If you wanted to see if an image is fake, or wanted to make one yourself, it usually starts with some kind of photo editing software. MIT’s GANpaint Studio is a tool that aims to make things a lot easier, thanks to artificial intelligence. For those who don’t know, a generative adversarial network (GAN) is a type of artificial intelligence machine learning technique made up of two nets that are in competition with one another in a zero-sum game framework. How does it work? Simply tell the tool where you want the object and it uses neural networks to insert one to match scene. Read more for a video demonstration and additional information.
David Bau, a PhD student at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL), touts the project as one of the first times computer scientists have been able to actually “paint with the neurons” of a neural network. It’s currently available online as an interactive demo and lets users upload an image of their choosing and modify multiple aspects of its appearance.
Whether it be changing the size of objects to adding completely new items, like buildings, GANpaint Studio should do the trick. “Right now, machine learning systems are these black boxes that we don’t always know how to improve, kind of like those old TV sets that you have to fix by hitting them on the side,” said David Bau, lead author on a related paper about the system with a team overseen by Torralba. Download the tool here.