An End-to-End Real-Time Face Identification and Attendance System using Convolutional Neural Networks

Carrying out the attendance process in any academic organization is a very significant task. However, the manual attendance process is very tedious and time-consuming. Hence, an automatic face attendance system using CCTV camera may be helpful by reducing the manpower and it also makes the attendance process faultless. There are some automated systems available commercially, but most of them deploy near frontal faces and processes them one by one, which is again a prolonged task. Some deep learning-based face attendance approaches have been proposed in the literature and improving the efficiency of the face attendance is still under research. In this paper, we propose an end-to-end face identification and attendance approach using Convolutional Neural Networks (CNN), which processes the CCTV footage or a video of the class and mark the attendance of the entire class in a single shot. One of the main advantages of the proposed solution is its robustness against usual challenges like occlusion (partially visible/covered faces), orientation, alignment and luminescence of the classroom. The proposed method obtained a real-time accuracy of 96.02% which is better than that of the existing end- to-end face attendance systems.