A Deep Learning-Based Approach Using MTCNN and FaceNet for Automated Attendance and Security Monitoring Systems

This study proposes a novel automated attendance and security monitoring system to overcome the limitations of conventional approaches is proposed. Three experiments were carried out using 150 facial images of three subjects and compared with state-of-the-art face recognition algorithms. The face detection and recognition modules’ performance were also compared. In experiment 1, LBPH was used for feature extraction and Haar cascade for face detection, while in experiment 2, LDA with PCA was used for feature extraction. It was found that this improved combination method can enhance the effectiveness of face detection and recognition. In experiment 3, two new deep learning based alternative algorithms were adopted: MTCNN for face detection and FileNet for face recognition. The best result was obtained in experiment 3 where MTCNN was used for face detection and FileNet for face recognition. The face recognition rate of $\mathbf{8 6 . 6 7 \%}$ and face detection rate of $86.91 \%$ were achieved by experiment 3. The robustness of the system was also tested in difficult conditions, such as poor illumination and different facial poses, and therefore, it can be used in real-life applications, such as attendance in class and security. From these results, it can be concluded that integrating MTCNN and FaceNet into an automated system is very effective and can provide a scalable and reliable solution for attendance and security management.