Small target detection algorithm based on improved YOLOv5

Yushan Li; Yuanjing Zhu; Peng Hu; Honglin Wang; Hongwei Ding

doi:10.1117/12.3034943

7 August 2024 Small target detection algorithm based on improved YOLOv5

Yushan Li, Yuanjing Zhu, Peng Hu, Honglin Wang, Hongwei Ding

Proceedings Volume 13224, 4th International Conference on Internet of Things and Smart City (IoTSC 2024); 132241Y (2024) https://doi.org/10.1117/12.3034943
Event: 4th International Conference on Internet of Things and Smart City, 2024, Hangzhou, China

Abstract

At present, the Internet and smart cities are developing rapidly, and all social services. Small target detection is widely used in these fields. This paper focuses on small target detection based on YOLOv5. The feature enhancement module is proposed to solve the problem of incomplete extraction of small target features by the model. An attention mechanism is added to the model, so that the detection model pays more attention to the region in the image that contains the target to be detected, increases the weight of the feature information in this part, enriches the feature space, and further improves the performance of small target detection. Finally, by combining the use of network pruning and knowledge distillation, the model is compressed to compress the model size and improve the detection speed under the premise of ensuring that the model detection accuracy is not affected. The experimental results show that our optimisation effectively improves small target detection in terms of both detection accuracy and detection speed, with FPS improved by 31.84 and AP improved by 1.7%.

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

Citation Download Citation

Yushan Li, Yuanjing Zhu, Peng Hu, Honglin Wang, and Hongwei Ding "Small target detection algorithm based on improved YOLOv5", Proc. SPIE 13224, 4th International Conference on Internet of Things and Smart City (IoTSC 2024), 132241Y (7 August 2024); https://doi.org/10.1117/12.3034943

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

Includes PDF, HTML & Video, when available