Here provides Super-Resolution with PyTorch Lightning.
Here is written in Python 3.7 and following packages are required(All of them are latest versions as of Aug. 29, 2019).
torch>=1.2
torchvision>=0.4
pytorch-lightning>=0.4.6
kornia>=0.1.3.post2
You can create virtual environment easily using pipenv like below:
pipenv install
[IMPORTANT] whl file of torch via pip requires CUDA9.2+.
To train SRGAN, run the following command. The other models will be added.
pipenv run python train.py --model srgan
You can set the other options, see train.py(common training settings) and models/srgan_model.py(model specific settings).
pipenv run python test.py --model srgan --ckpt ./logs/srgan/default/version_foo/media/_ckpt_epoch_bar.ckpt