Designig a real-time face recognition attendance using machine learning

Abstract

Systems for managing attendance are crucial for all organizations. One of the most often utilized biometrics for verifying human identity is the face. Most businesses in Rwanda utilize logbooks, cards, or fingerprints to track employees' attendance at work. However, because of COVID-19, which is an infectious disease, attendance has been discontinued due to concerns over its spread. This study describes real-time Face recognition attendance using Machine Learning, an Internet of Things (IoT)based biometric face recognition solution. To capture the live streaming video, a high-quality camera with a Sony IMX477 sensor and a 16mm 10MP Telephoto lens connected to a Raspberry Pi 4 Model B, can send frames at a time to the cloud. The pre-trained FaceNet model is employed by the system to extract features from a face image after using MTCNN (Multi-Task Cascaded Convolutional Neural Networks) to recognize facial landmarks on images. Real-time image processing is done in the cloud, and attendance is recorded on a dashboard that is accessible from anywhere. The system sends the email to the employee using SMTP Protocol in case of arriving late /absent without permission. The result reveals that the system is safe, dependable, trustworthy, and does not require physical touch.