Physics-Based People Tracking

Research Overview

We consider a very rich class of generative model for 3D hand tracking. This incorporates a parameterized mesh to model the shape of the hand and non-rigid deformations, the hand albedo, and the scene illumination. We simultaneously estimate hand pose,the lighting parameters, and an update to the albedo function over the surface of the hand. Tracking is accomplished through a carefully derived gradient-based form of optimization. The resulting model gives state-of-the-art hand tracking performance.

Related Publications

Return to David Fleet's home page.