Reproduce “From Egocentric Human Video to Robot Manipulation in the Era of Embodied Intelligence” 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
Yi Wang, Jingqiu Wu, Lei Yao, Zhaoshuo Li, Haoyu Zhu, Yirui Zhou, Huaiyuan Xu, Ying Zheng and 2 more
The authors review learning from egocentric human video (EHV) for robot manipulation, organizing approaches into four paradigms: embodiment transfer, action and task supervision, visual representation pretraining, and latent action and dynamics modeling. Methods extract interaction information through retargeting, supervision, learned features, or prediction. Robot demonstrations and synthetic data can provide grounding or augmentation. The review does not establish EHV’s benefit.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.