Research map / 3D vision / Gaussian splatting
GS-ID: Illumination Decomposition on Gaussian Splatting via Adaptive Light Aggregation and Diffusion-Guided Material Priors
Kang Du, Zhihao Liang, Yulin Shen, Zeyu Wang
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 diffusion normal priors (ablation) | GS-ID · TensoIR Synthetic | NVS PSNR: 37.962 -> 39.130 | Improved |
| add diffusion material priors (ablation) | GS-ID · TensoIR Synthetic | Albedo PSNR: 29.585 -> 33.493 | Improved |
| add both normal and material priors (ablation) | GS-ID · TensoIR Synthetic | Albedo PSNR: 29.101 -> 33.493 | Improved |
| add spherical Gaussian mixtures for local lighting (ablation) | GS-ID · TensoIR Synthetic | NVS PSNR: 36.670 -> 39.130 | Improved |
| add learnable environment map (ablation) | GS-ID · TensoIR Synthetic | NVS PSNR: 38.180 -> 39.130 | Improved |
| add per-splat shadow field (ablation) | GS-ID · TensoIR Synthetic | Albedo PSNR: 30.280 -> 33.493 | Improved |
| use grid of 3^3 SGMs instead of a single SGM (ablation) | GS-ID · TensoIR Synthetic | NVS PSNR: 38.670 -> 39.130 | Improved |
| add SGMs to lighting model (ablation) | GS-ID · TensoIR Synthetic | Normal MAE: 5.916 -> 4.602 | Improved |
| remove environment map (ablation) | GS-ID · TensoIR Synthetic | Normal MAE: 4.701 -> 4.602 | Mixed |
| use deferred shading instead of forward shading (ablation) | GS-ID · Mip-NeRF 360 (garden) | training time: 299min -> 68min | Improved |
| use CUDA implementation for deferred shading (ablation) | GS-ID · Mip-NeRF 360 (garden) | GPU memory: 36.8G -> 21.4G | Improved |
| GS-ID instead of RelightGS | GS-ID · ADT dataset | Albedo PSNR: 27.223 -> 29.133 | Improved |
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.
- GS-IR — GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024) · better on 5 of 5 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
- GI-GS: Global Illumination Decomposition on Gaussian Splatting for Inverse Rendering (ICLR 2025 · similarity 0.93)
- GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024 · similarity 0.92)
- Ref-GS: Directional Factorization for 2D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference (NeurIPS 2025 · similarity 0.90)
- MetaGS: A Meta-Learned Gaussian-Phong Model for Out-of-Distribution 3D Scene Relighting (NeurIPS 2025 · similarity 0.90)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.90)
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.
