Research map / Image generation / Super-resolution
Unpaired Image Super-Resolution Using Pseudo-Supervision
Shunta Maeda
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 |
|---|---|---|---|
| Use pseudo-clean LR images for paired SR supervision instead of directly using clean downscaled LR (ablation) | Our method · DIV2K realistic-wild validation, ×4 | PSNR: 21.09 -> 21.32 | Improved |
| Use pseudo-clean LR images for paired SR supervision instead of directly using clean downscaled LR (ablation) | Our method · DIV2K realistic-wild validation, ×4 | SSIM: 0.5312 -> 0.5541 | Improved |
| Train on pseudo-clean LR rather than degraded generated LR (ablation) | Our method · DIV2K realistic-wild validation, ×4 | PSNR: 20.84 -> 21.32 | Improved |
| Train on pseudo-clean LR rather than degraded generated LR (ablation) | Our method · DIV2K realistic-wild validation, ×4 | SSIM: 0.5500 -> 0.5541 | Improved |
| Add the HR discriminator D_X↑ (ablation) | Our method · DIV2K realistic-wild validation, ×4 | PSNR: 21.29 -> 21.32 | Improved |
| Add the HR discriminator D_X↑ (ablation) | Our method · DIV2K realistic-wild validation, ×4 | SSIM: 0.5532 -> 0.5541 | Improved |
| Replace L1 reconstruction loss with perceptual, relativistic adversarial, and content losses | Our method · Widerface real-world LR faces | FID: 14.89 -> 13.57 | Improved |
| Add standard self-ensemble | Our method · AIM 2019 Real-World SR Challenge Track 2 | PSNR: 22.88 -> 23.01 | Improved |
| Add the geometric ensemble loss (ablation) | Our method · DOTA aerial image validation | geometric consistency | Improved |
Closest papers by meaning
- Unsupervised Real-World Super-Resolution: A Domain Adaptation Perspective (ICCV 2021 · similarity 0.92)
- To learn image super-resolution, use a GAN to learn how to do image degradation first (ECCV 2018 · similarity 0.91)
- Learning Coupled Dictionaries from Unpaired Data for Image Super-Resolution (CVPR 2024 · similarity 0.91)
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network (CVPR 2017 · similarity 0.90)
- Toward Real-World Single Image Super-Resolution: A New Benchmark and a New Model (ICCV 2019 · similarity 0.90)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.88)
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.
