Research map / Video / Video generation
Simulating Fluids in Real-World Still Images
Siming Fan, Jingtan Piao, Chen Qian, Hongsheng Li, Kwan-Yee Lin
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 |
|---|---|---|---|
| replace convolution with partial convolution in single-layer fluid decoder | Reproduced Holynski · Holynski common validation set | PSNR (all region): 25.03 -> 24.75 | Hurt |
| use two-layer Surface-based Layered Representation instead of single-layer | Modified Holynski · CLAW test set | PSNR (all region): 19.97 -> 20.79 | Improved |
| use two-layer SLR instead of single-layer | Modified Holynski · Holynski common validation set | LPIPS (all region): 0.0793 -> 0.0834 LPIPS, PSNR 24.75 -> 25.14 | Mixed |
| naive uniform blending of stage 1 surface layer and stage 2 background (no joint finetune) (ablation) | SLR · CLAW test set | LPIPS (all region): 0.2040 -> 0.2411 | Hurt |
| joint finetuning of stages for final SLR vs stage 1 only (ablation) | SLR · CLAW test set | LPIPS (all region): 0.2143 -> 0.2040 | Improved |
| two-layer representation judged by users vs single layer | SLR · CLAW user study | preference (%): 20.5% vs 52.3% (27.2% neither) | Improved |
| solve Navier-Stokes on 2.5D surface mesh instead of 2D plane (ablation) | fluid motion prediction · qualitative | motion realism (qualitative) | Improved |
Closest papers by meaning
- FluidNexus: 3D Fluid Reconstruction and Prediction from a Single Video (CVPR 2025 · similarity 0.88)
- Optimizing 4D Gaussians for Dynamic Scene Video from Single Landscape Images (ICLR 2025 · similarity 0.86)
- DreamMesh4D: Video-to-4D Generation with Sparse-Controlled Gaussian-Mesh Hybrid Representation (NeurIPS 2024 · similarity 0.86)
- Gaussian Splashing: Unified Particles for Versatile Motion Synthesis and Rendering (CVPR 2025 · similarity 0.85)
- MotionCraft: Physics-Based Zero-Shot Video Generation (NeurIPS 2024 · similarity 0.85)
- Gaussian-Flow: 4D Reconstruction with Dynamic 3D Gaussian Particle (CVPR 2024 · similarity 0.85)
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.
