Face Recognition Attendance System / Ahmad Ikram Sharipudin

This project deals with the design of a Face Recognition Attendance System using deep learning algorithms and the Raspberry Pi platform. It is supposed to automate the identification and matching processes of attendance against a preregistered database through face recognition. Other traditional methods include manual attendance or fingerprint detection, but this system will try to provide an effective and time saving solution using advanced image processing. Other organizations are using conventional ways of taking attendance: manual attendance tracking, QR code-based, and Google Form-based attendance systems. The paper compares three models of face recognition: Facenet512 & RetinaFace, Facenet & RetinaFace, and VGG-Face & RetinaFace. In the experiment, the recognition accuracy of the Facenet512 & RetinaFace model is 100% (450/450 images), while those of the other two models were 96.67% and 92.67%, respectively. These findings confirm that Facenet512 & RetinaFace is indeed one of the best in face recognition accuracy. The selected model is then deployed on a Raspberry Pi, considering the feasibility and resource constraints of the platform. Python will be used as the embedded system in integrating it with an image processing library, such as OpenCV, and model implementation, like TensorFlow. Testing of the system on the Raspberry Pi confirms that it is doing well in real-time face recognition.