Reproduce “UniffordInteractor: Reasoning the Unified Affordance for Human and Multi-Object Interaction” 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
Chen Su, Xingqun Qi, Xurui Hu, Bingkun Yang, Muyi Sun, Man Zhang, Zhenan Sun
UniffordInteractor predicts human-object and object-object affordances before generating multi-object motion. Its 3D multimodal language model estimates interaction regions, while a hierarchical generator produces root, body, hand, and object motion. On HIMO’s three-object setting, it achieves FID 2.422, versus 4.706 for HIMO-Gen’s specialized model. The system uses three object slots, and motion quality degrades on unseen geometries.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.