Reproduce “FingerViP: Learning Real-World Dexterous Manipulation with Fingertip Visual Perception” 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
Zhen Zhang, Weinan Wang, Hejia Sun, Qingpeng Ding, Xiangyu Chu, Guoxin Fang, K. W. Samuel Au
FingerViP combines five fingertip cameras with a fixed third-view camera, camera-pose information, and per-finger joint currents to guide a transformer-based diffusion policy for dexterous manipulation. Across four task families, it achieved 80.8% average success, versus 51.6% for a third-view-plus-wrist-camera baseline. The system ran at a lower frame rate and higher policy computation, and results cover only tested tasks and variations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.