%Aigaion2 BibTeX export from Idiap Publications %Sunday 22 December 2024 04:08:11 AM @INPROCEEDINGS{monayACM:2004, author = {Monay, Florent and Gatica-Perez, Daniel}, projects = {Idiap}, title = {PLSA-based Image Auto-Annotation: Constraining the Latent Space}, booktitle = {Proc. ACM Int. Conf. on Multimedia (ACM MM)}, year = {2004}, note = {IDIAP-RR 04-30}, crossref = {monay02}, abstract = {We address the problem of unsupervised image auto-annotation with probabilistic latent space models. Unlike most previous works, which build latent space representations assuming equal relevance for the text and visual modalities, we propose a new way of modeling multi-modal co-occurrences, constraining the definition of the latent space to ensure its consistency in semantic terms (words,',','), while retaining the ability to jointly model visual information. The concept is implemented by a linked pair of Probabilistic Latent Semantic Analysis (PLSA) models. On a 16000-image collection, we show with extensive experiments and using various performance measures, that our approach significantly outperforms previous joint models.}, pdf = {https://publications.idiap.ch/attachments/papers/2004/monay-acm-1568937089.pdf}, postscript = {ftp://ftp.idiap.ch/pub/papers/2004/monay-acm-1568937089.ps.gz}, ipdmembership={vision}, } crossreferenced publications: @TECHREPORT{monay02, author = {Monay, Florent and Gatica-Perez, Daniel}, projects = {Idiap}, title = {PLSA-based Image Auto-Annotation: Constraining the Latent Space}, type = {Idiap-RR}, number = {Idiap-RR-30-2004}, year = {2004}, institution = {IDIAP}, note = {Published in ``Proc. ACM Multimedia 2004'', 2004}, abstract = {We address the problem of unsupervised image auto-annotation with probabilistic latent space models. Unlike most previous works, which build latent space representations assuming equal relevance for the text and visual modalities, we propose a new way of modeling multi-modal co-occurrences, constraining the definition of the latent space to ensure its consistency in semantic terms (words,',','), while retaining the ability to jointly model visual information. The concept is implemented by a linked pair of Probabilistic Latent Semantic Analysis (PLSA) models. On a 16000-image collection, we show with extensive experiments and using various performance measures, that our approach significantly outperforms previous joint models.}, pdf = {https://publications.idiap.ch/attachments/reports/2004/rr04-30.pdf}, postscript = {ftp://ftp.idiap.ch/pub/reports/2004/rr04-30.ps.gz}, ipdmembership={vision}, }