Research map / Image generation / Super-resolution
Efficient Non-local Contrastive Attention for Image Super-resolution
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 ENLCA modules to EDSR | EDSR · Set5 ×4 | PSNR: 32.46 vs 32.67 | Helped |
| Add ENLA to the baseline (ablation) | EDSR baseline · Set5 ×4 | PSNR: 32.21 -> 32.37 | Helped |
| Add amplification factor k to ENLA (ablation) | ENLA · Set5 ×4 | PSNR: 32.37 -> 32.44 | Helped |
| Add contrastive learning to ENLA without k (ablation) | ENLA · Set5 ×4 | PSNR: 32.37 -> 32.41 | Helped |
| Add both amplification factor k and contrastive learning to ENLA (ablation) | ENLA · Set5 ×4 | PSNR: 32.21 -> 32.48 | Helped |
| Increase contrastive positive/negative selection percentage n1 from 1% to 2% at n2=8% (ablation) | ENLCA · Set5 ×4 | PSNR: 32.42 -> 32.48 | Helped |
| Increase n1 from 2% to 3% at n2=8% (ablation) | ENLCA · Set5 ×4 | PSNR: 32.48 -> 32.46 | Hurt |
| Increase irrelevant-feature start index n2 from 4% to 8% at n1=2% (ablation) | ENLCA · Set5 ×4 | PSNR: 32.45 -> 32.48 | Helped |
| Increase n2 from 8% to 20% at n1=2% (ablation) | ENLCA · Set5 ×4 | PSNR: 32.48 -> 32.42 | Hurt |
| Increase random feature count from m=2 to m=128 (ablation) | ENLCA · Urban100 ×4 | PSNR: 26.88 -> 27.12 | Helped |
| Increase random feature count from m=128 to m=256 (ablation) | ENLCA · Urban100 ×4 | PSNR: 27.12 -> 27.12 | No clear effect |
| Use content-relevance sparsity instead of location-fixed criss-cross aggregation | non-local attention · Urban100 ×4 | PSNR: 26.85 -> 27.12 | Helped |
Closest papers by meaning
- Recursive Generalization Transformer for Image Super-Resolution (ICLR 2024 · similarity 0.85)
- Efficient Attention-Sharing Information Distillation Transformer for Lightweight Single Image Super-Resolution (AAAI 2025 · similarity 0.84)
- ACL: Activating Capability of Linear Attention for Image Restoration (CVPR 2025 · similarity 0.84)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · similarity 0.84)
- Transcending the Limit of Local Window: Advanced Super-Resolution Transformer with Adaptive Token Dictionary (CVPR 2024 · similarity 0.84)
- Progressive Focused Transformer for Single Image Super-Resolution (CVPR 2025 · similarity 0.84)
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.