Facial recognition based on a 3D model that was built from an insufficient training set

Abstract

Many facial recognition algorithms exist, but many of them have a problem with an insufficient training set. A new method of facial recognition is proposed in this paper, which relies on the information that was acquired using a 3D model of the head, which allows to expand the training set. The proposed algorithm uses control points that are placed on the image of the face and model, to determine the class that the image belongs to.