Research map / Image generation / Super-resolution
Feedback Network for Image Super-Resolution
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 top-down feedback instead of disconnecting losses at earlier iterations (ablation) | SRFBN-L · Set5 ×4 | PSNR: 32.09 -> 32.11 | Helped |
| Share weights across iterations (ablation) | SRFBN-L · Set5 ×4 | PSNR: 31.82 -> 32.11 | Helped |
| Provide the LR input at every iteration (ablation) | SRFBN-L · Set5 ×4 | PSNR: 0.17dB | Helped |
| Use up/down-sampling layers in the feedback block (ablation) | SRFBN-L · Set5 ×4 | PSNR: 31.41 -> 32.05 | Helped |
| Add dense skip connections to the feedback block (ablation) | SRFBN-L with UDSL · Set5 ×4 | PSNR: 32.05 -> 32.11 | Helped |
| Use the designed feedback block instead of ConvLSTM (ablation) | SRFBN-L · Set5 ×4 | PSNR: 31.26 -> 32.11 | Helped |
| Use the designed feedback block instead of projection units (ablation) | SRFBN-L · Set5 ×4 | PSNR: 32.07 -> 32.11 | Helped |
| Use the designed feedback block instead of RDB (ablation) | SRFBN-L · Set5 ×4 | PSNR: 32.01 -> 32.11 | Helped |
| Train with curriculum targets ordered from easy to hard (ablation) | SRFBN on BD · Set5 ×4 | PSNR: 29.78 -> 29.96 | Helped |
| Train with curriculum targets ordered from easy to hard (ablation) | SRFBN on DN · Set5 ×4 | PSNR: 26.92 -> 26.93 | No clear effect |
| Fine-tune from a BI-pretrained model rather than train from scratch (ablation) | SRFBN with curriculum · Set5 ×4, BD | PSNR: 29.96 -> 30.03 | Helped |
| Fine-tune from a BI-pretrained model rather than train from scratch (ablation) | SRFBN with curriculum · Set5 ×4, DN | PSNR: 26.93 -> 26.98 | Helped |
Closest papers by meaning
- Image Super-Resolution via Dual-State Recurrent Networks (CVPR 2018 · similarity 0.91)
- Deep Back-Projection Networks for Super-Resolution (CVPR 2018 · similarity 0.89)
- Deeply-Recursive Convolutional Network for Image Super-Resolution (CVPR 2016 · similarity 0.88)
- Accurate Image Super-Resolution Using Very Deep Convolutional Networks (CVPR 2016 · similarity 0.88)
- FSR: A General Frequency-Oriented Framework to Accelerate Image Super-resolution Networks (AAAI 2023 · similarity 0.88)
- Closed-Loop Matters: Dual Regression Networks for Single Image Super-Resolution (CVPR 2020 · 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.