Research map / Image generation / Image restoration
Not Just Streaks: Towards Ground Truth for Single Image Deraining
Yunhao Ba, Howard Zhang, Ethan Yang, Akira Suzuki, Arnold Pfahnl, Chethan Chinder Chandrappa, Celso M. de Melo, Suya You, Stefano Soatto, Alex Wong, Achuta Kadambi
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 the rain-robust contrastive feature loss (ablation) | deraining network · GT-RAIN test (overall) | PSNR: 21.82 -> 22.53 | Improved |
| add the rain-robust contrastive feature loss (ablation) | deraining network · GT-RAIN test (overall) | SSIM: 0.7148 -> 0.7304 | Improved |
| add the rain-robust loss to raise rainy/clean latent correlation (ablation) | deraining network · GT-RAIN test (overall) | normalized latent correlation: .85 ± .10 -> .95 ± .03 | Improved |
| retrain on GT-RAIN instead of original synthetic/semi-real data | RCDNet · GT-RAIN test (overall) | PSNR: 20.26 -> 20.94 | Improved |
| retrain on GT-RAIN instead of original synthetic data | EDR · GT-RAIN test (overall) | PSNR: 19.81 -> 21.44 | Improved |
| retrain on GT-RAIN instead of original synthetic data | MPRNet · GT-RAIN test (overall) | PSNR: 20.09 -> 21.56 | Improved |
| retrain on GT-RAIN: SSIM of RCDNet | RCDNet · GT-RAIN test (overall) | SSIM: 0.6881 -> 0.7091 | Improved |
| fine-tune on GT-RAIN after synthetic training instead of original weights | MPRNet · GT-RAIN test (overall) | PSNR: 20.09 -> 22.24 | Improved |
| fine-tune on GT-RAIN after synthetic training instead of training only on GT-RAIN | MPRNet · GT-RAIN test (overall) | PSNR: 21.56 -> 22.24 | Improved |
| pretrain on synthetic data then fine-tune on GT-RAIN | Ours · GT-RAIN test (overall) | PSNR: 22.53 -> 22.68 | Improved |
| train on GT-RAIN instead of other datasets | Ours vs MPRNet · GT-RAIN test (dense rain accumulation) | PSNR: 3.40 dB | Improved |
| train on real paired GT-RAIN rather than multiple synthetic datasets | Ours vs MPRNet · GT-RAIN test (overall) | PSNR: 2.44 dB | Improved |
Closest papers by meaning
- Spatial Attentive Single-Image Deraining With a High Quality Real Rain Dataset (CVPR 2019 · similarity 0.94)
- From Sky to the Ground: A Large-scale Benchmark and Simple Baseline Towards Real Rain Removal (ICCV 2023 · similarity 0.93)
- Rethinking Image Deraining via Rain Streaks and Vapors (ECCV 2020 · similarity 0.93)
- Harnessing Joint Rain-/Detail-aware Representations to Eliminate Intricate Rains (ICLR 2024 · similarity 0.91)
- Bidirectional Multi-Scale Implicit Neural Representations for Image Deraining (CVPR 2024 · similarity 0.91)
- Boosting Image De-Raining via Central-Surrounding Synergistic Convolution (AAAI 2025 · similarity 0.91)
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.
