Fig. 1From: Modelling urban networks using Variational AutoencodersVariational Autoencoder takes as input an image of the street network (left), condenses the image to a lower-dimensional encoding (middle) and finally reconstructs the image given the encoding (right)Back to article page