Timezone: »

 
Poster
Large Margin Learning of Upstream Scene Understanding Models
Jun Zhu · Li-Jia Li · Li Fei-Fei · Eric Xing

Wed Dec 08 12:00 AM -- 12:00 AM (PST) @ None #None

Upstream supervised topic models have been widely used for complicated scene understanding. However, existing maximum likelihood estimation (MLE) schemes can make the prediction model learning independent of latent topic discovery and result in an imbalanced prediction rule for scene classification. This paper presents a joint max-margin and max-likelihood learning method for upstream scene understanding models, in which latent topic discovery and prediction model estimation are closely coupled and well-balanced. The optimization problem is efficiently solved with a variational EM procedure, which iteratively solves an online loss-augmented SVM. We demonstrate the advantages of the large-margin approach on both an 8-category sports dataset and the 67-class MIT indoor scene dataset for scene categorization.

Author Information

Jun Zhu (Tsinghua University)
Li-Jia Li (Stanford University)
Li Fei-Fei (Stanford University)
Eric Xing (Petuum Inc. / Carnegie Mellon University)

More from the Same Authors