This project was done for UIUC’s Computer Vision course (CS 543). Bias in vision datasets is a well known issue that has been explored previously, but we thought it could be novel to explore using Generative Adversarial Networks (GANs) for generating an augmented dataset. Since GANs generate brand new data from current data, it would be great if a GAN could be used to generate new minority images that could be added to a biased dataset to make it less biased.