Research map / 3D vision / NeRF and radiance fields
F2-NeRF: Fast Neural Radiance Field Training With Free Camera Trajectories
Peng Wang, Yuan Liu, Zhaoxi Chen, Lingjie Liu, Ziwei Liu, Taku Komura, Christian Theobalt, Wenping Wang
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 |
|---|---|---|---|
| Use perspective warping instead of inverse-sphere warping (ablation) | F²-NeRF with exponential sampling · Free dataset, Pillar | PSNR: 27.86 -> 28.65 | Improved |
| Use perspective sampling instead of exponential sampling (ablation) | F²-NeRF with perspective warping · Free dataset, Pillar | PSNR: 28.65 -> 28.76 | Improved |
| Use inverse-sphere warping instead of no warping (ablation) | F²-NeRF with disparity sampling · Free dataset, Pillar | PSNR: 26.81 -> 27.09 | Improved |
| Replace disparity sampling with exponential sampling (ablation) | F²-NeRF with inverse-sphere warping · Free dataset, Pillar | PSNR: 27.09 -> 27.86 | Improved |
| Increase training from 20k to 50k steps (ablation) | F²-NeRF · Free dataset | PSNR: 26.32 -> 26.85 | Improved |
| Increase hash table size from 2^19 to 2^20 (ablation) | F²-NeRF trained for 50k steps · Free dataset | PSNR: 26.85 -> 27.19 | Improved |
| Train for 50k instead of 20k steps (ablation) | Instant-NGP · Free dataset | PSNR: 24.43 -> 25.07 | Improved |
| Use perspective warping instead of NDC warping (ablation) | MLP-based NeRF · LLFF | PSNR: 26.31 -> 26.29 | No clear effect |
| Use perspective warping instead of no warping (ablation) | MLP-based NeRF · Lego view extrapolation | PSNR: 37.15 -> 37.63; 30.35 -> 30.19; 27.03 -> 27.35 | Mixed |
| Use a single global hash table with multiple hash functions rather than separate hash tables per octree node (ablation) | F²-NeRF · Free dataset | rendering quality | Improved |
| Use disparity sampling rather than exponential sampling (ablation) | F²-NeRF with inverse-sphere warping · Free dataset, Pillar | PSNR: 27.86 -> 27.09 | Hurt |
| Use perspective warping and sampling rather than inverse-sphere warping and exponential sampling (ablation) | F²-NeRF · Free dataset, Pillar | SSIM: 0.751 -> 0.798 | Improved |
Closest papers by meaning
- PyNeRF: Pyramidal Neural Radiance Fields (NeurIPS 2023 · similarity 0.91)
- CF-NeRF: Camera Parameter Free Neural Radiance Fields with Incremental Learning (AAAI 2024 · similarity 0.91)
- Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022 · similarity 0.90)
- Forward Flow for Novel View Synthesis of Dynamic Scenes (ICCV 2023 · similarity 0.89)
- Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video (CVPR 2025 · similarity 0.88)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.88)
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.
