Reproduce “Surfsvr: 2D Surface Priors as 3D Geometric Regularizers for Sparse Voxel Reconstruction” 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

Yan Di, Chengxi Li, Yaoxing Wang, Mengge Liu, Zhigang Li, Ruida Zhang, Mingyang Li, Pengyuan Wang and 2 more

SURFSVR introduces a novel sparse voxel reconstruction paradigm that leverages coherent 2D surface regions, constructed by refining superpixels with geometric cues and adaptive polynomial models, as explicit 3D geometric regularizers. The method achieves a mean Chamfer distance of 0.45 on the DTU dataset and an F1-score of 0.62 on Tanks and Temples, producing more complete surfaces and effectively suppressing floating artifacts.

Key insights

  • Coherent 2D surface regions provide a more robust and informative interface for integrating image-space priors into dense 3D geometry than individual pixel-wise cues.
  • An adaptive approach to surface modeling, combining planar and quadratic fits, allows for flexible and reliable representation of diverse surface geometries.
  • Embedding structured surface-level coherence throughout the entire sparse voxel reconstruction process, from initial voxel classification to final floater suppression, leads to improved geometric fidelity.

Summary by alphaXiv.

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