摘要
为了提升海下垃圾检测精度,降低漏检频率,提出了一种基于改进YOLOv5s的检测算法。该算法添加了ECA模块来关注感兴趣区域,构造了四级多尺度特征融合使提取的特征图信息更加丰富,引用损失函数使得边框回归更加合适与准确。改进的YOLOv5s算法在TrashCan数据集上的mAP由原来的87.3%提高到92.1%,具有较高的精度。
In order to improve the accuracy of undersea garbage detection and reduce the occurrence of missed detection,this paper proposes a detection algorithm based on improved YOLOv5s.The algorithm adds an ECA module to focus on the region of interest,constructs a four-level multi-scale feature fusion to make the extracted feature map more informative,and references the loss function to make the edge regression more appropriate and accurate.The improved YOLOv5s algorithm improves the mAP on the TrashCan dataset from 87.3%to 92.1%with high accuracy.
作者
吴观茂
王涛
WU Guanmao;WANG Tao(College of Computer Science and Engineering,Anhui University of Science and Technology,Huainan Anhui 232001)
出处
《湖北理工学院学报》
2024年第1期47-51,共5页
Journal of Hubei Polytechnic University
基金
安徽省自然科学基金面上项目(项目编号:1908085MF189)
安徽省重点研究与开发计划项目(项目编号:202004b11020029)。
