I have a method for defining different regions in grayscale images that labels each instance of a region with a unique ID integer that ranges from 0 to N_objects - 1.
I would like to train a Mask RCNN model on this data and do predictions to perform instance segmentation.
I have been trying to use the imantics package to do this (see the minimal worked example in this GitHub question), but I'm running into this error when trying to add multiple Image objects to a Dataset object: "RecursionError: maximum recursion depth exceeded in comparison".
What do I need to do? If it is indeed possible to do in imantics, how can I do that? What other packages do people use for converting between masks and COCO style annotations?
Thank you!!