Research map / 3D vision / Face and shape
Rotation-Equivariant Conditional Spherical Neural Fields for Learning a Natural Illumination Prior
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a learned RENI natural-illumination prior instead of SH | Inverse-rendering illumination fitting · Inverse rendering, increasing Blinn-Phong specularity | render reconstruction performance | Helped |
| Use RENI instead of spherical harmonics | Natural illumination fitting · Natural illumination test set, D=27, 108, 147, 300 | mean PSNR: 17.02 vs 12.80; 19.58 vs 15.89; 19.97 vs 16.44; 20.47 vs 17.07 | Helped |
| Use RENI instead of SG | Natural illumination fitting · Natural illumination test set, D=27, 108, 147 | mean PSNR: 17.02 vs 16.04; 19.58 vs 18.44; 19.97 vs 19.26 | Helped |
| Add built-in equivariance rather than relying on rotation augmentation (ablation) | RENI · Natural illumination test set, D=27, 108, 147 | mean PSNR: 11.32 -> 17.02; 15.85 -> 19.58; 14.64 -> 19.97 | Helped |
| Restrict equivariance to SO(2) instead of using full SO(3) (ablation) | RENI · Natural illumination test set, D=27, 108, 147 | mean PSNR: 14.00 -> 17.02; 18.27 -> 19.58; 17.45 -> 19.97 | Helped |
| Use 5 SIREN layers instead of 3 (ablation) | RENI · Natural illumination test set, D=27, 108, 147 | mean PSNR: 16.25 -> 17.02; 18.29 -> 19.58; 18.57 -> 19.97 | Helped |
| Use 5 SIREN layers instead of 7 (ablation) | RENI · Natural illumination test set, D=27, 108, 147 | mean PSNR: 16.38 -> 17.02; 18.13 -> 19.58; 18.15 -> 19.97 | Helped |
| Increase latent dimension from D=27 to D=108, D=147, or D=300 | RENI · Natural illumination test set | mean PSNR: 17.02 -> 19.58 -> 19.97 -> 20.47 | Helped |
Closest papers by meaning
- VENI: Variational Encoder for Natural Illumination (CVPR 2026 · similarity 0.92)
- Grounding Continuous Representations in Geometry: Equivariant Neural Fields (ICLR 2025 · similarity 0.86)
- BerfScene: Bev-conditioned Equivariant Radiance Fields for Infinite 3D Scene Generation (CVPR 2024 · similarity 0.86)
- Few-Shot Neural Radiance Fields under Unconstrained Illumination (AAAI 2024 · similarity 0.85)
- GoLF-NRT: Integrating Global Context and Local Geometry for Few-Shot View Synthesis (CVPR 2025 · similarity 0.85)
- NeRFPrior: Learning Neural Radiance Field as a Prior for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.85)
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.