Abstract: Semantic segmentation is a popular theme in recent years. The two-dimensional face image segmentation technology is important for the field of robot industry, the head pose prediction, three-dimensional face recognition, face segmentation, facial animation and other fields. Because of the shortage of the existing segmentation algorithm, it proposes a face region segmentation method based on deep network. The experimental results show that the algorithm is of high precision and good robustness, which has practical application significance.

Key words: semantic segmentation, two dimensional face, region segmentation, deep network

摘要： 语义分割是近年来比较热的一个主题，而其中对二维人脸图片的区域分割技术的研究，对机器人应用，人脸头部姿势预测，三维人脸识别，分割，动画等方面有重要促进意义。由于目前的人脸区域分割算法在精度上存在一定不足，提出了基于深度网络的人脸区域分割方法，并进行了实验。实验结果表明该算法相较于以前的一些方法精度更高，鲁棒性好，有实际应用意义。

关键词: 语义分割, 二维人脸, 区域分割, 深度网络