I just had a few questions on the Face Recognition package of the Intel Real Sense Camera:
How are the user ids generated for face recognition? It seems to start at 100, but can it be customized?
We’re looking into enhancing/extending the face recognition authentication that the Intel Real Sense Camera provides. There are options such as voice recognition already built in for example. Do you have any suggestions on how to increase the accuracy of the face recognition interface beyond checking for pictures (we have considered asking for the user to blink, or check for their heartbeat), and what other factors can be incorporated into this system?
As a follow up to the last question, we are currently checking for if a face is a picture or real, by capturing all the landmarks and then getting the average of the z-values after x frames. And then creating a threshold to figure out what the average z-value should be. Are there better/more efficient ways of determining if a face is a picture or real? I have asked something similar before on this forum, but wanted to get some more opinions.