A team at UC Berkeley has proposed a new technique to leverage deep networks and AI, which allows novices, even those with limited artistic ability, to quickly produce reasonable results. They have written a software that knows the common colors for different objects and makes appropriate recommendations to the user. It then enables the user to provide guidance by adding colored points, or “hints”, which the system then propagates to the rest of the image. Even with minimal training and limited time—just one minute per image—users quickly learned how to produce colorizations that often fooled real human judges in a real vs. fake test scenario.
Though old B&W family photos have a charm all their own, it’s always interesting to add colors to them. Adding colors to skin, eyes and hair (where the colors are known) can be pretty straightforward (though still a time-consuming process that requires expertise), but adding colors to clothing becomes a guessing game.