Research map / Image generation / Super-resolution
Activating More Pixels in Image Super-Resolution Transformer
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 |
|---|---|---|---|
| Increase SwinIR self-attention window from (8,8) to (16,16) (ablation) | SwinIR · Urban100 ×4 | PSNR: 27.45 -> 27.81 | Helped |
| Add OCAB (ablation) | HAT baseline · Urban100 ×4 | PSNR: 27.81dB -> 27.91dB | Helped |
| Add CAB (ablation) | HAT baseline · Urban100 ×4 | PSNR: 27.81dB -> 27.91dB | Helped |
| Add both OCAB and CAB (ablation) | HAT baseline · Urban100 ×4 | PSNR: 27.81dB -> 27.97dB | Helped |
| Add channel attention to CAB (ablation) | HAT CAB · Urban100 ×4 | PSNR: 27.92dB -> 27.97dB | Helped |
| Set CAB weighting α to 0.01 instead of 0, 1, or 0.1 (ablation) | HAT · Urban100 ×4 | PSNR: α=0: 27.81dB; α=1: 27.86dB; α=0.1: 27.90dB; α=0.01: 27.97dB | Helped |
| Set OCAB overlap ratio γ to 0.5 rather than 0, 0.25, or 0.75 (ablation) | HAT · Urban100 ×4 | PSNR: γ=0: 27.85dB; γ=0.25: 27.81dB; γ=0.5: 27.91dB; γ=0.75: 27.86dB | Mixed |
| Increase CAB squeeze factor β from 1 to 2, 3, or 6 (ablation) | HAT · Urban100 ×4 | PSNR: β=1: 27.97dB; β=2: 27.92dB; β=3: 27.91dB; β=6: 27.88dB | Mixed |
| Use same-task ImageNet pre-training and fine-tuning (ablation) | HAT · Urban100 ×4 | PSNR: 27.97 -> 28.37 | Helped |
| Use same-task rather than multi-related-task pre-training (ablation) | HAT · Urban100 ×4, fine-tuning | PSNR: 28.21 -> 28.28 | Helped |
| Increase model depth to HAT-L and use same-task pre-training (ablation) | HAT · Urban100 ×4 | PSNR: 28.37 -> 28.60 | Helped |
Closest papers by meaning
- Recursive Generalization Transformer for Image Super-Resolution (ICLR 2024 · similarity 0.90)
- Transcending the Limit of Local Window: Advanced Super-Resolution Transformer with Adaptive Token Dictionary (CVPR 2024 · similarity 0.90)
- CFAT: Unleashing Triangular Windows for Image Super-resolution (CVPR 2024 · similarity 0.90)
- Efficient Attention-Sharing Information Distillation Transformer for Lightweight Single Image Super-Resolution (AAAI 2025 · similarity 0.89)
- Progressive Focused Transformer for Single Image Super-Resolution (CVPR 2025 · similarity 0.88)
- You Only Need Less Attention at Each Stage in Vision Transformers (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.