Research map / Architectures and efficiency / Attention and transformers
Efficient Concertormer for Image Deblurring and Beyond
Pin-Hung Kuo, Jinshan Pan, Shao-Yi Chien, Ming-Hsuan Yang
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 |
|---|---|---|---|
| replace MLP FFN with gated-dconv MLP (ablation) | Concertormer baseline · GoPro | PSNR: 26.28 -> 32.35 | Improved |
| add spatial Ripieno to spatial Concertino (ablation) | Concertormer-lite · GoPro | PSNR: 32.58 -> 33.11 | Improved |
| add channel Ripieno to channel Concertino (ablation) | Concertormer-lite · GoPro | PSNR: 32.65 -> 33.00 | Improved |
| add full spatial and channel CSA to gdMLP baseline (ablation) | Concertormer-lite · GoPro | PSNR: 32.35 -> 33.20 | Improved |
| add simplified channel attention (ablation) | Concertormer-lite · GoPro | PSNR: 33.20 -> 33.31 | Improved |
| add multi-head attention (ablation) | Concertormer-lite · GoPro | PSNR: 33.31 -> 33.35 | Improved |
| add cross-dimensional communication (ablation) | Concertormer-lite · GoPro | PSNR: 33.35 -> 33.53 | Improved |
| use single-stage gdMLP instead of two-stage CSA plus FFN (ablation) | Concertormer-lite · GoPro | PSNR: 31.90 -> 33.53 | Improved |
| use concatenation with linear projection (Concerto) instead of addition (Prototype) (ablation) | CSA · GoPro | PSNR: 32.50 -> 33.16 | Improved |
| omit the global mean in CDC (ablation) | CSA · GoPro | PSNR: 33.23 -> 33.30 | Improved |
| use shifted windows with spatial Concertino (ablation) | CSA · GoPro | SSIM: 0.953 -> 0.950 | Mixed |
| replace Restormer transposed self-attention with CSA | Restormer · GoPro | PSNR: 32.92 -> 33.32 | Improved |
Closest papers by meaning
- MLP Can Be A Good Transformer Learner (CVPR 2024 · similarity 0.86)
- PADRe: A Unifying Polynomial Attention Drop-in Replacement for Efficient Vision Transformer (ICLR 2025 · similarity 0.86)
- Sharing Key Semantics in Transformer Makes Efficient Image Restoration (NeurIPS 2024 · similarity 0.86)
- Attention-Only Transformers via Unrolled Subspace Denoising (ICML 2025 · similarity 0.86)
- Transformer-VQ: Linear-Time Transformers via Vector Quantization (ICLR 2024 · similarity 0.86)
- Skip-Attention: Improving Vision Transformers by Paying Less Attention (ICLR 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.
