Research map / Image generation / Super-resolution
GLEAN: Generative Latent Bank for Image Super-Resolution and Beyond
Kelvin C. K. Chan, Xiangyu Xu, Xintao Wang, Jinwei Gu, Chen Change Loy
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 |
|---|---|---|---|
| use LightGLEAN design (remove coarse feature connections, learned fixed latent codes) instead of GLEAN (ablation) | GLEAN · 16x face SR | parameters: 175.6M -> 36.3M | Improved |
| use LightGLEAN design instead of GLEAN (ablation) | GLEAN · 16x face SR | FLOPs: 277.5G -> 98.24G | Improved |
| use LightGLEAN design instead of GLEAN (ablation) | GLEAN · 16x face SR | PSNR: 26.84 -> 26.85 | No clear effect |
| use LightGLEAN design instead of GLEAN (ablation) | GLEAN · Car 16x SR | PSNR: 19.74 -> 19.46 | Hurt |
| reduce RRDB blocks from 23 to 10 (ablation) | LightGLEAN · 16x face SR | PSNR: 0.02 dB drop | No clear effect |
| add generative latent bank (GLEAN) over larger ESRGAN+ | ESRGAN+ · Cat 16x SR | PSNR: 19.99 -> 20.92 | Improved |
| add generative latent bank (GLEAN) over ESRGAN+ | ESRGAN+ · Tower 16x SR | LPIPS: 0.3132 -> 0.2850 | Improved |
| add generative latent bank (GLEAN) over ESRGAN+ | ESRGAN+ · Bedroom 16x SR | PSNR: 19.47 -> 19.44 | Hurt |
| add BigGAN latent bank (GLEAN) over ESRGAN+ | ESRGAN+ · ImageNet 8x SR | PSNR: 20.04 -> 20.62 | Improved |
| finetune the latent bank (ablation) | GLEAN · face SR | quality | No clear effect |
| add multi-resolution encoder convolutional features to the latent bank (ablation) | GLEAN (latent vectors only) · face SR | visual fidelity | Improved |
| pass latent bank features to the decoder (ablation) | GLEAN · face SR | visual quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PULSE — PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models (CVPR 2020) · better on 1 of 1 shared result
- SinGAN — SinGAN: Learning a Generative Model From a Single Natural Image (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GLEAN: Generative Latent Bank for Large-Factor Image Super-Resolution (CVPR 2021 · similarity 0.97)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.89)
- GSDD: Generative Space Dataset Distillation for Image Super-resolution (AAAI 2024 · similarity 0.88)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.87)
- Uncertainty-Aware GAN for Single Image Super Resolution (AAAI 2024 · similarity 0.87)
- SeD: Semantic-Aware Discriminator 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.
