Research map / Image generation / Super-resolution
Meta-SR: A Magnification-Arbitrary Network for Super-Resolution
Xuecai Hu, Haoyuan Mu, Xiangyu Zhang, Zilei Wang, Tieniu Tan, Jian Sun
Check 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 Meta-Upscale to predict scale- and coordinate-conditioned filter weights instead of fixed-scale upscaling | RDN · B100, arbitrary scales X1.5–X4.0 | PSNR: 35.57 -> 35.86 (X1.5); 32.08 -> 32.35 (X2.0); 29.03 -> 29.30 (X3.0); 27.47 -> 27.75 (X4.0) | Improved |
| Use the learned weight prediction network rather than a shared fixed convolution filter (ablation) | BiConv · B100, arbitrary scales X1.1–X4.0 | PSNR: 31.85 -> 32.18 (X2.0); 28.91 -> 29.22 (X3.0) | Improved |
| Add location projection and direct feature mapping rather than interpolate the feature maps (ablation) | Meta-Bi · B100, arbitrary scales X1.1–X4.0 | PSNR: 32.18 -> 32.35 (X2.0); 29.22 -> 29.30 (X3.0) | Improved |
| Fine-tune Meta-RDN separately for each integer scale (ablation) | Meta-RDN · Single integer scale factors | PSNR | No clear effect |
| Increase the weight-prediction network hidden width beyond 256 neurons (ablation) | Meta-Upscale · Super-resolution | performance | No clear effect |
| Use two fully connected layers in the weight-prediction network (ablation) | Meta-Upscale · Super-resolution | performance and speed balance | Improved |
| Use a 3 × 3 rather than a 5 × 5 convolution kernel (ablation) | Meta-Upscale · Super-resolution | kernel-size performance choice | Improved |
Closest papers by meaning
- Learning a Single Network for Scale-Arbitrary Super-Resolution (ICCV 2021 · similarity 0.90)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.89)
- Arbitrary-Scale Image Generation and Upsampling using Latent Diffusion Model and Implicit Neural Decoder (CVPR 2024 · similarity 0.87)
- TSD-SR: One-Step Diffusion with Target Score Distillation for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.87)
- Chain-of-Zoom: Extreme Super-Resolution via Scale Autoregression and Preference Alignment (NeurIPS 2025 · similarity 0.87)
- Continuous Optical Zooming: A Benchmark for Arbitrary-Scale Image Super-Resolution in Real World (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.
