I am testing the “detecnet” code of the jetson-inference repository.
Is it the classification confidence?, or the localization confidence (like IoU of the boundng boxes)?
Can somebody explain how that value works? Is it related to the output of the last layer in the network?
Thanks in advance!