Research map / Image generation / Super-resolution
Content-Aware Local GAN for Photo-Realistic Super-Resolution
JoonKyu Park, Sanghyun Son, Kyoung Mu Lee
Read the paperCheck 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 |
|---|---|---|---|
| replace plain GAN discriminator with CAL-GAN mixture of classifiers | EDSR · DIV2K (val) x4 SR | LPIPS: 0.220 -> 0.074 | Improved |
| replace plain GAN discriminator with CAL-GAN | SwinIR · DIV2K (val) x4 SR | LPIPS: 0.093 -> 0.087 | Improved |
| add balancing loss for routing (ablation) | CAL-GAN · DIV2K (val) x4 SR | LPIPS: 0.100 -> 0.091 | Improved |
| add orthogonal loss on classifier projections (ablation) | CAL-GAN · DIV2K (val) x4 SR | LPIPS: 0.109 -> 0.091 | Improved |
| add distribution (LDA-style) loss (ablation) | CAL-GAN · DIV2K (val) x4 SR | FID: 14.046 -> 11.772 | Improved |
| use 12 classifiers instead of 1 (ablation) | CAL-GAN · DIV2K (val) x4 SR | FID: 13.953 -> 11.772 | Improved |
| increase classifiers from 12 to 16 (ablation) | CAL-GAN · DIV2K (val) x4 SR | FID: 11.772 -> 12.520 | Hurt |
| use patch size 4 instead of 1 for class assignment (ablation) | CAL-GAN · DIV2K (val) x4 SR | FID: 14.116 -> 11.772 | Improved |
| use learned content-based routing instead of segmentation-based routing (ablation) | CAL-GAN · DIV2K (val) x4 SR | LPIPS: 0.137 -> 0.091 | Improved |
| CAL-GAN vs LDL on BSD100 FID with RRDB backbone | RRDB · BSD100 x4 SR | FID: 43.217 (CAL-GAN) vs 40.806 (LDL) | Mixed |
| CAL-GAN vs LDL on distortion metrics | RRDB · Urban100 x4 SR | PSNR (dB): 25.498 (LDL) -> 25.290 (CAL-GAN) | Hurt |
Closest papers by meaning
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network (CVPR 2017 · similarity 0.91)
- SRFeat: Single Image Super-Resolution with Feature Discrimination (ECCV 2018 · similarity 0.91)
- SeD: Semantic-Aware Discriminator for Image Super-Resolution (CVPR 2024 · similarity 0.90)
- Exploring Semantic Feature Discrimination for Perceptual Image Super-Resolution and Opinion-Unaware No-Reference Image Quality Assessment (CVPR 2025 · similarity 0.89)
- Uncertainty-Aware GAN for Single Image Super Resolution (AAAI 2024 · similarity 0.89)
- GSDD: Generative Space Dataset Distillation for Image Super-resolution (AAAI 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.
