Broad Area Colloquium For AI-Geometry-Graphics-Robotics-Vision


Shape Recipies

William T. Freeman
Artificial Intelligence Laboratory
Massachusetts Institute of Technology
Monday, January 27, 2003, 4:15PM
TCSeq 201
http://robotics.stanford.edu/ba-colloquium/

Abstract

The goal of low-level vision is to estimate an underlying scene, given an observed image. Real-world scenes (eg, shapes, or patterns of reflectance) can be complex, conventionally requiring high dimensional representations which can be hard to estimate and to store. But in many situations, scene and image are closely related and it is possible to learn a functional relationship between them. The scene information can then be represented in reference to the image, where the functional specifies how to translate the image into the associated scene. We illustrate the use of this representation for encoding shape information. We show that this representation has appealing properties such as low dimensionality, locality, and slow variation across space and scale. These properties allow us to improve initial shape estimates from modalities such as stereo. I'll also show two 5 minute "shorts": presentations showing results I'm excited about on the topics of non-parametric belief propagation and separating shading and reflectance.

About the Speaker

William T. Freeman is an Associate Professor of Electrical Engineering and Computer Science at the Artificial Intelligence Laboratory at MIT, joining the faculty in September, 2001. From 1992 - 2001 he worked at Mitsubishi Electric Research Labs (MERL), most recently as Sr. Research Scientist and Associate Director. He studied computer vision for his PhD in 1992 from the Massachussetts Institute of Technology, and received a BS in physics and MS in electrical engineering from Stanford in 1979, and an MS in applied physics from Cornell in 1981. His current research interests include machine learning applied to computer vision and graphics, and Bayesian models of visual perception. Previous research topics include steerable filters and pyramids, the generic viewpoint assumption, color constancy, computer vision for computer games, and separating "style and content". Hobbies include kite aerial photography (see http://www.ai.mit.edu/people/wtf/ ) and ping-pong.

References

Shape recipies:
ftp://publications.ai.mit.edu/ai-publications/2002/AIM-2002-016.pdf
ftp://publications.ai.mit.edu/ai-publications/2002/AIM-2002-019.pdf

Non-parametric belief propagation:
ftp://publications.ai.mit.edu/ai-publications/2002/AIM-2002-020.pdf

Separating shading and reflectance
ftp://publications.ai.mit.edu/ai-publications/2002/AIM-2002-015.pdf
Contact: bac-coordinators@cs.stanford.edu

Back to the Colloquium Page