For testing I built a temporary housing for the B5T which also allowed me to add a tripod mount which makes things easier
I have 4 faces currently registered in the album (the B5T allows for up to 500 faces to be registered) each with only 2 recognition templates (you can have up to 10 templates per user) and the B5T has never yet made a false recognition! I use photographs of people for face recognition testing as it saves me calling people into the lab all the time for these tests. One photo I am using is John Connor (from the Sarah Connor Chronicles TV show) which makes a good test subject. Face recognition on the B5T is superb and even works well in low light. The body and hand detection also work well as does face detection and direction. Age, gender and gaze estimation all work great and the icing on the cake is the face expression estimation function, now the ALTAIR robots will know when I am not happy with them!
I have used a microcontroller as a data buffer for the B5T (this being a PIC18F14K50) as there is a lot of high speed data comms required if you want to use all of the many B5T features and a direct connection to the v4 may flood the data channel. The PIC gets all the current data then compresses it into a small data packet that then gets sent to the v4. The PIC buffer board plugs into the B5T (UART sensor board) the same way as the USB dev board does.
This is the B5T/PIC output with just me in view
Here is the B5T/PIC output with me and John Connor in view
Here are the boards outside of the enclosure
It is amazing to see the humble PIC microcontroller being a part of advanced face/gesture recognition! admittedly it is with the help of the awesome Omron HVC chipset, the B5T sensor board is one great piece of kit!