Tra cứu thông tin sinh viên qua ảnh khuôn mặt

Abstract

To more conveniently support and speed up student conduct score query, this article presents a facial recognition method to perform student conduct score query. First, the authors used multitask cascaded convolutional networks (MTCNN) to detect and mark faces. Then, use the HOG feature extraction method to extract the feature vector of the face image and the SVM (Support vector machine) classifier to train the face recognition model. Experimental results and model application give an accuracy of 98.46% on the YaleFace image data set, 98.44% on the YaleFaceB image data set, and 86.2% on the student data set, respectively. From there, build an application to identify students through facial images, perform searches in the database, and return students' conduct results.