Research map / Image generation / Super-resolution
UCAN: Unified Convolutional Attention Network for Expansive Receptive Fields in Lightweight Super-Resolution
Cao Thien Tan, Phan Thi Thu Trang, Do Nghiem Duc, Ho Ngoc Anh, Hanyang Zhuang, Nguyen Duc Dung
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 HPA block (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 32.90 -> 33.22 | Improved |
| use 32x32 HPA window instead of 16x16 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.04 -> 33.22 | Improved |
| use WMHA window 16x16 instead of 8x8 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.02 -> 33.22 | Improved |
| use Hedgehog feature map instead of ReLU in DFL (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.16 -> 33.22 | Improved |
| use Hedgehog feature map instead of ELU+1 in DFL (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.16 -> 33.22 | Improved |
| use default LKD kernel size instead of 5x5 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.12 -> 33.22 | Improved |
| use default LKD kernel size instead of 47x47 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.15 -> 33.22 | Improved |
| use semi-sharing instead of full sharing (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 32.89 -> 33.22 | Improved |
| increase hybrid block depth from 1 to 3 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 32.71 -> 33.05 | Improved |
| increase LKD depth from 1 to 4 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 32.99 -> 33.12 | Improved |
| increase LKD kernel size from 5 to 65 (ablation) | UCAN · Urban100 x2 (DIV2K 400K iters ablation) | PSNR: 33.12 -> 33.19 | Improved |
| increase LKD kernel size from 5 to 65 (ablation) | UCAN · Set5 x2 (DIV2K 400K iters ablation) | PSNR: 38.33 -> 38.33 | No clear effect |
Closest papers by meaning
- Emulating Self-attention with Convolution for Efficient Image Super-Resolution (ICCV 2025 · similarity 0.89)
- Single Image Super-Resolution via a Holistic Attention Network (ECCV 2020 · similarity 0.89)
- Image Super-Resolution Using Very Deep Residual Channel Attention Networks (ECCV 2018 · similarity 0.88)
- Efficient Attention-Sharing Information Distillation Transformer for Lightweight Single Image Super-Resolution (AAAI 2025 · similarity 0.88)
- Image Super-Resolution Using Deep Convolutional Networks (TPAMI 2016 · similarity 0.88)
- Transcending the Limit of Local Window: Advanced Super-Resolution Transformer with Adaptive Token Dictionary (CVPR 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.
