#artificial intelligence
Jun 2026
HAT-4D: Lifting Monocular Video for 4D Multi-Object Interactions via Human-Agent Collaboration
HAT-4D is proposed, the first agentic framework designed to reconstruct the 3D geometry, temporal dynamics, and physical interactions of multiple objects from a single video, accompanied by a novel multi-dimensional evaluation protocol focused on physical plausibility and temporal consistency.
Jia-Xin Li, Yu-Xiang Wu, Zhenkai Zhang et al.
· arXiv.org · 0 citations