Strategies for General Recognition

One of the major challenges in computer vision is general recognition, the ability to infer useful properties of objects within sight. It's a slippery problem whose solution depends on the details of a particular situation — what do I want to know, what kind of object, how similar to those I have seen before. I'll present the work from my group over the past few years on how to organize and infer knowledge of objects, with a focus on generalizing to new types of objects, providing details, and learning models of parts. I will also summarize the most important directions for future work in this area.
Derek Hoiem is an assistant professor in Computer Science at the University of Illinois at Urbana-Champaign. He received his PhD in Robotics from Carnegie Mellon University in 2007 and completed a Beckman Postgraduate Fellowship in 2008. Derek's research in visual scene understanding and object recognition has been recognized with an ACM Doctoral Dissertation Award honorable mention, Intel Early Career Faculty award, and Sloan Fellowship. Likes include aged manchego and peaty whisky (especially together), scenic day hikes, and all kinds of games.