HSImul3R (ECCV 2026): simulation-ready human-scene reconstruction judged by physics, not looks

jiqizhixin · x · 2026-09-28

Xiaoxiao Robot, with NTU S-Lab and Shanghai AI Laboratory, present HSImul3R, accepted at ECCV 2026, targeting the long-standing "perception-to-simulation gap": reconstructions that look correct often collapse or clip through scenes in physics simulators.

Original post →

More from Embodied

Embodied channel →