Webinaires

Analysis of deforming 3D shapes

Annonce

Résumé

This talk presents our recent results on data-driven representations and analyses of deforming 3D shapes. Three classes of shapes will be discussed. First, we focus on human body shapes, where we consider motion sequences digitized by a discrete number of frames, each captured densely as 3D shape. Our recent works allow to generate human motion sequences using retargeting, where a target body shape is animated with the motion extracted from a source sequence of body shapes. We further studied how the resulting 3D human body motion sequences can be evaluated in a way that correlates with human perception. Second, we focus on clothing worn by humans during motion, where we discuss how diffusion models can be used to generate detailed deforming 3D clothing based on human body shape, motion, and material parameters of the cloth. Finally, we focus on 3D plants, where we use an encoder-decoder model to infer parameterized representations of plants from raw 3D scans. 

Rediffusion chaîne GdR IASIS