Reproduce “Estimating Accurate Hand Pose in Camera Space with Vision Transformer” 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
Kaiwen Ren, Yiran Jiang, Yongjing Ye, Shihong Xia
CS-ViT estimates hand pose and wrist location in camera coordinates from a monocular RGB image. It uses camera intrinsics and crop information through a perspective embedding, and trains latent features with transformation-isomorphism supervision linking image changes to 3D pose changes. On HO3D, its MDT setup reaches 31.7 mm CS-MJE versus HandDGP’s 50.3 mm, but relies on supplied calibration and crops.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.