Research map / Architectures and efficiency / Attention and transformers
Devil is in the Uniformity: Exploring Diverse Learners within Transformer for Image Restoration
Shihao Zhou, Dayu Li, Jinshan Pan, Juncheng Zhou, Jinglei Shi, Jufeng 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 |
|---|---|---|---|
| use HMHA instead of W-MSA (ablation) | HINT · LOL-v2-syn | PSNR: 24.19 -> 27.17 | Improved |
| use HMHA instead of MDTA (ablation) | HINT · LOL-v2-syn | PSNR: 26.42 -> 27.17 | Improved |
| use similarity reranking instead of random shuffle before the split (ablation) | HMHA · LOL-v2-syn | PSNR: 26.54 -> 27.17 | Improved |
| add reranking and hierarchical split versus no ranking (ablation) | HMHA · LOL-v2-syn | PSNR: 26.42 -> 27.17 | Improved |
| add QKCU (IntraCache and InterCache) (ablation) | HINT · LOL-v2-syn | PSNR: 26.47 -> 27.17 | Improved |
| add IntraCache only (ablation) | HINT · LOL-v2-syn | PSNR: 26.47 -> 26.67 | Improved |
| add InterCache only (ablation) | HINT · LOL-v2-syn | PSNR: 26.47 -> 26.72 | Improved |
| add InterCache on top of IntraCache (ablation) | HINT · LOL-v2-syn | PSNR: 26.67 -> 27.17 | Improved |
| add IntraCache on top of InterCache (ablation) | HINT · LOL-v2-syn | PSNR: 26.72 -> 27.17 | Improved |
| add QKCU modules (ablation) | HINT · LOL-v2-syn | parameters (M): 21.34 -> 24.87 | Hurt |
| add reranking to HMHA (ablation) | HMHA · LOL-v2-syn | parameters (M): 24.87 -> 24.87 | No clear effect |
| apply HINT enhancement before detector | YOLO-v3 · ExDark | mean AP: 45.1 -> 52.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MambaLLIE — MambaLLIE: Implicit Retinex-Aware Low Light Enhancement with Global-then-Local State Space (NeurIPS 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Sharing Key Semantics in Transformer Makes Efficient Image Restoration (NeurIPS 2024 · similarity 0.87)
- ReFIR: Grounding Large Restoration Models with Retrieval Augmentation (NeurIPS 2024 · similarity 0.87)
- ACL: Activating Capability of Linear Attention for Image Restoration (CVPR 2025 · similarity 0.87)
- Complexity Experts are Task-Discriminative Learners for Any Image Restoration (CVPR 2025 · similarity 0.86)
- UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement (CVPR 2026 · similarity 0.86)
- Boosting Image Restoration via Priors from Pre-trained Models (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.
