Reproduce “SGIP: Learning Scene-Grounded Interaction Priors for Scene-Aware Human Motion Prediction” 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
Donghyeon Soon, Jiil Yoon, Hyeondong Kim, Daehee Park
SGIP predicts future human paths and poses by assigning scene destinations, including objects and free space. Training distills an intent-aware teacher’s textual and geometric grounding to a student using only motion and scene at inference. On GIMO, its observation-only student’s trajectory-path error is 549 ± 5.0 mm versus 642 mm for gaze-free GAP3DS*, within benchmark protocols, not proof of true intent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.