Reproduce “Mesh-Agnostic Refinement Models for Latent Space Encoding of 3D Human Body Shapes” 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

Chaima Zouaoui, Jonathan Cerqueira, Sylvain Arguillere, Martin Bauer, Nicolas Charon, Mohamed Daoudi, Emmanuel Hartman

The authors present a body-mesh refinement method that compares a decoded parametric body directly with a target surface, without requiring matching vertices at inference. It refines model parameters, predicts vertex displacements, or optimizes those displacements using geometric residuals. On AMASS with OmniFit initialization, Vertex-VaRefine lowers mean vertex-to-vertex error from 1.181 cm to 0.342 cm, a 71.0% reduction; results assume a reasonably close, especially pose-accurate, initialization.

Summary by alphaXiv.

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