Improved YOLOv5-based small object detection method for UAVs

Sairu Liu; Jiarui Ni; Xiaoyang Hu

doi:10.1117/12.3025436

28 February 2024 Improved YOLOv5-based small object detection method for UAVs

Sairu Liu, Jiarui Ni, Xiaoyang Hu

Author Affiliations +

Proceedings Volume 13071, International Conference on Mechatronic Engineering and Artificial Intelligence (MEAI 2023); 130713E (2024) https://doi.org/10.1117/12.3025436
Event: International Conference on Mechatronic Engineering and Artificial Intelligence (MEAI 2023), 2023, Shenyang, China

Abstract

Many factors cause false or missed detection of small UAV objects, such as large changes in object scale due to illumination, dense objects, complex backgrounds and occlusions that lead to low model detection accuracy. To solve the above problems, an improved UAV small object detection method is proposed, based on the YOLOv5.Replace the original conv2d detection head with Adaptive Spatial Feature Fusion, add Attentional Convolutional Mixtures, and replace the original regression loss function with F-EIOU. Extensive experiments are conducted on the Visdrone2019 dataset. The experimental results show that the improved YOLOv5 increases the mAP@0.5 by 6.1%, the mAP@0.5:0.95 by 2.7%, the recall by 5.7% and the precision by 3.2% on the Visdrone2019 dataset, meeting the practical needs of UAV small object detection in complex scenarios.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Sairu Liu, Jiarui Ni, and Xiaoyang Hu "Improved YOLOv5-based small object detection method for UAVs", Proc. SPIE 13071, International Conference on Mechatronic Engineering and Artificial Intelligence (MEAI 2023), 130713E (28 February 2024); https://doi.org/10.1117/12.3025436

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

;

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE