Generative models are trained to generate more data similar to the one they are trained on, and adversarial models are trained to distinguish the real versus fake data by providing adversarial examples.
The Generative Adversarial Networks (GAN) combine the features of both the models. The GANs have two components:
- A generative model that learns how to generate similar data
- A discriminative model that learns how to distinguish between the real and generated data (from the generative model)
GANs have been successfully applied to various complex problems such as:
- Generating photo-realistic resolution images from low-resolution images
- Synthesizing images from the text
- Style transfer
- Completing the incomplete images and videos
In this chapter, we shall study the following topics for learning how to implement GANs in TensorFlow and Keras:
- Generative...
