Behavioral Engagement Assessment in University Classrooms via Deep Learning-based Video Object Detection.

Area of Science:

• Educational Technology
• Computer Science

Background:

• Assessing student learning engagement is crucial for effective pedagogy.
• Traditional methods of engagement assessment are often subjective and time-consuming.
• Automated analysis of classroom behavior offers a scalable solution.

Purpose of the Study:

• To develop and validate a deep learning-based system for real-time assessment of student learning engagement.
• To identify key classroom behaviors indicative of learning engagement.
• To compare the performance of Faster R-CNN and YOLOv5s for classroom behavior detection.

Main Methods:

• Correlation analysis to identify seven key classroom behaviors linked to learning engagement.
• Supervised machine learning training using manually annotated video data.
• Comparative performance evaluation of Faster R-CNN and YOLOv5s algorithms.
• Development of a three-level learning engagement scoring model using focus group data.

Main Results:

• YOLOv5s demonstrated superior accuracy and efficiency in classroom behavior detection compared to Faster R-CNN.
• The developed model enables automatic, real-time scoring of learning engagement.
• The system can assess engagement at both individual student and class levels.

Conclusions:

• Deep learning-based video object detection provides an effective tool for objective learning engagement assessment.
• The YOLOv5s algorithm is well-suited for real-time classroom behavior recognition.
• This automated system has the potential to enhance pedagogical strategies and improve educational outcomes.