FACE PATTERN RECOGNITION USING MULTI-TASK CASCADED CONVOLUTIONAL NEURAL NETWORKS

Abstract

The article provides 2 steps of identifying a human face including taking out the faces in the image; extracting the features of that face and from the information obtained after analyzing, and verifying the identity of the person.  The first step is based on the model of Multi-Task Cascaded Convolutional Networks (MTCNN) and the second step is based on the Convolutional Neural Network (CNN). The method of detecting faces by MTCNN due to the results after identification with high accuracy even if faces are obscured.