Confidence-Based Mesh Extraction from 3D Gaussians

Abstract summary

CoMe extracts surfaces from Gaussian scene representations using learned confidence values to balance geometric and image-based supervision. Additional regularizers constrain primitive colour and normal variation, while changes to the appearance loss improve consistency. The method focuses on efficient reconstruction in large, unbounded scenes, including cases where view-dependent effects make geometry ambiguous.

Abstract on arXiv

arXiv PDF DOI Website GitHub

Cite this work

@misc{confidencebasedmeshextractionfrom3dgaussians,
  title = {Confidence-Based Mesh Extraction from 3D Gaussians},
  author = {Lukas Radl and Felix Windisch and Andreas Kurz and Thomas Köhler and Michael Steiner and Markus Steinberger},
  year = {2026},
  note = {ECCV},
  url = {https://doi.org/10.1007/978-3-032-37435-6_17}
}

Download