Hard-hat Detection using YOLOv4

The safety of workers in the building environment is essential. Image and video monitoring for the detection of hard-hat is a new and innovative solution to this safety problem. For hard-hat detection and safety concern at construction sites, object detection algorithm You Only Look Once (YOLOv4) is used. A neural network is made up of input, minimum one hidden and output layer. Object dataset, which made up of classes of hard-hat, head and person captured in colored and in monochrome images. The dataset is made with varying light condition. YOLOv4 model is implemented for hard-hat detection in image and video. Acquired output shows that the algorithm successfully found the hard-hat with an approximate precision of 99% for images and 98% for videos.