We factor action realization and robot appearance out of the world model by rendering actions as visible robot geometry, yielding a shared visual interface that transfers across viewpoints and unseen robot embodiments.

HOPE estimates temporally evolving per-vertex pressure and contact on a hand mesh directly from monocular RGB video, transferring from gloved-hand supervision to bare-hand egocentric and in-the-wild interactions.

HRDexDB is a large-scale, multi-modal dataset of 2.1K paired human and robotic hand grasping trials with high-precision 3D annotations, serving as a benchmark for cross-embodiment dexterous manipulation.

We present DWM, a scene-action-conditioned video diffusion model that simulates dexterous human interactions in static 3D scenes.

Vanast tackles virtual try-on with human image animation via synthetic triplet supervision, generating identity-preserving, pose-driven try-on videos from a person image and one or more garment references.

The first method for generating portrait animation videos with facial attribute transfer from a given reference image to a target portrait in a zero-shot manner.

We present HairCUP, a universal prior model for 3D head avatars with hair compositionality, which enables hairstyle swapping and efficient personalization.

We present GALA, a framework that takes as input a single-layer clothed 3D human mesh and decomposes it into complete multi-layered 3D assets.

We present, PEGASUS, a method for constructing personalized generative 3D face avatars from monocular video sources.

We present Guess The Unseen, a method to reconstruct the world and multiple dynamic humans in 3D from a monocular video input.

We propose Chupa, a 3D human generation pipeline that combines the generative power of diffusion models and neural rendering techniques to create diverse, realistic 3D humans.

We propose a novel self-supervised learning framework for snow points removal in LiDAR point clouds.