Research on Multi Pose Facial Feature Recognition Based on Deep Learning

Abstract

A multi pose facial feature recognition method based on deep learning is proposed. Firstly, the face direction is determined by eyes and mouth combined with face features, and then the face is segmented according to the face direction to get the approximate forward candidate region of the face, which is then verified by the layered structure depth learning algorithm. Multi pose face detection algorithm based on face feature and deep learning algorithm. Most of the background regions are eliminated quickly by using skin color features, and the candidate regions are segmented according to the face direction determined by face geometry features, and the candidate regions are classified by using depth learning algorithm. The candidate regions are classified by deep learning algorithm, and the multi pose facial features are accurately recognized. This method can avoid the problem of increasing the complexity of the algorithm, such as multi pose face extended feature or pose detector, so as to improve the search speed and accuracy. The experimental results show that the multi pose facial feature recognition method based on deep learning can detect the multi pose face quickly and has good robustness to expression and occlusion.