cv2.namedWindow(myFloat)

I’d recommend to run samples/python/mosse.py, and then some DNN trackers (especially samples/dnn/*siam*.py). also give general object detection DNNs a look.

I’ve given the trackers module a few glances but don’t like any of them. the Tracker class is too restrictive for my use cases (can’t feed corrections back in, AFAICS). I took the mosse.py example and hacked it for my purposes, combining it with an object detector to remove drift.