Generative Adversarial Networks Projects

The adversarial model

To create an adversarial model, take both the generator and discriminator networks and create a new Keras model.

1. Start by creating three input layers to feed inputs to the network:

def build_adversarial_model(gen_model, dis_model):    input_layer = Input(shape=(1024,))    input_layer2 = Input(shape=(100,))    input_layer3 = Input(shape=(4, 4, 128))

2. Then, use the generator network to generate low-resolution images:

    # Get output of the generator model    x, mean_logsigma = gen_model([input_layer, input_layer2])    # Make discriminator trainable false    dis_model.trainable = False

3. Next, use the discriminator network to get the probabilities:

    # Get output of the discriminator models    valid = dis_model([x, input_layer3])

4. Finally, ...