REAL-TIME AI-POWERED ATTENTION MONITORING USING FACIAL LANDMARK ANALYSIS AND EYE TRACKING

With the rapid proliferation of digital environments—including online learning, remote work, and virtual 
collaboration—sustaining human attention has become a growing challenge. This paper presents an AI
powered real-time attention monitoring system that employs computer vision techniques for non-intrusive 
assessment of user focus levels. The system utilises MediaPipe Face Mesh for facial landmark detection and 
OpenCV for video processing, computing the Eye Aspect Ratio (EAR) to detect blinks and employing gaze
tracking algorithms to determine gaze direction. An adaptive attention score, ranging from 0 to 100, is 
dynamically computed based on blink frequency, gaze deviation, and prolonged eye closure. When the score 
falls below a configurable threshold, the system activates audio-visual alerts to prompt the user to refocus. A 
Flask-based web interface with SQLite-backed session management facilitates user authentication and session 
analytics. Experimental results indicate blink detection accuracy of approximately 95%, gaze detection 
accuracy of approximately 92%, and sub-100 ms per-frame processing latency. The proposed system 
demonstrates that open-source, webcam-based tools can deliver practical, cost-effective attention monitoring 
applicable across educational, corporate, healthcare, and automotive safety domains.