Reproduce “EgoContact: Evaluating Physical Intelligence with Egocentric Video” 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
Adam Deryło, Alexey Gavryushin, Daniel Cremers, Marc Pollefeys, Xi Wang
The authors introduce EgoContact, a benchmark for predicting action-conditioned contact points from human video. It contains 22,523 verified interactions and tests predictions against contact locations. Models often miss the object or precise contact, even with multiple guesses. Replacing the memory bank of a fixed grasp policy with EgoContact improves success in simulation and real-robot tests, but does not isolate contact precision.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.