Research map / Label-efficient and robust learning / Self-supervised and contrastive
Unsupervised Deraining: Where Asymmetric Contrastive Learning Meets Self-Similarity
Yi Chang, Yun Guo, Yuntong Ye, Chang‐Feng Yu, Lin Zhu, Xi-Le Zhao, Luxin Yan, Yonghong Tian
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 asymmetric contrastive loss (ablation) | NLCL · RainCityscapes | PSNR: 26.46 -> 27.42 | Improved |
| add layer contrastive loss (ablation) | ANLCL · RainCityscapes | PSNR: 25.83 -> 27.42 | Improved |
| add location contrastive loss (ablation) | ANLCL · RainCityscapes | PSNR: 26.45 -> 27.42 | Improved |
| add adversarial loss (ablation) | ANLCL · RainCityscapes | PSNR: 22.79 -> 27.42 | Improved |
| add L1 sparse loss (ablation) | ANLCL · RainCityscapes | PSNR: 27.07 -> 27.42 | Improved |
| non-local sampling for positives and negatives instead of random (ablation) | NLCL · RainCityscapes | PSNR: 25.83 -> 26.46 | Improved |
| neighbour sampling instead of random (ablation) | NLCL · RainCityscapes | PSNR: 25.83 -> 26.03 | Improved |
| use discriminator as contrastive encoder instead of image generator (ablation) | NLCL · RainCityscapes | PSNR: 24.86 -> 26.46 | Improved |
| use 32 positives instead of 8 (256 negatives) (ablation) | NLCL · RainCityscapes | PSNR: 26.46 -> 25.40 | Hurt |
| increase negatives from 64 to 256 (8 positives) (ablation) | NLCL · RainCityscapes | PSNR: 24.55 -> 26.46 | Improved |
| embed ANLCL loss into PReNet (ablation) | PReNet · supervised deraining | PSNR/SSIM: 35.16 / 0.9762 -> 36.34 / 0.9814 | Improved |
| embed ANLCL loss into RCDNet | RCDNet · supervised deraining | PSNR/SSIM: 36.65 / 0.9805 -> 37.46 / 0.9869 | Improved |
Closest papers by meaning
- Unsupervised Deraining: Where Contrastive Learning Meets Self-Similarity (CVPR 2022 · similarity 0.97)
- Harnessing Joint Rain-/Detail-aware Representations to Eliminate Intricate Rains (ICLR 2024 · similarity 0.92)
- Multi-Decoding Deraining Network and Quasi-Sparsity Based Training (CVPR 2021 · similarity 0.91)
- Single Image Deraining: From Model-Based to Data-Driven and Beyond (TPAMI 2020 · similarity 0.91)
- Not Just Streaks: Towards Ground Truth for Single Image Deraining (ECCV 2022 · similarity 0.91)
- Channel Consistency Prior and Self-Reconstruction Strategy Based Unsupervised Image Deraining (CVPR 2025 · similarity 0.88)
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.
