I want to get the detection confidence in the Kafka metadata stream. I made changes to the code to send confidence in the metadata. While I can see the confidence scores in the metadata stream when I use Yolo, I am not able to get the confidence when I use the Resnet10 model in Deepstream. This is the Resnet10 model that can detect cars, bikes, persons and roadsigns. Is this model designed to not put out confidence? Am I missing something?
Usually, ResNet10 is a classifier and the output flow will be different to Yolo, which is a detector.
So please also send the confidence into meta in the classifier part.