I’m following the tutorial of object detection using mrcnn on the website. And I noticed that there is a ‘score’ as one of the outputs of model. So I wonder how does this ‘score’ being calculated? Is it concerned about one or more of object, classification, mask or anything else?
It is classification score of objects.
Thank you very much!
I fed this model with my own dataset and here is what I got:
‘labels’: tensor([1, 1, 1, 1, 1, 1], device=‘cuda:0’),
‘scores’: tensor([0.9990, 0.9987, 0.9987, 0.9889, 0.2024, 0.0583]], device=‘cuda:0’).
Why the labels of the last 2 are positive while their scores are so low? If the the score comes from the output of softmax classification, shouldn’t it be at least over 0.5 when only 2 classes, target and background?