TeacherActivityNet: A Personalized Assistive System for Enhancing Cognitive Support and Improvement of Work-Life Balance in Academic Settings

Abstract

This research establishes a new idea that supports cognitive enhancement and work-life balance for academic office-based faculty members. The improvements in computer vision led to the creation of work-based wellness solutions which assist educational settings to manage their staff’s mental health requirements between intellectual requirements. Workload management methods and self-evaluation approaches from traditional times demand significant manual efforts while creating efficiency problems thus demonstrating the value of automated video-based support systems as better alternatives. Scientists have achieved notable improvements in Human Activity Recognition (HAR) detection yet studies about HAR based personalized workplace support systems designed for faculty office use are scarce. Current research examines student performance in classrooms rather than providing essential support for faculty cognitive processes and wellbeing. Therefore there exists an important deficiency in cognitive and wellbeing assistance for educators. The paper fills this gap through TeacherActivityNet, our proposed model with the help of newly created Workplace Activity Recognition Dataset which represents a new video database designed to detect nine different faculty office activities such as Arriving, Counselling, Eating, Idle, Leaving, Sleeping, Talking, Using Phone and Working. We adapted the YOLOv8n architecture to develop TeacherActivityNet for our model implementation that achieved 0.841 mAP50 through fine-tuning with our generated dataset. Comparison against existing systems reveals that TeacherActivityNet brings potential advantages to personal assistive systems by supporting better academic cognitive performance and life-work balance within educational institutions.