%Aigaion2 BibTeX export from Idiap Publications
%Saturday 20 December 2025 07:37:57 PM

@INPROCEEDINGS{grangier:2006:icann,
                      author = {Grangier, David and Bengio, Samy},
                    projects = {Idiap},
                       title = {A Neural Network to Retrieve Images from Text Queries},
                   booktitle = {International Conference on Artificial Neural Networks {(ICANN)}},
                      volume = {2},
                        year = {2006},
                    crossref = {grangier:2006:idiap-06-33},
                    abstract = {This work presents a neural network for the retrieval of images from text queries. The proposed network is composed of two main modules: the first one extracts a global picture representation from local block descriptors while the second one aims at solving the retrieval problem from the extracted representation. Both modules are trained jointly to minimize a loss related to the retrieval performance. This approach is shown to be advantageous when compared to previous models relying on unsupervised feature extraction: average precision over Corel queries reaches 26.2\\% for our model, which should be compared to 21.6\\% for PAMIR, the best alternative.},
                         pdf = {https://publications.idiap.ch/attachments/reports/2006/grangier_icann06.pdf},
                  postscript = {ftp://ftp.idiap.ch/pub/reports/2006/grangier_icann06.ps.gz},
ipdmembership={learning},
}



crossreferenced publications: 
@TECHREPORT{grangier:2006:idiap-06-33,
                      author = {Grangier, David and Bengio, Samy},
                    projects = {Idiap},
                       title = {A Neural Network to Retrieve Images from Text Queries},
                        type = {Idiap-RR},
                      number = {Idiap-RR-33-2006},
                        year = {2006},
                 institution = {IDIAP},
                    abstract = {This work presents a neural network for the retrieval of images from text queries. The proposed network is composed of two main modules: the first one extracts a global picture representation from local block descriptors while the second one aims at solving the retrieval problem from the extracted representation. Both modules are trained jointly to minimize a loss related to the retrieval performance. This approach is shown to be advantageous when compared to previous models relying on unsupervised feature extraction: average precision over Corel queries reaches 26.2\\% for our model, which should be compared to 21.6\\% for PAMIR, the best alternative.},
                         pdf = {https://publications.idiap.ch/attachments/reports/2006/grangier_rr06-33.pdf},
                  postscript = {ftp://ftp.idiap.ch/pub/reports/2006/grangier_rr06-33.ps.gz},
ipdmembership={learning},
}