Generative Adversarial Networks (GANs) are a class of machine learning frameworks that automatically studies the regularities in the input data and uses the pattern to generate new datasets from the original one. They are composed of two parts: a generator model and a discriminator model. The purpose of a generator model is to generate data, while the purpose of a discriminator model is to distinguish between the generated and real data. An adversarial network is formed by connecting the discriminator outputs to the generator, allowing the models to compete.

Though a relatively new technology, GANs have already made revolutionary impacts in areas such as image content creation and artistic editing. However, most of these GANs are limited to image-to-image transformations, which often require large datasets and high processing power, putting them out of reach from ordinary users. In this project, we attempt to customize many such GAN models (including a novel sketch-to-image GAN) to an accessible mobile iOS format so that users can easily interact with them via our app.

Team Members

Savi Hanagud, Abigail Hu, Cathy Le, Saaketh Rao, Jeffrey Wallace, Ming Zhao