Reproduce “Camera-to-Quadruped Robot Pose Estimation from Sequential Images” 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
Seobin Hwang, Ji-Seong Beom, Yeong-Jun Cho
QuadPose estimates a quadruped’s 18-DoF state from monocular video, including its base pose and 12 leg-joint angles. It uses role-specific keypoint tokens, skeleton-aware connections, and temporal modeling, with a robot model, camera intrinsics, and bounding boxes as inputs. On synthetic QuadSyn, it achieves 69.49% ADD@10 cm versus 52.08% for the strongest listed baseline; this does not establish real-world 3D accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.