Describing visual scenes using transformed dirichlet processes

Describing visual scenes using transformed dirichlet processes (2005)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Erik B. Sudderth , Antonio Torralba , William T. Freeman , Alan S. Willsky

Venue:	Advances in Neural Information Processing Systems 18
Citations:	47 - 6 self

Versions

Version History

Metadata Version 0

Datum	Value	Source
TITLE	Describing Visual Scenes using Transformed Dirichlet Processes	SVM HeaderParse 0.2
AUTHOR NAME	Erik B. Sudderth	SVM HeaderParse 0.2
AUTHOR AFFIL	Department of Electrical Engineering and Computer Science; Massachusetts Institute of Technology	SVM HeaderParse 0.2
AUTHOR NAME	Antonio Torralba	SVM HeaderParse 0.2
AUTHOR AFFIL	Department of Electrical Engineering and Computer Science; Massachusetts Institute of Technology	SVM HeaderParse 0.2
AUTHOR NAME	William T. Freeman	SVM HeaderParse 0.2
AUTHOR AFFIL	Department of Electrical Engineering and Computer Science; Massachusetts Institute of Technology	SVM HeaderParse 0.2
AUTHOR NAME	Alan S. Willsky	SVM HeaderParse 0.2
AUTHOR AFFIL	Department of Electrical Engineering and Computer Science; Massachusetts Institute of Technology	SVM HeaderParse 0.2
ABSTRACT	Motivated by the problem of learning to detect and recognize objects with minimal supervision, we develop a hierarchical probabilistic model for the spatial structure of visual scenes. In contrast with most existing models, our approach captures the intrinsic uncertainty in the number and identity of objects depicted in a given image. Our scene model is based on the transformed Dirichlet process (TDP), a novel extension of the hierarchical DP in which a set of stochastically transformed mixture components are shared between multiple groups of data. For visual scenes, mixture components describe the spatial structure of visual features in an object–centered coordinate frame, while transformations model the object positions in a particular image. Learning and inference in the TDP, which has many potential applications beyond computer vision, is based on an empirically effective Gibbs sampler. Applied to a dataset of partially labeled street scenes, we show that the TDP’s inclusion of spatial structure improves detection performance, and allows unsupervised discovery of object categories. 1	SVM HeaderParse 0.2
CITATIONS	17 found	ParsCit 1.0