Data Scarcity - What is data augmentation in deep learning?

The goal of data augmentation is to systematically increase the size of the training set by generating new, modified datasets from the original data. Modifying the dataset slightly or creating additional data points using deep learning are two possible approaches.