- torch>=0.4.0
- torchvision
- matplotlib
- numpy
- scipy
- pillow
- urllib3
- scikit-image
Pretrained model can be downloaded by link :
https://pan.baidu.com/s/1wW_3dlWMcSgkb4zXGW9azQ
password :up7m
The pretrained model should put in the folder 'saved_models'.
Model testing samples are in the folder 'test'
You can test the pretrained model by runnning
python model_test.py
the result is in the folder test_result. In addtion, the first image is the dirty image, the middle image is the deconved image and the last one is the ground truth.
If you want to train the model on your own dataset, you can split your data into three folder named train, val and test.
Run example:
python solar_deconv.py --dataset 'your dataset name'
Our code is based on the https://github.com/eriklindernoren/PyTorch-GAN
Thanks for the author.