Currently there are two ways to produce segmentation labels for various structures in the images that are loaded: