Research map / Image generation / Image restoration
Focal Network for Image Restoration
Yuning Cui, Wenqi Ren, Xiaochun Cao, Alois Knoll
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 spatial selection module (SSM) (ablation) | FocalNet tiny baseline · SOTS-Indoor | PSNR: 31.33 -> 33.51 | Improved |
| add frequency selection module (FSM) (ablation) | FocalNet tiny baseline · SOTS-Indoor | PSNR: 31.33 -> 32.44 | Improved |
| combine SSM and FSM (DSM) (ablation) | FocalNet tiny baseline · SOTS-Indoor | PSNR: 31.33 -> 35.56 | Improved |
| add MResBlock to the DSM model in tiny setting (ablation) | FocalNet tiny · SOTS-Indoor | PSNR: 35.56 -> 35.60 | Improved |
| use SSM instead of CBAM (ablation) | FocalNet tiny · SOTS-Indoor | PSNR: 32.71 -> 33.71 | Improved |
| replace SSM with simple gate or supervised attention (ablation) | FocalNet tiny · SOTS-Indoor | PSNR: 33.71 -> 32.23 / 32.29 | Hurt |
| add FSM to CBAM (ablation) | CBAM · SOTS-Indoor | PSNR: 32.71 -> 33.23 | Improved |
| apply SSM before FSM rather than FSM before SSM (ablation) | FocalNet tiny · SOTS-Indoor | PSNR: 35.17 -> 35.60 | Improved |
| add MResBlock in first scale of final model (ablation) | FocalNet · SOTS-Indoor | PSNR: 40.69 -> 40.82 | Improved |
| use MResBlock in all scales (FocalNet-S) (ablation) | FocalNet · SOTS-Indoor | PSNR: 40.82 -> 40.32 | Hurt |
| replace Restormer with FocalNet on defocus deblurring | DPDD combined · DPDD combined | LPIPS: 0.178 -> 0.210 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Restormer — Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Focal Modulation for Image Restoration (IJCV 2025 · similarity 0.93)
- Revitalizing Convolutional Network for Image Restoration (TPAMI 2024 · similarity 0.88)
- A Comparative Study of Image Restoration Networks for General Backbone Network Design (ECCV 2024 · similarity 0.88)
- Selective Frequency Network for Image Restoration (ICLR 2023 · similarity 0.88)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · similarity 0.87)
- Adapt or Perish: Adaptive Sparse Transformer with Attentive Feature Refinement for Image Restoration (CVPR 2024 · 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.
