Research on Face Segmentation Network Based on FCN

This paper is an automatic recognition method of face region segmentation based on FCN network. Compared with other face segmentation methods, in addition to improving the face recognition rate and face integrity in the samples, it also extracts the RGB value, position information and shape information of the sample features through five channels, making the extraction and recognition of various parts of the face in complex samples more accurate and rapid. Compared with the general face segmentation method, the recognition speed and accuracy have been greatly improved, and the recognition effect is better. Through the experiment, for all the main data sets of the mainstream, the success rate of face recognition accuracy and face recognition integrity can reach more than 95%, and more importantly, the accuracy rate of face segmentation has reached more than 90%.