A Comparative Study of YOLOv5 and YOLOv8 for Helmet Detection Inside ATMs

Automated helmet detection inside automated teller machine (ATM) cabin for cash dispense can enhance security as well as serves as a means to prevent heist and vandalism. This study presents a comparative study on helmet detection inside ATMs using two prominent object detection you look only once (YOLO) v5 and YOLOv8. The system facilitates detection of helmets from closed circuit television (CCTV) footages to help in identifying the people entering into the ATM by wearing helmets and also in avoiding potential intrusion. In addition to precision and recall, mean average precision (mAP), a standard object detection metric to measure the accuracy of localization is used in this paper to assess the performance of helmet detection models constructed using YOLOv5 and YOLOv8. Experimental evaluation carried out in this study elucidates that YOLOv8 performs better than YOLOv5. By integrating the proposed model with IoT techniques, this model improves the safety and security. In addition, the incorporation of other alert models with this system offers a safer environment for the users inside the ATM cabin.