Research map / 3D vision / NeRF and radiance fields
FastNeRF: High-Fidelity Neural Rendering at 200FPS
Stephan J. Garbin, Marek Kowalski, Matthew Johnson, Jamie Shotton, Julien Valentin
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 |
|---|---|---|---|
| cache the network outputs to a grid (1024^3) (ablation) | FastNeRF · NeRF Synthetic 360 | PSNR (dB): 29.155 -> 29.97 | Improved |
| cache the network outputs to a grid (768^3) (ablation) | FastNeRF · LLFF | PSNR (dB): 27.958 -> 26.035 | Hurt |
| cache the network outputs to a grid (1024^3) (ablation) | FastNeRF · Chair 800x800 | render time (ms): 28.2K -> 2.7 | Improved |
| use FastNeRF with cache instead of NeRF | NeRF · Chair 800x800 | render time (ms): 17.5K -> 2.7 | Improved |
| increase components from 8 to 16 (ablation) | FastNeRF · Ship scene, cache 768^3 | PSNR (dB): 27.72 -> 28.3 | Improved |
| increase components from 4 to 8 (ablation) | FastNeRF · Ship scene, cache 768^3 | PSNR (dB): 27.54 -> 27.72 | Improved |
| increase components from 6 to 8 (ablation) | FastNeRF · Horns scene, cache 768^3 | PSNR (dB): 26.85 -> 26.76 | No clear effect |
| increase cache resolution from 256^3 to 768^3 (ablation) | FastNeRF · Ship scene, 8 components | PSNR (dB): 24.89 -> 27.72 | Improved |
| use Fourier encoding L=1 instead of L=3 for view directions (ablation) | FastNeRF · NeRF Synthetic average | PSNR (dB): 29.449 -> 29.748 | Improved |
| use a smaller direction-dependent cache (ablation) | FastNeRF · Lego scene (32^3 direction cache) | PSNR (dB): 32.275 -> 32.155 | No clear effect |
| use a smaller direction-dependent cache (ablation) | FastNeRF · Ficus scene (32^3 direction cache) | PSNR (dB): 27.792 -> 28.193 | Improved |
| add grid-based sample pruner to the deformation network (ablation) | FastNeRF telepresence · Face rendering | runtime of deformation network: more than a factor of 2 | Improved |
Closest papers by meaning
- SteerNeRF: Accelerating NeRF Rendering via Smooth Viewpoint Trajectory (CVPR 2023 · similarity 0.92)
- Learning Neural Duplex Radiance Fields for Real-Time View Synthesis (CVPR 2023 · similarity 0.92)
- CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis (AAAI 2025 · similarity 0.91)
- A General Implicit Framework for Fast NeRF Composition and Rendering (AAAI 2024 · similarity 0.90)
- Ced-NeRF: A Compact and Efficient Method for Dynamic Neural Radiance Fields (AAAI 2024 · similarity 0.90)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.90)
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.
