Bay Area Vision Meeting: Learning Representations for Real-world Recognition

Bay Area Vision Meeting (more info below) Learning Representations for Real-world Recognition Trevor Darrell March 7, 2011 ABSTRACT Methods for visual recognition had made dramatic strides in recent years on various online benchmarks, but performance in the real world still often falters. Classic bag-of-local-feature models make overly simplistic assumptions regarding image appearance statistics, both locally and globally. Recent progress suggests that new learning-based representations can improve recognition by devices that are embedded in a physical world. I'll review new methods for domain adaptation which capture the visual domain shift between environments, and improve recognition of objects in specific places when trained from generic online sources. I'll also present recent results learning hierarchical local image representations based on recursive probabilistic topic models, and on learning strong object color models from sets of uncalibrated views, using a new multi-view color constancy paradigm. Finally, I'll mention other efforts in my lab, including efforts to exploit Primesense/KINECT sensors for indoor object category recognition. Prof. Trevor Darrell is on the faculty of the UC Berkeley EECS department, and leads the computer vision group at the International Computer Science Institute in Berkeley, CA. He obtained the PhD from MIT in 1996, and worked at Interval Research in Palo Alto from 1996 to 1999. From 1999 to 2008 he served on the faculty of the MIT EECS department, where he led the CSAIL Vision Interfaces group. In 2008 he returned to the west coast, and since then has been pursuing his interests in visual perception, machine learning, multimodal interfaces, and entrepreneurial activity in Berkeley. Prof. Darrell has authored over 100 journal and conference papers, has served on the editorial boards of PAMI and the Artificial Intelligence Journal, and served as Program Chair for CVPR 2010. He also serves on the advisory boards of local start-ups IQ Engines and BotSquare.