I have a Python script that detects objects in the video stream from an LTV-3CNB50-M0560 IP camera. Currently, an RTSP stream is being used. When an object is detected in the frame, it turns out that the current time is behind the real event time, i.e., the moment when the camera generated the frame, by a random amount. I need to know the exact time of the frame, and thus the exact time of the event. Right now, I plan to use Tesseract to recognize the timestamp on the video overlay, but this approach doesn’t seem rational. What’s the best way to solve this problem?