Reproduce “MetricHand: Geometry-Aware Flow Matching for 4D Hand Reconstruction in Metric World Space” 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
Yibo Bai, Likui Zhang, Jinglei Zhang, Zhengbin Pang, Xingshen Song, Xiaoqing Yin, Jinsheng Deng
MetricHand reconstructs egocentric hand motion in a metric world frame by combining appearance, metric depth, and nearby hand–object surfaces associated with anatomical joints. It carries this geometry through flow generation to preserve placement and articulation. On HOT3D, it reduces absolute world-space joint error by 28.7% and acceleration error by 44.2% versus HaWoR, though camera-motion and monocular-depth errors can still cause offsets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.