Adventures in GANLand
Compete against neural networks to learn how they create realistic images
What is a GAN?
Generative adversarial networks (GANs) are machine learning algorithms comprised of dueling neural networks. Their competition leads to the creation of complex outputs, such as photorealistic images. If you’ve ever seen a deepfake, you’ve seen what GANs are capable of. There are two neural networks within a GAN: a generator that generates images, and a discriminator that attempts to distinguish real images from generated ones when shown them.
Become the Discriminator
Your task is to take on the role of the discriminator, distinguishing between real photos of faces and fake ones created by the generator. In each round of the game, you will be shown a series of images and asked to distinguish the photos from the fakes. After each round, you will be shown how well you did as the discriminator using three key metrics, displayed in the following way —
Your sensitivity indicates the percentage of photos you identified correctly
Your specificity indicates what percentage of GANs you identified correctly
Your accuracy indicates what percent of your guesses were correct overall
The Feedback Loop
The feedback you receive is useful to both you as the discriminator, and to the generator itself. The better your performance, the more the generator will learn to improve its images in the next round. However, worse your performance, the less the GAN will improve over time. This means that how well the GAN does in this experience relies, in part, on your abilities! The rounds will keep going until you’ve ‘finished’ training the GAN, and they will get harder as train your GAN further. Your goal is to be the best discriminator you can be, so that the GAN can eventually produce truly photorealistic images. The better feedback you provide, the better the GANs output will be — how far will you get?