Research map / Image generation / Super-resolution
Robust Reference-Based Super-Resolution With Similarity-Aware Deformable Convolution
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SSEN to align and extract reference features | RefSR on CUFED5 · CUFED5 | PSNR: 26.24 -> 26.78 | Helped |
| use SSEN for self-similarity SR | RCAN · Urban100 | PSNR: 26.61 -> 26.71 | Helped |
| use SSEN for self-similarity SR | RCAN · Set5 | PSNR: 32.46 -> 32.48 | No clear effect |
| use SSEN for self-similarity SR | RCAN · Set14 | PSNR: 28.79 -> 28.84 | No clear effect |
| use SSEN for self-similarity SR | RCAN · B100 | PSNR: 27.69 -> 27.72 | No clear effect |
| use SSEN instead of SRNTT-ℓ2 | RefSR · CUFED5 at XH, H, M, L, and XL similarity | PSNR: 26.06 -> 26.78; 25.97 -> 26.52; 25.90 -> 26.48; 25.88 -> 26.42; 25.87 -> 26.41 | Helped |
| remove all deformable convolution layers (ablation) | SSEN · CUFED5 XH | PSNR: 26.42 -> 26.78 | Helped |
| use 1 deformable convolution layer rather than 3 (ablation) | SSEN · CUFED5 XH | PSNR: 26.78 -> 26.59 | Hurt |
| use 2 deformable convolution layers rather than 3 (ablation) | SSEN · CUFED5 XH | PSNR: 26.78 -> 26.64 | Hurt |
| use 4 deformable convolution layers rather than 3 (ablation) | SSEN · CUFED5 XH | PSNR: 26.78 -> 26.70 | Hurt |
| add non-local blocks to the dynamic offset estimator (ablation) | SSEN · CUFED5 XH | PSNR: 26.68 -> 26.78 | Helped |
Closest papers by meaning
- MASA-SR: Matching Acceleration and Spatial Adaptation for Reference-Based Image Super-Resolution (CVPR 2021 · similarity 0.90)
- Transformer-Based Selective Super-resolution for Efficient Image Refinement (AAAI 2024 · similarity 0.86)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.86)
- See More Details: Efficient Image Super-Resolution by Experts Mining (ICML 2024 · similarity 0.85)
- Universal Robustness via Median Randomized Smoothing for Real-World Super-Resolution (CVPR 2024 · similarity 0.85)
- Beyond Image Super-Resolution for Image Recognition with Task-Driven Perceptual Loss (CVPR 2024 · 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.