Secure Mobile Facial Recognition Attendance System with Edge Computing: A Comparative Study of Face Recognition Models for Android Application Integration

Attendance systems are designed to monitor and record the presence of individuals within an organization, such as a company or educational institution. These systems are usually in the form of biometric scanners or RFID cards. This research focuses on the development of a secure mobile based facial recognition attendance system by employing an edge computing approach. More specifically, the purpose of the research is to compare different facial recognition models and their suitability in an android application without sending confidential data such as images of the user's faces into a separate server. The final model chosen will be taking into account the model size, F1 score, and inference time. This model will be used to correctly identify a user in the android application strictly for attendance purposes. This research method involves comparing the performance of FaceNet, VGGFace, and MobileFaceNet for facial recognition. This research will also focus on designing and implementing a backend database system in order to facilitate the mobile based attendance application. The results of this study shows that the MobileFaceNet model was the best choice for this application as it is significantly smaller in size compared to the other two models while still having high F1 score. The model will then be integrated to the android application and the required classes and functions to facilitate this will be developed in Kotlin.