Research map / 3D vision / NeRF and radiance fields
R2L: Distilling Neural Radiance Field to Neural Light Field for Efficient Novel View Synthesis
Huan Wang, Jian Ren, Zeng Huang, Kyle Olszewski, Menglei Chai, Yun Fu, Sergey Tulyakov
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 |
|---|---|---|---|
| train the NeRF network on 10k pseudo images instead of 0.1k original images (ablation) | NeRF network as light field · Lego | test PSNR: 19.81 -> 26.67 | Improved |
| use the deep residual R2L network instead of the 11-layer NeRF network at 10k pseudo images (ablation) | R2L W181D88 · Lego | test PSNR: 26.67 -> 29.50 | Improved |
| add original real images to pseudo data (ablation) | R2L W181D88 · Lego | test PSNR: 29.50 -> 30.09 | Improved |
| add original real images to pseudo data (ablation) | R2L W256D88 · NeRF synthetic | PSNR: 30.48 -> 31.87 | Improved |
| add original real images to pseudo data (ablation) | R2L W256D88 · DONeRF synthetic | PSNR: 32.67 -> 35.45 | Improved |
| add original real images to pseudo data (ablation) | R2L W256D88 · NeRF real-world | PSNR: 27.58 -> 27.79 | Improved |
| add original real images to pseudo data | R2L W256D88 · NeRF real-world | LPIPS: 0.0733 -> 0.0968 | Hurt |
| add dropout to NeRF network on original data (ablation) | NeRF network · Lego | test PSNR: 19.81 -> 19.83 | No clear effect |
| add batch norm to NeRF network on original data (ablation) | NeRF network · Lego | test PSNR: 19.81 -> 19.76 | No clear effect |
| add residual connections to the deep MLP (ablation) | R2L · Lego | test PSNR | Improved |
| increase pseudo sample size beyond 5k images (ablation) | R2L W181D88 · Lego | test PSNR | No clear effect |
| add hard example mining (r > 0) to training batches (ablation) | R2L W181D88 · Lego | train and test PSNR | 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.
- KiloNeRF — KiloNeRF: Speeding Up Neural Radiance Fields With Thousands of Tiny MLPs (ICCV 2021) · 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
- Efficient View Synthesis with Neural Radiance Distribution Field (ICCV 2023 · similarity 0.96)
- NeRFLight: Fast and Light Neural Radiance Fields Using a Shared Feature Grid (CVPR 2023 · similarity 0.94)
- Neural Radiance Fields From Sparse RGB-D Images for High-Quality View Synthesis (TPAMI 2022 · similarity 0.94)
- NeRF-VPT: Learning Novel View Representations with Neural Radiance Fields via View Prompt Tuning (AAAI 2024 · similarity 0.92)
- A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields (AAAI 2024 · similarity 0.91)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.91)
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.
