Generative Adversarial Networks Projects

The discriminator network

The discriminator network is a classifier network; it contains a set of downsampling layers and classifies whether a given image is real or fake.

Let's start by writing the code for the network:

1. Start by creating an input layer to feed input to the network:

input_layer = Input(shape=(64, 64, 3))

2. Next, add a 2D convolutional layer with the following parameters:
   • Filters = 64
   • Kernel size = (4, 4)
   • Strides = 2
   • Padding = 'same'
   • Use bias = False
   • Activation = LeakyReLU with alpha=0.2

stage1_dis = Conv2D(64, (4, 4),                    padding='same', strides=2,                    input_shape=(64, 64, 3), use_bias=False)(input_layer)stage1_dis = LeakyReLU(alpha=0.2)(stage1_dis)

3. After that, add two convolutional layers, each followed by a batch normalization ...