This is an official implementation of RealDAN
If this repo works for you, please cite our paper
@article{luo2023end,
title={End-to-end Alternating Optimization for Real-World Blind Super Resolution},
author={Luo, Zhengxiong and Huang, Yan and Li, Shang and Wang, Liang and Tan, Tieniu},
journal={International Journal of Computer Vision (IJCV)},
year={2023}
}
This repo is buid on the basis of BasicSR
Download the checkpoints of RealDAN from BaiduYun(password: ig96).
Put the downloaded checkpoints into checkpoints
The model weights and datasets are also available at huggingface(https://huggingface.co/lzxlog/RealDAN) now.
For inference on Real-World images
cd codes/config/RealDAN
python3 inference.py \
--opt options/test/dan_edsr_gan_real.yml \
--input_dir=/dir/of/input/images \
--output_dir=/dir/of/saved/outputsFor inference on blurry images
cd codes/config/KernelDAN
python3 inference.py \
--opt options/test/x4.yml \
--input_dir=/dir/of/input/images \
--output_dir=/dir/of/saved/outputsFor evaluation on DIV2K-Real, please download the dataset to your own path, and run
cd codes/config/RealDAN
python3 test.py \
--opt options/test/dan_edsr_gan_syn.ymland
cd codes/config/RealDAN
python3 test.py \
--opt options/test/dan_edsr_syn.ymlFor evaluation on DIV2KRK, please download the dataset to your own path, and run
cd codes/config/KernelDAN
python3 test.py \
--opt options/test/x2.yml and
cd codes/config/KernelDAN
python3 test.py \
--opt options/test/x4.yml