Research map / Image generation / Editing and style transfer
IntrinsicWeather: Controllable Weather Editing in Intrinsic Space
Yixin Zhu, Zuo-Liang Zhu, Jian Yang, Miloš Hašan, Jin Xie, Beibei Wang
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 intrinsic map-aware attention (IMAA) to the inverse renderer (ablation) | inverse renderer (SD3.5 DiT) · WeatherSynthetic test, normal | MAE: 6.33 -> 4.24 | Improved |
| add IMAA (ablation) | inverse renderer · WeatherSynthetic test, irradiance | PSNR: 26.99 -> 29.66 | Improved |
| add IMAA (ablation) | inverse renderer · WeatherSynthetic test, albedo | PSNR: 26.78 -> 27.99 | Improved |
| add IMAA (ablation) | inverse renderer · WeatherSynthetic test, metallicity | LPIPS: 0.05 -> 0.04 | Mixed |
| add WeatherReal real-world data to forward-renderer training (ablation) | forward renderer · qualitative real-world re-rendering | realism of lighting and weather | Improved |
| replace intrinsic maps with a clean-weather image as intermediate (ablation) | weather editing pipeline · qualitative | scene detail preservation | Hurt |
| finetune IID and RGB↔X on WeatherSynthetic | IID, RGB↔X · WeatherSynthetic test, albedo | PSNR: IID 7.80 -> 11.55; RGB↔X 9.66 -> 11.35 | Improved |
| apply IntrinsicWeather re-rendering before detection | DETR · ACDC val | AP0.75: 13.15 -> 24.60 | Improved |
| apply IntrinsicWeather re-rendering before segmentation | Segformer · ACDC val | mIoU: 24.13 -> 30.05 | Improved |
| edit weather in intrinsic space versus pixel-space Flux-Kontext | weather editing · Weather editing | DINO-S: 85.50 -> 73.63 | Hurt |
| edit weather in intrinsic space versus Qwen-Image-Edit | weather editing · Weather editing | PickScore: 22.43 -> 22.32 (Snowy) | Hurt |
| edit weather in intrinsic space versus Qwen-Image-Edit | weather editing · Weather editing | DINO-S: 53.70 -> 73.63 | Improved |
Closest papers by meaning
- AutoWeather4D: Autonomous Driving Video Weather Conversion via G-Buffer Dual-Pass Editing (ECCV 2026 · similarity 0.89)
- WeatherCity: Urban Scene Reconstruction with Controllable Multi-Weather Transformation (CVPR 2026 · similarity 0.89)
- Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start (NeurIPS 2025 · similarity 0.85)
- Genuine Knowledge from Practice: Diffusion Test-Time Adaptation for Video Adverse Weather Removal (CVPR 2024 · similarity 0.84)
- Continuous Adverse Weather Removal via Degradation-Aware Distillation (CVPR 2025 · similarity 0.83)
- STORM: Spatio-TempOral Reconstruction Model For Large-Scale Outdoor Scenes (ICLR 2025 · similarity 0.83)
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.
