Computational Learning Techniques for Animation

Tuesday, May 15, 2001 - 11:00am - 12:00pm
Keller 3-180
Demetri Terzopoulos (University of Toronto)
We have been investigating the promising role that computational theories of learning can play in the domain of computer graphics. I will present the motivations and results of our research in the context of physics and biology based modeling for animation. In particular, the presentation will cover:

Efficient neural network emulators of physical dynamics, dubbed NeuroAnimators, that may be trained to produce physically realistic motions by observing simulated physical systems in action.

Biomechanically modeled artificial animals capable of acquiring, through sensor-guided reinforcement learning, motor controllers that effectuate lifelike, muscle-actuated locomotion.

Support vector machine learning methods to determine the functional competencies of composable motor controllers for the physics-based animation of articulated, anthropomorphic figures.

(References: 1. Grzeszczuk/Terzopoulos/Hinton SIGGRAPH 98. 2. Grzeszczuk/Terzopoulos SIGGRAPH 95. 3. Faloutsos/van de Panne/Terzopoulos SIGGRAPH 01. See also Terzopoulos CACM 42(8):1999.)