Construction site safety detection based on object detection with channel-wise attention

Construction safety has always been a vital problem all around the world. With the development of big data, 5G communication, artificial intelligence and other technologies, unmanned management has become a reality. In this paper, we proposed an algorithm based on YOLOv5 for safety detection on construction site. First, we establish datasets of helmet and safety vest for model training. Secondly, we construct a safety detection model by adding channel-wise attention to the YOLOv5 network. Finally, we conduct extensive experiments and comparisons based on established dataset. Experiments demonstrate that our model achieves favorable speed-accuracy trade-off compared to the other models.