Research map / 3D vision / 3D generation
SmokeSVD: Smoke Reconstruction from A Single View via Progressive Novel View Synthesis and Refinement with Diffusion Models
Chen Li, Shanshan Dong, Sheng Qiu, Jianmin Han, Yibo Zhao, Zan Gao, Taku Komura, Kemeng Huang
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 velocity loss to SvDiff (ablation) | SmokeSVD · ScalarFlow input view | PSNR: 41.6814 -> 44.5518 | Improved |
| add divergence term to SvDiff (ablation) | SmokeSVD · ScalarFlow input view | PSNR: 40.9043 -> 44.5518 | Improved |
| add noise threshold for guidance (ablation) | SmokeSVD · ScalarFlow input view | PSNR: 41.8412 -> 44.5518 | Improved |
| add rendered density feedback (3D reconstruction) to SvDiff (ablation) | SmokeSVD · ScalarFlow input view | PSNR: 41.4763 -> 44.5518 | Improved |
| add gradient loss (ablation) | SmokeSVD · ScalarFlow input view | PSNR: 42.0804 -> 44.5518 | Improved |
| velocity loss effect on input-view LPIPS (ablation) | SmokeSVD · ScalarFlow input view | LPIPS: 0.0069 -> 0.0075 | Hurt |
| add novel view refinement (ablation) | SmokeSVD · ScalarFlow novel views (views 0,3 input) | PSNR/LPIPS: PSNR 18.7490 -> 18.7978; LPIPS 0.1808 -> 0.1757 | Mixed |
| use multi-stage progressive instead of single-pass refinement (ablation) | SmokeSVD · ScalarFlow novel views | LPIPS: LPIPS 0.1704 -> 0.1757; PSNR 18.7902 -> 18.7978 | Mixed |
| add residual loss in NvRef (ablation) | SmokeSVD · ScalarFlow novel views | PSNR: 18.5066 -> 18.7978 | Improved |
| SmokeSVD vs GlobTrans on speed | SmokeSVD · ScalarFlow | time for 120 steps: >30h -> 15mins | Improved |
| SmokeSVD vs FluidNexus on novel views | SmokeSVD · ScalarFlow novel views | PSNR: 25.3569 -> 23.4393 | Hurt |
Closest papers by meaning
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.85)
- DMV3D: Denoising Multi-view Diffusion Using 3D Large Reconstruction Model (ICLR 2024 · similarity 0.85)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.84)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.84)
- DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos (NeurIPS 2025 · similarity 0.84)
- Motion Matters: Compact Gaussian Streaming for Free-Viewpoint Video Reconstruction (NeurIPS 2025 · similarity 0.84)
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.
