Research map / Image generation / Image restoration
Dynamic Attentive Graph Learning for Image Restoration
Chong Mou, Jian Zhang, Zhuoyuan Wu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add dynamic KNN module (adaptive neighbor threshold) (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 31.28 -> 31.39 | Improved |
| add graph attention weighting instead of averaging (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 30.77 -> 31.39 | Improved |
| use patch-wise graph instead of single-head pixel-wise non-local (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 30.73 -> 31.39 | Improved |
| use patch-wise graph instead of multi-head pixel-wise non-local (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 30.92 -> 31.39 | Improved |
| add GFAM modules (4 heads, 3 stages) to plain ResNet (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 30.43 -> 31.39 | Improved |
| increase heads from 1 to 4 at 3 stages (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 31.21 -> 31.39 | Improved |
| increase heads from 4 to 5 at 3 stages (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 31.39 -> 31.41 | No clear effect |
| increase stages from 3 to 4 at 4 heads (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 31.39 -> 31.42 | No clear effect |
| increase stages from 1 to 3 at 4 heads (ablation) | DAGL · Urban100 denoising (sigma=25) | PSNR: 31.05 -> 31.39 | Improved |
| DAGL versus GCDN graph denoiser at sigma=50 | DAGL · Urban100 gray denoising (sigma=50) | PSNR: 27.41 -> 27.97 | Improved |
| DAGL versus RNAN on demosaicing | DAGL · Urban100 demosaicing | PSNR: 39.75 -> 40.20 | Improved |
| DAGL versus DeamNet on real denoising | DAGL · DND real denoising | PSNR: 39.70 -> 39.83 | Improved |
Closest papers by meaning
- Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration (ECCV 2020 · similarity 0.88)
- Non-Local Recurrent Network for Image Restoration (NeurIPS 2018 · similarity 0.86)
- Cross-Scale Internal Graph Neural Network for Image Super-Resolution (NeurIPS 2020 · similarity 0.86)
- DRCNet: Dynamic Image Restoration Contrastive Network (ECCV 2022 · similarity 0.86)
- Residual Non-local Attention Networks for Image Restoration (ICLR 2019 · similarity 0.86)
- Adaptive Dynamic Filtering Network for Image Denoising (AAAI 2023 · similarity 0.85)
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.
