MATIONS FOR DATA AUGMENTATION

This work learns data augmentation directly from the training data by learning to transform images with an encoderdecoder architecture combined with a spatial transformer network.