Self-mentoring: A new deep learning pipeline to train a self-supervised U-net for few-shot learning of bio-artificial capsule segmentation

•A deep learning pipeline is introduced for segmentation from very few annotated images.•A referee network is trained on purely synthetic data.•A referee network leverages the structure of binary mask images for self-supervision.•A trainee network is meant to produce referee-approved segmentation masks.•On micro-capsule data, self-mentoring outperforms data augmentation and self training.