Face recognition is attractive as a biometric as it is cheap and unintrusive.  However, building practical systems is extremely challenging.  In particular, most systems fail when the pose, lighting
or expression of the gallery face differs from that of the probe face.  In this series of talks, I will discuss the history of face recognition,  characterizing the various models as either generative
or discriminative, before assessing the pros and cons of these two approaches.   I will discuss the  state of the art in face recognition in both controlled and uncontrolled viewing conditions.   Finally, I
will discuss the main approaches to dealing with variation in viewing conditions;  to attempt to decrease the variation in the original images, to build models which describe the variation, or to
pre-process the images to make the invariant to this variation.

Simon J.D. Prince:  Dr. Prince received his Ph.D. in 1999 from the University of Oxford for work concerning human stereo vision. He has a diverse background in biological and computing sciences and has published over 50 papers across the fields of biometrics, psychology, physiology, medical imaging, computer vision, computer graphics and HCI. He worked as a research scientist in Oxford,
Singapore and Toronto. Prince is currently a Reader (associate professor) in Computer Science at University College London and also works for Aurora Computer Services Ltd, a face recognition company based in the UK.  He has written a major new textbook on computer vision ("Computer vision: models, learning and inference") which will be published by Cambridge University Press in 2010.