Deep Learning for Detecting Quilt Status of Sleeping Children

Abstract

Sleeping is important for the physical growth of infants and toddlers. However, because of the growth of their bodies, they may kick off the quilt when they fall asleep and catch a cold as a result. Because it is difficult to extract the all-appropriate features that insists the quilt is not covered in the images taken by the home camera manually, which is suitable for the old types of symbolic AI to recognize, there has not been a suitable solution to this problem in the past. However, thanks to the rapid development of artificial intelligence and neural networks, intelligent target detection, which was previously difficult to achieve, has finally been realized and several approaches to intelligent target detection have emerged, leading to many different perspectives and algorithms. In this paper, by analyzing and comparing the advantages and disadvantages of different algorithms in different application scenarios, we found that YOLO (v5 in this paper) is very suitable for quilt status detection and realized the program by applying YOLOv5 neural network and cosine difference algorithm simultaneously.