---

U Net

Autoencoder (AE) model, using Restnet/Conv layers followed by downsampling layers to encode to latent space, and upsampling layers followed by Resnet/Conv to decode.

Layers of the same resolution are passing information via residual connections.