Reproduce “Learning elementary structures for 3D shape generation and matching” 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
Theo Deprelle, Thibault Groueix, Matthew Fisher, Vladimir G. Kim, Bryan C. Russell, Mathieu Aubry
The authors learn shared 3D surface structures from shape collections, then adjust and combine them to reconstruct or match new shapes. On 13-category ShapeNet Core, point translation with MLP adjustment reports symmetric Chamfer distance 1.22 × 10⁻³ versus AtlasNet’s 1.45 × 10⁻³. On 200 FAUST scans, deformation learning reports correspondence error 2.76 versus 3D-CODED’s 2.96; the best structure-learning variant depends on the task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.