I'm interested in writing software that depends on being able to identify that there is a face in a picture (or video frame). It doesn't have to ID the face - so no metrics other than:
- Is there a human face in the picture (or more than one)
- Where, approximately, are the eyes and mouth or nose tip (whatever it keyed on.)
It's popping up in cheap digital cameras now, so I expect there are open source packages that do this, or research that is easy enough to follow when rolling your own.