Reproduce “PolyFlow: Continuous Topology Embedding Flow Matching for Artist-style Mesh Generation” 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

Chunshi Wang, Haohan Weng, Junliang Ye, Biwen Lei, Yang Li, Zibo Zhao, Zeqiang Lai, Kaiyi Zhang and 4 more

PolyFlow is a Transformer-based flow-matching framework that generates 3D polygonal meshes with artist-like topology in a continuous and parallel manner. The method achieves state-of-the-art geometric fidelity with a Chamfer Distance of 0.008, and reduces generation time to 5.88 seconds for a 4,000-vertex mesh, representing a substantial speedup over autoregressive approaches.

Key insights

  • Discrete mesh topology can be effectively represented in a continuous embedding space, making it amenable to parallel, continuous generative processes like flow matching.
  • Parallel flow-matching significantly reduces mesh generation time from minutes to seconds compared to sequential autoregressive methods, without sacrificing quality.
  • Jointly generating vertex geometry (positions, normals) and a continuous topology embedding allows for coherent, high-quality, artist-like mesh outputs with explicit resolution control.

Summary by alphaXiv.

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