1. GENERATIVE AI

Generative Adversarial Networks (GANs)

The fake images on the previous page were generated by a type of Generative Adversarial Network (GAN) called StyleGAN. Over the years, many artificial neural network architectures have been developed to train models capable of generating content. One of the most notable is the architecture of Generative Adversarial Networks (GANs). Generative Adversarial Networks (GANs) are used to create new instances of images, music, text, and videos.


!    

Interact with the following explanation to understand how computers learn a model capable of generating content such as realistic images using the GAN architecture. Also try out the new full-screen mode (top left).

 

 

 
 
*The above explanation refers to a basic GAN architecture. However, there are many variations of this architecture, such as StyleGAN.
 
 

Nowadays, there are other well-known artificial neural network architectures for image generation, besides GANs. Among the most popular are Variational Autoencoders and Diffusion Models.