Attendease: A Smart AI-Based Classroom Attendance System Using Facial Recognition and Late Entry Detection

Traditional classroom attendance systems rely on manual roll calls or simple digital check in methods which are time consuming and prone to proxy attendance. This paper presents Attendease, an artificial intelligence based attendance system that uses real time facial recognition to automatically identify students and record their attendance. The system uses the InsightFace framework with an ArcFace embedding model and RetinaFace detector to extract 512 dimensional facial features from video frames and performs identification using cosine distance similarity with a threshold of 0.45. The system processes video streams from an RTSP enabled camera and records attendance automatically while also detecting late arrivals based on predefined session cutoff times. The prototype was evaluated using 100 facial samples captured from registered students under classroom conditions using a Samsung Galaxy S22 camera configured as an RTSP stream source. Experimental results show a recognition accuracy of approximately 97 percent with reliable identification at distances up to 5 meters. Performance evaluation indicates an average pipeline processing time of about 50 milliseconds enabling operation at approximately 40 frames per second for near real time attendance monitoring. The system was deployed on a test server equipped with an Intel Core i7 thirteenth generation processor, an NVIDIA RTX 5050 graphics processor with 8 GB video memory, and 16 GB system memory, with a backend implemented using FastAPI and a database using MongoDB. The current evaluation was conducted using a smartphone based RTSP camera and the system has not yet been tested with standard CCTV cameras. The results demonstrate that artificial intelligence driven facial recognition can provide an effective and scalable solution for automated classroom attendance while reducing manual effort and minimizing attendance fraud.