Reproduce “Hierarchical Kinematic Probability Distributions for 3D Human Shape and Pose Estimation from Images in the Wild” 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

Akash Sengupta, Ignas Budvytis, Roberto Cipolla

The authors predict a structured distribution over human body shape and pose from an RGB image, rather than one mesh. Shape uses a Gaussian, while joint rotations use parent-conditioned matrix-Fisher distributions. A reprojection loss matches sampled bodies to visible 2D joints. On 3DPW, mode-pose MPJPE-SC is 84.7 mm versus 91.3 mm for an independent-joint model; SSP-3D shape results use up to five images.

Summary by alphaXiv.

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