Long Beach, CA – Researchers at MIT and the Qatar Computing Research Institute have taught artificial intelligence to figure out the order in which a pizza should be constructed.
In the paper, “How to make a pizza:Learning a compositional layer-based GAN model,”How to make a pizza:Learning a compositional layer-based GAN model,” the researchers proposed PizzaGAN, a generative model that mirrors the pizza-making procedure.
They learned composable module operations (implemented with GANs) to add or remove a particular ingredient or even cook or uncook the input pizza.
The paper concluded, “Our experiments on both synthetic and real pizza images showed that our model (1) detects and segments the pizza toppings in a weakly-supervised fashion without any pixel-wise supervision, (2) fills in what has been occluded with what is underneath, and (3) infers the ordering of the toppings without any depth or-dering supervis ion.”
They believe a similar approach is promising for other types of foods that are naturally layered such as burgers, sandwiches and salads. It also may be applicable to digital fashion shopping assistants, where a key operation is the virtual combination of different layers of clothes.
Results were presented in a paper last week at an AI conference in Long Beach, Calif.
Print this page
