Tag
VideoMDM trains 3D human motion priors from 2D poses using a diffusion framework with 2D reprojection loss and 3D motion regularizers, achieving near-3D supervised performance without requiring 3D ground truth.
AnchorWorld is a framework for egocentric world simulation that enhances interaction integrity and enables flexible world customization through 3D human motion and anchor view definitions, outperforming state-of-the-art baselines.