AcFR: Active Face Recognition Using Convolutional Neural Networks

We propose AcFR, an active face recognition system that employs a convolutional neural network and acts consistently with human behaviors in common face recognition scenarios. AcFR comprises two main components-a recognition module and a controller module. The recognition module uses a pre-trained VGG-Face CNN to extract facial image features, along with a nearest-neighbor identity recognition criterion. The controller module can make three different decisions based on the results-greet a recognized individual, disregard an unknown individual, or acquire a different viewpoint from which to reassess the subject, which are natural reactions when people observe passers-by. Evaluated on the CMU PIE face database, our recognition module yields higher accuracy on images acquired at angles more similar to those saved in memory. The view-dependent accuracy provides evidence for the proper design of the controller module.