MC-YOLO: A Context Aware Down Sampling Algorithm for Construction Site Object Detection

Safety accidents at construction sites not only pose a threat to the lives of workers but also lead to significant economic losses. Therefore, strengthening safety management at construction sites has become an issue that urgently needs to be addressed. Traditional construction site safety relies on manual inspection methods with low efficiency. Computer vision makes it possible to conduct automatic target detection in construction site. However, the complexity of the construction site environment and the diversity of detection targets pose challenges to detection. This paper proposes an improved YOLOv8 model called MC-YOLO for object detection in construction site. To overcome the problem of information loss caused by down-sampling in YOLOv8, a CADS (Content Aware Down-Sampling) block is proposed to replace the original down-sampling block. Meanwhile, in order to enhance the multi-scale object detection ability of YOLOv8, an improved C2f block named C2f-MSC (C2f with Multi-Scale Convolution) block is introduced. The addition of these two blocks not only improves the model’s feature extraction ability but also significantly reduces the number of model parameters. The experimental results show that the MC-YOLO model performs well on the CSS and SODA datasets. Specifically, The MC-YOLOs model achieves a mAP@0.5 of 80.8% and a precision of 91.9%, which are 1.5% and 3.9% higher than the original algorithm respectively, and the number of model parameters is reduced by 21.09%.