University of Illinois at Urbana-Champaign

 

Discriminative Latent Variable Models for Visual Recognition

Discriminative latent variable models, in particular those involving structured latent variables, are emerging as a class of powerful tools for solving many computer vision problems that involve complex structures. In this talk, I will give an overview of several of our recent work in adapting these models for solving various problems in visual recognition, including part-based models for human action recognition, latent pose for still image action recognition, group activity recognition and image annotation. I will also mention applications in other areas(e.g. NLP) that can be addressed by these models. This is joint with Greg Mori, Weilong Yang and Tian Lan at Simon Fraser University, Canada.

Slides <link