Research map / Image generation / Super-resolution
Hybrid Pixel-Unshuffled Network for Lightweight Image Super-resolution
Read the paperCheck 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 |
|---|---|---|---|
| Add the input feature to the depthwise-convolution output (ablation) | DSC in the SR component comparison · Set5 | PSNR: 37.65 -> 37.73 | Helped |
| Use depthwise separable convolution instead of pointwise convolution only (ablation) | SR component comparison · Set5 | PSNR: 37.36 -> 37.65 | Helped |
| Combine pixel-unshuffled downsampling with Self-Residual DSC (ablation) | SR component comparison · Set5 | PSNR: 37.73 -> 37.76 | Helped |
| Use pixel-unshuffled downsampling alone instead of spatial operations (ablation) | SR component comparison · Set5 | PSNR: 37.65 -> 35.68 | Hurt |
| Use max-pooling with bilinear upsampling rather than average-pooling with bilinear upsampling (ablation) | PUD operation · Set5 ×2 | PSNR: 37.83 -> 37.87 | Helped |
| Use bilinear rather than nearest upsampling with max-pooling (ablation) | PUD operation · Set5 ×2 | PSNR: 37.85 -> 37.87 | Helped |
| Increase model size from HPUN-S to HPUN-L | HPUN · Urban100 ×4 BI | PSNR: 25.86 -> 26.36 | Helped |
| Apply self-ensemble | HPUN-L · Urban100 ×4 BI | PSNR: 26.36 -> 26.48 | Helped |
| Use HPUN-L rather than LatticeNet-CL+ | ×4 BI super-resolution · Urban100 | PSNR: 26.29 -> 26.36 | Helped |
| Use HPUN-L rather than LatticeNet-CL+ | ×4 BI super-resolution · Set5 | PSNR: 32.39 -> 32.38 | Hurt |
| Integrate a standard convolution into PUB to form HPUN-L (ablation) | HPUN · Set5 ×4 BI | PSNR | Helped |
| Use PUD rather than other downsampling operations (ablation) | SR downsampling module · Reported downsampling comparison | PSNR | Helped |
Closest papers by meaning
- Residual Dense Network for Image Super-Resolution (CVPR 2018 · similarity 0.88)
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network (CVPR 2016 · similarity 0.88)
- Learning a Single Convolutional Super-Resolution Network for Multiple Degradations (CVPR 2018 · similarity 0.87)
- Unified Dynamic Convolutional Network for Super-Resolution With Variational Degradations (CVPR 2020 · similarity 0.87)
- ShuffleMixer: An Efficient ConvNet for Image Super-Resolution (NeurIPS 2022 · similarity 0.87)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (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.