건설 근로자 안전모 실시간 감지를 위한 딥러닝 적용 연구

This study explored the applicability of deep learning models for real-time safety helmet detection of construction workers. The performance and speed of RCNN-based model and YOLO model, which are representative models of object recognition among deep learning models, were compared. Faster-RCNN model of RCNN series was used, and Yolov3 and Yolov5 of YOLO model were applied. As a result, the Yolov5 model showed the highest performance and fastest processing speed. Among them, Yolov5x showed the highest performance, and Yolov5n showed the fastest processing speed. As a result of this experiment, Yolov5x can be fully utilized for real-time detection of safety helmet.