Research map / Architectures and efficiency / RNNs and normalization
Progressive Image Deraining Networks: A Better and Simpler Baseline
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a recurrent LSTM layer across progressive stages (ablation) | PRN · Rain100H | PSNR: 28.07 -> 29.46 | Helped |
| Add a recurrent LSTM layer across progressive stages (ablation) | PRN · Rain100H | SSIM: 0.884 -> 0.899 | Helped |
| Use single negative SSIM loss instead of single MSE loss (ablation) | PReNet T=6 · Rain100H | PSNR: 29.08 -> 29.32 | Helped |
| Use single negative SSIM loss instead of single MSE loss (ablation) | PReNet T=6 · Rain100H | SSIM: 0.880 -> 0.898 | Helped |
| Apply negative SSIM supervision at every stage rather than only the final stage (ablation) | PReNet T=6 · Rain100H | PSNR: 29.32 -> 29.12 | Hurt |
| Apply negative SSIM supervision at every stage rather than only the final stage (ablation) | PReNet T=6 · Rain100H | SSIM: 0.898 -> 0.895 | Hurt |
| Increase progressive stages from 2 to 6 (ablation) | PReNet · Rain100H | PSNR: 27.27 -> 29.32 | Helped |
| Increase progressive stages from 6 to 7 (ablation) | PReNet · Rain100H | PSNR: 29.32 -> 29.24 | Hurt |
| Remove the original rainy image from each stage's input (ablation) | PReNet · Rain100H | PSNR: 29.46 -> 28.91 | Hurt |
| Remove the original rainy image from each stage's input (ablation) | PReNet · Rain100H | SSIM: 0.899 -> 0.895 | Hurt |
| Add residual learning to predict the rain layer instead of directly predicting clean background (ablation) | PReNet-LSTM · Rain100H | PSNR: 29.32 -> 29.46 | Helped |
| Replace LSTM with GRU as the recurrent layer (ablation) | PReNet · Rain100H | PSNR: 29.46 -> 29.08 | Hurt |
Closest papers by meaning
- Dreaming To Prune Image Deraining Networks (CVPR 2022 · similarity 0.90)
- Networks are Slacking Off: Understanding Generalization Problem in Image Deraining (NeurIPS 2023 · similarity 0.87)
- Residual Distillation: Towards Portable Deep Neural Networks without Shortcuts (NeurIPS 2020 · similarity 0.87)
- Are ResNets Provably Better than Linear Predictors? (NeurIPS 2018 · similarity 0.87)
- Dr2Net: Dynamic Reversible Dual-Residual Networks for Memory-Efficient Finetuning (CVPR 2024 · similarity 0.85)
- Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers (NeurIPS 2025 · 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.