Introduction to GANs

Objectives

• Construct your first GAN.
• Develop intuition behind GANs and their components.
• Examine real life applications of GANs.

Deep Convolutional GANs

Objectives

• Be able to explain the components of a Deep Convolutional GAN.
• Compose a Deep Convolutional GAN using these components.
• Examine the difference between upsampling and transposed convolutions.

Wasserstein GANs with Gradient Penalty

Objectives

• Examine the cause and effect of an issue in GAN training known as mode collapse.
• Implement a Wasserstein GAN with Gradient Penalty to remedy mode collapse.
• Understand the motivation and condition needed for Wasserstein-Loss.

Conditional GAN and Controllable Generation

• Control GAN generated outputs by adding conditional inputs.
• Control GAN generated outputs by manipulating z-vectors.
• Be able to explain disentanglement in a GAN.