Reproduce “Learning to Estimate Robust 3D Human Mesh from In-the-Wild Crowded Scenes” 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
Hongsuk Choi, Gyeongsik Moon, JoonKyu Park, Kyoung Mu Lee
3DCrowdNet recovers a target person’s 3D mesh when people overlap. It uses a 2D pose to guide image features, then samples local evidence to estimate 3D joints and body parameters. On 3DPW-Crowd, it reports MPJPE of 86.8, compared with 124.8 for Pose2Mesh using the same pose predictions; it can fail when poses are inaccurate and people look similar.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.