Deform360 — A Massive Multi-view Visuotactile Dataset for Deformable World Models (ECCV 2026)

Deform360 provides 215.7 cumulative multi-view hours across 41 cameras and 198 deformable objects for benchmarking 2D video and 3D particle world models. ECCV 2026.

关键词：Deform360, deformable objects, visuotactile dataset, world models, robot manipulation, 3D Gaussian Splatting, multi-view dataset, tactile sensing, deformable dynamics, ECCV 2026