Individual face detection and identification technologies have improved greatly in terms of accuracy, but there exist chaUenges of adoption in Consumer Electronic devices due to the lack of large training databases pertaining to the individual. With 1 or 2 images to train about the current user, it is nearly impossible to identify a person's face robustly. So, in this paper we propose the use of synthetic images to increase the database size. We will be using a Neural Network trained on these synthetic images. Observing that increasing the database alone does not solve the problem of varying illumination, background, pose and facial expression, a Gabor fIlter along with preprocessing is used to increase the robustness of the network. Experimental results show that using synthetic and real images together for training gives better results than using either of them alone.

No abstract