Research map / Image generation / Super-resolution
Tackling the Ill-Posedness of Super-Resolution Through Adaptive Target Generation
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 |
|---|---|---|---|
| Train RRDB with adaptive targets instead of only the paired ground-truth target | RRDB blind super-resolution · DIV2KRK | LPIPS: 0.3856 -> 0.3394 | Helped |
| Train ESRGAN with adaptive targets | ESRGAN · Set5, bicubic | LPIPS: 0.0736 -> 0.0644 | Helped |
| Train ESRGAN with adaptive targets | ESRGAN · Set14, bicubic | LPIPS: 0.1324 -> 0.1299 | Helped |
| Train ESRGAN with adaptive targets | ESRGAN · BSDS100, bicubic | LPIPS: 0.1643 -> 0.1611 | Helped |
| Train ESRGAN with adaptive targets | ESRGAN · Urban100, bicubic | LPIPS: 0.1236 -> 0.1105 | Helped |
| Train ESRGAN with adaptive targets | ESRGAN · Manga109, bicubic | LPIPS: 0.0645 -> 0.0549 | Helped |
| Use translation-only transformations rather than translation plus rotation (ablation) | Adaptive target generator · DIV2KRK | SSIM: 0.7854 -> 0.7832 | Hurt |
| Use integer-step nearest-neighbor search instead of learned sub-pixel affine localization (ablation) | Adaptive target generator · DIV2KRK | PSNR: 28.42 -> 27.75 | Hurt |
| Enlarge the adaptive-target search space to s = 11 (ablation) | Adaptive target generator · DIV2KRK | PSNR: 28.42 -> 28.27 | Hurt |
| Add shear transformations to translation and rotation (ablation) | Adaptive target generator · DIV2KRK | PSNR: 28.42 -> 28.39 | Hurt |
| Add shear and scale transformations to translation and rotation (ablation) | Adaptive target generator · DIV2KRK | PSNR: 28.42 -> 28.26 | Hurt |
| Use adaptive targets without GAN training | RRDB · Bicubic downsampling, five SR testsets | PSNR | Mixed |
Closest papers by meaning
- Explorable Super Resolution (CVPR 2020 · similarity 0.90)
- Learning the Non-Differentiable Optimization for Blind Super-Resolution (CVPR 2021 · similarity 0.89)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.89)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.89)
- Navigating Beyond Dropout: An Intriguing Solution towards Generalizable Image Super Resolution (CVPR 2024 · similarity 0.88)
- AdaBM: On-the-Fly Adaptive Bit Mapping for Image Super-Resolution (CVPR 2024 · 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.