Research map / Architectures and efficiency / Attention and transformers
KNN Local Attention for Image Restoration
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use k-NN local attention instead of conventional restoration attention | KiT · SIDD | PSNR: 39.75 -> 39.80 | Helped |
| Use k-NN local attention instead of conventional restoration attention | KiT · DND | PSNR: 39.96 -> 39.96 | No clear effect |
| Use k-NN local attention instead of conventional restoration attention | KiT · GoPro | PSNR: 32.68 -> 32.70 | Helped |
| Use k-NN local attention instead of conventional restoration attention | KiT · HIDE | PSNR: 30.29 -> 30.98 | Helped |
| Use KiT for deraining | KiT · Average deraining | PSNR: 32.91 -> 32.81 | Hurt |
| Use KiT for deraining | KiT · Average deraining | SSIM: 0.926 -> 0.929 | Helped |
| Increase chunk size from k=1 to k=4 (ablation) | KiT · SIDD | PSNR: 38.79 -> 39.80 | Helped |
| Increase chunk size from k=4 to k=16 (ablation) | KiT · SIDD | PSNR: 39.80 -> 39.80 | No clear effect |
| Increase hash rounds from h=1 to h=4 with k=4 (ablation) | KiT · SIDD | PSNR: 39.69 -> 39.80 | Helped |
| Increase hash rounds from h=4 to h=16 with k=4 (ablation) | KiT · SIDD | PSNR: 39.80 -> 39.80 | No clear effect |
| Do not share query and key projections (ablation) | KiT · SIDD | PSNR: 39.75 -> 39.80 | Helped |
| Do not share query and key projections (ablation) | KiT · SIDD | SSIM: 0.969 -> 0.972 | Helped |
Closest papers by meaning
- Residual Non-local Attention Networks for Image Restoration (ICLR 2019 · similarity 0.89)
- 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 0.84)
- MaIR: A Locality- and Continuity-Preserving Mamba for Image Restoration (CVPR 2025 · similarity 0.83)
- Boosting Image Restoration via Priors from Pre-trained Models (CVPR 2024 · similarity 0.83)
- Dual-Scale Transformer for Large-Scale Single-Pixel Imaging (CVPR 2024 · similarity 0.83)
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.