Research map / Architectures and efficiency / CNN design
Accurate Image Super-Resolution Using Very Deep Convolutional Networks
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 |
|---|---|---|---|
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · Set5 ×2 | PSNR: 36.66 -> 37.53 | Helped |
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · Set5 ×3 | PSNR: 32.75 -> 33.66 | Helped |
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · Set5 ×4 | PSNR: 30.48 -> 31.35 | Helped |
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · Set14 ×2, ×3, and ×4 | PSNR: 32.42 -> 33.03; 29.28 -> 29.77; 27.49 -> 28.01 | Helped |
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · B100 ×2, ×3, and ×4 | PSNR: 31.36 -> 31.90; 28.41 -> 28.82; 26.90 -> 27.29 | Helped |
| Use VDSR instead of the strongest listed baseline | Single-image super-resolution · Urban100 ×2, ×3, and ×4 | PSNR: 29.54 -> 30.76; 26.44 -> 27.14; 24.79 -> 25.18 | Helped |
| Predict residuals instead of HR images directly (ablation) | 10-layer SR network · Set5 ×2, initial learning rate 0.1, epoch 10 | PSNR: 27.42 -> 36.90 | Helped |
| Use residual learning with initial learning rate 0.1 rather than 0.001 (ablation) | 10-layer SR network · Set5 ×2, epoch 80 | PSNR: 36.52 -> 37.05 | Helped |
| Train one network on scales ×2, ×3, and ×4 rather than scale ×2 only (ablation) | VDSR · Set5 ×2 | PSNR: 37.10 -> 37.06 | No clear effect |
| Train one network on multiple scales including ×3 rather than scale ×3 only (ablation) | VDSR · Set5 ×3 | PSNR: 32.89 -> 33.27 | Helped |
| Train one network on multiple scales including ×4 rather than scale ×4 only (ablation) | VDSR · Set5 ×4 | PSNR: 30.84 -> 30.95 | Helped |
| Test a single-scale model at a scale different from its training scale (ablation) | VDSR · Set5 ×2 test, trained at ×3 or ×4 | PSNR: 37.10 -> 30.05; 37.10 -> 28.13 | Hurt |
Closest papers by meaning
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network (CVPR 2016 · similarity 0.91)
- Deeply-Recursive Convolutional Network for Image Super-Resolution (CVPR 2016 · similarity 0.91)
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network (CVPR 2017 · similarity 0.90)
- ClassSR: A General Framework to Accelerate Super-Resolution Networks by Data Characteristic (CVPR 2021 · similarity 0.90)
- SeD: Semantic-Aware Discriminator for Image Super-Resolution (CVPR 2024 · similarity 0.88)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 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.