Weakly Supervised Scene and Activity Understanding via Nonparametric Learning
Abstract
Bayesian nonparametric (BNP) models define distributions on infinite-dimensional spaces of functions, partitions, or other combinatorial structures. They lead to flexible, data-driven unsupervised learning algorithms, and models whose internal structure continually grows and adapts to new observations. Applied to computer vision problems, BNP methods have lead to segmentation algorithms which adapt their resolution to each image, learning algorithms which discover objects and activities from videos, and low-level vision systems with adaptive local appearance dictionaries. BNP models provide a practical alternative to unwieldy model selection methods, and reduce the need for manual algorithm tuning, parameter specification, and data annotation.Continuing and substantially extending the work completed in the first three years of this project, we propose further advances to contemporary spatio-temporal BNP models and inference algorithms, which may significantly advance the state-of-the-art in weakly supervised scene and activity understanding. Algorithms for semantic understanding of indoor and outdoor environments in three dimensions (3D), as well as human pose and activity understanding from video sequences, will be our primary focus.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 04, 2017
- Source ID
- N000141712094
Entities
People
- Erik Sudderth
Organizations
- Brown University
- Office of Naval Research
- United States Navy