Automated recognition of facial expressions and gender in humans implemented on mobile devices

Abstract

This paper presents the implementation on mobile devices of an automated system which can recognize 7 basic facial expressions each of them associated to an emotion: happy, sad, angry, disgust, surprise, fear, and neutrality alongside a person’s gender from a facial image using convolutional neural networks and transfer learning. The human faces are extracted from images generated by the camera of the device which runs the application in real time. The server side of the system was built in two versions: desktop which has good performances regarding recognition and processing speed and web which can run on any device which provides a browser and a camera.