EngageSense AI: A Rule-Based System for Real-Time Student Engagement Detection Using Head Pose and Eye Gaze

EngageSense AI is a lightweight rule-based system for real-time student engagement detection using head pose estimation and eye gaze tracking. The system utilizes MediaPipe FaceMesh, OpenCV, and Dlib to classify engagement levels into Engaged, Partially Engaged, and Not Engaged without relying on deep learning models. The proposed approach focuses on interpretable behavioral cues instead of emotion-based inference, enabling transparent decision-making and efficient CPU-only execution. A manually annotated dataset of 1,000 frames was used for evaluation, achieving an overall accuracy of 82.2% at approximately 25 FPS. This repository contains the research paper submitted as an academic research project.