Research map / 3D vision / Gaussian splatting
IR-HGP: Physically-Aware Gaussian Inverse Rendering for High-Illumination Scenes via Generative Priors
Qingan Zhang, Wensheng Li, Chengying Gao
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 2D Gaussian surfels in HVD (ablation) | IR-HGP · NeRF Synthetic: Ficus | PSNR: 36.14 -> 36.86 | Improved |
| add mesh ray-traced visibility in HVD (ablation) | IR-HGP · NeRF Synthetic: Ficus | PSNR: 35.58 -> 36.86 | Improved |
| add Generative Illumination Field Prior (ablation) | IR-HGP · NeRF Synthetic: Ficus | PSNR: 34.77 -> 36.86 | Improved |
| add Physics-Aware Radiance Correction (ablation) | IR-HGP · NeRF Synthetic: Ficus | PSNR: 35.15 -> 36.86 | Improved |
| add Generative Illumination Field Prior (ablation) | IR-HGP · Shiny Blender: Car | PSNR: 32.95 -> 34.63 | Improved |
| add mesh visibility (ablation) | IR-HGP · Shiny Blender: Car | PSNR: 33.78 -> 34.63 | Improved |
| add PARC (ablation) | IR-HGP · Shiny Blender: Car | PSNR: 33.43 -> 34.63 | Improved |
| learnable exposure beta in ACES mapping instead of fixed ACES (ablation) | PARC · albedo visual quality | albedo cleanliness (qualitative) | Improved |
| IR-HGP versus DiscretizedSDF on mean relighting | IR-HGP · RelightObj mean | PSNR: 32.12 -> 33.61 | Improved |
| IR-HGP versus DiscretizedSDF on Chair | IR-HGP · NeRF Synthetic: Chair | PSNR: 37.54 -> 35.10 | Hurt |
| include GIFP and mesh extraction in training | IR-HGP · RelightObj | training time: 0.6h (GS-IR) vs 1.5h | Hurt |
| IR-HGP on real-world Mip-NeRF 360 versus the best PSNR baseline | IR-HGP · Mip-NeRF 360 | PSNR: PSNR 26.92, SSIM 0.807, LPIPS 0.196 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GS-IR — GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024) · better on 1 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
- SVG-IR: Spatially-Varying Gaussian Splatting for Inverse Rendering (CVPR 2025 · similarity 0.90)
- GI-GS: Global Illumination Decomposition on Gaussian Splatting for Inverse Rendering (ICLR 2025 · similarity 0.90)
- GS-IR: 3D Gaussian Splatting for Inverse Rendering (CVPR 2024 · similarity 0.89)
- MaterialRefGS: Reflective Gaussian Splatting with Multi-view Consistent Material Inference (NeurIPS 2025 · similarity 0.89)
- Inverse Rendering using Multi-Bounce Path Tracing and Reservoir Sampling (ICLR 2025 · similarity 0.89)
- RobIR: Robust Inverse Rendering for High-Illumination Scenes (NeurIPS 2024 · similarity 0.89)
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.
