I am currently running OpenCV on a raspberry pi 3 for facial recognition after trying https://github.com/ageitgey/face_recognition but finding it too slow. I liked the system for returning the name of the person, however, I was coming across 2 minute compute times.
I followed the tutorial by Toni DiCola and have got it working nicely after some tweaking of thresholds and plenty of training images.
https://learn.adafruit.com/raspberry-pi ... x?view=all
I was wondering how I would go about further developing this for use with another or multiple people? I feel like adding more people to the training images file would just skew the eigenface or am I wrong here? Could I add separate files to the training images file and create multiple eigenfaces and training.xml files or would this not work? My end goal ideally is to integrate it into an access system which will greet the person who is trying to access the box by name.
Any ideas, recommendations for this current setup or another setup are all welcome.