My current research focuses on incorporating prior knowledge into deep neural networks and exploring how to make neural segmentation nets more shape-aware.