Research map / 3D vision / Gaussian splatting
GlassGS: Geometry and Concept-Aware 3D Gaussian Splatting for Reflective Enclosures
Mingxuan Cui, Yunrui Zhu, Wuqi Wang, Di Lin, Jianhua Zhang, Jie Zhang, Ming-Ming Cheng, Shengyong Chen, Qing Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use GlassGS instead of vanilla 3DGS | 3DGS · GlassFlash3D | Mask MAE: 32.59 -> 24.68 | Improved |
| use GlassGS instead of vanilla 3DGS | 3DGS · GlassFlash3D | Mask PSNR: 17.17 -> 18.94 | Improved |
| use GlassGS instead of SpotLessSplats | 3DGS · GlassFlash3D | Mask MAE: 30.82 -> 24.68 | Improved |
| 2D reflection removal preprocessing (DSRNet, YTMT) before 3DGS | 3DGS · GlassFlash3D | Mask PSNR: 17.17 -> 16.91 (DSRNet), 17.10 (YTMT) | No clear effect |
| add geometry-consistency perception (ablation) | GlassGS · GlassFlash3D | Mask MAE: 28.22 -> 24.68 | Improved |
| add reflection-free concept learning (ablation) | GlassGS · GlassFlash3D | Mask MAE: 30.31 -> 24.68 | Improved |
| use Cons weighting in 3DGS loss (Eq. 2) (ablation) | GlassGS · GlassFlash3D | MAE: 36.73 -> 31.99 | Improved |
| use Cons weighting in textual inversion loss (Eq. 3) (ablation) | GlassGS · GlassFlash3D | Mask MAE: 24.95 -> 24.68 | Improved |
| add x0 loss in ControlNet training (ablation) | GlassGS · GlassFlash3D | Mask MAE: 28.81 -> 24.68 | Improved |
| iterate self-cycling refinement (ablation) | GlassGS · GlassFlash3D (one scene) | MAE/SSIM | Improved |
Closest papers by meaning
- SpecTRe-GS: Modeling Highly Specular Surfaces with Reflected Nearby Objects by Tracing Rays in 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- PlanarGS: High-Fidelity Indoor 3D Gaussian Splatting Guided by Vision-Language Planar Priors (NeurIPS 2025 · similarity 0.88)
- IndoorGS: Geometric Cues Guided Gaussian Splatting for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.88)
- GS-2DGS: Geometrically Supervised 2DGS for Reflective Object Reconstruction (CVPR 2025 · similarity 0.88)
- MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference (NeurIPS 2025 · similarity 0.88)
- CoCoGaussian: Leveraging Circle of Confusion for Gaussian Splatting from Defocused Images (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
