Reproduce “Fast and Accurate Body Estimation in Pixels, World Space and Scenes” 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

Jinkun Cao

FAB connects image-based body estimation to world scenes and robot motion by predicting dense body points directly, fitting SMPL to them, and using predicted contacts downstream. On 3DPW, FAB-L reports 32.7 mm PA-MPJPE versus 33.7 mm for SAM 3D Body, improving on most camera-space benchmarks. Its robot-tracking evidence comes from MuJoCo simulation, not a physical robot.

Summary by alphaXiv.

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