Reproduce “Light-SQ: Structure-aware Shape Abstraction with Superquadrics for Generated Meshes” 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

Yuhan Wang, Weikai Chen, Zeyu Hu, Runze Zhang, Yingda Yin, Ruoyu Wu, Keyang Luo, Shengju Qian and 14 more

Light-SQ introduces a superquadric-based optimization framework for structure-aware shape abstraction of 3D meshes generated by image-to-3D models. The method produces low-overlap, structurally aligned, and compact primitive representations, demonstrating superior editability and competitive fidelity on generative geometry benchmarks.

Key insights

  • Explicitly enforcing "structure-awareness," including low overlap, part alignment, and compactness, during superquadric optimization is essential for creating truly editable 3D abstractions from generative models.
  • Combining geometric analysis for robust convex decomposition with a multi-stage primitive fitting strategy (Block-Regrow-Fill) effectively balances structural alignment with overall shape fidelity.
  • A new benchmark, 3DGen-Prim, with generative meshes and novel primitive-level editability metrics, is crucial for evaluating and advancing research in shape abstraction for user-generated content (UGC).

Summary by alphaXiv.

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