Research map / 3D vision / NeRF and radiance fields
DiffusioNeRF: Regularizing Neural Radiance Fields With Denoising Diffusion Models
Jamie Wynn, Daniyar Turmukhambetov
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 an RGBD diffusion prior to the geometric baseline (ablation) | Instant NGP NeRF · DTU all views | Mean Chamfer-L1: 1.36 -> 1.21 | Improved |
| Add an RGBD diffusion prior to the geometric baseline (ablation) | Instant NGP NeRF · LLFF 3-view | Average: 0.118 -> 0.127 | Hurt |
| Add foreground loss to photometric loss (ablation) | Instant NGP NeRF · DTU all views | Mean Chamfer-L1: 2.87 -> 1.71 | Improved |
| Add foreground loss to photometric loss (ablation) | Instant NGP NeRF · LLFF 3-view | Average: 0.210 -> 0.210 | No clear effect |
| Add frustum loss (ablation) | Photo + foreground loss NeRF · LLFF 3-view | Average: 0.210 -> 0.135 | Improved |
| Add density compactness loss (ablation) | Photo + foreground + frustum loss NeRF · LLFF 3-view | Average: 0.135 -> 0.118 | Improved |
| Use 24x24 instead of 48x48 DDM patches (ablation) | DiffusioNeRF · LLFF 3-view | Average: 0.127 -> 0.126 | Improved |
| Use 24x24 instead of 48x48 DDM patches (ablation) | DiffusioNeRF · DTU 3-view | Average: 0.135 -> 0.195 | Hurt |
| Increase DDM weight to 4x (ablation) | 24x24-patch DiffusioNeRF · DTU 3-view | Average: 0.195 -> 0.260 | Hurt |
| Do not provide training-image patches to the DDM 25% of the time (ablation) | DiffusioNeRF · DTU all views | Mean Chamfer-L1: 1.21 -> 1.91 | Hurt |
| Train the DDM on only 20% of Hypersim scenes (ablation) | DiffusioNeRF · DTU all views | Mean Chamfer-L1: 1.21 -> 1.65 | Hurt |
| Use an RGB-only DDM prior instead of an RGBD prior (ablation) | DiffusioNeRF · DTU all views | Mean Chamfer-L1: 1.21 -> 1.31 | Hurt |
Later papers that compared with this method
4 later accepted papers report DiffusioNeRF as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Sparse auto-regressive modeling for scene generation from multi-view images (ECCV 2026 · better on 3 of 3 shared results)
- ExtraNeRF: Visibility-Aware View Extrapolation of Neural Radiance Fields with Diffusion Models (CVPR 2024 · better on 1 of 1 shared result)
- ReconFusion: 3D Reconstruction with Diffusion Priors (CVPR 2024 · better on 1 of 1 shared result)
- Diffusion-Generated Pseudo-Observations for High-Quality Sparse-View Reconstruction (ECCV 2024 · 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
- RegNeRF: Regularizing Neural Radiance Fields for View Synthesis From Sparse Inputs (CVPR 2022 · similarity 0.94)
- Taming Latent Diffusion Model for Neural Radiance Field Inpainting (ECCV 2024 · similarity 0.93)
- Reducing Shape-Radiance Ambiguity in Radiance Fields with a Closed-Form Color Estimation Method (NeurIPS 2023 · similarity 0.92)
- ExtraNeRF: Visibility-Aware View Extrapolation of Neural Radiance Fields with Diffusion Models (CVPR 2024 · similarity 0.90)
- Bright-NeRF: Brightening Neural Radiance Field with Color Restoration from Low-Light RAW Images (AAAI 2025 · similarity 0.90)
- Optimize the Unseen - Fast NeRF Cleanup with Free Space Prior (NeurIPS 2025 · 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.
