Reproduce “CISyn: Human Mesh Recovery in Close Interactions with Physics-Simulated Synthetic Data” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yufan Deng, Lorenza Prospero, Christian Rupprecht

The authors introduce CISyn, a dataset of physics-simulated interactions between people, and train CISYN-HMR with occlusion weighting and per-joint uncertainty for ambiguous hidden-joint labels. On Hi4D, CISYN-HMR achieves 52.3 mm MPJPE, versus 54.8 mm for MA-HMR with the same training mix. This measures joint-position accuracy; simulated actors are passive, and predictions are not constrained against interpenetration.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.