We propose a implementation of a face recognition method, and an adaptation of this method to face verification, used for access control. We use the classical eigenface method, followed by a classification step. For this step, we compare three methods, based on a neural network, on euclidian distance, and a geometric classifier (using a set of hyperrectangles).

Finally, we present a application software, verifiying a self localised face at 8 im/s, and we plan to implement the face localisation using a real time pattern matching algorithm.