Reproduce “Hand in Silhouette: Synthetic-to-Real Multi-View Hand Reconstruction via Binary Masks” 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
Leiyao Cui, Lecheng Ruan, Zihang Zhao
SilHand reconstructs articulated 3D hands from calibrated multi-view binary masks, using projected mask and camera-ray features fused across a metric grid to estimate pose and shape. It is trained on synthetic configurations, without real-world fine-tuning. With annotation-rendered masks, its mean per-vertex position error is 49.2%–63.4% lower than the best RGB-based baseline across five benchmarks; this comparison uses different input modalities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.