作業員のヘルメットと安全ベストの着用判別

抄録

The safety of workers is an issue at construction sites because of the danger of accidents. Therefore, the objective of this study was to detect helmets and safety vests in the state of being worn by YOLOv5, an object detection algorithm using deep learning, in order to confirm their wearing condition and ensure the safety of workers. We used multiple Data Augmentation during training to increase the training data and compare and verify the usefulness of each method by investigating false positives and omissions. The results showed the effectiveness of combining multiple Data Augmentation and the strengths and weaknesses of each method.