Doing the video stream and Classification separately is doable but I don’t understand how to run them simultaneously, in real time…since a webcam can’t be used for more than one application. If there is a way, please let me know.
The best option may be to create a virtual camera. You program can take the frames and perform classification and then send the frame to a virtual camera used by Zoom.
Let me know if that would work for your use case. If you’re able to provide more information on the use case and implementation that should help too.