There are posted examples of facial recognition using the Raspberry Pi, for example https://www.pyimagesearch.com/2018/06/2 ... cognition/
don't expect it to work very fast, as this needs a lot of computation. He mentions 1 or 2 FPS for face recognition, but you can get 8 or 10 fps if you're just doing tracking, or face detection
instead of specific individual recognition
There are "AI accelerator" add-on USB devices that can speed up specific kinds of deep learning networks by a factor of maybe 10x on the Pi. Not sure if they would be specifically applicable to this task or not.
Coral from Google: https://coral.withgoogle.com/products/accelerator/
Neural Compute Stick 2 from Intel: https://www.amazon.com/Intel-Neural-Com ... 07KT6361R/