In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. Olaf Ronneberger, et al., "U-Net: Convolutional Networks for Biomedical I…