Include the detected object class in the classification process.
Reason: We already have that information from the object detector and right now, the network has to learn from the mask to do exactly the same thing over again. So, adapt the network architecture to allow feeding in the classes of the two objects that should be classified to also be used as input.
Include the detected object class in the classification process.
Reason: We already have that information from the object detector and right now, the network has to learn from the mask to do exactly the same thing over again. So, adapt the network architecture to allow feeding in the classes of the two objects that should be classified to also be used as input.