Facial and Behavioural Analysis for Classroom Management using Computer Vision

In the contemporary academic system, regular attendance and activity of students in the classroom play a notable role in performance evaluation and quality monitoring. Conventional attendance monitoring techniques like procuring the count of students manually have challenges such as increased human intervention, and higher time consumption. Installation of Biometric attendance systems or RFID tags for attendance management raises the complexity and cost of the system. Traditional Behaviour analysis techniques such as gesture or posture recognition become a complicated task as the population of the classroom increases. This paper aims to eliminate the complexity involved in the existing classroom management systems and presents an automated system for attendance and behaviour monitoring ensuring convenience and data reliability. The proposed methodology takes a video input and marks the students' attendance, employing the concepts of Computer vision. Facial Encoding and the YOLOv4 Tiny algorithm are used to procure the attendance of the students. The data is further processed to obtain the behavioural analysis by deploying Haar Cascade Eye XML code to analyze the Sleep Status and YOLOv4 Tiny algorithm to detect the Mobile Phone usage to collectively obtain the attentiveness of students in the classroom. The attendance data and analyzed behavioural reports of each student are stored in Google Firebase. Long-term performance can be monitored by utilizing this data for continuous assessment of every student.