Abstract:. In this talk, I will present our recent results related to aligning and analyzing data sequences consisting of 3D frames that capture a surface undergoing a motion. We focus on the motion of human body and face shapes, which are especially important in various applications including tele-presence and design tasks. Aligning these motion sequences, i.e. finding meaningful point-to-point correspondences between all 3D frames, is challenging due to large non-rigid deformations during the motion and due to occlusion by clothing, hair, and accessories. We tackle this problem with the help of statistical shape spaces that model shape variations across individuals and across posture separately and are built in a data-driven way. This allows us to analyze the data, e.g. by estimating the motion of the human body shape under clothing.
Geometric Modeling of 3D Environments
