Recognition using Boosted Binary Features},
type = {Idiap-RR},
number = {Idiap-RR-35-2011},
year = {2011},
institution = {Idiap},
abstract = {A novel parts-based binary-valued feature termed Boosted Binary Feature (BBF) was recently proposed for ASR. Such features look at specific pairs of time-frequency bins in the spectro-temporal plane. The most discriminative of these features are selected by boosting and integrated into a standard HMM-based system using multilayer perceptron (MLP) and single layer perceptron (SLP). Previous studies on TIMIT phoneme recognition task showed that BBF yields similar or better performance compared to cepstral features. In this work, this study is extended to continuous speech recognition task on the DARPA Resource Management database. Results show that BBF achieves comparable word error rate (5.5\%) on this task with respect to standard cepstral features (5.1\%) using MLP. Using SLP, the error rate for BBF shows much lower degradation (from 5.5\% to 7.1\%) compared to cepstral features (from 5.1\% to 14.7\%). In addition, it is found that BBF features can be selected well using auxiliary data.},
pdf = {https://publications.idiap.ch/attachments/reports/2011/Roy_Idiap-RR-35-2011.pdf}
}
@ARTICLE{Roy_IEEETRANS.IFS_2011,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
month = feb,
title = {A Fast Parts-based Approach to Speaker Verification using Boosted Slice Classifiers},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {7},
number = {1},
year = {2012},
pages = {241-254},
abstract = {Speaker verification on portable devices like smartphones is gradually becoming popular. In this context, two issues need to be considered: 1) such devices have relatively limited computation resources, and 2) they are liable to be used everywhere, possibly in very noisy, uncontrolled environments. This work aims to address both these issues by proposing a computationally efficient yet robust speaker verification system. This novel parts-based system draws inspiration from face and
object detection systems in the computer vision domain. The system involves boosted ensembles of simple threshold-based classifiers. It uses a novel set of features extracted from speech
spectra, called “slice features”. The performance of the proposed system was evaluated through extensive studies involving a wide range of experimental conditions using the TIMIT, HTIMIT and
MOBIO corpus, against standard cepstral features and Gaussian Mixture Model-based speaker verification systems.},
pdf = {https://publications.idiap.ch/attachments/papers/2011/Roy_IEEETRANS.IFS_2011.pdf}
}
@INPROCEEDINGS{Roy_IJCB2011_2011,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
title = {Fast Speaker Verification on Mobile Phone data using Boosted Slice Classifiers},
booktitle = {IAPR IEEE International Joint Conference on Biometrics},
year = {2011},
location = {Washington DC},
abstract = {In this work, we investigate a novel computationally efficient speaker verification (SV) system involving boosted ensembles of simple threshold-based classifiers. The system is based on a novel set of features called “slice features”. Both the system and the features were inspired by
the recent success of pixel comparison-based ensemble approaches in the computer vision domain. The performance of the proposed system was evaluated through speaker verification
experiments on the MOBIO corpus containing mobile phone speech, according to a challenging protocol. The system was found to perform reasonably well, compared to multiple state-of-the-art SV systems, with the benefit of significantly lower computational complexity. Its dual characteristics
of good performance and computational efficiency could be important factors in the context of SV system implementation on portable devices like mobile phones.},
pdf = {https://publications.idiap.ch/attachments/papers/2011/Roy_IJCB2011_2011.pdf}
}
@INPROCEEDINGS{Roy_MLSLP_2012,
author = {Roy, Anindya and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {IM2, MOBIO, SNSF-MULTI},
month = sep,
title = {Boosting localized binary features for speech recognition},
booktitle = {Symposium on Machine Learning in Speech and Language Processing (MLSLP)},
year = {2012},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Roy_MLSLP_2012.pdf}
}
@INPROCEEDINGS{S.Luevano_IJCB2025_2025,
author = {S. Luevano, Luis and Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {SENTINEL-UK},
mainresearchprogram = {Human-AI Teaming},
title = {Identity-Preserving Aging and De-Aging of Faces in the StyleGAN Latent Space},
booktitle = {2025 IEEE International Joint Conference on Biometrics (IJCB)},
year = {2025},
publisher = {IEEE},
abstract = {Face aging or de-aging with generative AI has gained significant attention for its applications in such fields like forensics, security, and media. However, most state of the art methods rely on conditional Generative Adversarial Networks (GANs), Diffusion-based models, or Visual Language Models (VLMs) to age or de-age faces based on predefined age categories and conditioning via loss functions, fine-tuning, or text prompts. The reliance on such conditioning leads to complex training requirements, increased data needs, and challenges in generating consistent results. Additionally,
identity preservation is rarely taken into accountor evaluated on a single face recognition system without any control or guarantees on whether identity would be preserved in a generated aged/de-aged face. In this paper, we propose to synthesize aged and de-aged faces via editing latent space of StyleGAN2 using a simple support vector modeling of aging/de-aging direction and several feature selection approaches. By using two state-of-the-art face recognition systems, we empirically find the identity preserving subspace within the StyleGAN2 latent space, so that an apparent age of a given face can changed while preserving the identity. We then propose a simple yet practical formula for estimating the limits on aging/de-aging parameters that ensures identity preservation for a given input face. Using
our method and estimated parameters we have generated a public dataset of synthetic faces at different ages that can
be used for benchmarking cross-age face recognition, age assurance systems, or systems for detection of synthetic images. Our code and dataset are available at the project page https://www.idiap.ch/paper/agesynth/},
pdf = {https://publications.idiap.ch/attachments/papers/2025/S.Luevano_IJCB2025_2025.pdf}
}
@TECHREPORT{Sahidullah_Idiap-RR-14-2019,
author = {Sahidullah, Md and Patino, Jose and Cornell, Samuele and Yin, Ruiqing and Sivasankaran, Sunit and Bredin, Herve and Korshunov, Pavel and Brutti, Alessio and Serizel, Romain and Vincent, Emmanuel and Evans, Nicholas and Marcel, S{\'{e}}bastien and Squartini, Stefano and Barras, Claude},
keywords = {diarization, DIHARD challenge, evaluation, single-channel and multi-channel speech},
projects = {Idiap, ODESSA},
month = {11},
title = {The Speed Submission to DIHARD II: Contributions & Lessons Learned},
type = {Idiap-RR},
number = {Idiap-RR-14-2019},
year = {2019},
institution = {Idiap},
note = {The paper on arXiv: https://arxiv.org/abs/1911.02388},
abstract = {This paper describes the speaker diarization systems developed for the Second DIHARD Speech Diarization Challenge (DIHARD II) by the Speed team. Besides describing the system, which considerably outperformed the challenge baselines, we also focus on the lessons learned from numerous approaches that we tried for single and multi-channel systems. We present several components of our diarization system, including categorization of domains, speech enhancement, speech activity detection, speaker embeddings, clustering methods, resegmentation, and system fusion. We analyze and discuss the effect of each such component on the overall diarization performance within the realistic settings of the challenge.},
pdf = {https://publications.idiap.ch/attachments/reports/2019/Sahidullah_Idiap-RR-14-2019.pdf}
}
@INPROCEEDINGS{sanders-icme03,
author = {Sanderson, Conrad and Bengio, Samy and Bourlard, Herv{\'{e}} and Mari{\'{e}}thoz, Johnny and Collobert, Ronan and BenZeghiba, Mohamed Faouzi and Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {7},
title = {{S}peech & {F}ace {B}ased {B}iometric Authentication at {IDIAP}},
booktitle = {{P}roceedings of the 2003 {IEEE} International {C}onference on {M}ultimedia & Expo ({ICME}-03)},
year = {2003},
address = {Baltimore, Maryland},
crossref = {sanders-rr-03-13},
ipdmembership={learning},
}
@TECHREPORT{Sarfjoo_Idiap-RR-10-2019,
author = {Sarfjoo, Seyyed Saeed and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
editor = {Sarfjoo, Seyyed Saeed},
projects = {Tesla, UNITS},
month = {9},
title = {Domain Adaptation and Investigation of Robustness of DNN-based Embeddings for Text-Independent Speaker Verification Using Dilated Residual Networks},
type = {Idiap-RR},
number = {Idiap-RR-10-2019},
year = {2019},
institution = {Idiap},
address = {Centre du Parc, Rue Marconi 19, P.O. Box 592, CH - 1920 Martigny},
abstract = {Robustness of extracted embeddings in cross-database scenarios is one of the main challenges in text-independent speaker verification (SV) systems. In this paper, we investigate this robustness via performing structural cross-database experiments with or without additive noise. This noise can be added from the seen set, where the noise type is similar to the noise which is used in data augmentation for training the SV model, or unseen set, where distribution of additive noise in train and evaluation sets are different. For extracting the robust embeddings, we investigate applying the time dilation in the ResNet architecture, so-called dilated residual network (DRN). Dimension and number of segment level layers are tuned in this architecture. The proposed model with time dilation significantly outperformed the ResNet model and is comparable with the state-of-the-art SV systems on Voxceleb1 dataset. In addition, this architecture showed significant robustness in out of domain scenarios.
Language mismatch is part of domain mismatch which recently is one of the main focuses of research in SV systems. Similar to image recognition field, we hypothesize that low-level convolutional neural network (CNN) layers are domain-specific features while high-level CNN layers are domain-independent and have more discriminative power. For adapting these domain-specific units, combination of triplet and intra-class losses are investigated. The adapted model on the evaluation part of the CMN2 dataset, relatively outperformed the DRN and x-vector SV systems without adaptation with 8.0 and 20.5 \%, respectively in equal error-rate.},
pdf = {https://publications.idiap.ch/attachments/reports/2019/Sarfjoo_Idiap-RR-10-2019.pdf}
}
@TECHREPORT{Sarfjoo_Idiap-RR-15-2019,
author = {Sarfjoo, Seyyed Saeed and Madikeri, Srikanth and Hajibabaei, Mahdi and Motlicek, Petr and Marcel, S{\'{e}}bastien},
keywords = {adaptation, batch normalization, speaker recognition},
projects = {Idiap, ODESSA, EC H2020-ROXANNE},
month = {11},
title = {Idiap submission to the NIST SRE 2019 Speaker Recognition Evaluation},
type = {Idiap-RR},
number = {Idiap-RR-15-2019},
year = {2019},
institution = {Idiap},
address = {Rue Marconi 19, 1920 Martigny},
abstract = {Idiap has made a submission to the conversational telephony speech (CTS) challenge of the NIST SRE 2019. The submission consists of six speaker verification (SV) systems: four extended TDNN (E-TDNN) and two TDNN x-vector systems. Employment of various training sets, loss functions, adaptation sets and extracted speech features is among the main differences of the submitted systems. Domain adaptation is represented by a supervised method (developed using a limited data) with transfer learning of the batch norm layers. \% was applied.
Mean shift and covariance estimation of batch norm allows to map the target domain to the source domain, alleviating the problem of over-fitting on the adaptation data.
The back-end of all the systems is represented by the conventional
Linear Discriminant Analysis (LDA) projection followed by Probabilistic LDA (PLDA) scoring for inference. The PLDA was also adapted unsupervisedly using the unlabelled part of the NIST SRE 2018 set. In addition, training the LDA and PLDA using in-domain data was investigated. The entire system was built around the Kaldi toolkit.},
pdf = {https://publications.idiap.ch/attachments/reports/2019/Sarfjoo_Idiap-RR-15-2019.pdf}
}
@INPROCEEDINGS{Sarfjoo_INTERSPEECH_2020,
author = {Sarfjoo, Seyyed Saeed and Madikeri, Srikanth and Motlicek, Petr and Marcel, S{\'{e}}bastien},
keywords = {batch norm, speaker recognition, speaker verification, supervised adaptation, transfer learning},
projects = {ODESSA, EC H2020-ROXANNE},
title = {Supervised domain adaptation for text-independent speaker verification using limited data},
booktitle = {Interspeech},
year = {2020},
pages = {3815-3819},
url = {http://www.interspeech2020.org/uploadfile/pdf/Thu-1-7-4.pdf},
abstract = {To adapt the speaker verification (SV) system to a target domain with limited data, this paper investigates the transfer learning of the model pre-trained on the source domain data. To that end, layer-by-layer adaptation with transfer learning from the initial and final layers of the pre-trained model is investigated. We show that the model adapted from the initial layers outperforms the model adapted from the final layers. Based on this evidence, and inspired by the works in image recognition field, we hypothesize that low-level convolutional neural network (CNN) layers characterize domain-specific component while high-level CNN layers are domain-independent and have more discriminative power. For adapting these domain-specific components, angular margin softmax (AMSoftmax) applied on the CNN-based implementation of the x-vector architecture. In
addition, to reduce the problem of over-fitting on the limited target data, transfer learning on the batch norm layers is investigated. Mean shift and covariance estimation of batch norm allows to map the represented components of the target domain to the source domain. Using TDNN and E-TDNN versions of the x-vectors as baseline models, the adapted models on the development set of NIST SRE 2018 outperformed the baselines
with relative improvements of 11.0 and 13.8 \%, respectively.},
pdf = {https://publications.idiap.ch/attachments/papers/2020/Sarfjoo_INTERSPEECH_2020.pdf}
}
@INPROCEEDINGS{Sarkar_ICASSP_2022,
author = {Sarkar, Eklavya and Korshunov, Pavel and Colbois, Laurent and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Face Recognition, Morphing Attack, StyleGAN 2, Vulnerability Analysis},
projects = {Idiap, Biometrics Center},
month = may,
title = {Are GAN-based Morphs Threatening Face Recognition?},
booktitle = {International Conference on Acoustics, Speech and Signal Processing},
year = {2022},
abstract = {Morphing attacks are a threat to biometric systems where the biometric reference in an identity document can be altered. This form of attack presents an important issue in applications relying on identity documents such as border security or access control. Research in generation of face morphs and their detection is developing rapidly, however very few datasets with morphing attacks and open-source detection toolkits are publicly available. This paper bridges this gap by providing two datasets and the corresponding code for four types of morphing attacks: two that rely on facial landmarks based on OpenCV and FaceMorpher, and two that use StyleGAN 2 to generate synthetic morphs. We also conduct extensive experiments to assess the vulnerability of four state-of-the-art face recognition systems, including FaceNet, VGG-Face, ArcFace, and ISV. Surprisingly, the experiments demonstrate that, although visually more appealing, morphs based on StyleGAN 2 do not pose a significant threat to the state to face recognition systems, as these morphs were outmatched by the simple morphs that are based facial landmarks.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Sarkar_ICASSP_2022.pdf}
}
@TECHREPORT{Sarkar_Idiap-RR-38-2020,
author = {Sarkar, Eklavya and Korshunov, Pavel and Colbois, Laurent and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Face Recognition, Morphing Attack, StyleGAN 2, Vulnerability Analysis},
projects = {Idiap},
month = {12},
title = {Vulnerability Analysis of Face Morphing Attacks from Landmarks and Generative Adversarial Networks},
type = {Idiap-RR},
number = {Idiap-RR-38-2020},
year = {2020},
institution = {Idiap},
address = {19 Rue Macroni, 1920 Martigny},
abstract = {Morphing attacks is a threat to biometric systems where the biometric reference in an identity document can be altered. This form of attack presents an important issue in applications relying on identity documents such as border security or access control. Research in face morphing attack detection is developing rapidly, however very few datasets with several forms of attacks are publicly available. This paper bridges this gap by providing a new dataset with four different types of morphing attacks, based on OpenCV, FaceMorpher, WebMorph and a generative adversarial network (Style-GAN), generated with original face images from three public face datasets. We also conduct extensive experiments to assess the vulnerability of the state-of-the-art face recognition systems, notably FaceNet, VGG-Face, and ArcFace. The experiments demonstrate that VGG-Face, while being less accurate face recognition system compared to FaceNet, is also less vulnerable to morphing attacks. Also, we observed that na{\i}̈ve morphs generated with a StyleGAN do not pose a significant threat.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/Sarkar_Idiap-RR-38-2020.pdf}
}
@TECHREPORT{sauquet-rr-05-49,
author = {Sauquet, Tiffany and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {{Multiview Face Detection}},
type = {Idiap-RR},
number = {Idiap-RR-49-2005},
year = {2005},
institution = {IDIAP},
abstract = {In this document, we address the problem of multiview face detection. This work extends the frontal face detection system developed at the IDIAP Research Institute to multiview face detection. The main state-of-the art techniques are reviewed and a novel architecture is presented, based on a pyramid of detectors that are trained for different views of faces. The proposed approach robustly detects faces rotated up to -67.5 degree in the image plane and up to -90 degree out of the image plane. The system is real-time and achieves high performances on benchmark test sets, comparable to some state-of-the-art approaches.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/rr-05-49.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/rr-05-49.ps.gz},
ipdmembership={vision},
language={English},
}
@INPROCEEDINGS{Sequeira_IJCB_2017,
author = {Sequeira, Ana and Chen, Lulu and Ferryman, James and Wild, Peter and Alonso-Fernandez, Fernando and Big{\"{u}}n, Josef and Raja, Kiran B. and Raghavendra, R. and Busch, Christoph and de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien and Behera, Sushree Sangeeta and Gour, Mahesh and Kanhangad, Vivek},
projects = {Idiap, HFACE},
month = oct,
title = {Cross-Eyed 2017: Cross-Spectral Iris/Periocular Recognition Competition.},
booktitle = {IEEE/IAPR International Joint Conference on Biometrics},
year = {2017},
publisher = {IEEE},
location = {Denver, Colorado, USA}
}
@ARTICLE{Shamsi_COMPUTERSPEECH&LANGUAGE_2022,
author = {Shamsi, Meysam and Larcher, Anthony and Barrault, Lo{\"{\i}}c and Meignier, Sylvain and Prokopalo, Yevhenii and Tahon, Marie and Mehrish, Ambuj and Petitrenaud, Simon and Galibert, Olivier and Gaist, Samuel and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien and Costa-Juss{\`{a}}, Marta},
projects = {Idiap},
month = jul,
title = {Towards Lifelong Human Assisted Speaker Diarization},
journal = {Computer Speech & Language},
year = {2022},
url = {https://www.sciencedirect.com/science/article/pii/S0885230822000638},
doi = {10.1016/j.csl.2022.101437},
abstract = {This paper introduces the resources necessary to develop and evaluate human assisted lifelong learning speaker diarization systems. It describes the ALLIES corpus and associated protocols, especially designed for diarization of a collection audio recordings across time. This dataset is compared to existing corpora and the performances of three baseline systems, based on x-vectors, i-vectors and VBxHMM, are reported for reference. Those systems are then extended to include an active correction process that efficiently guides a human annotator to improve the automatically generated hypotheses. An open-source simulated human expert is provided to ensure reproducibility of the human assisted correction process and its fair evaluation. An exhaustive evaluation, of the human assisted correction shows the high potential of this approach. The ALLIES corpus, a baseline system including the active correction module and all evaluation tools are made freely available to the scientific community.}
}
@ARTICLE{Sizov_IEEETIFS_2015,
author = {Sizov, Aleksandr and Khoury, Elie and Kinnunen, Tomi and Wu, Zhizheng and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI},
month = apr,
title = {Joint Speaker Verification and Anti-Spoofing in the i-Vector Space},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {10},
number = {4},
year = {2015},
pages = {821-832},
doi = {10.1109/TIFS.2015.2407362},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Sizov_IEEETIFS_2015.pdf}
}
@INPROCEEDINGS{Subburaman_ECCVWORKSHOP-2_2010,
author = {Subburaman, Venkatesh Bala and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2},
month = {9},
title = {Fast Bounding Box Estimation based Face Detection},
booktitle = {ECCV, Workshop on Face Detection: Where we are, and what next?},
year = {2010},
url = {http://vis-www.cs.umass.edu/fdWorkshop/papers/W03.010.pdf},
crossref = {Subburaman_Idiap-RR-38-2010},
abstract = {The sliding window approach is the most widely used technique
to detect an object from an image. In the past few years, classifiers
have been improved in many ways to increase the scanning speed. Apart
from the classifier design (such as cascade,',','),
the scanning speed also depends
on number of different factors (such as grid spacing, and scale at
which the image is searched). When the scanning grid spacing is larger
than the tolerance of the trained classifier it suffers from low detections.
In this paper we present a technique to reduce the number of miss detections
while increasing the grid spacing when using the sliding window
approach for object detection. This is achieved by using a small patch
to predict the bounding box of an object within a local search area. To
achieve speed it is necessary that the bounding box prediction is comparable
or better than the time it takes in average for the object classifier
to reject a subwindow. We use simple features and a decision tree as
it proved to be efficient for our application. We analyze the effect of
patch size on bounding box estimation and also evaluate our approach
on benchmark face database (CMU+MIT). We also report our results on
the new FDDB dataset [1]. Experimental evaluation shows better detection
rate and speed with our proposed approach for larger grid spacing
when compared to standard scanning technique.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/Subburaman_ECCVWORKSHOP-2_2010.pdf}
}
@INPROCEEDINGS{Subburaman_ICASSP_2010,
author = {Subburaman, Venkatesh Bala and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO},
title = {An Alternative Scanning Strategy to Detect Faces},
booktitle = {Proceedings IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2010},
location = {Dallas, USA},
abstract = {The sliding window approach is the most widely used technique to detect faces in an image. Usually a classifier is applied on a regular grid and to speed up the scanning, the grid spacing is increased, which increases the number of miss detections. In this paper we propose an alternative scanning method which minimizes the number of misses, while improving the speed of detection. To achieve this we use an additional classifier that predicts the bounding box of a face within a local search area. Then a face/non-face classifier is used to verify the presence or absence of a face. We propose a new combination of binary features which we term as u-Ferns for bounding box estimation, which performs comparable or better than former techniques. Experimental evaluation on benchmark database show that we can achieve 15-30\% improvement in detection rate or speed when compared to the standard scanning technique.},
pdf = {https://publications.idiap.ch/attachments/papers/2010/Subburaman_ICASSP_2010.pdf}
}
@TECHREPORT{Subburaman_Idiap-Com-01-2011,
author = {Subburaman, Venkatesh Bala and Marcel, S{\'{e}}bastien},
editor = {Subburaman, Venkatesh Bala},
keywords = {Cascade, Face Detection, Ferns},
projects = {Idiap, IM2},
month = {12},
title = {Face Detection using Ferns},
type = {Idiap-Com},
number = {Idiap-Com-01-2011},
year = {2011},
institution = {Idiap},
address = {Centre du Parc, Rue Marconi 19, Case Postale 592, CH-1920 Martigny, Switzerland, tel. +41 27 721 77 11, fax +41 27 721 77 12, infor@idiap.ch, www.idiap.ch},
abstract = {This paper discusses the use of ferns (a set of binary features) for face detection.
The binary feature used here is the sign of pixel intensity difference. Ferns were first introduced for keypoint recognition and showed good performance, and improving the speed of recognition. Keypoint recognition deals with classification of few hundred different classes, while face detection is a two-class problem with an unbalanced data. For keypoint recognition random pixel pairs proved to be good enough while we used conditional mutual information criteria to select a small subset of informative binary feature to build class conditional densities and a Naive Bayesian classifier is used for face and non-face classification. We compared our approach with boosted haar-like features, modified census transform (MCT,',','),
and local binary pattern on a single stage classifier. Results shows that ferns when compared to haar-like features are robust to illumination changes and comparable to boosted MCT feature. Finally a cascade of classifiers was built and the performance on cropped face images and the localization results using Jesorsky measure are reported on XM2VTS and BANCA database.},
pdf = {https://publications.idiap.ch/attachments/reports/2008/Subburaman_Idiap-Com-01-2011.pdf}
}
@INPROCEEDINGS{Tapia_IJCB2025_2025,
author = {Tapia, Juan E. and Mario, Nieto and Espin, Juan and Sanchez, Alvaro and Damer, Naser and Busch, Christoph and Ivanovska, Marija and Todorov, Leon and Khizbullin, Renat and Grishin, Aleksei and Lazarevic, Lazar and Schulz, Daniel and Gonzalez, Sebastian and Mohammadi, Amir and Kotwal, Ketan and Marcel, S{\'{e}}bastien and Mudgalgundurao, Raghavendra and Raja, Kiran B. and Schuch, Patrick and Couto, Pedro and Pinto, Joao and Xavier, Mariana and Valenzuela, Andres and Batagelj, Borut and Barrachina, Javier and Peterlin, Marko and Peer, Peter and Muhammed, Ajnas and Nunes, Diogo and Gon{\c c}alves, Nuno and Patwardhan, Sushrut and Ramachandra, Raghavendra},
projects = {ROSALIND},
mainresearchprogram = {Sustainable & Resilient Societies},
additionalresearchprograms = {AI for Everyone},
month = sep,
title = {Second Competition on Presentation Attack Detection on ID Card},
booktitle = {IEEE International Joint Conference on Biometrics (IJCB)},
year = {2025},
abstract = {This work summarises and reports the results of the second Presentation Attack Detection competition on ID cards. This new version includes new elements compared to the previous one. (1) An automatic evaluation platform was enabled for automatic benchmarking; (2) Two tracks were proposed in order to evaluate algorithms and datasets respectively; and (3) A new ID card dataset was shared with Track 1 teams to serve as the baseline dataset for the training and optimisation.
The Hochschule Darmstadt, Fraunhofer-IGD, and Facephi company organised jointly this challenge.
20 teams were registered, and 74 submitted models were evaluated. For Track 1, the "Dragons" team reached first place with an Average Ranking and Equal Error rate (EER) of AV_Rank of 40.48\% and 11.44\% EER, respectively. For the more challenging approach in Track 2, the "Incode" team reached the best results with an AV_Rank of 14.76\% and 6.36\% EER, improving on the results of the first edition of 74.30\% and 21.87\% EER, respectively. These results suggest that PAD on ID cards is improving, but it is still a challenging problem related to the number of images, especially of bona fide images.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Tapia_IJCB2025_2025.pdf}
}
@INPROCEEDINGS{Tome_BIOSIG_2015,
author = {Tome, Pedro and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT},
month = sep,
title = {Palm Vein Database and Experimental Framework for Reproducible Research},
booktitle = {IEEE International Conference of the Biometrics Special Interest Group},
year = {2015},
pages = {1-7},
url = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7314614},
doi = {10.1109/BIOSIG.2015.7314614},
abstract = {A palm vein database acquired by a contactless sensor together with an experimental framework freely available for fair reproducible research purposes are described. The palm vein recognition system uses automatic palm region segmentation and circular Gabor filter approach to enhance the veins in the preprocessing, LBP features and histogram intersection as matching. Results are presented comparing two automatic segmentation using the ROI-1 region proportioned by the acquisition sensor and the ROI-2 region generated by the recognition software developed. Complete benchmark results using popular methods and the source code are attached to the database as a reference for other researchers.},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Tome_BIOSIG_2015.pdf}
}
@INPROCEEDINGS{Tome_ICB2015-SpoofingPalmvein,
author = {Tome, Pedro and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT, TABULA RASA},
month = may,
title = {On the Vulnerability of Palm Vein Recognition to Spoofing Attacks},
booktitle = {The 8th IAPR International Conference on Biometrics (ICB)},
year = {2015},
pages = {319 - 325},
url = {http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=7139056},
doi = {10.1109/ICB.2015.7139056},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Tome_ICB2015-SpoofingPalmvein.pdf}
}
@INPROCEEDINGS{Tome_ICB2015_AntiSpoofFVCompetition,
author = {Tome, Pedro and Raghavendra, Ramachandra and Busch, Christoph and Tirunagari, Santosh and Poh, Norman and Shekar, B. H. and Gragnaniello, Diego and Sansone, Carlo and Verdoliva, Luisa and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT, TABULA RASA},
month = may,
title = {The 1st Competition on Counter Measures to Finger Vein Spoofing Attacks},
booktitle = {The 8th IAPR International Conference on Biometrics (ICB)},
year = {2015},
pages = {513 - 518},
url = {http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7139067},
doi = {10.1109/ICB.2015.7139067},
pdf = {https://publications.idiap.ch/attachments/papers/2015/Tome_ICB-2015_2015.pdf}
}
@INPROCEEDINGS{Tome_IEEEBIOSIG2014,
author = {Tome, Pedro and Vanoni, Matthias and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Finger vein, Spoofing Attacks},
projects = {Idiap, BEAT},
month = sep,
title = {On the Vulnerability of Finger Vein Recognition to Spoofing},
booktitle = {IEEE International Conference of the Biometrics Special Interest Group (BIOSIG)},
volume = {230},
year = {2014},
pages = {1 - 10},
publisher = {IEEE},
location = {Darmstadt, Germay},
isbn = {978-3-88579-624-4},
abstract = {The vulnerability of finger vein recognition to spoofing is studied in this paper. A collection of spoofing finger vein images has been created from real finger vein samples. Finger vein images are printed using a commercial printer and then, presented at an open source finger vein sensor. Experiments are carried out using an extensible framework, which allows fair and reproducible benchmarks. Experimental results lead to a spoofing false accept rate of 86\%, thus showing that finger vein biometrics is vulnerable to spoofing attacks, pointing out the importance to investigate countermeasures against this type of fraudulent actions.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Tome_IEEEBIOSIG2014.pdf}
}
@INPROCEEDINGS{Unnervik_BIOSIG2022_2022,
author = {Unnervik, Alexander and Marcel, S{\'{e}}bastien},
keywords = {anomaly detection, Backdoor attack, Biometrics, CNN, Face Recognition, security, trojan attack},
projects = {Idiap, Biometrics Center},
month = sep,
title = {An anomaly detection approach for backdoored neural networks: face recognition as a case study},
booktitle = {21st International Conference of the Biometrics Special Interest Group (BIOSIG 2022)},
year = {2022},
location = {Darmstadt, Germany},
crossref = {Unnervik_Idiap-RR-08-2022},
abstract = {Backdoor attacks allow an attacker to embed functionality jeopardizing proper behavior of any algorithm, machine learning or not. This hidden functionality can remain inactive for normal use of the algorithm until activated by the attacker. Given how stealthy backdoor attacks are, conse- quences of these backdoors could be disastrous if such networks were to be deployed for applications as critical as border or access control. In this paper, we propose a novel backdoored network detec- tion method based on the principle of anomaly detection, involving access to the clean part of the training data and the trained network. We highlight its promising potential when considering various triggers, locations and identity pairs, without the need to make any assumptions on the nature of the backdoor and its setup. We test our method on a novel dataset of backdoored networks and report detectability results with perfect scores.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Unnervik_BIOSIG2022_2022.pdf}
}
@INPROCEEDINGS{Unnervik_NEURIPS_2024,
author = {Unnervik, Alexander and Otroshi Shahreza, Hatef and George, Anjith and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN},
title = {Model Pairing Using Embedding Translation for Backdoor Attack Detection on Open-Set Classification Tasks},
booktitle = {NeurIPS Safe Generative AI Workshop 2024},
year = {2024},
abstract = {Backdoor attacks allow an attacker to embed a specific vulnerability in a machine learning algorithm, activated when an attacker-chosen pattern is presented, causing a specific misprediction. The need to identify backdoors in biometric scenarios has led us to propose a novel technique with different trade-offs. In this paper we propose to use model pairs on open-set classification tasks for detecting backdoors. Using a simple linear operation to project embeddings from a probe model's embedding space to a reference model's embedding space, we can compare both embeddings and compute a similarity score. We show that this score, can be an indicator for the presence of a backdoor despite models being of different architectures, having been trained independently and on different datasets. This technique allows for the detection of backdoors on models designed for open-set classification tasks, which is little studied in the literature. Additionally, we show that backdoors can be detected even when both models are backdoored. The source code is made available for reproducibility purposes.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Unnervik_NEURIPS_2024.pdf}
}
@PHDTHESIS{Unnervik_THESIS_2024,
author = {Unnervik, Alexander},
editor = {Marcel, S{\'{e}}bastien},
keywords = {anomaly detection, Backdoor attack, backdoor attack detection, Convolutional neural network., embedding translation, Face Recognition, model pairing, trojan attack},
projects = {Idiap, TRESPASS-ETN},
month = jun,
title = {Performing And Detecting Backdoor Attacks on Face Recognition Algorithms},
year = {2024},
school = {Ecole Polytechnique F{\'{e}}d{\'{e}}rale de Lausanne},
abstract = {The field of biometrics, and especially face recognition, has seen a wide-spread adoption the last few years, from access control on personal devices such as phones and laptops, to automated border controls such as in airports. The stakes are increasingly higher for these applications and thus the risks of succumbing to attacks are rising. More sophisticated algorithms typically require more data samples and larger models, leading to the need for more compute and expertise. These add up to making deep learning algorithms more a service provided by third parties, meaning more control and oversight of these algorithms are relinquished.
When so much depends on these models working right, with nefarious actors gaining so much from them being circumvented, how does one then verify their integrity? This is the conundrum of integrity which is at the heart of the work presented here.
One way by which face recognition algorithms (or more generally speaking, deep learning algorithms) fail, is by being vulnerable to backdoor attacks (BA): a type of attack involving a modification of the training set or the network weights to control the output behavior when exposed to specific samples. The detection of these backdoored networks (which we refer to as backdoor attack detection (BAD) is a challenging task, which is still an active field of research, particularly so when considering the constraints within which the literature considers the challenge (e.g. little to no consideration of open-set classification algorithms).
In this thesis, we demonstrate that BAs can be performed on large face recognition algorithms and further the state of the art in BAD by providing with the following contributions: first, we study the vulnerability of face recognition algorithms to backdoor attacks and identify backdoor attack success with respect to the choice of identities and other variables. Second, we propose a first method by which backdoor attacks can be detected by studying weights distribution of clean models and comparing an unknown model to such distributions. This method is based on the principle of anomaly detection. Third, we propose a method for safely deploying models to make use of their clean behavior and detecting the activation of backdoors with a technique we call model pairing.},
pdf = {https://publications.idiap.ch/attachments/papers/2024/Unnervik_THESIS_2024.pdf}
}
@INPROCEEDINGS{Unnervik_WACVW_2024,
author = {Unnervik, Alexander and George, Anjith and Ecabert, Christophe and Rahimi, Parsa and Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {TRESPASS-ETN, SAFER},
title = {FRCSyn Challenge at WACV 2024: Face Recognition Challenge in the Era of Synthetic Data},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops},
year = {2024},
pages = {892-901},
url = {https://openaccess.thecvf.com/content/WACV2024W/FRCSyn/html/Melzi_FRCSyn_Challenge_at_WACV_2024_Face_Recognition_Challenge_in_the_WACVW_2024_paper}
}
@INPROCEEDINGS{vanderMeer_ACL2025_2025,
author = {van der Meer, Michiel and Korshunov, Pavel and Marcel, S{\'{e}}bastien and van der Plas, Lonneke},
projects = {Idiap, FACTCHECK},
mainresearchprogram = {Sustainable & Resilient Societies},
month = jul,
title = {HintsOfTruth: A Multimodal Checkworthiness Detection Dataset with Real and Synthetic Claims},
booktitle = {The 63rd Annual Meeting of the Association for Computational Linguistics},
year = {2025},
abstract = {Misinformation can be countered with fact-checking, but the process is costly and slow. Identifying checkworthy claims is the first step, where automation can help scale fact-checkers’ efforts. However, detection methods struggle with content that is (1) multimodal, (2) from diverse domains, and (3) synthetic. We introduce HINTSOFTRUTH, a public dataset for multimodal checkworthiness detection with 27K real-world and synthetic image/claim pairs. The mix of real and synthetic data makes this dataset unique and ideal for benchmarking detection methods. We compare fine-tuned and prompted Large Language Models (LLMs). We find that well-configured lightweight text-based encoders perform comparably to multimodal models but the former only focus on identifying non-claim-like content. Multimodal LLMs can be more accurate but come at a significant computational cost, making them impractical for large-scale applications. When faced with synthetic data, multimodal models perform more robustly.},
pdf = {https://publications.idiap.ch/attachments/papers/2025/vanderMeer_ACL2025_2025.pdf}
}
@INPROCEEDINGS{Vanoni_BIOMS_2014,
author = {Vanoni, Matthias and Tome, Pedro and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT},
month = oct,
title = {Cross-Database Evaluation With an Open Finger Vein Sensor},
booktitle = {IEEE Workshop on Biometric Measurements and Systems for Security and Medical Applications (BioMS)},
year = {2014},
pages = {30-35},
publisher = {IEEE},
location = {Rome, Italy},
isbn = {978-1-4799-5175-8},
doi = {10.1109/BIOMS.2014.6951532},
abstract = {Finger vein recognition is a recent biometric application, which relies on the use of human finger vein patterns beneath the skin's surface. While several methods have been proposed in the literature, its applicability to uncontrolled scenarios has not yet been shown. To this purpose this paper first introduces the VERA database, a new challenging publicly available database of finger vein images. This corpus consists of 440 index finger images from 110 subjects collected with an open device in an uncontrolled way. Second, an evaluation of state-of-the-art finger vein recognition systems is performed, both on the controlled UTFVP database and on the new VERA database. This is achieved using a new open source and extensible framework, which allows fair and reproducible benchmarks. Experimental results show that challenging recording conditions such as misalignments of the fingers lead to an absolute degradation in equal error rate of 2.75\% up to 24.10\% on VERA when compared to the best performances on UTFVP.},
pdf = {https://publications.idiap.ch/attachments/papers/2014/Vanoni_BIOMS_2014.pdf}
}
@INPROCEEDINGS{Vidit_SRBS BMVCWORKSHOP_2025,
author = {Vidit, Vidit and Korshunov, Pavel and Mohammadi, Amir and Ecabert, Christophe and Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {ROSALIND},
mainresearchprogram = {AI for Everyone},
additionalresearchprograms = {AI for Everyone},
title = {Detecting Text Manipulation in Images using Vision Language Models},
booktitle = {36th British Machine Vision Conference 2025},
year = {2025},
pdf = {https://publications.idiap.ch/attachments/papers/2025/Vidit_SRBSBMVCWORKSHOP_2025.pdf}
}
@ARTICLE{Wallace_IEEETRANSACTIONSONINFORMATIONFORENSICSANDSECURITY_2012,
author = {Wallace, Roy and McLaren, Mitchell and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Cross-pollination of normalisation techniques from speaker to face authentication using Gaussian mixture models},
journal = {IEEE Transactions on Information Forensics and Security},
volume = {7},
number = {2},
year = {2012},
pages = {553 -- 562},
crossref = {Wallace_Idiap-RR-03-2012},
pdf = {https://publications.idiap.ch/attachments/papers/2012/Wallace_IEEETRANSACTIONSONINFORMATIONFORENSICSANDSECURITY_2012.pdf}
}
@INPROCEEDINGS{Wallace_IJCB_2011,
author = {Wallace, Roy and McLaren, Mitchell and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Inter-session Variability Modelling and Joint Factor Analysis for Face Authentication},
booktitle = {International Joint Conference on Biometrics},
year = {2011},
crossref = {Wallace_Idiap-RR-28-2011}
}
@INCOLLECTION{Z.Li_SPRINGER_2014,
author = {Z.Li, Stan and Galbally, Javier and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
editor = {Marcel, S{\'{e}}bastien and Nixon, Mark and Z.Li, Stan},
projects = {Idiap},
title = {Evaluation Databases},
booktitle = {Handbook of Biometric Anti-Spoofing},
chapter = {A},
year = {2014},
pages = {247-278},
publisher = {Springer-Verlag},
isbn = {978-1-4471-6523-1},
doi = {10.1007/978-1-4471-6524-8}
}
crossreferenced publications:
@TECHREPORT{Anjos_Idiap-RR-25-2012,
author = {Anjos, Andr{\'{e}} and El Shafey, Laurent and Wallace, Roy and G{\"{u}}nther, Manuel and McCool, Chris and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, computer vision, machine learning, Open Source, pattern recognition, signal processing},
projects = {Idiap},
month = {7},
title = {Bob: a free signal processing and machine learning toolbox for researchers},
type = {Idiap-RR},
number = {Idiap-RR-25-2012},
year = {2012},
institution = {Idiap},
note = {Submitted to the ACM MM 2012 Open Source Software Competition},
abstract = {Bob is a free signal processing and machine learning toolbox originally developed by the Biometrics group at Idiap Research Institute, Switzerland. The toolbox is designed to meet the needs of researchers by reducing development time and efficiently processing data. Firstly, Bob provides a researcher-friendly Python environment for rapid development. Secondly, efficient processing of large amounts of multimedia data is provided by fast C++ implementations of identified bottlenecks. The Python environment is integrated seamlessly with the C++ library, which ensures the library is easy to use and extensible. Thirdly, Bob supports reproducible research through its integrated experimental protocols for several databases. Finally, a strong emphasis is placed on code clarity, documentation, and thorough unit testing. Bob is thus an attractive resource for researchers due to this unique combination of ease of use, efficiency, extensibility and transparency. Bob is an open-source library and an ongoing community effort.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Anjos_Idiap-RR-25-2012.pdf}
}
@TECHREPORT{bengio:2001:idiap-01-38,
author = {Bengio, Samy and Marcel, Christine and Marcel, S{\'{e}}bastien and Mari{\'{e}}thoz, Johnny},
projects = {Idiap},
title = {Confidence Measures for Multimodal Identity Verification},
type = {Idiap-RR},
number = {Idiap-RR-38-2001},
year = {2001},
institution = {IDIAP},
pdf = {https://publications.idiap.ch/attachments/reports/2001/rr01-38.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2001/rr01-38.ps.gz},
ipdmembership={speech, learning, vision},
}
@TECHREPORT{Cardinaux02RR,
author = {Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Verification using {MLP} and {SVM}},
type = {Idiap-RR},
number = {Idiap-RR-21-2002},
year = {2002},
institution = {IDIAP},
abstract = {The performance of machine learning algorithms has steadily improved over the past few years, such as MLP or more recently SVM. In this paper, we compare two successful discriminant machine learning algorithms apply to the problem of face verification: MLP and SVM. These two algorithms are tested on a benchmark database, namely XM2VTS. Results show that a MLP is better than a SVM on this particular task.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr-02-21.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr-02-21.ps.gz},
ipdmembership={vision},
language={English},
}
@TECHREPORT{Chingovska_Idiap-RR-19-2013,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {Anti-spoofing, Counter-Measures, recognition, security, verification},
projects = {Idiap},
month = {5},
title = {Anti-spoofing in action: joint operation with a verification system},
type = {Idiap-RR},
number = {Idiap-RR-19-2013},
year = {2013},
institution = {Idiap},
abstract = {Besides the recognition task, today's biometric systems need to cope with additional problem: spoofing attacks. Up to date, academic research considers spoofing as a binary classification problem: systems are trained to discriminate between real accesses and attacks. However, spoofing counter-measures are not designated to operate stand-alone, but as a part of a recognition system they will protect. In this paper, we study techniques for decision-level and score-level fusion to integrate a recognition and anti-spoofing systems, using an open-source framework that handles the ternary classification problem (clients, impostors and attacks) transparently. By doing so, we are able to report the impact of different counter-measures, fusion techniques and thresholding on the overall performance of the final recognition system. For a specific use-case covering face verification, experiments show to what extent simple fusion improves the trustworthiness of the system when exposed to spoofing attacks.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Chingovska_Idiap-RR-19-2013.pdf}
}
@TECHREPORT{Chingovska_Idiap-RR-18-2013,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {2D attacks, Competition, Counter-Measures, Replay-Attack, Spoofing},
projects = {Idiap, TABULA RASA},
month = {5},
title = {The 2nd Competition on Counter Measures to 2D Face Spoofing Attacks},
type = {Idiap-RR},
number = {Idiap-RR-18-2013},
year = {2013},
institution = {Idiap},
abstract = {As a crucial security problem, anti-spoofing in biometrics, and particularly for the face modality, has achieved great progress in the recent years. Still, new threats arrive in form of better, more realistic and more sophisticated spoofing attacks. The objective of the 2nd Competition on Counter Measures to 2D Face Spoofing Attacks is to challenge researchers to create counter measures effectively detecting a variety of attacks. The submitted propositions are evaluated on the Replay-Attack database and the achieved results are presented in this paper.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Chingovska_Idiap-RR-18-2013.pdf}
}
@TECHREPORT{Chingovska_Idiap-RR-19-2012,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
keywords = {Biometrics, Counter-Measures, Local Binary Patterns, Spoofing Attacks},
projects = {TABULA RASA},
month = {7},
title = {On the Effectiveness of Local Binary Patterns in Face Anti-spoofing},
type = {Idiap-RR},
journal = {IEEE BIOSIG 2012},
number = {Idiap-RR-19-2012},
year = {2012},
institution = {Idiap},
abstract = {Spoofing attacks are one of the security traits that biometric recognition systems are proven to be vulnerable to. When spoofed, a biometric recognition system is bypassed by presenting a copy of the biometric evidence of a valid user. Among all biometric modalities, spoofing a face recognition system is particularly easy to perform: all that is needed is a simple photograph of the user.
In this paper, we address the problem of detecting face spoofing attacks. In particular, we inspect the potential of texture features based on Local Binary Patterns (LBP) and their variations on three types of attacks: printed photographs, and photos and videos displayed on electronic screens of different sizes. For this purpose, we introduce REPLAY-ATTACK, a novel publicly available face spoofing database which contains all the mentioned types of attacks. We conclude that LBP, with ~15\% Half Total Error Rate, show moderate discriminability when confronted with a wide set of attack types.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Chingovska_Idiap-RR-19-2012.pdf}
}
@TECHREPORT{Chingovska_Idiap-RR-12-2014,
author = {Chingovska, Ivana and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA, BEAT},
month = {8},
title = {Biometrics Evaluation under Spoofing Attacks},
type = {Idiap-RR},
number = {Idiap-RR-12-2014},
year = {2014},
institution = {Idiap},
note = {IEEE Transactions of Information Forensics and Security: in minor revision},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Chingovska_Idiap-RR-12-2014.pdf}
}
@TECHREPORT{Chingovska_Idiap-RR-18-2020,
author = {Chingovska, Ivana and Erdogmus, Nesli and Anjos, Andr{\'{e}} and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {9},
title = {Face Recognition Systems Under Spoofing Attacks},
type = {Idiap-RR},
number = {Idiap-RR-18-2020},
year = {2020},
institution = {Idiap},
note = {Submitted for as a book-chapter for: Face Recognition Across the Electromagnetic Spectrum (Springer)},
abstract = {In this chapter we give an overview of spoofing attacks and spoofing
counter-measures for face recognition systems, in particular in a verification sce-
nario. We focus on 2D and 3D attacks to Visible Spectrum systems (VIS), as well
as Near Infrared (NIR) and multispectral systems. We cover the existing types of
spoofing attacks and report on their success to bypass several state-of-the-art face
verification systems. The results on two different face spoofing databases with VIS
attacks and one newly developed face spoofing database with VIS and NIR attacks,
show that spoofing attacks present a significant security threat for face verification
systems in any part of the spectrum. The risk is partially reduced when using mul-
tispectral systems. We also give a systematic overview of the existing anti-spoofing
techniques, with an analysis of their advantages and limitations and prospectives for
future work.},
pdf = {https://publications.idiap.ch/attachments/reports/2015/Chingovska_Idiap-RR-18-2020.pdf}
}
@TECHREPORT{Colbois_Idiap-RR-07-2022,
author = {Colbois, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
month = {8},
title = {On the detection of morphing attacks generated by GANs},
type = {Idiap-RR},
number = {Idiap-RR-07-2022},
year = {2022},
institution = {Idiap},
address = {Rue Marconi 19},
note = {Accepted for publication at BIOSIG 2022 conference},
abstract = {Recent works have demonstrated the feasibility of GAN-based morphing attacks that reach similar success rates as more traditional landmark-based methods. This new type of "deep" morphs might require the development of new adequate detectors to protect face recognition systems.
We explore simple deep morph detection baselines based on spectral features and LBP histograms features, as well as on CNN models, both in the intra-dataset and cross-dataset case. We observe that simple LBP-based systems are already quite accurate in the intra-dataset setting, but struggle with generalization, a phenomenon that is partially mitigated by fusing together several of those systems at score-level. We conclude that a pretrained ResNet effective for GAN image detection is the most effective overall, reaching close to perfect accuracy. We note however that LBP-based systems maintain a level of interest : additionally to their lower computational requirements and increased interpretability with respect to CNNs, LBP+ResNet fusions sometimes also showcase increased performance versus ResNet-only, hinting that LBP-based systems can focus on meaningful signal that is not necessarily picked up by the CNN detector.},
pdf = {https://publications.idiap.ch/attachments/reports/2022/Colbois_Idiap-RR-07-2022.pdf}
}
@TECHREPORT{Colbois_Idiap-RR-07-2023,
author = {Colbois, Laurent and Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center, TRESPASS-ETN},
month = {7},
title = {Approximating Optimal Morphing Attacks using Template Inversion},
type = {Idiap-RR},
number = {Idiap-RR-07-2023},
year = {2023},
institution = {Idiap},
note = {Submited to the International Joint Conference on Biometrics (IJCB 2023)},
abstract = {Recent works have demonstrated the feasibility of inverting face recognition systems, enabling to recover convincing face images using only their embeddings. We leverage such template inversion models to develop a novel type of deep morphing attack based on inverting a theoretical optimal morph embedding, which is obtained as an average of the face embeddings of source images. We experiment with two variants of this approach : the first one exploits a fully self-contained embedding-to-image inversion model, while the second leverages the synthesis network of a pretrained StyleGAN for increased morph realism. We generate morphing attacks from several source datasets and study the effectiveness of those attacks against several face recognition networks. We showcase that our method can compete with and regularly beat the previous state of the art for deep-learning based morph generation in terms of effectiveness, both in white-box and black-box attack scenarios, and is additionally much faster to run. We hope this might facilitate the development of large scale deep morph datasets for training detection models.},
pdf = {https://publications.idiap.ch/attachments/reports/2023/Colbois_Idiap-RR-07-2023.pdf}
}
@TECHREPORT{ElShafey_Idiap-RR-37-2011,
author = {El Shafey, Laurent and Wallace, Roy and Marcel, S{\'{e}}bastien},
keywords = {Face Recognition, Gabor, Gaussian Mixture Models (GMM)},
projects = {Idiap},
month = {12},
title = {Face Verification using Gabor Filtering and Adapted Gaussian Mixture Models},
type = {Idiap-RR},
number = {Idiap-RR-37-2011},
year = {2011},
institution = {Idiap},
abstract = {The search for robust features for face recognition in uncontrolled environments is an important topic of research. In particular, there is a high interest in Gabor-based features which have invariance properties to simple geometrical transformations. In this paper, we first reinterpret Gabor filtering as a frequency decomposition into bands, and analyze the influence of each band separately for face recognition. Then, a new face verification scheme is proposed, combining the strengths of Gabor filtering with Gaussian Mixture Model (GMM) modelling. Finally, this new system is evaluated on the BANCA database with respect to well known face recognition algorithms and using both manual and automatic face localization. The proposed system demonstrates up to 47\% relative improvement in verification error rate compared to a standard GMM approach, and comparable results with the state-of-the-art Local Gabor Binary Pattern Histogram Sequence (LGBPHS) technique for four of the seven BANCA face verification protocols.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/ElShafey_Idiap-RR-37-2011.pdf}
}
@TECHREPORT{ElShafey_Idiap-RR-07-2013,
author = {El Shafey, Laurent and McCool, Chris and Wallace, Roy and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {3},
title = {A Scalable Formulation of Probabilistic Linear Discriminant Analysis: Applied to Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-07-2013},
year = {2013},
institution = {Idiap},
note = {Accepted for publication},
url = {https://pypi.python.org/pypi/xbob.paper.tpami2013},
abstract = {In this paper we present a scalable and exact solution for probabilistic linear discriminant analysis (PLDA). PLDA is a probabilistic model that has been shown to provide state-of-the-art performance for both face and speaker recognition. However, it has one major drawback, at training time estimating the latent variables requires the inversion and storage of a matrix whose size grows quadratically with the number of samples for the identity (class). To date two approaches have been taken to deal with this problem, to: i) use an exact solution which calculates this large matrix and is obviously not scalable with the number of samples or ii) derive a variational approximation to the problem.
We present a scalable derivation which is theoretically equivalent to the previous non-scalable solution and so obviates the need for a variational approximation. Experimentally, we demonstrate the efficacy of our approach in two ways. First, on Labelled Faces in the Wild we illustrate the equivalence of our scalable implementation with previously published work. Second, on the large Multi-PIE database, we illustrate the gain in performance when using more training samples per identity (class), which is
made possible by the proposed scalable formulation of PLDA.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/ElShafey_Idiap-RR-07-2013.pdf}
}
@TECHREPORT{Erdogmus_Idiap-RR-42-2013,
author = {Erdogmus, Nesli and Marcel, S{\'{e}}bastien},
keywords = {3D Masks, Spoofing Attacks},
projects = {Idiap, TABULA RASA},
month = {12},
title = {Spoofing Attacks To 2D Face Recognition Systems With 3D Masks},
type = {Idiap-RR},
number = {Idiap-RR-42-2013},
year = {2013},
institution = {Idiap},
address = {Centre du Parc - rue Marconi 19 CH-1920 Martigny, Suisse},
abstract = {Vulnerability to spoofing attacks is a serious drawback for many biometric systems. Among all biometric traits, face is the one that is exposed to the most serious threat, since it is exceptionally easy to access. The limited work on fraud detection capabilities for face mainly shapes around
2D attacks forged by displaying printed photos or replaying recorded videos on mobile devices. A significant portion of this work is based on the flatness of the facial surface in front of the sensor. In this study, we complicate the spoofing problem further by introducing the 3rd dimension and ex-
amine possible 3D attack instruments. A small database is constructed with six different types of 3D facial masks and it is utilized to conduct experiments on state-of-the-art 2D face recognition systems. Spoofing performance for each type of mask is assessed and analysed thoroughly.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Erdogmus_Idiap-RR-42-2013.pdf}
}
@TECHREPORT{Erdogmus_Idiap-RR-27-2013,
author = {Erdogmus, Nesli and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA},
month = {7},
title = {Spoofing in 2D Face Recognition with 3D Masks and Anti-spoofing with Kinect},
type = {Idiap-RR},
number = {Idiap-RR-27-2013},
year = {2013},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Erdogmus_Idiap-RR-27-2013.pdf}
}
@TECHREPORT{George_Idiap-RR-08-2023,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {11},
title = {Attacking Face Recognition with T-shirts: Database, Vulnerability Assessment and Detection},
type = {Idiap-RR},
number = {Idiap-RR-08-2023},
year = {2023},
institution = {Idiap},
abstract = {Face recognition systems are widely deployed for biometric authentication. Despite this, it is well-known that, without any safeguards, face recognition systems are highly vulnerable to presentation attacks. In response to this security issue, several promising methods for detecting presentation attacks have been proposed which show high performance on existing benchmarks. However, an ongoing challenge is the generalization of presentation attack detection methods to unseen and new attack types. To this end, we propose a new T-shirt Face Presentation Attack (TFPA) database of 1,608 T-shirt attacks using 100 unique presentation attack instruments. In an extensive evaluation, we show that this type of attack can compromise the security of face recognition systems and that some state-of-the-art attack detection mechanisms trained on popular benchmarks fail to robustly generalize to the new attacks. Further, we propose three new methods for detecting T-shirt attack images, one which relies on the statistical differences between depth maps of bona fide images and T-shirt attacks, an anomaly detection approach trained on features only extracted from bona fide RGB images, and a fusion approach which achieves competitive detection performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2023/George_Idiap-RR-08-2023.pdf}
}
@TECHREPORT{George_Idiap-RR-08-2025,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, Biometrics Center},
additionalresearchprograms = {AI for Everyone},
month = {8},
title = {EdgeDoc: Hybrid CNN-Transformer Model for Accurate Forgery Detection and Localization in ID Documents},
type = {Idiap-RR},
number = {Idiap-RR-08-2025},
year = {2025},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2025/George_Idiap-RR-08-2025.pdf}
}
@TECHREPORT{George_Idiap-RR-30-2020,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
month = {11},
title = {On the Effectiveness of Vision Transformers for Zero-shot Face Anti-Spoofing},
type = {Idiap-RR},
number = {Idiap-RR-30-2020},
year = {2020},
institution = {Idiap},
abstract = {The vulnerability of face recognition systems to presentation attacks has limited their application in security-critical scenarios. Automatic methods of detecting such malicious attempts are essential for the safe use of facial recognition technology. Although various methods have been suggested for detecting such attacks, most of them over-fit the training set and fail in generalizing to unseen attacks and environments. In this work, we use transfer learning from the vision transformer model for the zero-shot anti-spoofing task. The effectiveness of the proposed approach is demonstrated through experiments in publicly available datasets. The proposed approach outperforms the state of the art methods in the zero-shot protocols in the HQ-WMCA and SiW-M datasets by a large margin. Besides, the model achieves a significant boost in cross-database performance as well.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/George_Idiap-RR-30-2020.pdf}
}
@TECHREPORT{George_Idiap-RR-03-2022,
author = {George, Anjith and Marcel, S{\'{e}}bastien},
projects = {Idiap, ODIN/BATL},
month = {3},
title = {Robust Face Presentation Attack Detection with Multi-channel Neural Networks},
type = {Idiap-RR},
number = {Idiap-RR-03-2022},
year = {2022},
institution = {Idiap},
abstract = {Vulnerability against presentation attacks remains a challenging issue
limiting the reliable use of face recognition systems. Though several methods have
been proposed in the literature for the detection of presentation attacks, the majority
of these methods fail in generalizing to unseen attacks and environments. Since the
quality of attack instruments keeps getting better, the difference between bonafide
and attack samples is diminishing making it harder to distinguish them using the
visible spectrum alone. In this context, multi-channel presentation attack detection
methods have been proposed as a solution to secure face recognition systems. Even
with multiple channels, special care needs to be taken to ensure that the model gener-
alizes well in challenging scenarios. In this chapter, we present three different strate-
gies to use multi-channel information for presentation attack detection. Specifically,
we present different architecture choices for fusion, along with ad-hoc loss func-
tions as opposed to standard classification objective. We conduct an extensive set
of experiments in the HQ-WMCA dataset, which contains a wide variety of attacks
and sensing channels together with challenging unseen attack evaluation protocols.
We make the protocol, source codes, and data publicly available to enable further
extensions of the work.},
pdf = {https://publications.idiap.ch/attachments/reports/2021/George_Idiap-RR-03-2022.pdf}
}
@TECHREPORT{Gunther_Idiap-RR-29-2012,
author = {G{\"{u}}nther, Manuel and Wallace, Roy and Marcel, S{\'{e}}bastien},
projects = {Idiap, FP 7},
month = {10},
title = {An Open Source Framework for Standardized Comparisons of Face Recognition Algorithms},
type = {Idiap-RR},
number = {Idiap-RR-29-2012},
year = {2012},
institution = {Idiap},
abstract = {In this paper we introduce the facereclib, the first software library that allows to compare a variety of face recognition algorithms on most of the known facial image databases and that permits rapid prototyping of novel ideas and testing of meta-parameters of face recognition algorithms. The facereclib is built on the open source signal processing and machine learning library Bob. It uses well-specified face recognition protocols to ensure that results are comparable and reproducible. We show that the face recognition algorithms implemented in Bob as well as third party face recognition libraries can be used to run face recognition experiments within the framework of the facereclib. As a proof of concept, we execute four different state-of-the-art face recognition algorithms: local Gabor binary pattern histogram sequences (LGBPHS), Gabor graph comparisons with a Gabor phase based similarity measure, inter-session variability modeling (ISV) of DCT block features, and the linear discriminant analysis on two different color channels (LDA-IR) on two different databases: The Good, The Bad, & The Ugly, and the BANCA database, in all cases using their fixed protocols. The results show that there is not one face recognition algorithm that outperforms all others, but rather that the results are strongly dependent on the employed database.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Gunther_Idiap-RR-29-2012.pdf}
}
@TECHREPORT{Gunther_Idiap-RR-36-2013,
author = {G{\"{u}}nther, Manuel and Costa-Pazo, Artur and Ding, Changxing and Boutellaa, Elhocine and Chiachia, Giovani and Zhang, Honglei and de Assis Angeloni, Marcus and Struc, Vitomir and Khoury, Elie and Vazquez-Fernandez, Esteban and Tao, Dacheng and Bengherabi, Messaoud and Cox, David and Kiranyaz, Serkan and de Freitas Pereira, Tiago and Zganec-Gros, Jerneja and Argones-R{\'{u}}a, Enrique and Pinto, Nicolas and Gabbouj, Moncef and Sim{\~{o}}es, Fl{\'{a}}vio and Dobrisek, Simon and Gonz{\'{a}}lez-Jim{\'{e}}nez, Daniel and Rocha, Anderson and Uliani Neto, M{\'{a}}rio and Pavesic, Nikola and Falc{\~{a}}o, Alexandre and Violato, Ricardo and Marcel, S{\'{e}}bastien},
projects = {Idiap, BBfor2, BEAT},
month = {11},
title = {The 2013 Face Recognition Evaluation in Mobile Environment},
type = {Idiap-RR},
number = {Idiap-RR-36-2013},
year = {2013},
institution = {Idiap},
abstract = {Automatic face recognition in unconstrained environments is a challenging task. To test current trends in face recognition algorithms, we organized an evaluation on face recognition in mobile environment. This paper presents the results of eight different participants using two verification metrics. Most submitted algorithms rely on on or more of three types of features: local binary patterns, Gabor wavelet responses including Gabor phases, and color information. The best results are obtained from UNILJ-ALP, which fused several image representations and feature types, and UC-HU, which learns ptimal features with a convolutional neural network. Additionally, we assess the usability of the algorithms in mobile devices with limited resources.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Gunther_Idiap-RR-36-2013.pdf}
}
@TECHREPORT{heusch:rr05-76,
author = {Heusch, Guillaume and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Local Binary Patterns as an Image Preprocessing for Face Authentication},
type = {Idiap-RR},
number = {Idiap-RR-76-2005},
year = {2005},
institution = {IDIAP},
abstract = {One of the major problem in face authentication systems is to deal with variations in illumination. In a \mbox{realistic} scenario, it is very likely that the lighting conditions of the probe image does not correspond to those of the gallery image, hence there is a need to handle such variations. In this work, we present a new preprocessing algorithm based on Local Binary Patterns (LBP): a texture representation is derived from the input face image before being forwarded to the classifier. The efficiency of the proposed approach is empirically demonstrated using both an appearance-based (LDA) and a feature-based (HMM) face authentication systems on two databases: BANCA and XM2VTS (with its darkened set). Conducted experiments show a significant improvement in terms of verification error rates and compare to results obtained with state-of-the-art preprocessing techniques.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/heusch-idiap-rr-05-76.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/heusch-idiap-rr-05-76.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{heusch:rr07-04,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Authentication with Salient Local Features and Static Bayesian Network},
type = {Idiap-RR},
number = {Idiap-RR-04-2007},
year = {2007},
institution = {IDIAP},
abstract = {In this paper, the problem of face authentication using salient facial features together with statistical generative models is adressed. Actually, classical generative models, and Gaussian Mixture Models in particular make strong assumptions on the way observations derived from face images are generated. Indeed, systems proposed so far consider that local observations are independent, which is obviously not the case in a face. Hence, we propose a new generative model based on Bayesian Networks using only salient facial features. We compare it to Gaussian Mixture Models using the same set of observations. Conducted experiments on the BANCA database show that our model is suitable for the face authentication task, since it outperforms not only Gaussian Mixture Models, but also classical appearance-based methods, such as Eigenfaces and Fisherfaces.},
pdf = {https://publications.idiap.ch/attachments/reports/2007/heusch-idiap-rr-07-04.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2007/heusch-idiap-rr-07-04.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{Heusch_Idiap-RR-27-2009,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {9},
title = {Bayesian Networks to Combine Intensity and Color Information in Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-27-2009},
year = {2009},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2008/Heusch_Idiap-RR-27-2009.pdf}
}
@TECHREPORT{Heusch_Idiap-RR-09-2019,
author = {Heusch, Guillaume and de Freitas Pereira, Tiago and Marcel, S{\'{e}}bastien},
projects = {FARGO},
month = {9},
title = {A Comprehensive Experimental and Reproducible Study on Selfie Biometrics in Multistream and Heterogeneous Settings},
type = {Idiap-RR},
number = {Idiap-RR-09-2019},
year = {2019},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2019/Heusch_Idiap-RR-09-2019.pdf}
}
@TECHREPORT{heusch:rr07-39,
author = {Heusch, Guillaume and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {A Novel Statistical Generative Model Dedicated To Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-39-2007},
year = {2007},
institution = {IDIAP},
abstract = {In this paper, a novel statistical generative model to describe a face is presented, and is applied on the face authentication task. Classical generative models used so far in face recognition, such as Gaussian Mixture Models (GMM) and Hidden Markov Models (HMM) for instance, are making strong assumptions on the observations derived from a face image. Indeed, such models usually assume that local observations are independent, which is obviously not the case in a face. The presented model hence proposes to encode relationships between salient facial features by using a static Bayesian Network. Since robustness against imprecisely located faces is of great concern in a real-world scenario, authentication results are presented using automatically localised faces. Experiments conducted on the XM2VTS and the BANCA databases showed that the proposed approach is suitable for this task, since it reaches state-of-the-art results. We compare our model to baseline appearance-based systems (Eigenfaces and Fisherfaces) but also to classical generative models, namely GMM, HMM and pseudo-2DHMM.},
pdf = {https://publications.idiap.ch/attachments/reports/2007/heusch-idiap-rr-07-39.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2007/heusch-idiap-rr-07-39.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{I.Mantasari_Idiap-RR-01-2014,
author = {I. Mantasari, Miranti and G{\"{u}}nther, Manuel and Wallace, Roy and Saedi, Rahim and Marcel, S{\'{e}}bastien and Van Leeuwen, David},
keywords = {calibration, forensic face recognition, likelihood ratio, linear score transformation.},
projects = {BBfor2},
month = {1},
title = {Score Calibration in Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-01-2014},
year = {2014},
institution = {Idiap},
abstract = {This paper presents an evaluation of verification and calibration performance of a face recognition system based on inter-session variability modeling. As an extension to the calibration through
linear transformation of scores, categorical calibration is introduced as a way to include additional
information of images to calibration. The cost of likelihood ratio, which is a well-known measure in
the speaker recognition field, is used as a calibration performance metric. Evaluated on the challenging MOBIO and SCface databases, the results indicate that through linear calibration the scores
produced by the face recognition system can be less misleading in its likelihood ratio interpretation. In addition, it is shown through the categorical calibration experiments that calibration can be
used not only to assure likelihood ratio interpretation of scores, but also improving the verification
performance of face recognition system.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/I.Mantasari_Idiap-RR-01-2014.pdf}
}
@TECHREPORT{just2003,
author = {Just, Agn{\`{e}}s and Bernier, O. and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Recognition of Isolated Complex Mono- and Bi-Manual 3D Hand Gestures},
type = {Idiap-RR},
number = {Idiap-RR-63-2003},
year = {2003},
institution = {IDIAP},
note = {Published in "Proceedings of the sixth International Conference on Automatic Face and Gesture Recognition", 2004},
abstract = {In this paper, we address the problem of the recognition of isolated complex mono- and bi-manual hand gestures. In the proposed system, hand gestures are represented by the 3D trajectories of blobs. Blobs are obtained by tracking colored body parts in real-time using the EM algorithm. In most of the studies on hand gestures, only small vocabularies have been used. In this paper, we study the results obtained on a more complex database of mono- and bi-manual gestures. These results are obtained by using a state-of-the-art sequence processing algorithm, namely Hidden Markov Models (HMMs,',','),
implemented within the framework of an open source machine learning library.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-63.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-63.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{just:rr06-02,
author = {Just, Agn{\`{e}}s and Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Hand Posture Classification and Recognition using the Modified Census Transform},
type = {Idiap-RR},
number = {Idiap-RR-02-2006},
year = {2006},
institution = {IDIAP},
note = {Published in Proc. of the {IEEE} Int. Conf. on Automatic Face and Gesture Recognition 2006},
abstract = {Developing new techniques for human-computer interaction is very challenging. Vision-based techniques have the advantage of being unobtrusive and hands are a natural device that can be used for more intuitive interfaces. But in order to use hands for interaction, it is necessary to be able to recognize them in images. In this paper, we propose to apply to the hand posture classification and recognition tasks an approach that has been successfully used for face detection~\cite{Froba04}. The features are based on the Modified Census Transform and are illumination invariant. For the classification and recognition processes, a simple linear classifier is trained, using a set of feature lookup-tables. The database used for the experiments is a benchmark database in the field of posture recognition. Two protocols have been defined. We provide results following these two protocols for both the classification and recognition tasks. Results are very encouraging.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/just-idiap-rr-06-02.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/just-idiap-rr-06-02.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{Khoury_Idiap-RR-35-2013,
author = {Khoury, Elie and G{\"{u}}nther, Manuel and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, BBfor2, SNSF-LOBI},
month = {11},
title = {On the Improvements of Uni-modal and Bi-modal Fusions of Speaker and Face Recognition for Mobile Biometrics},
type = {Idiap-RR},
number = {Idiap-RR-35-2013},
year = {2013},
institution = {Idiap},
abstract = {The MOBIO database provides a challenging test-bed for speaker and face recognition systems because it includes voice and face samples as they would appear in forensic scenarios. In this paper, we investigate uni-modal and bi-modal multi-algorithm fusion using logistic regression. The source speaker and face recognition systems were taken from the 2013 speaker and face recognition evaluations that were held in the context of the last International Conference on Biometrics (ICB-2013).
Using the unbiased MOBIO protocols, the employed evaluation measures are the equal error rate (EER), the half-total error rate (HTER) and the detection error trade-off (DET). The results show that by uni-modal algorithm fusion, the HTER's of the speaker recognition system are reduced by around 35\%, and of the face recognition system by between 15\% and 20\%. Bi-modal fusion drastically boosts recognition by a relative gain of 65\% - 70\% of performance compared to the best uni-modal system.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Khoury_Idiap-RR-35-2013.pdf}
}
@TECHREPORT{Khoury_Idiap-RR-32-2013,
author = {Khoury, Elie and Vesnicer, Bostjan and Franco-Pedroso, Javier and Violato, Ricardo and Boulkenafet, Zenelabidine and Mazaira Fernandez, Luis-Miguel and Diez, Mireia and Kosmala, Justina and Khemiri, Houssemeddine and Cipr, Tomas and Saedi, Rahim and G{\"{u}}nther, Manuel and Zganec-Gros, Jerneja and Zazo Candil, Ruben and Sim{\~{o}}es, Fl{\'{a}}vio and Bengherabi, Messaoud and Alvarez Marquina, Augustin and Penagarikano, Mikel and Abad, Alberto and Boulayemen, Mehdi and Schwarz, Petr and Van Leeuwen, David and Gonzalez-Dom{\i}nguez, Javier and Uliani Neto, M{\'{a}}rio and Boutellaa, Elhocine and Gomez Vilda, Pedro and Varona, Amparo and Petrovska-Delacretaz, Dijana and Matejka, Pavel and Gonzalez-Rodr{\i}guez, Joaquin and de Freitas Pereira, Tiago and Harizi, Farid and Rodriguez-Fuentes, Luis Javier and El Shafey, Laurent and de Assis Angeloni, Marcus and Bordel, German and Chollet, G{\'{e}}rard and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {11},
title = {The 2013 Speaker Recognition Evaluation in Mobile Environment},
type = {Idiap-RR},
number = {Idiap-RR-32-2013},
year = {2013},
institution = {Idiap},
crossref = {Khoury_ICB2013_2013},
abstract = {This paper evaluates the performance of the twelve primary systems submitted to the evaluation on speaker verification in the context of a mobile environment using the MOBIO database. The mobile environment provides a challenging and realistic test-bed for current state-of-the-art speaker verification techniques. Results in terms of equal error rate (EER), half total error rate (HTER) and detection error tradeoff (DET) confirm that the best performing systems are based on total variability modeling, and fuse several subsystems. Nevertheless, the old good UBM-GMM based systems are still competitive. The results also show that the use of additional data for training as well as gender-dependent features can be helpful.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Khoury_Idiap-RR-32-2013.pdf}
}
@INPROCEEDINGS{Khoury_ICB2013_2013,
author = {Khoury, Elie and Vesnicer, Bostjan and Franco-Pedroso, Javier and Violato, Ricardo and Boulkenafet, Zenelabidine and Mazaira Fernandez, Luis-Miguel and Diez, Mireia and Kosmala, Justina and Khemiri, Houssemeddine and Cipr, Tomas and Saedi, Rahim and G{\"{u}}nther, Manuel and Zganec-Gros, Jerneja and Zazo Candil, Ruben and Sim{\~{o}}es, Fl{\'{a}}vio and Bengherabi, Messaoud and Alvarez Marquina, Augustin and Penagarikano, Mikel and Abad, Alberto and Boulayemen, Mehdi and Schwarz, Petr and Van Leeuwen, David and Gonzalez-Dom{\i}nguez, Javier and Uliani Neto, M{\'{a}}rio and Boutellaa, Elhocine and Gomez Vilda, Pedro and Varona, Amparo and Petrovska-Delacretaz, Dijana and Matejka, Pavel and Gonzalez-Rodr{\i}guez, Joaquin and de Freitas Pereira, Tiago and Harizi, Farid and Rodriguez-Fuentes, Luis Javier and El Shafey, Laurent and Angeloni, Marcus and Bordel, German and Chollet, G{\'{e}}rard and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BEAT},
month = jun,
title = {The 2013 Speaker Recognition Evaluation in Mobile Environment},
booktitle = {The 6th IAPR International Conference on Biometrics},
year = {2013},
crossref = {Khoury_Idiap-RR-32-2013},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Khoury_ICB2013_2013.pdf}
}
@TECHREPORT{Khoury_Idiap-RR-30-2013,
author = {Khoury, Elie and El Shafey, Laurent and McCool, Chris and G{\"{u}}nther, Manuel and Marcel, S{\'{e}}bastien},
projects = {Idiap, SNSF-LOBI, BBfor2, BEAT},
month = {10},
title = {Bi-Modal Biometric Authentication on Mobile Phones in Challenging Conditions},
type = {Idiap-RR},
number = {Idiap-RR-30-2013},
year = {2013},
institution = {Idiap},
abstract = {This paper examines the issue of face, speaker and bi-modal authentication in mobile environments when there is significant condition mismatch. We introduce this mismatch by enrolling client models on high quality biometric samples obtained on a laptop computer and authenticating them on lower quality biometric samples acquired with a mobile phone. To perform these experiments we develop three novel authentication protocols for the large publicly available MOBIO database. We evaluate state-of-the-art face, speaker and bi-modal authentication techniques and show that inter-session variability modelling using Gaussian mixture models provides a consistently robust system for face, speaker and bi-modal
authentication. It is also shown that multi-algorithm fusion provides a consistent performance improvement for face, speaker and bi-modal authentication. Using this bi-modal multi-algorithm system we derive a state-of-the-art authentication system that obtains a half total error rate of 6.3\% and 1.9\% for Female and Male trials, respectively.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Khoury_Idiap-RR-30-2013.pdf}
}
@TECHREPORT{Korshunov_Idiap-RR-25-2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, BEAT},
month = {10},
title = {Joint Operation of Voice Biometrics and Presentation Attack Detection},
type = {Idiap-RR},
number = {Idiap-RR-25-2016},
year = {2016},
institution = {Idiap},
address = {Niagara Falls, NY, USA},
note = {published in BTAS 2016},
url = {https://pypi.python.org/pypi/bob.paper.btas_j2016},
abstract = {Research in the area of automatic speaker verification (ASV) has advanced enough for the industry to start using ASV systems in practical applications. However, as it was also shown for fingerprints, face, and other verification systems, ASV systems are highly vulnerable to spoofing or presentation attacks, limiting their wide practical deployment. Therefore, to protect against such attacks, effective anti-spoofing detection techniques, more formally known as presentation attack detection (PAD) systems, need to be developed. These techniques should be then seamlessly integrated into existing ASV systems for practical all-in-one solutions. In this paper, we focus on the integration of PAD and ASV systems. We consider the state of the art i-vector and ISV-based ASV systems and demonstrate the effect of score-based integration with a PAD system on the verification and attack detection accuracies. In our experiments, we rely on AVspoof database that contains realistic presentation attacks, which are considered by the industry to be the threat to practical ASV systems. Experimental results show a significantly increased resistance of the joint ASV-PAD system to the attacks at the expense of slightly degraded performance for scenarios without spoofing attacks. Also, an important contribution of the paper is an open source and an online-based implementations of the separate and joint ASV-PAD systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2016/Korshunov_Idiap-RR-25-2016.pdf}
}
@TECHREPORT{Korshunov_Idiap-RR-24-2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien and Muckenhirn, Hannah and Gon{\c c}alves, A. R. and Mello, A. G. Souza and Violato, R. P. Velloso and Sim{\~{o}}es, Fl{\'{a}}vio and Uliani Neto, M{\'{a}}rio and de Assis Angeloni, Marcus and Stuchi, J. A. and Dinkel, H. and Chen, N. and Qian, Yanmin and Paul, D. and Saha, G. and Sahidullah, Md},
projects = {Idiap, BEAT},
month = {10},
title = {Overview of BTAS 2016 Speaker Anti-spoofing Competition},
type = {Idiap-RR},
number = {Idiap-RR-24-2016},
year = {2016},
institution = {Idiap},
address = {Niagara Falls, NY, USA},
note = {Open source software for the paper: https://pypi.python.org/pypi/bob.paper.btas_c2016},
url = {https://pypi.python.org/pypi/bob.paper.btas_c2016},
abstract = {This paper provides an overview of the Speaker Anti-spoofing Competition organized by Biometric group at Idiap Research Institute for the IEEE International Conference on Biometrics: Theory, Applications, and Systems (BTAS 2016). The competition used AVspoof database, which contains a comprehensive set of presentation attacks, including, (i) direct replay attacks when a genuine data is played back using a laptop and two phones (Samsung Galaxy S4 and iPhone 3G), (ii) synthesized speech replayed with a laptop, and (iii) speech created with a voice conversion algorithm, also replayed with a laptop.
The paper states competition goals, describes the database and the evaluation protocol, discusses solutions for spoofing or presentation attack detection submitted by the participants, and presents the results of the evaluation.},
pdf = {https://publications.idiap.ch/attachments/reports/2016/Korshunov_Idiap-RR-24-2016.pdf}
}
@TECHREPORT{Korshunov_Idiap-RR-18-2018,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {Idiap, SAVI},
month = {12},
title = {DeepFakes: a New Threat to Face Recognition? Assessment and Detection},
type = {Idiap-RR},
number = {Idiap-RR-18-2018},
year = {2018},
institution = {Idiap},
abstract = {It is becoming increasingly easy to automatically replace a face of one person in a video with the face of another person by using a pre-trained generative adversarial network (GAN). Recent public scandals, e.g., the faces of celebrities being swapped onto pornographic videos, call for automated ways to detect these Deepfake videos. To help developing such methods, in this paper, we present the first publicly available set of Deepfake videos generated from videos of VidTIMIT database. We used open source software based on GANs to create the Deepfakes, and we emphasize that training and blending parameters can significantly impact the quality of the resulted videos. To demonstrate this impact, we generated videos with low and high visual quality (320 videos each) using differently tuned parameter sets. We showed that the state of the art face recognition systems based on VGG and Facenet neural networks are vulnerable to Deepfake videos, with 85.62\% and 95.00\% false acceptance rates respectively, which means methods for detecting Deepfake videos are necessary. By considering several baseline approaches, we found that audio-visual approach based on lip-sync inconsistency detection was not able to distinguish Deepfake videos. The best performing method, which is based on visual quality metrics and is often used in presentation attack detection domain, resulted in 8.97\% equal error rate on high quality Deepfakes. Our experiments demonstrate that GAN-generated Deepfake videos are challenging for both face recognition systems and existing detection methods, and the further development of face swapping technology will make it even more so.},
pdf = {https://publications.idiap.ch/attachments/reports/2018/Korshunov_Idiap-RR-18-2018.pdf}
}
@TECHREPORT{Korshunov_Idiap-RR-23-2016,
author = {Korshunov, Pavel and Marcel, S{\'{e}}bastien},
projects = {BEAT, SWAN},
month = {10},
title = {Cross-database evaluation of audio-based spoofing detection systems},
type = {Idiap-RR},
number = {Idiap-RR-23-2016},
year = {2016},
institution = {Idiap},
note = {Open source software package for the paper: https://pypi.python.org/pypi/bob.paper.interspeech_2016},
url = {https://pypi.python.org/pypi/bob.paper.interspeech_2016},
abstract = {Since automatic speaker verification (ASV) systems are highly vulnerable to spoofing attacks, it is important to develop mechanisms that can detect such attacks. To be practical, however, a spoofing attack detection approach should have (i) high accuracy, (ii) be well-generalized for practical attacks, and (iii) be simple and efficient. Several audio-based spoofing detection methods have been proposed recently but their evaluation is limited to less realistic databases containing homogeneous data. In this paper, we consider eight existing presentation attack detection (PAD) methods and evaluate their performance using two major publicly available speaker databases with spoofing attacks: AVspoof and ASVspoof. We first show that realistic presentation attacks (speech is replayed to PAD system) are significantly more challenging for the considered PAD methods compared to the so called `logical access' attacks (speech is presented to PAD system directly). Then, via a cross-database evaluation, we demonstrate that the existing methods generalize poorly when different databases or different types of attacks are used for training and testing. The results question the efficiency and practicality of the existing PAD systems, as well as, call for creation of databases with larger variety of realistic speech presentation attacks.},
pdf = {https://publications.idiap.ch/attachments/reports/2016/Korshunov_Idiap-RR-23-2016.pdf}
}
@TECHREPORT{Kotwal_Idiap-RR-10-2020,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
projects = {Tesla},
month = {5},
title = {CNN Patch Pooling for Detecting 3D Mask Presentation Attacks in NIR},
type = {Idiap-RR},
number = {Idiap-RR-10-2020},
year = {2020},
institution = {Idiap},
abstract = {Presentation attacks using 3D masks pose a serious threat to face recognition systems. Automatic detection of these attacks is challenging due to hyper-realistic nature of masks. In this work, we consider presentations acquired in near infrared (NIR) imaging channel for detection of mask-based attacks. We propose a patch pooling mechanism to learn complex textural features from lower layers of a convolutional neural network (CNN). The proposed patch pooling layer can be used in conjunction with a pretrained face recognition CNN without fine-tuning or adaptation. The pretrained CNN, in fact, can also be trained from visual spectrum data. We demonstrate efficacy of the proposed method on mask attacks in NIR channel from WMCA and MLFP datasets. It achieves near perfect results on WMCA data, and outperforms existing benchmark on MLFP dataset by a large margin.},
pdf = {https://publications.idiap.ch/attachments/reports/2020/Kotwal_Idiap-RR-10-2020.pdf}
}
@TECHREPORT{Kotwal_Idiap-RR-04-2024,
author = {Kotwal, Ketan and {\"{O}}zbulak, G{\"{o}}khan and Marcel, S{\'{e}}bastien},
month = {7},
title = {Assessing the Reliability of Biometric Authentication on Virtual Reality Devices},
type = {Idiap-RR},
number = {Idiap-RR-04-2024},
year = {2024},
institution = {Idiap},
abstract = {Recent developments in Virtual Reality (VR) headsets have unlocked a plethora of innovative use-cases, many of which were previously unimaginable. However, as these use-cases, such as personalized immersive experiences, necessitate user authentication, ensuring robustness and resistance to spoofing attacks becomes imperative. The absence of appropriate dataset has constrained our understanding and assessment of VR devices’ vulnerability to presentation attacks. To address this research gap, we introduce a new periocular video dataset acquired from a VR headset (Meta Quest Pro), comprising 900 genuine and 996 presentation attack videos, each spanning 10 seconds. The bona-fide videos consist of variations in terms of gaze and glasses; while the attacks are constructed with 6 different types of instruments. Additionally, we evaluate the performance of two prominent CNN architectures trained using various configurations for detecting presentation attacks in the newly created dataset, VRPAD. Our benchmarking on VRPAD reveals the presence of spoofing threats in VR headsets. While baseline models exhibit considerable efficacy in attack detection, substantial scope exists for improvement in detecting attacks on periocular videos. Our dataset will be a useful resource for researchers aiming to enhance the security and reliability of VR-based authentication systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2024/Kotwal_Idiap-RR-04-2024.pdf}
}
@TECHREPORT{Kotwal_Idiap-RR-01-2025,
author = {Kotwal, Ketan and Marcel, S{\'{e}}bastien},
keywords = {Demographic bias, Face Recognition},
projects = {Idiap, SAFER, Biometrics Center},
month = {2},
title = {Review of Demographic Bias in Face Recognition},
type = {Idiap-RR},
number = {Idiap-RR-01-2025},
year = {2025},
institution = {Idiap},
abstract = {Demographic bias in face recognition (FR) has emerged as a critical area of research, given its impact on fairness, equity, and reliability across diverse applications. As FR technologies are increasingly deployed globally, disparities in performance across demographic groups- such as race, ethnicity, and gender- have garnered significant attention. These biases not only compromise the credibility of FR systems but also raise ethical concerns, especially when these technologies are employed in sensitive domains. This review consolidates extensive research efforts providing a comprehensive overview of the multifaceted aspects of demographic bias in FR.
We systematically examine the primary causes, datasets, assessment metrics, and mitigation approaches associated with demographic disparities in FR. By categorizing key contributions in these areas, this work provides a structured approach to understanding and addressing the complexity of this issue. Finally, we highlight current advancements and identify emerging challenges that need further investigation. This article aims to provide researchers with a unified perspective on the state-of-the-art while emphasizing the critical need for equitable and trustworthy FR systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2025/Kotwal_Idiap-RR-01-2025.pdf}
}
@ARTICLE{Krivokuca_ARXIV-3_2022,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques},
journal = {arXiv},
year = {2022},
note = {Version 4 -- Corresponds to TIFS accepted manuscript, fixes some broken URLs},
url = {https://arxiv.org/abs/2110.05044v4},
crossref = {Krivokuca_IEEETIFS_2022},
abstract = {As automated face recognition applications tend towards ubiquity, there is a growing need to secure the sensitive face data used within these systems. This paper presents a survey of biometric template protection (BTP) methods proposed for securing face templates (images/features) in neural-network-based face recognition systems. The BTP methods are categorised into two types: Non-NN and NN-learned. Non-NN methods use a neural network (NN) as a feature extractor, but the BTP part is based on a non-NN algorithm, whereas NN-learned methods employ a NN to learn a protected template from the unprotected template. We present examples of Non-NN and NN-learned face BTP methods from the literature, along with a discussion of their strengths and weaknesses. We also investigate the techniques used to evaluate these methods in terms of the three most common BTP criteria: recognition accuracy, irreversibility, and renewability/unlinkability. The recognition accuracy of protected face recognition systems is generally evaluated using the same (empirical) techniques employed for evaluating standard (unprotected) biometric systems. However, most irreversibility and renewability/unlinkability evaluations are found to be based on theoretical assumptions/estimates or verbal implications, with a lack of empirical validation in a practical face recognition context. So, we recommend a greater focus on empirical evaluations to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice. Additionally, an exploration of the reproducibility of the studied BTP works, in terms of the public availability of their implementation code and evaluation datasets/procedures, suggests that it would be difficult to faithfully replicate most of the reported findings. So, we advocate for a push towards reproducibility, in the hope of advancing face BTP research.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_ARXIV-3_2022.pdf}
}
@ARTICLE{Krivokuca_IEEETIFS_2022,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques},
journal = {IEEE Transactions on Information Forensics and Security},
year = {2022},
crossref = {Krivokuca_ARXIV-3_2022},
abstract = {As automated face recognition applications tend towards ubiquity, there is a growing need to secure the sensitive face data used within these systems. This paper presents a survey of biometric template protection (BTP) methods proposed for securing face "templates'' (images/features) in neural-network-based face recognition systems. The BTP methods are categorised into two types: Non-NN and NN-learned. Non-NN methods use a neural network (NN) as a feature extractor, but the BTP part is based on a non-NN algorithm applied at either image-level or feature-level. In contrast, NN-learned methods specifically employ a NN to learn a protected template from the unprotected face image/features. We present examples of Non-NN and NN-learned face BTP methods from the literature, along with a discussion of the two categories' comparative strengths and weaknesses. We also investigate the techniques used to evaluate these BTP methods, in terms of the three most common BTP criteria: "recognition accuracy'', "irreversibility'', and "renewability/unlinkability''. As expected, the recognition accuracy of protected face recognition systems is generally evaluated using the same (empirical) techniques employed for evaluating standard (unprotected) biometric systems. On the contrary, most irreversibility and renewability/unlinkability evaluations are found to be based on theoretical assumptions/estimates or verbal implications, with a lack of empirical validation in a practical face recognition context. We recommend, therefore, a greater focus on empirical evaluation strategies, to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice. Additionally, an exploration of the reproducibility of the studied BTP works, in terms of the public availability of their implementation code and evaluation datasets/procedures, suggests that it would currently be difficult for the BTP community to faithfully replicate (and thus validate) most of the reported findings. So, we advocate for a push towards reproducibility, in the hope of furthering our understanding of the face BTP research field.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_IEEETIFS_2022.pdf}
}
@ARTICLE{Krivokuca_arxiv_PolyProtect_v2,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Towards Protecting Face Embeddings in Mobile Face Verification Scenarios},
journal = {arXiv},
year = {2021},
note = {Version 2 -- After revisions to IEEE T-BIOM manuscript},
url = {https://arxiv.org/abs/2110.00434v2},
crossref = {Krivokuca_arxiv_PolyProtect_v3},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_arxiv_PolyProtect_v2.pdf}
}
@ARTICLE{Krivokuca_arxiv_PolyProtect_v3,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Towards Protecting Face Embeddings in Mobile Face Verification Scenarios},
journal = {arXiv},
year = {2022},
note = {Version 3 -- Accepted for publication in IEEE T-BIOM},
url = {https://arxiv.org/abs/2110.00434v3},
crossref = {Krivokuca_IEEET-BIOM_2022},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_arxiv_PolyProtect_v3.pdf}
}
@ARTICLE{Krivokuca_IEEET-BIOM_2022,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
month = jan,
title = {Towards Protecting Face Embeddings in Mobile Face Verification Scenarios},
journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},
volume = {4},
number = {1},
year = {2022},
pages = {117-134},
issn = {2637-6407},
doi = {10.1109/TBIOM.2022.3140472},
crossref = {Krivokuca_arxiv_PolyProtect_v3},
abstract = {This paper proposes PolyProtect, a method for protecting the sensitive face embeddings that are used to represent
people’s faces in neural-network-based face verification systems. PolyProtect transforms a face embedding to a more secure template, using a mapping based on multivariate polynomials parameterised by user-specific coefficients and exponents. In this work, PolyProtect is evaluated on two open-source face recognition systems in a cooperative-user mobile face verification context, under the toughest threat model that assumes a fully-informed attacker with complete knowledge of the system and all its parameters. Results indicate that PolyProtect can be tuned to achieve a satisfactory trade-off between the recognition accuracy of the PolyProtected face verification system and the irreversibility of the PolyProtected templates. Furthermore, PolyProtected templates are shown to be effectively unlinkable, especially if the user-specific parameters employed in the PolyProtect mapping are selected in a non-naive manner. The evaluation is conducted using practical methodologies with tangible results, to present realistic insight into the method’s robustness as a face embedding protection scheme in practice. This work is fully reproducible using the publicly available code at:
https://gitlab.idiap.ch/bob/bob.paper.polyprotect_2021.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_IEEET-BIOM_2022.pdf}
}
@ARTICLE{Krivokuca_arxiv_survey_2022,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques},
journal = {arXiv},
year = {2022},
note = {Version 2 -- After addition of BTP vs B-PET explanation + 2 new references},
url = {https://arxiv.org/abs/2110.05044v2},
crossref = {Krivokuca_ARXIV-2_2022},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_arxiv_survey_2022.pdf}
}
@ARTICLE{Krivokuca_ARXIV-2_2022,
author = {Krivokuca, Vedrana and Marcel, S{\'{e}}bastien},
projects = {CITeR},
title = {Biometric Template Protection for Neural-Network-based Face Recognition Systems: A Survey of Methods and Evaluation Techniques},
journal = {arXiv},
year = {2022},
note = {Version 3 -- After incorporating revisions suggested by TIFS reviewers},
url = {https://arxiv.org/abs/2110.05044v3},
crossref = {Krivokuca_ARXIV-3_2022},
abstract = {As automated face recognition applications tend towards ubiquity, there is a growing need to secure the sensitive face data used within these systems. This paper presents a survey of biometric template protection (BTP) methods proposed for securing face templates (images/features) in neural-network-based face recognition systems. The BTP methods are categorised into two types: Non-NN and NN-learned. Non-NN methods use a neural network (NN) as a feature extractor, but the BTP part is based on a non-NN algorithm, whereas NN-learned methods employ a NN to learn a protected template from the unprotected template. We present examples of Non-NN and NN-learned face BTP methods from the literature, along with a discussion of their strengths and weaknesses. We also investigate the techniques used to evaluate these methods in terms of the three most common BTP criteria: recognition accuracy, irreversibility, and renewability/unlinkability. The recognition accuracy of protected face recognition systems is generally evaluated using the same (empirical) techniques employed for evaluating standard (unprotected) biometric systems. However, most irreversibility and renewability/unlinkability evaluations are found to be based on theoretical assumptions/estimates or verbal implications, with a lack of empirical validation in a practical face recognition context. So, we recommend a greater focus on empirical evaluations to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice. Additionally, an exploration of the reproducibility of the studied BTP works, in terms of the public availability of their implementation code and evaluation datasets/procedures, suggests that it would be difficult to faithfully replicate most of the reported findings. So, we advocate for a push towards reproducibility, in the hope of advancing face BTP research.},
pdf = {https://publications.idiap.ch/attachments/papers/2022/Krivokuca_ARXIV-2_2022.pdf}
}
@TECHREPORT{Marcel02-36IRR,
author = {Marcel, S{\'{e}}bastien and Marcel, Christine and Bengio, Samy},
projects = {Idiap},
title = {A State-of-the-art Neural Network for Robust Face Verification},
type = {Idiap-RR},
number = {Idiap-RR-36-2002},
year = {2002},
institution = {IDIAP},
note = {Published in the Proceedings of the COST275 Workshop on The Advent of Biometrics on the Internet, Rome, Italy, 7-8 November, 2002},
abstract = {The performance of face verification systems has steadily improved over the last few years, mainly focusing on models rather than on feature processing. State-of-the-art methods often use the gray-scale face image as input. In this paper, we propose to use an additional feature to the face image: the skin color. The new feature set is tested on a benchmark database, namely XM2VTS, using a simple discriminant artificial neural network. Results show that the skin color information improves the performance and that the proposed model achieves robust state-of-the-art results.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr02-36.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr02-36.ps.gz},
ipdinar={2002},
ipdmembership={learning, vision},
language={English},
}
@TECHREPORT{Marcel01-44IRR,
author = {Marcel, S{\'{e}}bastien and Bengio, Samy},
projects = {Idiap},
title = {Improving Face Verification using Skin Color Information},
type = {Idiap-RR},
number = {Idiap-RR-44-2001},
year = {2001},
institution = {IDIAP},
note = {Published in the Proceedings of the International {C}onference on {P}attern {R}ecognition, Quebec City, Canada, 2002},
abstract = {The performance of face verification systems has steadily improved over the last few years, mainly focusing on models rather than on feature processing. State-of-the-art methods often use the gray-scale face image as input. In this paper, we propose to use an additional feature to the face image: the skin color. The new feature set is tested on a benchmark database, namely XM2VTS, using a simple discriminant artificial neural network. Results show that the skin color information improves significantly the performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2001/rr01-44.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2001/rr01-44.ps.gz},
ipdinar={2001},
ipdmembership={vision},
language={English},
}
@TECHREPORT{marcel:symfacelda:2003,
author = {Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {A Symmetric Transformation for LDA-based Face Verification},
type = {Idiap-RR},
number = {Idiap-RR-67-2003},
year = {2003},
institution = {IDIAP},
abstract = {One of the major problem in face verification is to deal with a few number of images per person to train the system. A solution to that problem is to generate virtual samples from an unique image by doing simple geometric transformations such as translation, scale, rotation and vertical mirroring. In this paper, we propose to use a symmetric transformation to generate a new virtual sample. This symmetric virtual sample is obtained by computing the average between the original image and the vertical mirrored image. The face verification system is based on LDA feature extraction, successfully used in previous studies, and MLP for classification. Experiments were carried out on a difficult multi-modal data\-base, namely BANCA. Results on this database show that our face verification system performs better that the state-of-the-art and also that the addition of the symmetric virtual sample improves the performance.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-67.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-67.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr05-81,
author = {Marcel, S{\'{e}}bastien and Mill{\'{a}}n, Jos{\'{e}} del R.},
projects = {Idiap},
title = {Person Authentication using Brainwaves (EEG) and Maximum A Posteriori Model Adaptation},
type = {Idiap-RR},
number = {Idiap-RR-81-2005},
year = {2005},
institution = {IDIAP},
note = {To appear in IEEE Transactions on Pattern Analysis and Machine Intelligence -- Special Issue on Biometrics 2007},
abstract = {In this paper, we investigate the use of brain activity for person authentication. It has been shown in previous studies that the brain-wave pattern of every individual is unique and that the electroencephalogram (EEG) can be used for biometric identification. EEG-based biometry is an emerging research topic and we believe that it may open new research directions and applications in the future. However, very little work has been done in this area and was focusing mainly on person identification but not on person authentication. Person authentication aims to accept or to reject a person claiming an identity, i.e comparing a biometric data to one template, while the goal of person identification is to match the biometric data against all the records in a database. We propose the use of a statistical framework based on Gaussian Mixture Models and Maximum A Posteriori model adaptation, successfully applied to speaker and face authentication, which can deal with only one training session. We perform intensive experimental simulations using several strict train/test protocols to show the potential of our method. We also show that there are some mental tasks that are more appropriate for person authentication than others.},
pdf = {https://publications.idiap.ch/attachments/reports/2005/marcel-idiap-rr-05-81.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2005/marcel-idiap-rr-05-81.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr06-34,
author = {Marcel, S{\'{e}}bastien and Rodriguez, Yann and Heusch, Guillaume},
projects = {Idiap},
title = {On the Recent Use of Local Binary Patterns for Face Authentication},
type = {Idiap-RR},
number = {Idiap-RR-34-2006},
year = {2006},
institution = {IDIAP},
note = {Submitted for publication},
abstract = {This paper presents a survey on the recent use of Local Binary Patterns (LBPs) for face recognition. LBP is becoming a popular technique for face representation. It is a non-parametric kernel which summarizes the local spacial structure of an image and it is invariant to monotonic gray-scale transformations. This is a very interesting property in face recognition. This probably explains the recent success of Local Binary Patterns in face recognition. In this paper, we describe the LBP technique and different approaches proposed in the literature to represent and to recognize faces. The most representatives are considered for experimental comparison on a common face authentication task. For that purpose, the XM2VTS and BANCA databases are used according to their respective experimental protocols.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/marcel-idiap-rr-06-34.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/marcel-idiap-rr-06-34.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{marcel:rr06-18,
author = {Marcel, S{\'{e}}bastien and Mari{\'{e}}thoz, Johnny and Rodriguez, Yann and Cardinaux, Fabien},
projects = {Idiap},
title = {Bi-Modal Face and Speech Authentication: a BioLogin Demonstration System},
type = {Idiap-RR},
number = {Idiap-RR-18-2006},
year = {2006},
institution = {IDIAP},
note = {To appear in Proceedings of the Second Workshop on Multimodal User Authentication},
abstract = {This paper presents a bi-modal (face and speech) authentication demonstration system that simulates the login of a user using its face and its voice. This demonstration is called BioLogin. It runs both on Linux and Windows and the Windows version is freely available for download. Bio\-Login is implemented using an open source machine learning library and its machine vision package.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/marcel-idiap-rr-06-18.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/marcel-idiap-rr-06-18.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{McCool_Idiap-RR-03-2009,
author = {McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO},
month = {3},
title = {Parts-Based Face Verification using Local Frequency Bands},
type = {Idiap-RR},
journal = {International Conference on Biometrics, 2009},
booktitle = {Proceedings of the International Conference on Biometrics, 2009},
number = {Idiap-RR-03-2009},
year = {2009},
institution = {Idiap},
note = {Submitted to ICB 2009},
abstract = {In this paper we extend the Parts-Based approach of face veri{\"{\i}}¬cation by performing a frequency-based decomposition. The Parts-Based approach divides the face into a set of blocks which are then considered to be separate observations, this is a spatial decomposition of the face. This pap er extends the Parts-Based approach by also dividing the face in the frequency domain and treating each frequency response from an observation separately. This can be expressed as forming a set of sub-images where each sub-image represents the response to a di{\"{\i}}¬€erent frequency of, for instance, the Discrete Cosine Transform. Each of these sub-images is treated separately by a Gaussian Mixture Model (GMM) based classi{\"{\i}}¬er. The classi{\"{\i}}¬ers from each sub-image are then combined using weighted summation with the weights b eing derived using linear logistic regression. It is shown on the BANCA database that this method improves the performance of the system from an Average Half Total Error Rate of 24.38\% to 15.17\% when compared to a GMM Parts-Based approach on Protocol P.},
pdf = {https://publications.idiap.ch/attachments/reports/2008/McCool_Idiap-RR-03-2009.pdf}
}
@TECHREPORT{McCool_Idiap-RR-13-2012,
author = {McCool, Chris and Marcel, S{\'{e}}bastien and Hadid, Abdenour and Pietikainen, Matti and Matejka, Pavel and Cernocky, Jan and Poh, Norman and Kittler, J. and Larcher, Anthony and Levy, Christophe and Matrouf, Driss and Bonastre, Jean-Fran{\c c}ois and Tresadern, Phil and Cootes, Timothy},
keywords = {bi-modal authentication, Face Recognition, mobile biometrics, speaker recognition},
projects = {Idiap, MOBIO, TABULA RASA},
month = {4},
title = {Bi-Modal Person Recognition on a Mobile Phone: using mobile phone data},
type = {Idiap-RR},
number = {Idiap-RR-13-2012},
year = {2012},
institution = {Idiap},
abstract = {This paper presents a novel fully automatic bi-modal, face and speaker, recognition system which runs in real-time on a mobile phone. The implemented system runs in real-time on a Nokia N900 and demonstrates the feasibility of performing both automatic face and speaker recognition on a mobile phone. We evaluate this recognition system on a novel publicly-available mobile phone database and provide a well defined evaluation protocol. This database was captured almost exclusively using mobile phones and aims to improve research into deploying biometric techniques to mobile devices. We show, on this mobile phone database, that face and speaker recognition can be performed in a mobile environment and using score fusion can improve the performance by more than 25\% in terms of error rates.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/McCool_Idiap-RR-13-2012.pdf}
}
@TECHREPORT{McCool_Idiap-RR-17-2013,
author = {McCool, Chris and Wallace, Roy and McLaren, Mitchell and El Shafey, Laurent and Marcel, S{\'{e}}bastien},
projects = {Idiap, TABULA RASA},
month = {5},
title = {Session Variability Modelling for Face Authentication},
type = {Idiap-RR},
number = {Idiap-RR-17-2013},
year = {2013},
institution = {Idiap},
abstract = {This paper examines session variability modelling for face authentication using Gaussian mixture models. Session variability modelling aims to explicitly model and suppress detrimental within-class (inter-session) variation. We examine two techniques to do this, inter-session variability modelling (ISV) and joint factor analysis (JFA), which were initially developed for speaker authentication. We present a self-contained description of these two techniques and demonstrate that they can be successfully applied to face authentication. In particular, we show that using ISV leads to significant error rate reductions of, on average, 22\% on the challenging and publicly-available databases SCface, BANCA, MOBIO, and Multi-PIE. Finally, we show that a limitation of both ISV and JFA for face authentication is that the session variability model captures and suppresses a significant portion of between-class variation.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/McCool_Idiap-RR-17-2013.pdf}
}
@TECHREPORT{Motlicek_Idiap-RR-18-2012,
author = {Motlicek, Petr and El Shafey, Laurent and Wallace, Roy and McCool, Chris and Marcel, S{\'{e}}bastien},
keywords = {face verification, Speaker identification},
projects = {Idiap, TABULA RASA},
month = {7},
title = {Bi-Modal Authentication in Mobile Environments Using Session Variability Modelling},
type = {Idiap-RR},
number = {Idiap-RR-18-2012},
year = {2012},
institution = {Idiap},
address = {Rue Marconi 19},
abstract = {We present a state-of-the-art bi-modal authentication
system for mobile environments, using session variability
modelling. We examine inter-session variability
modelling (ISV) and joint factor analysis (JFA) for
both face and speaker authentication and evaluate our
system on the largest bi-modal mobile authentication
database available, the MOBIO database, with over 61
hours of audio-visual data captured by 150 people in
uncontrolled environments on a mobile phone. Our system
achieves 2.6\% and 9.7\% half total error rate for
male and female trials respectively – relative improvements
of 78\% and 27\% compared to previous results.},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Motlicek_Idiap-RR-18-2012.pdf}
}
@TECHREPORT{Muckenhirn_Idiap-RR-30-2017,
author = {Muckenhirn, Hannah and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, UNITS},
month = {11},
title = {Towards directly modeling raw speech signal for speaker verification using CNNs},
type = {Idiap-RR},
number = {Idiap-RR-30-2017},
year = {2017},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2017/Muckenhirn_Idiap-RR-30-2017.pdf}
}
@TECHREPORT{Muckenhirn_Idiap-RR-11-2018,
author = {Muckenhirn, Hannah and Abrol, Vinayak and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, UNITS},
month = {7},
title = {Gradient-based spectral visualization of CNNs using raw waveforms},
type = {Idiap-RR},
number = {Idiap-RR-11-2018},
year = {2018},
institution = {Idiap},
abstract = {Modeling directly raw waveform through neural networks for speech processing is gaining more and more attention. Despite its varied success, a question that remains is: what kind of information are such neural networks capturing or learning for different tasks from the speech signal? Such an insight is not only interesting for advancing those techniques but also for understanding better speech signal characteristics. This paper takes a step in that direction, where we develop a gradient based approach to estimate the relevance of each speech sample input on the output score. We show that analysis of the resulting ``relevance signal" through conventional speech signal processing techniques can reveal the information modeled by the whole network. We demonstrate the potential of the proposed approach by analyzing raw waveform CNN-based phone recognition and speaker identification systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2018/Muckenhirn_Idiap-RR-11-2018.pdf}
}
@TECHREPORT{Muckenhirn_Idiap-RR-11-2017,
author = {Muckenhirn, Hannah and Korshunov, Pavel and Magimai-Doss, Mathew and Marcel, S{\'{e}}bastien},
projects = {Idiap, SWAN, UNITS},
month = {3},
title = {Long Term Spectral Statistics for Voice Presentation Attack Detection},
type = {Idiap-RR},
number = {Idiap-RR-11-2017},
year = {2017},
institution = {Idiap},
abstract = {Automatic speaker verification systems can be spoofed through recorded, synthetic or voice converted speech of target speakers. To make these systems practically viable, the detection of such attacks, referred to as presentation attacks, is of paramount interest. In that direction, this paper investigates two aspects: (a) a novel approach to detect presentation attacks where, unlike conventional approaches, no speech signal related assumptions are made, rather the attacks are detected by computing first order and second order spectral statistics and feeding them to a classifier, and (b) generalization of the presentation attack detection systems across databases. Our investigations on Interspeech 2015 ASVspoof challenge dataset and AVspoof dataset show that, when compared to the approaches based on conventional short-term spectral processing, the proposed approach with a linear discriminative classifier yields a better system, irrespective of whether the spoofed signal is replayed to the microphone or is directly injected into the system software process. Cross-database investigations show that neither the short-term spectral processing based approaches nor the proposed approach yield systems which are able to generalize across databases or methods of attack. Thus, revealing the difficulty of the problem and the need for further resources and research.},
pdf = {https://publications.idiap.ch/attachments/reports/2016/Muckenhirn_Idiap-RR-11-2017.pdf}
}
@INPROCEEDINGS{OtroshiShahreza_ICLR_2025,
author = {Otroshi Shahreza, Hatef and Marcel, S{\'{e}}bastien},
projects = {SAFER},
title = {HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere},
booktitle = {The Thirteenth International Conference on Learning Representations},
year = {2025},
url = {https://openreview.net/pdf?id=4YzVF9isgD},
abstract = {Face recognition datasets are often collected by crawling Internet and without individuals' consents, raising ethical and privacy concerns. Generating synthetic datasets for training face recognition models has emerged as a promising alternative. However, the generation of synthetic datasets remains challenging as it entails adequate inter-class and intra-class variations. While advances in generative models have made it easier to increase intra-class variations in face datasets (such as pose, illumination, etc.), generating sufficient inter-class variation is still a difficult task. In this paper, we formulate the dataset generation as a packing problem on the embedding space (represented on a hypersphere) of a face recognition model and propose a new synthetic dataset generation approach, called HyperFace. We formalize our packing problem as an optimization problem and solve it with a gradient descent-based approach. Then, we use a conditional face generator model to synthesize face images from the optimized embeddings. We use our generated datasets to train face recognition models and evaluate the trained models on several benchmarking real datasets. Our experimental results show that models trained with HyperFace achieve state-of-the-art performance in training face recognition using synthetic datasets. Project page: https://www.idiap.ch/paper/hyperface}
}
@TECHREPORT{Poh_03_faceVerif,
author = {Poh, Norman and Marcel, S{\'{e}}bastien and Bengio, Samy},
projects = {Idiap},
title = {Improving Face Authetication Using Virtual Samples},
type = {Idiap-RR},
number = {Idiap-RR-40-2002},
year = {2002},
institution = {IDIAP},
note = {Published in ICASSP'03},
abstract = {In this paper, we present a simple yet effective way to improve a face verification system by generating multiple virtual samples from the unique image corresponding to an access request. These images are generated using simple geometric transformations. This method is often used during training to improve accuracy of a neural network model by making it robust against minor translation, scale and orientation change. The main contribution of this paper is to introduce such method during testing. By generating $N$ images from one single image and propagating them to a trained network model, one obtains $N$ scores. By merging these scores using a simple mean operator, we show that the variance of merged scores is decreased by a factor between 1 and $N$. An experiment is carried out on the XM2VTS database which achieves new state-of-the-art performances.},
pdf = {https://publications.idiap.ch/attachments/reports/2002/rr02-40.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2002/rr02-40.ps.gz},
ipdmembership={learning},
}
@TECHREPORT{rodrig2003,
author = {Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Boosting Pixel-based Classifiers for Face Verification},
type = {Idiap-RR},
number = {Idiap-RR-65-2003},
year = {2003},
institution = {IDIAP},
note = {Published in BIOAW Workshop of ECCV, 2004},
abstract = {The performance of face verification systems has steadily improved over the last few years. State-of-the-art methods use the projection of the gray-scale face image into a Linear Discriminant subspace as input of a classifier such as Support Vector Machines or Multi-layer Perceptrons. Unfortunately, these classifiers involve thousands of parameters that are difficult to store on a smart-card for instance. Recently, boosting algorithms has emerged to boost the performance of simple (weak) classifiers by combining them iteratively. The famous AdaBoost algorithm have been proposed for object detection and applied successfully to face detection. In this paper, we investigate the use of AdaBoost for face verification to boost weak classifiers based simply on pixel values. The proposed approach is tested on a benchmark database, namely XM2VTS. Results show that boosting only hundreds of classifiers achieved near state-of-the-art results. Furthermore, the proposed approach outperforms similar work on face verification using boosting algorithms on the same database.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rodrig-idiap-rr-03-65.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rodrig-idiap-rr-03-65.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{rodrig2003-2,
author = {Popovici, Vlad and Rodriguez, Yann and Thiran, Jean-Philippe and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {On Performance Evaluation of Face Detection and Localization Algorithms},
type = {Idiap-RR},
number = {Idiap-RR-80-2003},
year = {2003},
institution = {IDIAP},
note = {Published in International Conference on Pattern Recognition, {ICPR}, 2004},
abstract = {When comparing different methods for face detection or localization, one realizes that just simply comparing the reported results is misleading as, even if the results are reported on the same dataset, different authors have different views of what a correct detection/localization means. This paper addresses exactly this problem, proposing an objective measure for the goodness of a detection/localization for the case of frontal faces. The usage of the proposed technique insures a fair and unbiased way of reporting the results, making the experiment repeatable, measurable, and comparable by anybody else.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rodrig-idiap-rr-03-80.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rodrig-idiap-rr-03-80.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{rodriguez:rr06-06,
author = {Rodriguez, Yann and Marcel, S{\'{e}}bastien},
projects = {Idiap},
title = {Face Authentication Using Adapted Local Binary Pattern Histograms},
type = {Idiap-RR},
number = {Idiap-RR-06-2006},
year = {2006},
institution = {IDIAP},
note = {Published in 9th European Conference on Computer Vision {ECCV}, 2006},
abstract = {In this paper, we propose a novel generative approach for face authentication, based on a Local Binary Pattern (LBP) description of the face. A generic face model is considered as a collection of LBP-histograms. Then, a client-specific model is obtained by an adaptation technique from this generic model under a probabilistic framework. We compare the proposed approach to standard state-of-the-art face authentication methods on two benchmark databases, namely XM2VTS and BANCA, associated to their experimental protocol. We also compare our approach to two state-of-the-art LBP-based face recognition techniques, that we have adapted to the verification task.},
pdf = {https://publications.idiap.ch/attachments/reports/2006/rodrig-idiap-rr-06-06.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2006/rodrig-idiap-rr-06-06.ps.gz},
ipdmembership={vision},
}
@TECHREPORT{Roy_Idiap-RR-29-2009,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI},
month = {11},
title = {Visual processing-inspired Fern-Audio features for Noise-Robust Speaker Verification},
type = {Idiap-RR},
number = {Idiap-RR-29-2009},
year = {2009},
institution = {Idiap},
abstract = {In this paper, we consider the problem of speaker verification as a two-class object detection problem in computer vision, but the object instances are 1-D short-time spectral vectors obtained from the speech signal. More precisely, we investigate the general problem of speaker verification in the presence of additive white Gaussian noise, which we consider as analogous to visual object detection under varying illumination conditions. Inspired by their recent success in illumination-robust object detection, we apply a certain class of binary-valued pixel-pair based features
called Ferns for noise-robust speaker verification. Intensive experiments on a benchmark database according to a standard evaluation protocol have shown the advantage of the proposed features in the presence of moderate to extremely high amounts of additive noise.},
pdf = {https://publications.idiap.ch/attachments/reports/2009/Roy_Idiap-RR-29-2009.pdf}
}
@TECHREPORT{Roy_Idiap-RR-28-2009,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2, MOBIO, SNSF-MULTI},
month = {9},
title = {Haar Local Binary Pattern Feature for Fast Illumination Invariant Face Detection},
type = {Idiap-RR},
number = {Idiap-RR-28-2009},
year = {2009},
institution = {Idiap},
abstract = {Face detection is the first step in many visual processing systems like face recognition, emotion recognition and lip reading. In this paper, we propose a novel feature called Haar Local Binary Pattern (HLBP) feature for fast and reliable face detection, particularly in adverse imaging conditions. This binary feature compares bin values of Local Binary Pattern histograms calculated over two adjacent image subregions. These subregions are similar to those in the Haar masks, hence the name of the feature. They capture the region-specific variations of local texture patterns and are boosted using AdaBoost in a framework similar to that proposed by Viola and Jones. Preliminary results obtained on several standard databases show that it competes well with other face detection systems, especially in adverse illumination conditions.},
pdf = {https://publications.idiap.ch/attachments/reports/2009/Roy_Idiap-RR-28-2009.pdf}
}
@TECHREPORT{Roy_Idiap-RR-13-2010,
author = {Roy, Anindya and Marcel, S{\'{e}}bastien},
projects = {Idiap, MOBIO, SNSF-MULTI, IM2},
month = {6},
title = {Crossmodal Matching of Speakers using Lip and Voice Features in Temporally Non-overlapping Audio and Video Streams},
type = {Idiap-RR},
number = {Idiap-RR-13-2010},
year = {2010},
institution = {Idiap},
abstract = {Person identification using audio (speech) and visual (facial appearance, static or dynamic) modalities, either independently or jointly, is a thoroughly investigated problem in pattern recognition. In this work, we explore a novel task : person identification in a cross-modal scenario, i.e., matching the speaker in an audio recording to the same speaker in a video recording, where the two recordings have been made during different sessions, using speaker specific information which is common to both the audio and video modalities. Several recent psychological studies have shown how humans can indeed perform this task with an accuracy significantly higher than chance. Here we propose two systems which can solve this task comparably well, using purely pattern recognition techniques. We hypothesize that such systems could be put to practical use in multimodal biometric and surveillance systems.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Roy_Idiap-RR-13-2010.pdf}
}
@INPROCEEDINGS{Saedi_INTERSPEECH_2013,
author = {Saedi, Rahim and Lee, Kong Aik and Kinnunen, Tomi and Hasan, Tawfik and Fauve, Benoit and Bousquet, Pierre-Michel and Khoury, Elie and Martinez, Pablo Luis Sordo and Kua, Jia Min Karen and You, Changhuai and Sun, Hanwu and Larcher, Anthony and Rajan, Padmanabhan and Hautam{\"{a}}ki, Ville and Hanilci, Cemal and Braithwaite, Billy and Rosa, Gonzalez-Hautam{\"{a}}ki and Sadjadi, Seyed Omid and Liu, Gang and Boril, Hynek and Shokouhi, Navid and Matrouf, Driss and El Shafey, Laurent and Mowlaee, Pejman and Epps, Julien and Thiruvaran, Tharmarajah and Van Leeuwen, David and Ma, Bin and Li, Haizhou and Hansen, John and Bonastre, Jean-Fran{\c c}ois and Marcel, S{\'{e}}bastien and Mason, John and Ambikairajah, Eliathamby},
projects = {Idiap, SNSF-LOBI},
month = aug,
title = {I4U Submission to NIST SRE 2012: a large-scale collaborative effort for noise-robust speaker verification},
booktitle = {INTERSPEECH},
year = {2013},
location = {Lyon, France},
crossref = {Saedi_Idiap-RR-34-2013},
abstract = {I4U is a joint entry of nine research Institutes and Universities across 4 continents to NIST SRE 2012. It started with a brief discussion during the Odyssey 2012 workshop in Singapore. An online discussion group was soon set up, providing a discussion platform for different issues surrounding NIST SRE’12. Noisy test segments, uneven multi-session training, variable enrollment duration, and the issue of open-set identification were actively discussed leading to various solutions integrated to the I4U submission. The joint submission and several of its 17 sub-systems were among top-performing systems. We summarize the lessons learnt from this large-scale effort.},
pdf = {https://publications.idiap.ch/attachments/papers/2013/Saedi_INTERSPEECH_2013.pdf}
}
@TECHREPORT{Saedi_Idiap-RR-34-2013,
author = {Saedi, Rahim and Lee, Kong Aik and Kinnunen, Tomi and Hasan, Tawfik and Fauve, Benoit and Bousquet, Pierre-Michel and Khoury, Elie and Martinez, Pablo Luis Sordo and Kua, Jia Min Karen and You, Changhuai and Sun, Hanwu and Larcher, Anthony and Rajan, Padmanabhan and Hautam{\"{a}}ki, Ville and Hanilci, Cemal and Braithwaite, Billy and Rosa, Gonzalez-Hautam{\"{a}}ki and Sadjadi, Seyed Omid and Liu, Gang and Boril, Hynek and Shokouhi, Navid and Matrouf, Driss and El Shafey, Laurent and Mowlaee, Pejman and Epps, Julien and Thiruvaran, Tharmarajah and Van Leeuwen, David and Ma, Bin and Li, Haizhou and Hansen, John and Bonastre, Jean-Fran{\c c}ois and Marcel, S{\'{e}}bastien and Mason, John and Ambikairajah, Eliathamby},
projects = {Idiap, SNSF-LOBI},
month = {11},
title = {I4U Submission to NIST SRE 2012: a large-scale collaborative effort for noise-robust speaker verification},
type = {Idiap-RR},
number = {Idiap-RR-34-2013},
year = {2013},
institution = {Idiap},
crossref = {Saedi_INTERSPEECH_2013},
abstract = {The submission of I4U, is a joint effort of nine research Institutes and
Universities across 4 continents for submitting speaker recognition
results to NIST SRE 2012. The joint efforts were started with a brief
discussion during the Odyssey 2012 workshop in Singapore. An online
discussion group was soon set up afterwards, providing a discussion
platform for different issues surrounding the NIST SRE’12. In
particular, noisy test segments, uneven multi-session training, variable
enrollment duration, and the issue of open-set identification
have been actively discussed. Various solutions were put in place
as part of the I4U submission. The submission of I4U as well as several
individual submissions from coalition members, was found to
be among top-performing systems submitted to SRE’12. This paper
summarizes the system components’ details for 17 systems included
in I4U submission.},
pdf = {https://publications.idiap.ch/attachments/reports/2013/Saedi_Idiap-RR-34-2013.pdf}
}
@TECHREPORT{sanders-rr-03-13,
author = {Sanderson, Conrad and Bengio, Samy and Bourlard, Herv{\'{e}} and Mari{\'{e}}thoz, Johnny and Collobert, Ronan and BenZeghiba, Mohamed Faouzi and Cardinaux, Fabien and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {2},
title = {{S}peech & {F}ace {B}ased {B}iometric Authentication at {IDIAP}},
type = {Idiap-RR},
number = {Idiap-RR-13-2003},
year = {2003},
institution = {IDIAP},
abstract = {We present an overview of recent research at IDIAP on speech & face based biometric authentication. This report covers user-customised passwords, adaptation techniques, confidence measures (for use in fusion of audio & visual scores,',','),
face verification in difficult image conditions, as well as other related research issues. We also overview the Torch machine-learning library, which has aided in the implementation of the above mentioned techniques.},
pdf = {https://publications.idiap.ch/attachments/reports/2003/rr03-13.pdf},
postscript = {ftp://ftp.idiap.ch/pub/reports/2003/rr03-13.ps.gz},
ipdmembership={learning},
}
@TECHREPORT{Subburaman_Idiap-RR-38-2010,
author = {Subburaman, Venkatesh Bala and Marcel, S{\'{e}}bastien},
projects = {Idiap, IM2},
month = {11},
title = {Fast Bounding Box Estimation based Face Detection},
type = {Idiap-RR},
number = {Idiap-RR-38-2010},
year = {2010},
institution = {Idiap},
abstract = {The sliding window approach is the most widely used technique to detect an object from an image. In the past few years, classifiers have been improved in many ways to increase the scanning speed. Apart from the classifier design (such as cascade,',','),
the scanning speed also depends on number of different factors (such as grid spacing, and scale at which the image is searched). When the scanning grid spacing is larger than the tolerance of the trained classifier it suffers from low detections. In this paper we present a technique to reduce the number of miss detections while increasing the grid spacing when using the sliding window approach for object detection. This is achieved by using a small patch to predict the bounding box of an object within a local search area. To achieve speed it is necessary that the bounding box prediction is comparable or better than the time it takes in average for the object classifier to reject a subwindow. We use simple features and a decision tree as it proved to be efficient for our application. We analyze the effect of patch size on bounding box estimation and also evaluate our approach on benchmark face database. Since perturbing the training data can have an affect on the final performance, we evaluate our approach for classifiers trained with and without perturbations and also compare with OpenCV. Experimental evaluation shows better detection rate and speed with our proposed approach for larger grid spacing when compared to standard scanning technique.},
pdf = {https://publications.idiap.ch/attachments/reports/2010/Subburaman_Idiap-RR-38-2010.pdf}
}
@TECHREPORT{Unnervik_Idiap-RR-08-2022,
author = {Unnervik, Alexander and Marcel, S{\'{e}}bastien},
keywords = {anomaly detection, Backdoor attack, Biometrics, CNN, Face Recognition, security, trojan attack},
projects = {Idiap, TRESPASS-ETN},
month = {8},
title = {An anomaly detection approach for backdoored neural networks: face recognition as a case study},
type = {Idiap-RR},
number = {Idiap-RR-08-2022},
year = {2022},
institution = {Idiap},
note = {Under review [BIOSIG 2022]},
url = {https://gitlab.idiap.ch/biometric/paper.backdoors_anomaly_detection.biosig2022},
abstract = {Backdoor attacks allow an attacker to embed functionality jeopardizing proper behavior
of any algorithm, machine learning or not. This hidden functionality can remain inactive for normal
use of the algorithm until activated by the attacker. Given how stealthy backdoor attacks are, conse-
quences of these backdoors could be disastrous if such networks were to be deployed for applications
as critical as border or access control. In this paper, we propose a novel backdoored network detec-
tion method based on the principle of anomaly detection, involving access to the clean part of the
training data and the trained network. We highlight its promising potential when considering various
triggers, locations and identity pairs, without the need to make any assumptions on the nature of the
backdoor and its setup. We test our method on a novel dataset of backdoored networks and report
detectability results with perfect scores.},
pdf = {https://publications.idiap.ch/attachments/reports/2022/Unnervik_Idiap-RR-08-2022.pdf}
}
@TECHREPORT{Wallace_Idiap-RR-03-2012,
author = {Wallace, Roy and McLaren, Mitchell and McCool, Chris and Marcel, S{\'{e}}bastien},
projects = {Idiap},
month = {1},
title = {Cross-pollination of normalisation techniques from speaker to face authentication using Gaussian mixture models},
type = {Idiap-RR},
journal = {IEEE Transactions on Information Forensics and Security},
number = {Idiap-RR-03-2012},
year = {2012},
institution = {Idiap},
pdf = {https://publications.idiap.ch/attachments/reports/2012/Wallace_Idiap-RR-03-2012.pdf}
}
@TECHREPORT{Wallace_Idiap-RR-28-2011,
author = {Wallace, Roy and McLaren, Mitchell and McCool, Chris and Marcel, S{\'{e}}bastien},
keywords = {2D Face Authentication, Inter-session Variability Modelling, Joint Factor Analysis},
projects = {Idiap, TABULA RASA},
month = {8},
title = {Inter-session Variability Modelling and Joint Factor Analysis for Face Authentication},
type = {Idiap-RR},
number = {Idiap-RR-28-2011},
year = {2011},
institution = {Idiap},
abstract = {This paper applies inter-session variability modelling
and joint factor analysis to face authentication using Gaussian
Mixture Models. These techniques, originally developed
for speaker authentication, aim to explicitly model and
remove detrimental within-client (inter-session) variation
from client models. We apply the techniques to face authentication
on the publicly-available BANCA, SCface and MOBIO
databases. We propose a face authentication protocol
for the challenging SCface database, and provide the first
results on the MOBIO still face protocol. The techniques
provide relative reductions in error rate of up to 44\%, using
only limited training data. On the BANCA database,
our results represent a 31\% reduction in error rate when
benchmarked against previous work.},
pdf = {https://publications.idiap.ch/attachments/reports/2011/Wallace_Idiap-RR-28-2011.pdf}
}