The generator is the fun part of this structure. The generator is going to take inputs from the latent space (a sample from a normal distribution in this recipe) and produce realistic looking data. The generator will also be added to the adversarial part of the training. The GAN will take in latent examples with labels and train on that until the generator itself is able to produce realistic looking images. We'll see some examples of the generated images in the near future.
Explaining your second GAN component – generator
Getting ready
As with the discriminator development, the important part of this recipe is that you have the appropriate folder structure and the discriminator.py file. Testing each of these components...
