Reproduce “Body of Evidence: Triangulating 3D Human Pose with Geometry-Conditioned Depth” 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

Erik Brorsson, Kilian Freitag, Josef Bengtson, Kristofer Bengtsson, Lars Hammarstrand, Knut Åkesson, Fredrik Kahl

BoE estimates per-view joint depths from image evidence and the person’s 3D pose, then uses them as constraints in calibrated triangulation. It iteratively updates pose context for depth estimates. On CMU0(3), trained on five-camera CMU0 without fine-tuning, it reaches 66.0 AP25, average precision at a 25-mm 3D-joint-error threshold, versus 55.4 for MV-SSM. It cannot recover joints occluded in every camera.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.