How can I utilize sublabels to train my mask R-CNN with sublabel accuracy?

In the videoLabeler, MATLAB allows sublabeling parts within an object, e.g. headlights on a vehicle.

How is this sublabelling utilized for training a network to recognize the objects and their parts?

Is there any implementation for the training with sublabels in MATLAB now?

Thank you!

Answer 1

[Jacob Buerkle]

Dear @Yi-Ping Hsueh, I have the same questions

Have you got anyanswers or found it out by your own?

Thank you!

Answer 2

[yanqi liu]

sir，i think may be use multi class label to replace the sub label