ML Kit object detection using custom model how to get bounding box coordinate?

Question

Custom object detection model: mobile_object_labeler_v1

outputs:

Name	Type	Description	Shape	Min/Max
probability	Feature <uint8>	Probabilities of the outputs classes.	[1,631]	[0]/[1]

enter image description here

there is only each class probabilities and no bounding box coordinate in outputs, so how to get the coordinate? Hope someone can help me!