Generating a New Reality: From Autoencoders and Adversarial Networks to Deepfakes

© Micheal Lanham 2021

M. LanhamGenerating a New Reality https://doi.org/10.1007/978-1-4842-7092-9_5

5. Image to Image Content Generation

Micheal Lanham¹  

(1)

Calgary, AB, Canada

 

In days of old, before cameras captured everything, if a crime were committed and it was particularly heinous, the police would send a sketch artist around to interview witnesses. The purpose of this was to build an image or likeness of the criminal and expedite an arrest.

Sketch artists would work by questioning witnesses about features, and the witness would answer yes or no to help the artist create a better likeness. This process is not unlike that of a GAN in that the artist is the generator and the witness is the discriminator. Over time, instead of relying on verbal ...