Face Recognition Using Segmentation Technology

Abstract

Face recognition technology has become a popular research topic in the fields of pattern recognition, image processing, and computer vision. However, the face recognition accuracy rate will be confronted with challenges if the technology is trained on the unconstrained images with disguised and covered objects on a face area. This paper aims to propose a state-of-the-art face recognition methodology which could be applied in the crime field, especially in the cases of face-disguised crime. The main idea is to extract face pixels from a background and send them to a trained system to eliminate noise decrease meaningless contribution from useless pixels. This research combines two technologies: Fully Convolutional Networks (FCN) for face extraction from the background and disguised part and deep convolution neural network for face training and testing. The process of this research is to build a training set of segmentation of the face area with FCN as an input of Convolution Neural Network (CNN), and then train the CNN network with transfer learning from VGG-Face. Lastly, the classification function SoftMax is applied for face probability distribution. This algorithm has been experimented on a challengeable face dataset. Our experimental results show that the accuracy of recognition achieves better results compared to recognition without face segmentation.