Multi-task Cascaded Convolutional Neural Networks for Real-Time Dynamic Face Recognition Method

摘要

Due to the variety of poses, lighting, and scenes, dynamic face detection and calibration pose a big challenge under unconstrained environment. In this paper, we use the inherent correlation between detection and calibration to enhance their performance in a deep multi-task cascaded convolutional neural network (MTCNN). In addition, we utilize Google's FaceNet framework to learn a mapping from face images to a compact Euclidean space, where distances directly correspond to a measure of face similarity to extract the performance of facial feature algorithms. In the practical application scenario, we set up a multi-camera real-time monitoring system to perform face matching and recognition of collected continuous frames from different angles in real time.