The Virtual Spectator is an "on-the-fly" face recognition system. This means that it needs not be trained in order to recognize the people it finds within its field of vision. Or more precisely, it trains itself in real time.
The Virtual Spectator aims at making a computer imitate the process of getting to know people that humans do naturally.
Based on the OpenCV library, the Virtual Spectator detects and tracks the faces that appear in the image. If that person is seen for the first time, the system automatically learns its facial features, incorporating them to its database, following a process which is completely transparent to the user.
Once that person is known to the system, the Virtual Spectator will automatically recognize it every time it appears on the scene. As time goes by, the Virtual Spectator creates a growing database of known people, just like we humans do.
The Virtual Spectator was developed as part of the MSc thesis of student Nil Molné, supervised by Xavier Sevillano.