AttendNet: A Real-Time Deep Learning Framework for Multi-Face Recognition and Automated Attendance Monitoring

Automated face recognition attendancesystems offer a contactless alternative to biometric andhardware-based roll-call and biometric attendancesystems. Nevertheless, the majority of current solutionsare based on single-frame recognition, making themunstable in dynamic scenarios like a classroom whereobjects are in constant motion, and the lighting is notconstant.In this paper, the author will introduce AttendNet AI,a temporal-sensitive real-time face recognition systemthat can be used to monitor attendance automatically.The system proposed is an identity matching model thatuses a distance-based similarity measure to matchidentities with a pre-trained deep metric model thatgenerates facial embeddings. To increase the reliability, amulti-frame temporal validation mechanism is added,where the recognition is reliable between consecutiveframes before marking attendance.The system is deployed as a full-stack applicationwith built-in web interface, secure administrationauthentication, student portal and database basedattendance logging. Experimental analysis shows goodperformance with an accuracy of 96.3, a precision of96.9, a recall of 95.0 and an F1-score of 95.6.The findings reveal the enhanced resilience to motion,occlusion, and transient detections, and the system canbe deployed in the real world in educational settings.