Research map / 3D vision / NeRF and radiance fields
Efficient NeRF Optimization - Not All Samples Remain Equally Hard
Juuso Korhonen, Goutham Rangu, Hamed Rezazadegan Tavakoli, Juho Kannala
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 |
|---|---|---|---|
| mine hard samples with an inference-mode first forward pass and backpropagate only b hard samples | Instant-NGP (PyTorch) · Mip-NeRF-360 | validation PSNR per training time: ~1 dB gain per same training time; ~2x faster to a given PSNR | Improved |
| build the computational graph only for hard samples | Instant-NGP (PyTorch) · Mip-NeRF-360 | network memory: ~1 GB -> ~570 MB | Improved |
| train on hard sample minibatch (b/B about 10-15%) | Instant-NGP (PyTorch) · Mip-NeRF-360 | iteration time: down to ~53-54% of baseline | Improved |
| hard sample mining versus baseline per iteration | Instant-NGP (PyTorch) · Mip-NeRF-360 | training loss per iteration | No clear effect |
| apply hard sample mining on scenes with uniformly hard content (room, garden, stump) | Instant-NGP (PyTorch) · Mip-NeRF-360 | PSNR gain: stump 8 min 23.4 vs 23.4; bonsai 29.7 vs 28.4 | Mixed |
| train longer (20 min) with hard sample mining | Instant-NGP (PyTorch) · Mip-NeRF-360 stump | validation PSNR: 23.28 vs 23.46 baseline | Hurt |
| apply hard sample mining to Nerfacto with proposal sampling | Nerfacto · Mip-NeRF-360 | PSNR per training time and memory: memory 300 MB -> 160 MB; PSNR gains smaller | Mixed |
Closest papers by meaning
- EfficientNeRF Efficient Neural Radiance Fields (CVPR 2022 · similarity 0.91)
- NerfAcc: Efficient Sampling Accelerates NeRFs (ICCV 2023 · similarity 0.91)
- NeRFLight: Fast and Light Neural Radiance Fields Using a Shared Feature Grid (CVPR 2023 · similarity 0.90)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.90)
- ProvNeRF: Modeling per Point Provenance in NeRFs as a Stochastic Field (NeurIPS 2024 · similarity 0.89)
- Neural NeRF Compression (ICML 2024 · similarity 0.89)
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.
