RankSRGAN: Generative Adversarial Networks with Ranker for Image Super-Resolution
- Python 3 (Recommend to use Anaconda)
- PyTorch >= 1.0.0
- NVIDIA GPU + CUDA
- Python packages:
pip install numpy opencv-python lmdb
- [option] Python packages:
pip install tensorboardX, for visualizing curves.
- We update the codes version based on mmsr. The old version can be downloaded from Google Drive
- This version is under testing. We will provide more details of RankSRGAN in near future.
How to Test
- Clone this github repo.
git clone https://github.com/WenlongZhang0724/RankSRGAN.git cd RankSRGAN
- Place your own low-resolution images in
- Download pretrained models from Google Drive. Place the models in
./experiments/pretrained_models/. We provide three Ranker models and three RankSRGAN models (see model list).
- Run test. We provide RankSRGAN (NIQE, Ma, PI) model and you can config in the
python test.py -opt options/test/test_RankSRGAN.yml
- The results are in
How to Train
- Download DIV2K and Flickr2K from Google Drive or Baidu Drive
- Generate rank dataset ./datasets/generate_rankdataset/
- Run command:
python train_rank.py -opt options/train/train_Ranker.yml
We use a PSNR-oriented pretrained SR model to initialize the parameters for better quality.
- Prepare datasets, usually the DIV2K dataset.
- Prerapre the PSNR-oriented pretrained model. You can use the
mmsr_SRResNet_pretrain.pthas the pretrained model that can be downloaded from Google Drive.
- Modify the configuration file
- Run command:
python train.py -opt options/train/train_RankSRGAN.yml
python train_niqe.py -opt options/train/train_RankSRGAN.yml
Using the train.py can output the convergence curves with PSNR; Using the train_niqe.py can output the convergence curves with NIQE and PSNR.