Motion Guided Deep Dynamic 3D Garments

Realistic dynamic garments on animated characters have many AR/VR applications. While authoring such garment geometry is still a challenging task, data-driven simulation provides an attractive alternative, especially if it can be controlled simply using the motion of underlying character. In this work, we focus guided 3D garments, for loose garments. setup, first learn generative space plausible geometries. Then, mapping to capture dependent deformations, conditioned previous state as well its relative position with respect body. Technically, model dynamics, driven input character motion, by predicting per-frame local displacements in canonical that enriched frame-dependent skinning weights bring global space. We resolve any remaining collisions residual displacements. The resultant used history enable iterative roll-out prediction. demonstrate generalization unseen body shapes and inputs, show improvements over multiple state-of-the-art alternatives. Code data released https://geometry.cs.ucl.ac.uk/projects/2022/MotionDeepGarment/