Reproduce “ART: Bridging the Embodiment Gap in 3D Human-to-Robot Affordance Transfer” 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

Zichen Zhang, Wei Zhai, Xuji Zhang, Yang Cao, Yu Kang

ART transfers coarse human affordance priors into action-conditioned robot contact maps by filtering unreliable points and reconstructing their spatial patterns for the gripper. On held-out HPAD categories, it achieves 29.70% IoU₀.₉, versus 21.82% for Cross-Attention. In physical tests, ART succeeds on 86.5% of 480 trials, versus 58.3% for direct prior use, on one parallel-jaw gripper.

Summary by alphaXiv.

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