Research map / Video / Video generation
Controllable Weather Synthesis and Removal with Video Diffusion Models
Chih-Hao Lin, Zian Wang, Ruofan Liang, Yuxuan Zhang, Sanja Fidler, Shenlong Wang, Zan Gojcic
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 video model with temporal modules instead of image-only variant (ablation) | WeatherWeaver · weather synthesis (qualitative) | transient effect quality | Improved |
| add simulation data to training (ablation) | WeatherWeaver · weather synthesis (qualitative) | effect control and shading | Improved |
| add generation (Prompt-to-Prompt image pairs) data (ablation) | WeatherWeaver · weather synthesis (qualitative) | generalization of specific effects | Improved |
| add auto-labeled real-world video data (ablation) | WeatherWeaver · weather synthesis (qualitative) | realism | Improved |
| use WeatherWeaver instead of specialized removal models | weather removal · Weather removal | Align. VLM: 22.79 -> 71.61 (WeatherDiffusion vs Ours) | Improved |
| compare structure preservation versus baselines | WeatherWeaver · Weather removal | DINO Struct.: 0.05 -> 2.09 (Histoformer vs Ours) | Hurt |
| split pipeline into removal model that pseudo-labels real videos and synthesis model | WeatherWeaver · Training data | paired real-video data: 4.6k video pairs | Improved |
| apply weather removal before Grounded-SAM detection | Grounded-SAM · dense fog example | object detection | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FRESCO — FRESCO: Spatial-Temporal Correspondence for Zero-Shot Video Translation (CVPR 2024) · better on 2 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
- Genuine Knowledge from Practice: Diffusion Test-Time Adaptation for Video Adverse Weather Removal (CVPR 2024 · similarity 0.89)
- WeatherEdit: Controllable Weather Editing with 4D Gaussian Field (AAAI 2026 · similarity 0.89)
- Semantic-Aware, Physics-Informed, Geometry-Grounded Weather Synthesis (ECCV 2026 · similarity 0.89)
- Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start (NeurIPS 2025 · similarity 0.86)
- RainyGS: Efficient Rain Synthesis with Physically-Based Gaussian Splatting (CVPR 2025 · similarity 0.85)
- Precipitation Downscaling with Spatiotemporal Video Diffusion (NeurIPS 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.
