Research map / 3D vision / Gaussian splatting
RefracGS: Novel View Synthesis Through Refractive Water Surfaces with 3D Gaussian Ray Tracing
Yiming Shao, Qiyu Dai, Chong Gao, Guanbin Li, Yequan Wang, He Sun, Qiong Zeng, Baoquan Chen, Wenzheng Chen
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 |
|---|---|---|---|
| add the explicit water height map instead of NeRFrac's refractive field (ablation) | RefracGS · NeRFrac real+synthetic ablation | PSNR: 27.782 -> 32.876 | Improved |
| add the explicit water height map instead of NeRFrac's refractive field (ablation) | RefracGS · NeRFrac real+synthetic ablation | training time: 22 min -> 11 min | Improved |
| add the explicit water height map instead of NeRFrac's refractive field (ablation) | RefracGS · NeRFrac real+synthetic ablation | FPS: 57 -> 223 | Improved |
| add normal smoothing (ablation) | RefracGS · NeRFrac real+synthetic ablation | PSNR: 32.643 -> 32.876 | Improved |
| add recursive subdivision intersection (ablation) | RefracGS · NeRFrac real+synthetic ablation | PSNR: 32.966 -> 32.876 | Hurt |
| add recursive subdivision intersection (ablation) | RefracGS · NeRFrac real+synthetic ablation | training time: 20 min -> 11 min | Improved |
| add opacity loss (ablation) | RefracGS · NeRFrac real+synthetic ablation | PSNR: 31.927 -> 32.876 | Improved |
| add opacity loss (ablation) | RefracGS · NeRFrac real+synthetic ablation | LPIPS: 0.125 -> 0.132 | Hurt |
| use RefracGS instead of NeRFrac on the strong-refraction dataset | RefracGS · RefracGS dataset | PSNR: 17.837 -> 30.224 | Improved |
| use RefracGS instead of NeRFrac for surface reconstruction | RefracGS · RefracGS dataset | surface RMSE (cm): 3.655 -> 0.115 | Improved |
| use RefracGS instead of NeRFrac on real NeRFrac scenes | RefracGS · NeRFrac Real | PSNR: 28.146 -> 30.671 | Improved |
| use RefracGS instead of NeRFrac for water removal | RefracGS · RefracGS dataset water removal | PSNR: 14.950 -> 23.151 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- NeRFrac — NeRFrac: Neural Radiance Fields through Refractive Surface (ICCV 2023) · better on 5 of 5 shared results
- Plenoxels — Plenoxels: Radiance Fields Without Neural Networks (CVPR 2022) · better on 3 of 3 shared results
- Mip-NeRF — Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields (ICCV 2021) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference (NeurIPS 2025 · similarity 0.87)
- Differentiable Neural Surface Refinement for Modeling Transparent Objects (CVPR 2024 · similarity 0.87)
- Subsurface Scattering for Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- Reflective Gaussian Splatting (ICLR 2025 · similarity 0.87)
- DySurface: Consistent 4D Surface Reconstruction via Bridging Explicit Gaussians and Implicit Functions (NeurIPS 2026 · similarity 0.87)
- SpecTRe-GS: Modeling Highly Specular Surfaces with Reflected Nearby Objects by Tracing Rays in 3D Gaussian Splatting (CVPR 2025 · similarity 0.87)
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.
