University Archives

Excessive smartphone use negatively impacts attention, productivity, and well-being in educational and workplace settings. This paper presents a computer vision–based system for real-time detection and quantification of smartphone use. The system consists of a laptop and external webcam that stream continuous video to a Python application for analysis. The core method employs Ultralytics YOLO, a single-stage object detection model that identifies persons and mobile devices within video frames. Multi-object tracking (BYTETracker) assigns persistent user IDs, enabling per-person usage estimates across time. The system captures bounding box coordinates, class predictions, track IDs, and frame-level phone usage events. A functional proof of concept supports live detection, annotation, and unique user counting via an OpenCV dashboard. The final system will aggregate per-user usage histories and provide interpretable metrics for monitoring smartphone use in shared environments. Performance evaluation uses accuracy, precision, recall, specificity, and F1 score to quantify detection reliability in real-world conditions.