基于SimAM-YOLOv4的自动驾驶目标检测算法  

Automatic driving target detection algorithm based on SimAM-YOLOv4

作　　者：刘丽伟[1] 侯德彪 侯阿临[1] 梁超[1] 郑贺伟 LIU Liwei;HOU Debiao;HOU Alin;LIANG Chao;ZHENG Hewei(School of Computer Science and Engineering,Changchun University of Technology,Changchun 130102,China)

机构地区：[1]长春工业大学计算机科学与工程学院,吉林长春130102

出　　处：《长春工业大学学报》2022年第3期244-250,共7页Journal of Changchun University of Technology

基　　金：吉林省科技发展计划重点项目(20200403037SF);吉林省教育厅科学技术研究规划项目(JJKH20210738KJ)。

摘　　要：针对自动驾驶场景下单阶段目标检测对小目标精度不足的问题,权衡精度与速度的共同需求,提出一种改进的YOLOv4目标检测算法。首先,在网络的残差模块中嵌入SimAM注意力模块,旨在提高网络对重要特征的提取能力,然后,利用ACON-C激活函数替换残差模块中的Mish激活函数,使残差模块可以自适应地激活,进而提升网络性能。在KITTI数据集上进行训练和测试,实验结果表明,该模型的平均精度均值达到91.16%,检测速度达到32帧/s,满足实时检测的要求。Aiming at the problem of insufficient precision for small targets in single-stage target detection in autonomous driving scenarios,and weighing the common requirements of accuracy and speed,an improved YOLOv4 target detection algorithm is proposed.First,the SimAM attention module is embedded in the residual module of the network to improve the network's ability to extract important features,and then the ACON-C activation function is used to replace the Mish activation function in the residual module,so that the residual module can adaptively activation,thereby improving the performance of the network.Training and testing are performed on the KITTI data set.The experimental results show that the average precision of the model reaches 91.16%,and the detection speed reaches 32 FPS,which meets the requirements of real-time detection.

关 键 词：目标检测 YOLOv4 SimAM注意力模块 ACON-C 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]