Hands-On Artificial Intelligence for Beginners

Discriminator network

The discriminator network in image-related GANs is a standard convolutional neural network. It takes in an image and outputs a single number that tells us whether the image is real or fake. The discriminator takes in an image, and learns the attributes of that image so that it may be a good judge vis-à-vis the outputs of the generator. In TensorFlow, we can create the discriminator as a function that we will then run in a TensorFlow session later on. This framework is more or less the same as you've seen in the previous sections with autoencoder and variational autoencoders; we'll use the higher level tf.layers api to create three main network layers and an output layer. After each of the main network layers, we'll add ...