Reproduce “HumanCOIN: A 12,500-Hour Dataset for Human Intent in Egocentric Manipulation” 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

Shihan Wu, You Liu, Chunyu He, Minghui Qin, Longhan Jiao, Jinbang Guo, Shaoxuan Xie, Kaixi Tian and 43 more

HumanCOIN records gaze, forearm muscle activity, and brain activity alongside video, pose, and action annotations to study signals preceding movement. Its three nested tiers contain 12,500 hours across 38 scene classes, with higher tiers adding surface EMG and EEG as tasks and settings become more controlled. These are proxies for intent; the dataset does not show improved prediction.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.