Extracting model bounding box output for another object detection model

Hello everyone, I want to build an alpr pipeline where I have a model that detects license plate and another model that recognizes license plate characters with Yolov5 (both models).I am using deepstream with python. I have been trying to find a way to extract license plate output from one model and then feed that bounding box area to character recognizer model. Any help and tips will be much appreciated.