Reproduce “3D-CODED : 3D Correspondences by Deep Deformation” 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

Thibault Groueix, Matthew Fisher, Vladimir G. Kim, Bryan C. Russell, Mathieu Aubry

3D-CODED matches points across scans by deforming a shared template to each scan, whose point identities link the shapes. An encoder predicts a shape feature, refined at test time by gradient descent on Chamfer distance. On supervised FAUST-inter, its configuration reaches 2.878 cm average correspondence error, versus 3.12 cm for Stitched Puppet; it depends on training correspondences and pose coverage.

Summary by alphaXiv.

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