Child Detection Model Using YOLOv5

Abstract

CCTV surveillance systems have been installed in public locations and are used to search for missing children and fight crime. The Penang City Council has deployed a face recognition CCTV monitoring system. As a result, the goal of this research is to identify children who were in the wrong area or at the wrong time and then notify authorities such as police and parents. According to child detection research, the average child loss rate is greater owing to a lack of child detection features. Existing research employs machine learning and deep learning across several platforms, yielding inaccurate accuracy findings. Using the YOLOv5 algorithm, this study will categorize images based on children detection in restricted locations. The datasets Coco, Coco128 and Pascal VOC were chosen because they are the standard datasets of YOLOv5 along with the public dataset INRIA Person. Annotations and augmentation techniques are employed in the pre-processing phase to acquire labelling in text file format and offer data for any object position. The YOLOv5s model then will be designed to make the proposed detector model. After training using YOLOv5s, child detector model is produced and evaluated on the dataset to acquire findings according to the performance metrics of recall, precision, and mean average precision (mAP). Finally, the performance metrics acquired from all four datasets are compared. The INRIA Person dataset performed the best, with a recall of 0.995, an accuracy of 0.998, and a mean Average Accuracy of 0.995. The findings for both models, YOLOv5s and the proposed model, are, nevertheless, quite close. This demonstrates that the proposed model can detect as well as YOLOv5s model.