Deep Learning Based Attention Inference System Using

Abstract

The project focuses on using AI and IoT for understanding student attention in classrooms. It employs Convolutional Neural Networks (CNNs) and IoT devices like Raspberry Pi kits with cameras to capture real-time data on student engagement. The system follows a ZeroTier architecture, leveraging decentralized networking for seamless communication between devices and the central server. A deep learning model analyzes eye-tracking data locally, achieving 90% accuracy in attention detection. Reports are generated every 5 minutes over a 40-minute session, demonstrating the potential for scalable, distributed systems in personalized education and behavioral analysis. The circumplex model is utilized to combine emotion detection and gaze tracking, enabling the extraction of students' attention levels within the classroom environment.