Research map / 3D vision / NeRF and radiance fields
NerfDiff: Single-image View Synthesis with NeRF-guided Distillation from 3D-aware Diffusion
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| NeRF-guided distillation (NGD) fine-tuning at test time (ablation) | NerfDiff-B · ShapeNet Cars | FID: 42.37 -> 18.09 | Helped |
| NeRF-guided distillation (NGD) fine-tuning at test time (ablation) | NerfDiff-B · ShapeNet Cars | LPIPS: 0.093 -> 0.082 | Helped |
| NeRF-guided distillation (NGD) fine-tuning at test time (ablation) | NerfDiff-B · ShapeNet Cars | PSNR: 23.81 -> 23.51 | Hurt |
| NeRF-guided distillation (NGD) fine-tuning at test time (ablation) | NerfDiff-B · ShapeNet Chairs and ABO | FID: 15.72 -> 5.65 (Chairs); 18.01 -> 7.77 (ABO) | Helped |
| Naive direct distillation from CDM virtual views instead of NGD (ablation) | NerfDiff fine-tuning · ShapeNet Cars | PSNR: 23.51 (NGD) -> 23.46 (direct) | Hurt |
| Score distillation sampling (SDS) used for fine-tuning instead of NGD (ablation) | NerfDiff fine-tuning · ShapeNet Cars | PSNR: 23.51 -> 20.32 | Hurt |
| Number of virtual views reduced to 5 (ablation) | NGD fine-tuning · ShapeNet Cars | PSNR: 23.51 (50 views) -> 22.86 | Hurt |
| 3D-aware CDM conditioning (rendered target-view from NeRF) vs concatenating input view (ablation) | CDM · ShapeNet Cars | PSNR: 20.72 (concat) -> 23.51 | Helped |
| Image encoder changed to UNet with camera-aligned triplanes (vs pixelNeRF-style projection) | NerfDiff single-image NeRF · ShapeNet Cars | PSNR: 23.17 (pixelNeRF) -> 23.81 (w/o NGD) | Helped |
| Scaling NerfDiff from base to large | NerfDiff · ShapeNet Cars | FID: 18.09 -> 15.49 | Helped |
| Using 3DiM geometry-free model instead of NerfDiff on ShapeNet Cars | geometry-free baseline · ShapeNet Cars | FID: 3DiM 8.99 vs NerfDiff-B 18.09 | Mixed |
Closest papers by meaning
- NeRF-VPT: Learning Novel View Representations with Neural Radiance Fields via View Prompt Tuning (AAAI 2024 · similarity 0.91)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.90)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.90)
- SIGNeRF: Scene Integrated Generation for Neural Radiance Fields (CVPR 2024 · similarity 0.90)
- NerfBaselines: Consistent and Reproducible Evaluation of Novel View Synthesis Methods (NeurIPS 2025 · similarity 0.90)
- CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis (AAAI 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.