Research map / Image generation / Image restoration
Random Shuffle Transformer for Image Restoration
Read the paperCheck 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 |
|---|---|---|---|
| Add random shuffle (CW-MSA) training with Monte-Carlo averaging at test instead of shifted window | ShuffleFormer · SIDD | PSNR: +0.11 dB (39.89 to 40.00) | Helped |
| Add random shuffle training with Monte-Carlo testing instead of shifted window | ShuffleFormer · SPA-Data | PSNR: +0.39 dB (48.80 to 49.19) | Helped |
| Add random shuffle training with Monte-Carlo testing instead of shifted window | ShuffleFormer · GoPro/HIDE/RealBlur average | PSNR: +0.24 dB (32.30 to 32.54) | Helped |
| Apply Monte-Carlo averaging at test without random shuffle training (SM vs SS) (ablation) | ShuffleFormer · SIDD | PSNR: -0.54 dB (39.89 to 39.35) | Hurt |
| Apply Monte-Carlo averaging at test without random shuffle training (SM vs SS) (ablation) | ShuffleFormer · SPA-Data | PSNR: -0.89 dB (48.80 to 47.91) | Hurt |
| Apply Monte-Carlo averaging at test without random shuffle training (SM vs SS) (ablation) | ShuffleFormer · GoPro/HIDE/RealBlur average | PSNR: -1.07 dB (32.30 to 31.23) | Hurt |
| Random shuffle training with shifted window only at test (CS) instead of Monte-Carlo random shuffle testing (ablation) | ShuffleFormer · SIDD | PSNR: -0.40 dB (40.00 to 39.60) | Hurt |
| Random shuffle training with shifted window only at test (CS) instead of Monte-Carlo random shuffle testing (ablation) | ShuffleFormer · SPA-Data | PSNR: -1.11 dB (49.19 to 48.08) | Hurt |
| Random shuffle training with shifted window only at test (CS) instead of Monte-Carlo random shuffle testing (ablation) | ShuffleFormer · GoPro/HIDE/RealBlur average | PSNR: -0.80 dB (32.54 to 31.74) | Hurt |
| Add depth-wise convolution for position modeling before shuffling (ablation) | ShuffleFormer-CM · SIDD | PSNR: +0.07 dB (39.93 to 40.00) | Helped |
| Random shuffle applied to CNN instead of Transformer window attention (ablation) | CNN · SIDD | PSNR | Hurt |
Closest papers by meaning
- HomoFormer: Homogenized Transformer for Image Shadow Removal (CVPR 2024 · similarity 0.85)
- Adaptive Window Pruning for Efficient Local Motion Deblurring (ICLR 2024 · similarity 0.84)
- Adapt or Perish: Adaptive Sparse Transformer with Attentive Feature Refinement for Image Restoration (CVPR 2024 · similarity 0.84)
- Sharing Key Semantics in Transformer Makes Efficient Image Restoration (NeurIPS 2024 · similarity 0.83)
- VolFormer: Explore More Comprehensive Cube Interaction for Hyperspectral Image Restoration and Beyond (CVPR 2025 · similarity 0.82)
- SeedVR: Seeding Infinity in Diffusion Transformer Towards Generic Video Restoration (CVPR 2025 · similarity 0.82)
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.