Automated Face Recognition based Attendance System using RetinaFace and FaceNet

Traditional approach for attendance in schools and colleges is professor calls the name or roll no. of students and record the attendance. The manual work included in the maintenance and management of the traditional attendance sheets is difficult and is a tedious work. This paper presents a system that automatically identifies and recognizes the individual in a live captured image and marks the attendance for that person. It is a web-based application in which RetinaFace algorithm has been used to detect the face in face image. FaceNet algorithm then extracts features from the image of a person's face and SVM classifier classifies the face based on extracted features. The classifier used here has been trained with 128 dimension values of each face. The developed system exhibits 99.76% accuracy on training data and 97.21% accuracy on validation data with Parameters of SVM as C=100, kernel=poly, degree=5 and probability=True. Other classifiers, namely, Random Forest, KNN and Logistic regression were also used for referencing with Accuracy, but among those, SVM classifier provided the best result