Research map / Image generation / Super-resolution
Rethinking Data Augmentation for Image Super-resolution: A Comprehensive Analysis and a New Strategy
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 |
|---|---|---|---|
| Apply CutBlur to EDSR | EDSR · RealSR ×4 | PSNR: +0.23 | Helped |
| Use the full mixture of augmentations | EDSR · RealSR ×4 | PSNR: +0.27 | Helped |
| Use CutBlur with RCAN | RCAN · RealSR test (×4) | PSNR: +0.27 | Helped |
| Use the proposed strategy with EDSR | EDSR · Urban100 (×4) | PSNR: +0.14 | Helped |
| Use the proposed strategy with CARN | CARN · Set14 (×4) | PSNR: +0.00 | No clear effect |
| Use the proposed strategy with SRCNN | SRCNN · DIV2K ×4 validation, 100% training data | PSNR: -0.02 | Hurt |
| Use pixel-level Cutout at 0.1% rather than the default 25% rectangular removal (ablation) | EDSR · DIV2K ×4 | PSNR: +0.01 dB | Helped |
| Remove 25% of pixels with original Cutout (ablation) | EDSR · DIV2K ×4 | PSNR: 0.1 dB | Hurt |
| Apply Mixup (ablation) | EDSR · DIV2K ×4 | PSNR: +0.05 | Helped |
| Apply CutMix (ablation) | EDSR · RealSR ×4 | PSNR: +0.00 | No clear effect |
| Use feature-domain augmentation (Manifold Mixup or ShakeDrop) (ablation) | EDSR · DIV2K ×4 validation | PSNR | Hurt |
| Use feature-domain augmentation (Manifold Mixup or ShakeDrop) (ablation) | RCAN · DIV2K ×4 validation | PSNR | Hurt |
Closest papers by meaning
- CutMIB: Boosting Light Field Super-Resolution via Multi-View Image Blending (CVPR 2023 · similarity 0.90)
- Accurate Image Super-Resolution Using Very Deep Convolutional Networks (CVPR 2016 · similarity 0.88)
- ADD: Attribution-Driven Data Augmentation Framework for Boosting Image Super-Resolution (CVPR 2025 · similarity 0.87)
- Navigating Beyond Dropout: An Intriguing Solution towards Generalizable Image Super Resolution (CVPR 2024 · similarity 0.87)
- Auto-Encoded Supervision for Perceptual Image Super-Resolution (CVPR 2025 · similarity 0.86)
- CutFreq: Cut-and-Swap Frequency Components for Low-Level Vision Augmentation (AAAI 2024 · similarity 0.86)
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.