Image-to-image
Starting from an existing image instead of a description. Posed as a general problem rather than a set of separate tasks: the network learns the mapping from input image to output image and, unusually, learns the loss function that trains it too.
Official source: Image-to-Image Translation with Conditional Adversarial Networks →
That second half is what let one approach cover problems — photos from label maps, objects from edge maps, colourisation — that each demanded their own hand-built objective.
Sources 1 source on record · Automated review to redo
Sources
What this page is based on — every source is verified, and links out whenever the document is still reachable.