Real Time Face Attendance System using Deep Learning

Face is an individual's unique representation, and therefore, we propose an automated system for student attendance using face recognition. Face recognition systems have significant applications, particularly in security control systems. For instance, the airport protection system relies on face recognition to identify potential suspects, while the Federal Bureau of Investigation (FBI) utilizes this technology for criminal investigations. Our proposed approach begins with video framing, initiated through a user-friendly interface. By employing the Viola-Jones algorithm, we detect and segment the region of interest (ROI) containing the face from the video frame. In the preprocessing stage, we perform image scaling as necessary to preserve information integrity. Next, we apply median filtering to eliminate noise and convert color images to grayscale. To enhance image contrast, we implement contrast-limited adaptive histogram equalization (CLAHE). In the face recognition stage, we utilize enhanced local binary pattern (LBP) and principal component analysis (PCA) to extract facial image features. Subsequently, we record the attendance of the recognized student, saving the data in an Excel file. Unregistered students have the opportunity to register on the spot, and notifications are triggered if a student signs in more than once. The recognition accuracy is 100% for high-quality images, 94.12% for low-quality images, and 95.76% for the Yale face database when training with two images per person.