Tag4You - About Us

Technology

Our core technology is face detection and face recognition.

Face detection can extract faces from a photo it finds a face without knowing who it is. This technology is very advanced and reaches high levels of detection. It finds almost all faces in the photo, it can find faces that are barley identified by people, very dark, small, unclear and faces with different expressions. Some false detection do occur and we will try to eliminate them but we feel it’s negligible and can easily be ignored. Keep in mind that in a single photo a face could appear anywhere and in any scale. So in a photo of resolution 330X330 if we scan all windows 30X30 in all scales we have more then 9 million sub-windows. If there aren’t any faces in that photo in order not to have any false positive we need to reject all the 9 million sub windows. Our algorithms are a little smarter, they don’t scan every sub window and have a few more tricks nevertheless we do need an extremely low false detection rate.

Face recognition can determine who the person in the photo is. How do these algorithms know who is in the photo? They usually take a few samples of a person’s face, when given a face the algorithm can determine who the face looks like. In order to recognize a person there must be enough details in the picture for the algorithms to work. Face detection can detect a face that will not be recognized by any known algorithm today (and probably ever). Our face recognition algorithms were optimized for general photos. Most research on face recognition is done with controlled or semi controlled conditions (lighting, background…). Although not as good as we would like, the algorithms after training can recognize most of the faces. As apposed to the face detection algorithms that perform very close to human’s perhaps even better, face recognition is far from reaching humans ability to recognize people. Our face recognition system is currently in a closed beta, if you are interested you can apply here.