Research map / 3D vision / 3D generation
Novel View Synthesis with Diffusion Models
Daniel Watson, William Chan, Ricardo Martin Brualla, Jonathan Ho, Andrea Tagliasacchi, Mohammad Norouzi
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 |
|---|---|---|---|
| Use stochastic conditioning rather than a fixed conditioning view during denoising (ablation) | 3DiM · SRN cars 3D consistency scoring | PSNR: 25.78 -> 28.53 | Improved |
| Use stochastic conditioning rather than a fixed conditioning view during denoising (ablation) | 3DiM · SRN cars 3D consistency scoring | FID: 30.51 -> 22.09 | Improved |
| Use stochastic conditioning rather than a fixed conditioning view during denoising (ablation) | 3DiM · SRN chairs* 3D consistency scoring | PSNR: 17.61 -> 18.84 | Improved |
| Use stochastic conditioning rather than a fixed conditioning view during denoising (ablation) | 3DiM · SRN chairs* 3D consistency scoring | FID: 116.16 -> 98.78 | Improved |
| Remove stochastic conditioning (ablation) | 3DiM · SRN cars | FID: 8.99 -> 1.87 | Improved |
| Remove stochastic conditioning (ablation) | 3DiM · SRN chairs* | FID: 3.30 -> 1.05 | Improved |
| Replace diffusion sampling with one-step regression (ablation) | 3DiM · SRN cars | FID: 8.99 -> 17.45 | Hurt |
| Replace diffusion sampling with one-step regression (ablation) | 3DiM · SRN cars | PSNR: 21.01 -> 22.55 | Improved |
| Replace X-UNet with channel-concatenation UNet (ablation) | 3DiM · SRN cars | FID: 8.99 -> 21.54 | Hurt |
| Replace X-UNet with channel-concatenation UNet (ablation) | 3DiM · SRN chairs* | FID: 3.30 -> 5.15 | Hurt |
| Increase model capacity from approximately 471M to approximately 1.3B parameters (ablation) | 3DiM · SRN cars 3D consistency scoring | FID: 22.09 -> 29.55 | Hurt |
| Increase model capacity from approximately 471M to approximately 1.3B parameters (ablation) | 3DiM · SRN chairs* 3D consistency scoring | FID: 98.78 -> 58.61 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- CodeNeRF — CodeNeRF: Disentangled Neural Radiance Fields for Object Categories (ICCV 2021) · better on 0 of 2 shared results
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
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.94)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.92)
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis (NeurIPS 2026 · similarity 0.91)
- Denoising Diffusion via Image-Based Rendering (ICLR 2024 · similarity 0.91)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.90)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (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.
