Publications

Selected research papers, conference proceedings, and preprints in computer vision and generative AI.

arXiv 20262026-06

DragMesh-2: Physically Plausible Dexterous Hand-Object Interaction with Articulated Objects

#2 Paper of the Day on Hugging Face

Tianshan Zhang, Yijia Duan, Yanjun Li, Zeyu Zhang, Hao Tang

DragMesh-2 is a contact-driven framework for physically plausible dexterous hand interaction with articulated objects. It extends articulated interaction from object-centric generation to hand-driven interaction, where target parts move through physical contact between a dexterous hand and the object handle.

IJCV 20262026-10

DragMesh: Interactive 3D Generation Made Easy

Tianshan Zhang, Zeyu Zhang, Hao Tang

DragMesh lets a user articulate a 3D object with a single drag. It separates kinematic reasoning from motion generation. The model first infers the joint type and axis implied by the drag, then a dual-quaternion generative model produces the motion, which can be applied to meshes or 3D Gaussian Splatting. It runs in real time, keeps the articulation physically valid, and shows some generalization to new objects.