Reproduce “GTSR: Subsurface Scattering Awared 3D Gaussians for Translucent Surface 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

Youwen Yuan, Xi Zhao

The GTSR pipeline from Xi’an Jiaotong University provides a 3D Gaussian Splatting-based method for reconstructing translucent object surfaces and synthesizing novel views. It achieves state-of-the-art surface geometry accuracy and improved novel view synthesis quality by separating surface and interior light transport and integrating physically-based rendering, completing training in approximately 2.5 hours.

Key insights

  • Decoupling the representation of surface geometry and interior scattering effects with dual Gaussian sets is crucial for simultaneously achieving accurate surface reconstruction and high-fidelity translucent appearance.
  • Integrating physically plausible light transport models, like the Fresnel term, within the 3DGS framework improves rendering realism and robustly guides geometric optimization for translucent surfaces.
  • A PBR module, implemented via deferred rendering, effectively provides pixel-level geometric constraints, enabling accurate decomposition of geometry and material properties even when traditional photometric consistency is invalid for translucent objects.

Summary by alphaXiv.

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