End to End Deep Learning for Single Step Real-Time Facial Expression Recognition

摘要

In recent years, a lot of research has been carried out in face detection and facial expression recognition. Very few of them are capable of achieving these at real time with a very high accuracy. In this paper we present a real time end to end, single step face and facial expression recognition technique which performs at a speed of more than 10 fps (frames per second). We use an end-to-end deep learning approach for localization and expression classification. On the CK+ [1] dataset we get a 10-fold validation accuracy of 94.8% on 640 * 480 images. We have also created a webcam interface, which classifies the emotion of a person at 10 fps, which proves our claim that facial expression recognition has approached real time speed with very decent accuracy.