Safety Helmet Detection based on YOLOV4-M

In the intelligent monitoring of the construction site, there is often the problem of not wearing a safety helmet. In order to effectively judge the wearing condition of the helmet, a detection method of YOLOV4-M is proposed. This method replaces the backbone network in YOLOV4 with MobileNetV3, uses depthwise separable convolution to reduce the amount of parameters of the backbone network, and uses H-swish to improve the performance of the model. Experiments show that the final detection accuracy rate is 98.2%, and the FPS reaches 40 frames, which can be well applied to the detection of helmets.