Research map / Image generation / Image restoration
Training Your Image Restoration Network Better with Random Weight Network as Optimization Function
Man Zhou, Naishan Zheng, Yuan Xu, Chun-Le Guo, Chongyi 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 |
|---|---|---|---|
| add Taylor unfolding network random-weight loss to L1/L2 loss | DRBN · LOL | PSNR: 19.8509 -> 20.1156 | Improved |
| add Taylor random-weight loss with per-epoch re-initialization (epochR) | DnCNN · SIDD | PSNR: 37.1992 -> 37.3719 | Improved |
| re-initialize random weights each epoch (epochR) instead of once (ablation) | DRBN +INN · LOL | PSNR: 19.8543 -> 20.1913 | Improved |
| add INN random-weight loss | SFINet · GaoFen2 | PSNR: 47.4712 -> 47.6316 | Improved |
| add INN random-weight loss | SFINet · WorldView-II | PSNR: 41.7244 -> 41.9521 | Improved |
| add random-weight loss manifolds | MPRnet · SIDD | PSNR: 39.2372 -> 39.3317 | No clear effect |
| replace CNN random weight network with Transformer (ablation) | DnCNN +Taylor+epochR · SIDD | PSNR: 37.3719 -> 37.3560 | No clear effect |
| replace CNN random weight network with Transformer (ablation) | DRBN +Taylor+epochR · LOL | PSNR: 20.2405 -> 20.1826 | No clear effect |
| increase random weight network depth to 3 layers (ablation) | DnCNN +CDC+epochR · SIDD | PSNR: 37.2784 -> 37.2218 | Hurt |
| increase random weight network depth to 7 layers (ablation) | DnCNN +CDC+epochR · SIDD | PSNR: 37.2784 -> 37.2923 | No clear effect |
| replace Kaiming with Xavier initialization (ablation) | DnCNN +CDC+epochR · SIDD | PSNR: 37.2784 -> 37.2567 | No clear effect |
| replace Kaiming with Xavier initialization (ablation) | SID +CDC+epochR · LOL | PSNR: 20.4750 -> 20.3271 | Hurt |
Closest papers by meaning
- A Comparative Study of Image Restoration Networks for General Backbone Network Design (ECCV 2024 · similarity 0.88)
- Revitalizing Convolutional Network for Image Restoration (TPAMI 2024 · similarity 0.88)
- Deep Generalized Unfolding Networks for Image Restoration (CVPR 2022 · similarity 0.88)
- Functional Neural Networks for Parametric Image Restoration Problems (NeurIPS 2021 · similarity 0.88)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.87)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · similarity 0.87)
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.
