Research map / Image generation / Super-resolution
Self-Supervised Learning for Real-World Super-Resolution from Dual Zoomed Observations
Zhilu Zhang, Ruohao Wang, Hongzhi Zhang, Yunjin Chen, Wangmeng Zuo
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 self-supervised fine-tuning instead of SRA fine-tuning (ablation) | SelfDZSR · Nikon | PSNR: 0.27 dB higher | Improved |
| train DCSR with our self-supervised approach (ablation) | DCSR · CameraFusion | PSNR: 0.31 dB PSNR gain | Improved |
| add auxiliary-LR alignment of LR (ablation) | SelfDZSR · Nikon | PSNR: 28.48 -> 28.67 | Improved |
| add Ref alignment with auxiliary-LR (ablation) | SelfDZSR · Nikon | PSNR: 28.48 -> 28.61 | Improved |
| use auxiliary-LR generator instead of bicubic-downsampled GT (ablation) | SelfDZSR · Nikon | PSNR: PSNR drops by 0.47 dB, LPIPS gets worse by 0.165 | Improved |
| set position preserving loss coefficient to 100 (ablation) | SelfDZSR · Nikon | PSNR: 28.22 -> 28.67 | Improved |
| raise position preserving coefficient from 100 to 10000 (ablation) | SelfDZSR · Nikon | PSNR: 28.67 -> 28.28 | Hurt |
| use AdaSTN instead of deformable convolution (ablation) | SelfDZSR · Nikon | PSNR: 28.52 -> 28.67 | Improved |
| use AdaSTN instead of STN (ablation) | SelfDZSR · Nikon | PSNR: 28.57 -> 28.67 | Improved |
| set P=0 in AdaSTN (ablation) | SelfDZSR · Nikon | inference time: increases speed by ~0.2 seconds without performance dropping | No clear effect |
| SW loss instead of perceptual+adversarial loss (ablation) | SelfDZSR · Nikon | PSNR: 28.45 -> 28.67 | Improved |
Closest papers by meaning
- Self-Supervised Learning for Real-World Super-Resolution From Dual and Multiple Zoomed Observations (TPAMI 2024 · similarity 0.97)
- Dual-Camera Super-Resolution With Aligned Attention Modules (ICCV 2021 · similarity 0.92)
- LMR: A Large-Scale Multi-Reference Dataset for Reference-Based Super-Resolution (ICCV 2023 · similarity 0.89)
- Robust Reference-Based Super-Resolution With Similarity-Aware Deformable Convolution (CVPR 2020 · similarity 0.88)
- Toward Real-World Single Image Super-Resolution: A New Benchmark and a New Model (ICCV 2019 · similarity 0.88)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.87)
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.
