Reproduce “InterPose: Learning to Generate Human-Object Interactions from Large-Scale Web Videos” 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
Yangsong Zhang, Abdul Ahad Butt, Gül Varol, Ivan Laptev
InterPose converts interaction-rich videos into human-motion sequences and captions for training spatially controllable motion generators. Its dataset contains 73.8K sequences from 45.8K videos, without reconstructed object motion. In an OMOMO evaluation using supplied object waypoints, MaskedMimic trained on InterPose achieves contact F1 of 0.7172 versus 0.6773 for supervised CHOIS, though its contact precision is slightly lower.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.