Research map / Image generation / GANs
Re-Boosting Self-Collaboration Parallel Prompt GAN for Unsupervised Image Restoration
Xin Lin, Yuyan Zhou, Jingtong Yue, Chao Ren, Kelvin C. K. Chan, Lu Qi, Ming-Hsuan Yang
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 BGM loss to unpaired GAN denoiser (V1 to V2) (ablation) | V1 · SIDD denoising ablation | PSNR: 34.52 -> 34.69 | Improved |
| add PL module (V2 to V3) (ablation) | V2 · SIDD denoising ablation | PSNR: 34.69 -> 34.92 | Improved |
| add self-synthesis constraint (V3 to V4) (ablation) | V3 · SIDD denoising ablation | PSNR: 34.92 -> 35.37 | Improved |
| use parallel two-branch P2GAN instead of V4 (ablation) | V4 SGabNet · SIDD denoising ablation | PSNR: 35.37 -> 35.90 | Improved |
| add self-collaboration iterations (ablation) | P2GAN · SIDD Validation | PSNR: gains exceeding 0.5 dB in the first iteration; final iteration only about 0.03 dB | Improved |
| add Reb-SC stage after SC converges (ablation) | SCP2GAN · SIDD Validation and Rain100L | PSNR: approximately 0.3 dB | Improved |
| add SC and Reb-SC (full RSCP2GAN) to P2GAN (ablation) | P2GAN · SIDD Validation and Rain100L | PSNR: 1.93 dB on the SIDD Validation and 1.95 dB on Rain100L | Improved |
| Reb-SC with N=2 augmentations vs none (ablation) | Reb-SC · SIDD Validation | PSNR: 37.51 -> 37.71 | Improved |
| Reb-SC with N=4 augmentations vs none (ablation) | Reb-SC · SIDD Validation | PSNR: 37.51 -> 37.83 | Improved |
| increase Reb-SC augmentations from N=4 to N=8 (ablation) | Reb-SC · SIDD Validation | PSNR: 37.83 -> 37.78 | No clear effect |
| apply SC to DnCNN restorer (ablation) | P2GAN with DnCNN · SIDD Validation | PSNR/SSIM: 0.96 dB/0.0085 | Improved |
| apply SC to UNet restorer (ablation) | P2GAN with UNet · SIDD Validation | PSNR/SSIM: 1.58 dB/0.0373 | Improved |
Closest papers by meaning
- Unsupervised Image Denoising in Real-World Scenarios via Self-Collaboration Parallel Generative Adversarial Branches (ICCV 2023 · similarity 0.88)
- PromptRestorer: A Prompting Image Restoration Method with Degradation Perception (NeurIPS 2023 · similarity 0.88)
- Exploiting the Potential of Standard Convolutional Autoencoders for Image Restoration by Evolutionary Search (ICML 2018 · similarity 0.88)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.86)
- Vision-Language Gradient Descent-driven All-in-One Deep Unfolding Networks (CVPR 2025 · similarity 0.86)
- IRPO: Boosting Image Restoration via Post-training GRPO (NeurIPS 2026 · similarity 0.86)
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.
