RealSRT

Paper | BibTex

PyTorch implementation of ECCV 2024 paper "Confidence-Based Iterative Generation for Real-World Image Super-Resolution"

Introduction

Visualizations of our confidence-based iterative generation process for real-world SR.

Method

Overview of RealSRT.

Installation

This implementation is based on BasicSR

git clone https://github.com/USTC-JialunPeng/RealSRT
cd RealSRT
pip install -r requirements.txt
python setup.py develop

Inference

1. Download the pre-trained model and place it in ./experiments/pretrained_models/

2. Download the test dataset (e.g., RealSR), place input images in /data/input/ and place target images (if available) in /data/target/

3. Testing

python inference_realsrt.py --input /data/input/ --output /data/results/ --model_path experiments/pretrained_models/net_g_80000.pth

4. To reproduce scores in Table 1, run

python calculate_metrics.py

Citing

If our method is useful for your research, please consider citing.

@inproceedings{peng2024confidence,
  title={Confidence-Based Iterative Generation for Real-World Image Super-Resolution},
  author={Peng, Jialun and Luo, Xin and Fu, Jingjing and Liu, Dong},
  booktitle={European Conference on Computer Vision},
  year={2024}
}