A helmet detection scheme based on improved YOLOv3
Жипенг Ли, Shaobo Liu, Fupeng Li, Jun Wang, Jianfeng Zhang, Yan Li · 2022
In order to solve the problem that the target detection model is difficult to deploy in edge devices in smart grid, an improved YOLOv3 algorithm SG-YOLOv3 based on depthwise separable convolution is proposed. Use depthwise separable convolution to replace up-sampling in darknet-53 and standard convolution in residual units, respectively. The experimental results show that compared with YOLOv3, although SG-YOLOv3 has a 6.52% decrease in mAP. In addition, SG-YOLOv3 reduces the model size by 74%, while only 1/5 the amount of model parameters.