Inference over video stream

[jayavanth]

Which demo code should I use if I want to do pose inference over a video stream? I see that there is one for webcam
python demo/inferencer_demo.py webcam and one for image/video files but is there one for a video stream?

Basically, I want to get the keypoints in a streaming fashion. I used --pred-out-dir to get a json file with the keypoints but how does this work on a video stream? Do I have to continuously check for updates to this file?