While taking the fast.ai course, I was learning about Variational Autoencoders (VAE) and began to wonder:
Is it possible to represent the latent space as an image, and then reconstruct the original image from that representation?
Well, here's the answer.