Keras Deep Learning Cookbook

Initializing the BGAN class

We created a custom class GAN, which is initialized with a generator and a discriminator. The steps followed are listed in the following points:

1. Initialize the variables img_rows, img_cols, channels, img_shape, and latent_dim
2. Initialize the optimizer; we are using the Adam optimizer in this case
3. Instantiate the discriminator:
   • Use build_discriminator()
   • Compile the discriminator with the loss function as binary_crossentropy, the optimizer as Adam, and the metrics as accuracy
4. Generator:
   • Instantiate using build_generator()
   • Get the generated image with noise as the input
5. Discriminator checks the validity of the images
6. Combined model: Used to fool the discriminator with the generator:
   • z with input shape (*, self.latent_dim) ...