AN OPTIMAL BACKPROPAGATION NETWORK FOR FACE IDENTIFICATION AND LOCALIZATION

摘要

Face identification and localization is the most focused research area of pattern recognition and computer vision. It has rapidly established itself as a promising tool for e-passport, surveillance and biometrics. The present paper designs and describes an optimal back propagation (BP) network for partially rotation invariant face identification and localization in single and multiple image frames. The optimal BP network is having single hidden layer, and it was observed that for an optimal number of hidden layer neurons, the identification rate is highest.