Helmet Wearing Detection Using Deep Learning: A Comparative Performance Evaluation of YOLOv12 and YOLOv11

Abstract

Object detection in images using deep learning methods has made significant progress in recent years. One of the most efficient approaches is YOLO (You Only Look Once), which, in addition to providing good accuracy, also offers high speed, making it practical in many applications. YOLO version 12 is the first version to extensively integrate the attention mechanism into its architecture as a method to improve the detection of small objects in images. To evaluate the performance of YOLOv12, a helmet-wearing detection model was developed and its accuracy and speed were compared with a model implemented using YOLOv11.