Research map / 3D vision / 3D generation
DreamSparse: Escaping from Plato’s Cave with 2D Diffusion Model Given Sparse Views
Paul Yoo, Jiaxian Guo, Yutaka Matsuo, Shixiang Shane Gu
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 frozen Stable Diffusion prior with spatial guidance on top of geometry module (training categories) (ablation) | DreamSparse · CO3D training categories avg | FID: 191.21 -> 81.8 | Improved |
| add diffusion prior (training categories) (ablation) | DreamSparse · CO3D training categories avg | LPIPS: 0.33 -> 0.21 | Improved |
| add diffusion prior (training categories) (ablation) | DreamSparse · CO3D training categories avg | PSNR: 21.95 -> 22.03 | Mixed |
| add diffusion prior (open-set categories) (ablation) | DreamSparse · CO3D open-set avg | FID: 219.64 -> 122.2 | Improved |
| add diffusion prior (open-set categories) (ablation) | DreamSparse · CO3D open-set avg | PSNR: 20.66 -> 20.19 | Hurt |
| add diffusion prior (scene-level, 5 views) (ablation) | DreamSparse · CO3D hydrant scene-level | FID: 216.34 -> 70.62 | Improved |
| add diffusion prior (scene-level, 5 views) (ablation) | DreamSparse · CO3D hydrant scene-level | PSNR: 17.16 -> 14.15 | Hurt |
| DreamSparse vs SparseFusion on training categories | DreamSparse · CO3D training categories avg | FID: 172.6 -> 81.8 | Improved |
| DreamSparse vs SparseFusion on open-set categories | DreamSparse · CO3D open-set avg | FID: 212.9 -> 122.2 | Improved |
| increase context views from 1 to 5 | DreamSparse · CO3D hydrant | FID: 75.63 -> 70.62 | Improved |
| increase context views from 1 to 5 | SparseFusion · CO3D hydrant | FID: 272.72 -> 231.73 | Improved |
| spatial guidance weight lambda=0 (no guidance) (ablation) | DreamSparse · qualitative | consistency | Hurt |
Closest papers by meaning
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.89)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.89)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.89)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.88)
- DreamComposer: Controllable 3D Object Generation via Multi-View Conditions (CVPR 2024 · similarity 0.88)
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis (NeurIPS 2026 · similarity 0.87)
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.
