Research map / Image generation / Super-resolution
Fast Adaptation to Super-Resolution Networks via Meta-Learning
Seobin Park, Jinsu Yoo, Donghyeon Cho, Jiwon Kim, Tae Hyun Kim
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 |
|---|---|---|---|
| add MLSR meta-learning and 100 test-time steps | SRCNN · Urban100 x2 bicubic | PSNR: 29.39 -> 29.54 | Improved |
| add MLSR meta-learning and 100 test-time steps | ENET · Urban100 x2 bicubic | PSNR: 30.38 -> 30.52 | Improved |
| add MLSR meta-learning and 100 test-time steps | IDN · Urban100 x2 bicubic | PSNR: 31.95 -> 32.23 | Improved |
| add MLSR meta-learning and 20 test-time steps | RCAN · Urban100 x2 bicubic | PSNR: 33.10 -> 33.32 | Improved |
| increase test-time steps from 20 to 100 (ablation) | RCAN + MLSR · DIV2K test x2 bicubic | PSNR: 35.75 -> 35.48 | Hurt |
| increase test-time steps from 5 to 100 (ablation) | IDN + MLSR · BSD100 x2 bicubic | PSNR: 32.17 -> 32.08 | Hurt |
| meta-train on Urban100 instead of DIV2K (ablation) | IDN-ML · Urban100 40-image test | PSNR: 32.37 -> 32.43 | Improved |
| increase test-time gradient steps from 0 to 100 (ablation) | IDN-ML · Urban100 unknown 5x5 kernel | PSNR: 23.48 -> 25.42 | Improved |
| fine-tune on the test image at test time | IDN · Urban100 x2 non-bicubic kernel | PSNR: 23.49 -> 25.93 | Improved |
| replace ZSSR with meta-trained IDN using 100 steps | IDN-ML · Urban100 unknown kernel | PSNR: 25.02 -> 25.42 | Improved |
| use large scale factors x3 and x4 with test-time fine-tuning | IDN · Urban100 x3 and x4 | PSNR: 0.04dB | No clear effect |
Closest papers by meaning
- Meta-Transfer Learning for Zero-Shot Super-Resolution (CVPR 2020 · similarity 0.91)
- Accelerating Image Super-Resolution Networks with Pixel-Level Classification (ECCV 2024 · similarity 0.90)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.89)
- ODE-Inspired Network Design for Single Image Super-Resolution (CVPR 2019 · similarity 0.89)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.88)
- Lightweight Image Super-Resolution via Flexible Meta Pruning (ICML 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.
