Tracking module part of face detection in video?

[Jack Reilly]

Hello team,

I'm interested in using face detection on pre-recorded videos (not live), and I'd like to take advantage of the continuous nature of face locations as a function of time in videos to improve recall/precision.

Is tracking both forward and backward in time part of the standard face-detection solution? If not, is there a way I can add tracking to face detection?

In other words, what optimizations are performed on videos for face detection that aren't available on single images?

Best,

Jack