AI Paper Atlas

Research map / Image generation / Image restoration

Scale-Wise Convolution for Image Restoration

Yuchen Fan, Jiahui Yu, Ding Liu, Thomas S. Huang

AAAI 2020 · Image generation › Image restoration · approach: Priors for restoration networks · 36 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use scale-wise convolution rather than U-Net-style multi-scale fusion (ablation)SCN · Super-resolution ×2PSNR: 34.56 -> 34.67Improved
Use scale-wise convolution rather than PSPNet-style multi-scale fusion (ablation)SCN · Super-resolution ×2PSNR: 34.58 -> 34.67Improved
Share weights across scales instead of using un-shared weights at 1.2M parameters (ablation)multi-scale model · Super-resolution ×2PSNR: 34.68 -> 34.77Improved
Use a larger un-shared model instead of the 1.2M shared model (ablation)multi-scale model · Super-resolution ×2PSNR: 34.77 -> 34.77No clear effect
Use bilinear resampling rather than convolution/deconvolution (ablation)SCN · Super-resolution ×2PSNR: 34.77 -> 34.80Improved
Use bilinear resampling rather than average/nearest resampling (ablation)SCN · Super-resolution ×2PSNR: 34.78 -> 34.80Improved
Use multi-scale modeling rather than single-scale modeling (ablation)SCN · Super-resolution ×2PSNR: 34.74 -> 34.67Improved
Increase the number of scales beyond the best setting (ablation)SCN · Image restorationPSNRHurt
Train with limited scales, then evaluate with a mismatched number of scales (ablation)SCN · Image restorationPSNRHurt
Train with enough scales and evaluate with additional scales (ablation)SCN · Image restorationPSNRImproved
Mismatch scale ratios between training and evaluation (ablation)SCN · Image restorationPSNRHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy