Jumping behavior Analysis after Identification of Daycare Children's Activities utilizing YOLOv8 Algorithm

Monitoring human activities to ensure the safety and welfare of humans is essential in the present age of mindfulness. Human Activity Recognition (HAR) offers a technical method to detect and identify human actions in real-time. We used YOLOv8s, a deep learning system, to recognize a kindergartener's behaviors and positions. After recognizing the student's movements, an assessment determines jumps per second and jump duration. We utilized our dataset, named Kid Action Recognition (KAR). The KAR (Kid Action Recognition) dataset consisted of 6489 photos. We manually labeled all photos using Makesense.ai. We analyzed and trained the data in a Google Colaboratory setting with an A-100 GPU. The model demonstrates an overall accuracy of 88.7% and an F1 score of 86.6%. The recall score is 89.5%, and the precision is 83.9%. The model excelled in all parameters and can effectively identify student activities, making it suitable for potential use in institutions such as daycare centers and schools.