We define a good quality data set as a dataset which has over 500 images, annotations, and clear images. A quality dataset will sometimes be split into a training, validation and testing dataset.