De-mark GAN
data_read.py is the data reading file.
gen_image.py uses the trained de-mark GAN model to generate clean ID photos.
module.py is the demark GAN model file.
pre_model.py is the pre-trained resnet model file. You can use the file trained your own resnet model on the mscelba dataset to compute the feature loss.
'main.py' is the train file.
first setp: use the pre_model.py train the resnet on the mscelba dataset
second step: change the path in 'main.py' and 'data_read.py' to train the De-mark GAN model in you own datasets.
Prerequisites
- python 2.7
- pytorch 2.0
- PIL
- pickle