Research map / Image generation / Super-resolution
Unsupervised Real-World Image Super Resolution via Domain-Distance Aware Training
Yunxuan Wei, Shuhang Gu, Yawei Li, Radu Timofte, Longcun Jin, Hengjie Song
Check 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 domain-gap aware target-domain adversarial training (ablation) | DASR SR network · AIM | LPIPS: 0.378 -> 0.359 | Improved |
| Add domain-gap aware target-domain adversarial training (ablation) | Bicubic-source SR network · AIM | LPIPS: 0.543 -> 0.390 | Improved |
| Add domain-distance weighted supervision (ablation) | DASR SR network · AIM | LPIPS: 0.359 -> 0.336 | Improved |
| Add domain-distance weighted supervision (ablation) | Generated-pair SR network without target training · AIM | LPIPS: 0.378 -> 0.348 | Improved |
| Combine target-domain training and weighted supervision (ablation) | DASR SR network · AIM | LPIPS: 0.378 -> 0.336 | Improved |
| Use HR images directly rather than bicubic LR as DSN input (ablation) | Down-sampling network · AIM validation | LPIPS: 0.110 -> 0.101 | Improved |
| Use wavelet-frequency adversarial loss instead of GBFS (ablation) | Down-sampling network with HR input · AIM validation | LPIPS: 0.101 -> 0.067 | Improved |
| Use wavelet-frequency adversarial loss instead of RGB adversarial loss (ablation) | Down-sampling network with HR input · AIM validation | LPIPS: 0.138 -> 0.067 | Improved |
| Use DASR rather than ZSSR | Real-world image SR · AIM | LPIPS: 0.630 -> 0.336 | Improved |
| Use DASR rather than FSSR | Real-world image SR · RealSR | PSNR: 25.992 -> 26.782 | Improved |
| Use DASR rather than FSSR | Real-world image SR · CameraSR | LPIPS: 0.180 -> 0.151 | Improved |
| Use DASR rather than supervised ESRGAN | Real-world image SR · RealSR | LPIPS: 0.199 -> 0.228 | Hurt |
Closest papers by meaning
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.90)
- Towards Realistic Data Generation for Real-World Super-Resolution (ICLR 2025 · similarity 0.90)
- Learning Coupled Dictionaries from Unpaired Data for Image Super-Resolution (CVPR 2024 · similarity 0.89)
- Unsupervised Diffusion-Based Degradation Modeling for Real-World Super-Resolution (AAAI 2025 · similarity 0.89)
- SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution (CVPR 2024 · similarity 0.88)
- Pixel-level and Semantic-level Adjustable Super-resolution: A Dual-LoRA Approach (CVPR 2025 · similarity 0.87)
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.
