On the role of abductive reasoning in semantic image segmentation

cover