Reproduce “MetricPose: Flexible 3D Human Pose Estimation with Depth and Geometric Cues” 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
Shifan Zhu, Sabad Modi, Hang Yang, Arushi Agrawal, Steven Gao, Meghan Huber, Nathan Wycoff, Donghyun Kim
MetricPose extends SAM 3D Body with RGB, RGB-Ground, and RGB-D modes to estimate camera-frame human pose and placement. RGB-D uses measured depth to refine pose and correct metric scale, while RGB-Ground uses floor geometry when contact conditions hold. On held-out HuMMan, RGB-D lowers camera-local MPJPE from 81.2 to 39.7 mm; simulated distance results do not establish real-world guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.