Research map / 3D vision / Gaussian splatting
Relightable 3D Gaussians: Realistic Point Cloud Relighting with BRDF Decomposition and Ray Tracing
Jian Gao, chun gu, Youtian Lin, Zhihao Li, Hao Zhu, Xun Cao, Li Zhang, Yao Yao
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 normal gradient densification (NGD) (ablation) | R3DG · NeRF synthetic Lego (qualitative) | normal map detail | Improved |
| use full light model (global environment map plus per-Gaussian indirect SH light with visibility) instead of only a global environment map (ablation) | R3DG · relighting (qualitative) | shadow effects | Improved |
| add depth distribution constraint L_u (ablation) | R3DG · NeRF synthetic (qualitative) | depth uncertainty and ambient occlusion plausibility | Improved |
| make 3D Gaussians relightable with PBR instead of plain 3DGS color | 3DGS · NeRF synthetic | PSNR: 33.88 -> 31.22 | Hurt |
| use per-Gaussian PBR with BVH ray-traced visibility precomputation instead of neural-field inverse rendering | R3DG · Synthetic4Relight | training time (hours): 3.24 -> 0.90 | Improved |
| R3DG versus best prior relighting method | R3DG · Synthetic4Relight (relighting) | PSNR: 29.69 -> 31.00 | Improved |
| R3DG versus TensorIR on albedo | R3DG · Synthetic4Relight (albedo) | PSNR: 30.58 -> 28.31 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- NeILF++ — NeILF: Neural Incident Light Field for Physically-Based Material Estimation (ECCV 2022) · better on 1 of 1 shared result
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
- Differentiable Point-based Inverse Rendering (CVPR 2024 · similarity 0.89)
- MetaGS: A Meta-Learned Gaussian-Phong Model for Out-of-Distribution 3D Scene Relighting (NeurIPS 2025 · similarity 0.89)
- RNG: Relightable Neural Gaussians (CVPR 2025 · similarity 0.88)
- GI-GS: Global Illumination Decomposition on Gaussian Splatting for Inverse Rendering (ICLR 2025 · similarity 0.88)
- Subsurface Scattering for Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- SVG-IR: Spatially-Varying Gaussian Splatting for Inverse Rendering (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.
