Reproduce “EgoHMP: Achieving Precise Human Motion Prediction in 3D Scenes via Egocentric Cues” 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
Xin Zhao, Pengzhan Zhou, Ziyi Li, Riheng Jia, Zhida Qin, Yu Liu
EgoHMP predicts future human motion in 3D scenes by inferring interaction intent from first-person video and head movement, without using gaze. It maps likely interaction regions onto the scene, then uses them to guide a global trajectory and contact-aware pose generation. On GIMO, it reports 114.9 mm MPJPE-Path versus 141.2 mm for GAP3DS; evidence is clearest on GIMO and GTA-IM.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.