Oct 18: Kristen Grauman: Capturing Human Insight for Large-Scale Visual Learning

CIS Seminar Series
Tuesday October 18, 2011, 1:00 pm
Clark 314
Kristen Grauman, PhD
Assistant Professor
Computer Science
University of Texas at Austin
Hosted by:
Dr. Rene Vidal
Capturing Human Insight for Large-Scale Visual Learning
How should visual recognition algorithms solicit and exploit human knowledge?  Existing approaches often manage human supervision in haphazard ways, and only allow a narrow, one-way channel of input from the annotator to the system.  We propose learning algorithms that steer human insight towards where it will have the most impact, and expand the manner in which recognition methods can assimilate that insight.
I will present an approach to actively seek annotators’ input when training an object recognition system.  Unlike traditional active learning methods, we target not only the example for which a label is most needed, but also the type of label (e.g., an image tag vs. full segmentation).  To allow large-scale selection, we introduce novel randomized hashing algorithms that can rapidly identify uncertain points within massive unlabeled pools of data.   Using these ideas, we have recently deployed a “live learning” system that autonomously refines its models by actively requesting crowd-sourced annotations on images crawled from the Web.  It yields state-of-the-art accuracy on some of the most challenging categories in the PASCAL object detection benchmark.  Finally, beyond “asking” the right questions, I will briefly describe how we can “listen” more deeply to annotators, learning implied cues about objects’ relative importance in images and videos.
This talk describes work with Sudheendra Vijayanarasimhan, Prateek Jain, Sung Ju Hwang, and Yong Jae Lee.
Kristen Grauman is a Clare Boothe Luce Assistant Professor in the Department of Computer Science at the University of Texas at Austin.  Her research in computer vision and machine learning focuses on visual search and object recognition.  Before joining UT-Austin in 2007, she received her Ph.D. in the MIT EECS department in the Computer Science and Artificial Intelligence Laboratory, and her B.A. in Computer Science from Boston College.  She is a Microsoft Research New Faculty Fellow, a recipient of a 2008 NSF CAREER award, and was named one of “AI’s Ten to Watch” by IEEE Intelligent Systems in 2010.

Comments are closed.