[wlug] Image-To-Image Translation In TensorFlow

This page <http://affinelayer.com/pixsrv/index.html> lets you create schematic renditions of various kinds of objects, and then synthesize corresponding photorealistic representations of those objects, based on pieces of images taken from the training databases. You can draw your own outlines, or just click the “Random” buttons to see what it does. Results range from entertaining to disturbing.