Constructing visual models with a latent space approach
Florent Monay, Pedro Quelhas, Daniel Gatica-Perez and Jean-Marc Odobez
In: Subspace, Latent Structure and Feature Selection techniques: Statistical and Optimisation perspectives Workshop, 23-25 Feb 2005, Bohinj, Slovenia.
We propose the use of latent space models applied to local invariant features for object classification. We investigate whether using latent space models enables to learn patterns of visual co-occurrence and if the learned visual models improve performance when less labeled data are available. We present and discuss results that support these hypotheses. Probabilistic Latent Semantic Analysis (PLSA) automatically identies aspects from the data with semantic meaning, producing unsupervised soft clustering. The resulting compact representation retains sufficient discriminative information for accurate object classiffcation, and improves the classification accuracy through the use of unlabeled data when less labeled training data are available. We perform experiments on a 7-class object database containing 1776 images.
|EPrint Type:||Conference or Workshop Item (Paper)|
|Project Keyword:||Project Keyword UNSPECIFIED|
|Deposited By:||Florent Monay|
|Deposited On:||01 January 2006|