Is it possible to perform Instance Segmentation on multiple objects of the same type?

I've been reading the Semantic Segmentation tutorial and was wondering if it's possible to also do Instance Segmentation where each animal or person is given a unique mask (like an R-CNN bounding box).

If not, is it possible to extend the trainRCNNObjectDetector in to a Mask R-CNN?

I am trying to detect multiple objects of the same type and want to be able to save their masks separately, so in the above image I'd want to save each of dog's mask. I

I've trained an R-CNN and can save each bounding box but would prefer to get a more exact object shape in order to help with background subtraction.

Thanks.