Research map / 3D vision / Depth and stereo
RENI++: A Rotation-Equivariant, Scale-Invariant, Natural Illumination Prior
James Gardner, Bernhard Egger, William A. P. Smith
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 scale invariance (ablation) | RENI with transformer decoder · Env map fitting D=300 | LDR PSNR (dB): 16.90 -> 22.10 | Improved |
| add scale invariance (ablation) | RENI with transformer decoder · Env map fitting D=27 | LDR PSNR (dB): 13.05 -> 18.02 | Improved |
| add data augmentation (ablation) | RENI (50k steps) · Env map fitting D=300 | LDR PSNR (dB): 16.53 -> 15.80 | No clear effect |
| replace SIREN decoder with transformer decoder (ablation) | RENI (50k steps) · Env map fitting D=300 | LDR PSNR (dB): 16.53 -> 16.90 | No clear effect |
| use SO(2) equivariance instead of no equivariance (ablation) | RENI++ · Env map fitting D=300 | mean PSNR (dB): 21.41 -> 22.10 | Improved |
| use SO(2) instead of SO(3) equivariance (ablation) | RENI++ · Env map fitting D=300 | mean PSNR (dB): 21.12 -> 22.10 | Improved |
| use SO(3) instead of SO(2) equivariance (ablation) | RENI++ · Env map fitting D=27 | mean PSNR (dB): 18.02 -> 18.09 | No clear effect |
| replace SH with RENI++ | SH · Env map fitting D=300 | LDR PSNR (dB): 19.15 -> 22.10 | Improved |
| replace SG with RENI++ | SG · Env map fitting D=300 | LDR PSNR (dB): 20.24 -> 22.10 | Improved |
| replace original RENI with RENI++ | RENI · Env map fitting D=300 | LDR PSNR (dB): 20.47 -> 22.10 | Improved |
| replace original RENI with RENI++ | RENI · Env map fitting D=27 | SSIM: 0.40 -> 0.39 | Hurt |
| rotate input by 90 degrees instead of 5 degrees | RENI++ · latent code rotation consistency | relative error: 0.582 -> 0.049 | Improved |
Closest papers by meaning
- Rotation-Equivariant Conditional Spherical Neural Fields for Learning a Natural Illumination Prior (NeurIPS 2022 · similarity 0.96)
- VENI: Variational Encoder for Natural Illumination (CVPR 2026 · similarity 0.91)
- Rendering-Aware HDR Environment Map Prediction from a Single Image (AAAI 2022 · similarity 0.89)
- Invertible Neural BRDF for Object Inverse Rendering (ECCV 2020 · similarity 0.88)
- Equivariant Neural Rendering (ICML 2020 · similarity 0.88)
- RobIR: Robust Inverse Rendering for High-Illumination Scenes (NeurIPS 2024 · 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.
