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


Modeling the Illusive Thing Called `Shape': a Mathematician's View


David Mumford
Brown University


Monday, November 11, 2002, 4:15PM
TCSeq 200
http://robotics.stanford.edu/ba-colloquium/

Abstract

When I first got involved with natural and computer vision, I discovered that no existing mathematical theory seems to express well our perception of two shapes being similar or to yield the features needed to recognize objects for vision applications. In the last decade, an interesting approach to deforming shapes was started by Miller and Grenander, using ideas from fluid mechanics. I want to outline some new results in this program and raise some problems.

About the Speaker

Professor Mumford received his PhD at Harvard University in 1961. He was a Higgins Professor of Mathematics with a joint appointment in Engineering and Applied Science at Harvard until 1997. Since then he has been a professor of Applied Mathematics at Brown University. Prof. Mumford has done work machine learning and natural intelligence, in particular the field of vision perception. Hi research focuses on statistical models for the variables of vision: direct models of the raw images, the object shapes, and the texture of object surfaces. Prof. Mumford has received numerous awards, among them the Fields Medal. He is a MacArthur Foundation Fellow and a member of the National Academy of Sciences.
Contact: bac-coordinators@cs.stanford.edu

Back to the Colloquium Page