Research map / Detection and segmentation / CNN features
Generative Status Estimation and Information Decoupling for Image Rain Removal
Di Lin, Xin WANG, Jia Shen, Renjie Zhang, Ruonan Liu, Miaohui Wang, Wuyuan Xie, Qing Guo, Ping Li
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 |
|---|---|---|---|
| increase number of generated kernels N from 1 to 8 (ablation) | SEIDNet · Rain100H | PSNR: 31.457 -> 33.217 | Improved |
| increase kernels from N=8 to 32 (ablation) | SEIDNet · Rain100H | PSNR: 33.217 -> 33.221 | No clear effect |
| replace both discriminative networks with CVAEs (status estimation and information decoupling) (ablation) | SEIDNet · Rain100H | PSNR: 29.43 -> 33.22 | Improved |
| use CVAE only for status estimation (ablation) | discriminative baseline · Rain100H | PSNR: 29.43 -> 29.62 | Improved |
| use CVAE only for information decoupling (ablation) | discriminative baseline · Rain100H | PSNR: 29.43 -> 31.42 | Improved |
| use generative CVAE instead of discriminative network for kernel estimation (ablation) | SEIDNet · Rain100H | PSNR: 29.91 -> 33.22 | Improved |
| average discriminative and generative kernel maps (ablation) | SEIDNet · Rain100H | PSNR: 33.22 -> 33.11 | Hurt |
| two factorized CVAEs for (K, R) instead of one CVAE for (K, R) (ablation) | SEIDNet · Rain100H | PSNR: 29.03 -> 33.22 | Improved |
| add status map conditioning (versus single CVAE for K only) (ablation) | SEIDNet · Rain100H | PSNR: 25.21 -> 33.22 | Improved |
| train with only 1K random SPA images (subset A) | SEIDNet vs MPR · SPA subset A | PSNR: 35.57 -> 39.07 | Improved |
| SEIDNet vs best prior on Rain100H | SPDNet · Rain100H | PSNR: 32.68 -> 33.22 | Improved |
| SEIDNet on desnowing vs DDMSNet | DDMSNet · Snow100K Overall | PSNR: 32.03 -> 32.77 | Improved |
Closest papers by meaning
- Singe Image Rain Removal with Unpaired Information: A Differentiable Programming Perspective (AAAI 2019 · similarity 0.89)
- Joint Bi-Layer Optimization for Single-Image Rain Streak Removal (ICCV 2017 · similarity 0.89)
- Both Diverse and Realism Matter: Physical Attribute and Style Alignment for Rainy Image Generation (ICCV 2023 · similarity 0.88)
- Erase or Fill? Deep Joint Recurrent Rain Removal and Reconstruction in Videos (CVPR 2018 · similarity 0.88)
- Rain Streak Removal Using Layer Priors (CVPR 2016 · similarity 0.88)
- From Rain Generation to Rain Removal (CVPR 2021 · 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.
