Synthetic images with very low aesthetic scores from a variety of dataset, compressed to comical levels for training against.