Has the system been trained and tested to even work with children's faces? Can it cope with having two different photos of a child taken a year apart? Or (as s

Has the system been trained and tested to even work with children's faces? Can it cope with having two different photos of a child taken a year apart? Or (as seems to be the case) are we just shoveling data into a system and hoping it can cope.