Reproduce “HumanMAC: Masked Motion Completion for 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
Ling-Hao Chen, Jiawei Zhang, Yewen Li, Yiren Pang, Xiaobo Xia, Tongliang Liu
HumanMAC predicts future human motion by denoising whole sequences, then uses a mask to preserve observed poses while generating future frames or completing transitions to target poses. It trains end-to-end with one noise-prediction loss. On HumanEva-I, at 0.25 seconds observed and 1 second predicted, it reports ADE 0.209 and FDE 0.223, versus DivSamp’s 0.220 and 0.234; MMFDE is worse, and 100 DDIM steps limit real-time use.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.