Lightweight Object Detection Networks for UAV Aerial Images Based on YOLO

被引:1
作者
Li, Yanshan [1 ,2 ]
Wang, Jiarong [1 ,2 ]
Zhang, Kunhua [1 ,2 ]
Yi, Jiawei [1 ,2 ]
Wei, Miaomiao [1 ,2 ]
Zheng, Lirong [1 ,2 ]
Xie, Weixin [1 ,2 ]
机构
[1] Shenzhen Univ, ATR Natl Key Lab Def Technol, Shenzhen 518000, Peoples R China
[2] Shenzhen Univ, Guangdong Key Lab Intelligent Informat Proc, Shenzhen 518000, Peoples R China
基金
中国国家自然科学基金;
关键词
Aerial images; Deep learning; You only look once; Lightweight network; Object detection;
D O I
10.23919/cje.2022.00.300
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Existing high-precision object detection algorithms for UAV (unmanned aerial vehicle) aerial images often have a large number of parameters and heavy weight, which makes it difficult to be applied to mobile devices. We propose three YOLO-based lightweight object detection networks for UAVs, named YOLO-L, YOLO-S, and YOLO-M, respectively. In YOLO-L, we adopt a deconvolution approach to explore suitable upsampling rules during training to improve the detection accuracy. The convolution-batch normalization-SiLU activation function (CBS) structure is replaced with Ghost CBS to reduce the number of parameters and weight, meanwhile Maxpool maximum pooling operation is proposed to replace the CBS structure to avoid generating parameters and weight. YOLO-S greatly reduces the weight of the network by directly introducing CSPGhostNeck residual structures, so that the parameters and weight are respectively decreased by about 15% at the expense of 2.4% mAP. And YOLO-M adopts the CSPGhostNeck residual structure and deconvolution to reduce parameters by 5.6% and weight by 5.7%, while mAP only by 1.8%. The results show that the three lightweight detection networks proposed in this paper have good performance in UAV aerial image object detection task.
引用
收藏
页码:997 / 1009
页数:13
相关论文
共 21 条
[1]   Drones for litter mapping: An inter-operator concordance test in marking beached items on aerial images [J].
Andriolo, Umberto ;
Goncalves, Gil ;
Rangel-Buitrago, Nelson ;
Paterni, Marco ;
Bessa, Filipa ;
Goncalves, Luisa M. S. ;
Sobral, Paula ;
Bini, Monica ;
Duarte, Diogo ;
Fontan-Bouzas, Angela ;
Goncalves, Diogo ;
Kataoka, Tomoya ;
Luppichini, Marco ;
Pinto, Luis ;
Topouzelis, Konstantinos ;
Velez-Mendoza, Anubis ;
Merlino, Silvia .
MARINE POLLUTION BULLETIN, 2021, 169
[2]  
Howard AG, 2017, Arxiv, DOI [arXiv:1704.04861, DOI 10.48550/ARXIV.1704.04861]
[3]   Monitoring and surveillance of urban road traffic using low altitude drone images: a deep learning approach [J].
Gupta, Himanshu ;
Verma, Om Prakash .
MULTIMEDIA TOOLS AND APPLICATIONS, 2022, 81 (14) :19683-19703
[4]   GhostNet: More Features from Cheap Operations [J].
Han, Kai ;
Wang, Yunhe ;
Tian, Qi ;
Guo, Jianyuan ;
Xu, Chunjing ;
Xu, Chang .
2020 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2020, :1577-1586
[5]   Searching for MobileNetV3 [J].
Howard, Andrew ;
Sandler, Mark ;
Chu, Grace ;
Chen, Liang-Chieh ;
Chen, Bo ;
Tan, Mingxing ;
Wang, Weijun ;
Zhu, Yukun ;
Pang, Ruoming ;
Vasudevan, Vijay ;
Le, Quoc V. ;
Adam, Hartwig .
2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, :1314-1324
[6]   Positional Precision Analysis of Orthomosaics Derived from Drone Captured Aerial Imagery [J].
Hung, I-Kuai ;
Unger, Daniel ;
Kulhavy, David ;
Zhang, Yanli .
DRONES, 2019, 3 (02) :1-10
[7]   AI-Enabled Object Detection in UAVs: Challenges, Design Choices, and Research Directions [J].
Jain, Ayush ;
Ramaprasad, Rohit ;
Narang, Pratik ;
Mandal, Murari ;
Chamola, Vinay ;
Yu, F. Richard ;
Guizan, Mohsen .
IEEE NETWORK, 2021, 35 (04) :129-135
[8]  
Jocher G., 2023, YOLOv5
[9]  
Li L. D., 2020, Chinese Journal of Electronics, V29, p19K
[10]   Hyperspectral Image Super-Resolution Based on Spatial-Spectral Feature Extraction Network [J].
Li Yanshan ;
Chen Shifu ;
Luo Wenhan ;
Zhou Li ;
Xie Weixin .
CHINESE JOURNAL OF ELECTRONICS, 2023, 32 (03) :415-428