Safety Helmet Detection Based on YOLOv5

As the most basic protection for workers, safety helmets have great significance to workers' lives. However, due to a lack of safety awareness, safety helmets are often not worn. With the continuous development of object detection technology, the YOLO series of algorithms with very high precision and speed has been used in various scene detection tasks. To establish a digital safety helmet monitoring system, we propose a safety helmet detection method based on YOLOv5 and annotate the 6045 collected data sets. Finally, we used the YOLOv5 model with different parameters for training and testing. The four models are compared and analyzed. Experimental results show that the average detection speed of YOLOv5s reaches 110 FPS. Fully meet the requirements of real-time detection. Using the trainable target detector's pre-training weight, the mAP of YOLOv5x reaches 94.7%, proving the effectiveness of helmet detection based YOLOv5.