Kid Activity Recognition: A Comprehensive study of kid activity recognition with monitoring activity level using YOLOv8s Algorithms

Recognizing human activities and behavior is a cutting-edge field of research because of the complexity and limited availability of data. Our research involves the use of the trending YOLOv8s deep learning algorithm for the classification of active states (sitting, standing, running, laying, and jumping). The research focuses on the detection of a child’s activities in the kindergarten playroom. To train the model, we created the KAR (Kids Action Recognition) dataset using our own sources. We labeled over 6,489 images in the dataset using Makesense.ai. Each image is annotated with activity labels and the location of the student (in a room or on a mat). This assisted in presenting insights on the student’s vitality level and physical involvement over a period of approximately an hour. We analyzed the whole activity and compared it with the times. Standing in the room and mat, standing in the room, and jumping in the room and mat are most of the activities of the kid. The model achieved an impressive overall accuracy of 88.7 %. We conducted all the training in a Google collaborative environment with A-100 GPU specifications. The results demonstrate that the model can identify and classify each activity with high precision.