Reproduce “CoT-Pose: Chain-of-Thought Reasoning for 3D Pose Generation from Abstract Prompts” 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
Junuk Cha, Jihyeon Kim
CoT-Pose turns abstract action requests into detailed descriptions of body configuration before generating 3D poses. It uses autoregressive text generation followed by pose-query tokens to produce SMPL poses. On a 50-image evaluation it slightly improves over ChatPose, with PFD 0.6162 versus 0.6769 and MPJPE 124.91 versus 126.63 mm. These results do not establish a numerical advantage on unseen prompts or independently annotated ground truth.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.