In this module, we will include the hyperparameters, preprocess the data, generate synthetic data, train the gan, train the CNN classifier, and import all the necessary functions from other modules.
Figure 14.17: Python module train_gan.py
You can use the same modules you have created to train on fashion MNIST data. All you have to do is replace line 11 in train_gan.py with (from keras.datasets import fashion_mnist) and replace line 28 with ((X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()). The results will be good but won't be excellent since the parameters set here work best on the MNIST digit data. This will be a good exercise for you to get incredible results without much effort.
Here is a resource on tips to train GAN's you must check out. https://github.com/soumith/ganhacks
Jupyter notebook for the above DCGAN MNIST inpainting - https://github.com/PacktPublishing/Python-Deep-Learning-Projects/blob/master/Chapter%2014/DCGAN_MNIST.ipynb
Jupyter notebook for the DCGAN Fashion MNIST inpainting - https://github.com/PacktPublishing/Python-Deep-Learning-Projects/blob/master/Chapter%2014/DCGAN_Fashion_MNIST.ipynb