Generative adversarial networks (GANs) are a class of CNN that learns to estimate the probability distribution of the data. A GAN consists of two competing connected neural networks called a generator and a discriminator. The generator generates an artificial image based on the noise input of the image feature and the discriminator compares the artificial image with the real image to determine the probability that the image is real. The probability information is passed to the image input to learn in the next stage. The following image illustrates the mechanism of a GAN:
The step-by-step explanation of the GAN algorithm is listed as follows:
- Given a training set, z, the generator network takes in a random vector representing the feature of the image and runs through a CNN to produce an artificial image, G(z).
- The discriminator network is a binary classifier...
