Wednesday, November 14, 2001
Hewlett Packard Auditorium, 306 Soda Hall
4:00-5:00 p.m.

Dr. Chris Bregler

Computer Science Department, Stanford University

Capturing Motion Models
for Animation

Abstract:

We will survey our current research efforts on vision based capture and animation techniques applied to animals, humans, and cartoon characters. We will present new capture techniques that are able to track and infer kinematic chain and 3D non-rigid blend-shape models directly from 2D video data without the use of pre-tracked features and prior models. Furthermore we demonstrate how to use such motion capture data to estimate statistical models for synthesis and how to retarget motion to new characters. We show several examples on capturing kangaroos, giraffes, human body deformations, facial expressions, animating hops and walks with natural fluctuations, and retargeting expressive cartoon motion.

This reports on joint work with Kathy Pullen, Lorie Loeb, Lorenzo Torressani, Danny Yang, Gene Alexander, Aaron Hertzmann, Henning Biermann