An ML Approach for Secure and Contactless Recognition System

However, attendance management still remains a crucial yet highly ineffective procedure both at educational institutions and companies. Manual procedures are lengthy, error-prone, and can be bypassed by proxy attendance. The use of biometric solutions (fingerprint, RFID) leads to such issues as poor hygiene and insecurity due to hardware reliance. In the following paper, a framework is suggested that utilizes machine learning techniques to create a contactless secure solution for real-time facial recognition and automated attendance management with SMS alerts. The system consists of three principal components: (i) creation of deep learning based face embeddings (512-dimensional vector) to make contactless identification precise; (ii) implementation of NLP chatbot to allow the user to check their attendance by conversing; and (iii) an automated SMS notification feature. As to architecture, FastAPI (microservices) serves as backend, PostgreSQL database extended with pgvector (IVFFlat indexation) extension is employed, WebSocket protocol is used to provide low latency video processing, and JWT authentication is used for security. According to the experiment, the algorithm recognizes faces accurately (95-98%) when there is enough light, response time does not exceed 1 second, and FAR is less than 0.5%. Comparison with other attendance management solutions suggests that proxy attendance is completely eliminated, efforts are reduced, and additional features such as a chatbot and SMS alert feature are included. Scalability tests were performed using 10,000 synthetic embeddings, leading to recognition accuracy of 95.8% and a match time not exceeding 50 ms. Thus, the research shows how recent advances in deep learning technology can be applied to create an intelligent solution along with the use of scalable database solutions.