The application needs you to be able to tag “strangers” as known people and should find away to use those new composites to become smarter. Otherwise I’m going to continue to get 3/4 or side shots of heads that are clearly known and if the system was starter could be more quickly ignored.
To identify who is a stranger is a very complicated algorithm. I have seen some systems, but never with 100% accuracy in identifying people in general. Not shaving for a week can change detection of a known person…
Absolutely, no argument there, but given the ability of say Google pictures to identify known faces across multiple years, there are possibilities. Also the more data you provide the easier it becomes to compare to a composite. As it is now unless the camera gets a head on shot of the person in question, it seems pointless, I doubt most folks are positioning a camera to capture nothing but head on approaches to entries.
Thanks for the suggestion. This is definitely a way we can make our system smarter!
I agree with this suggestion. Great idea
I was wondering if there any been any development on this front? I get face shots that are identified as Stranger, it would be nice to assign them to known users, to better train the system.