Real-Time Attendance System with Liveness Detection and Multi-Face Recognition using Transfer Learning

Attendance is an integral part of maintaining the records and activities of an individual. The face, like the iris and the biometric is the identity of a person. The ways by which the human body has unique imprints in its physical feature is a great chance to apply image processing techniques. The documenting of attendance plays an essential role in recording the in and out time of an individual in a particular organization or a college. The traditional approach for attendance in an educational institution includes the professor calling student name and then, recording the attendance while maintaining manually for the same. It is a rigorous task each day to maintain the records of attendance. It approximately requires 5 - $\mathbf{1 0}$ minutes to record attendance manually. To avoid the time and energy loss, the paper proposes a method which uses automatic attendance process which is done by image processing. The paper proposes a system wherein the dataset i.e., the images would be captured, the face is detected by Single Shot Detector (SSD) and the person is recognized using PyTorch model of OpenFace. It aims to identify individuals accurately while ensuring their liveness. It employs facial landmarks detection, eye aspect ratio calculation, and deep learning-based face embedding. Therefore, by utilizing this automated system, time is saved and maintaining attendance becomes a hassle-free process. With the help of the system outlined in this paper, it is convenient to record attendance of multiple faces in real-time. The method proposes to detect and recognize multiple faces and maintain a record of the attendance automatically.