How it works...
The generator (net) generates new image instances, while the discriminator evaluates them for authenticity.
The adversarial training focuses on the system's weaknesses, forcing each other to improve over time (the generator gets better at generating realistic images and the discriminator gets better at differentiating the real from the fake images), until the fake images generated by the generator become indistinguishable from the real images and the discriminator can't differentiate them (at this point, the generator has learned how to generate a good image).
The discriminator is just a binary classifier that gets two batches of images as input (from real training data and from the generator) and figures out whether an image ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access