Empowering Real-Time Attendance Management with Facial Recognition and Computer Vision

Abstract

This paper describes an automated system of attendance that uses facial recognition and computer vision to address the challenges that come with manual attendance, fake attendance, and time wasted. Using state-of-the-art deep learning models and fine-tuned preprocessing, the proposed system achieves high accuracy, extensibility, and adaptability to changes in the operating conditions. The system uses the ArcFace model, which is a very accurate model and also a feature extractor and smoothing and sharpening filters to remove the noise in the image and to retain the edges of the image. The performance of the model is shown in the experimental results which indicates that it can detect faces of more than 20 per frame with a speed of 28-62 frames per second with 95.1% accuracy while compared with other models like FaceNet 94.3% and VGG-Face 92.5%. The preprocessing stages enhanced the recognition rates due to solutions of lighting changes, motion blur, and occlusion, making reliable detection possible in various scenarios. The conclusions are pointing at the versatility and effectiveness of the system which can be applied to educational facilities, offices and public areas where the accountability is a priority as well as the time issues. Lightweight version to support IoT gadgets and edge computing systems to facilitate work in settings with limited resources but with acceptable performance. Besides, the lack of accuracy, scalability, and dynamic adaptability, which is important in real-world applications, is also solved in this research, and new achievements in the field of AI-based biometric systems are given, which can be a basis for further developments in automated attendance, surveillance, and identity verification systems.