Improved 2D target detection with YoloV5 based on attention mechanism

Yong Zhou; Mei Zhang; Fujin Hou; Bin Lv; Jianqing Wu

doi:10.1117/12.2643639

9 October 2022 Improved 2D target detection with YoloV5 based on attention mechanism

Yong Zhou, Mei Zhang, Fujin Hou, Bin Lv, Jianqing Wu

Proceedings Volume 12246, 2nd International Conference on Signal Image Processing and Communication (ICSIPC 2022); 122460S (2022) https://doi.org/10.1117/12.2643639
Event: 2nd International Conference on Signal Image Processing and Communication (ICSIPC 2022), 2022, Qingdao, China

Abstract

For 2D image data target detection, this paper proposes an improved detection algorithm for YoloV5s based on the attention mechanism. By replacing the Focus module in YoloV5s with a CBL module, it facilitates model export. The four attention modules, SELayer, EcaLayer, CBAM and CoordAtt are added to the convolutional network of YoLoV5s. Before the P4 layer, before the P5 layer, before the SPPF layer and before the last layer of the original backbone network, respectively. By experimenting with algorithms that include YoloV5s, It was concluded that among the five algorithms tested for evaluation. The proposed YoloV5s-CoordAtt has the best performance level in terms of accuracy, with a 4.48% improvement compared to the original algorithm.

Citation Download Citation

Yong Zhou, Mei Zhang, Fujin Hou, Bin Lv, and Jianqing Wu "Improved 2D target detection with YoloV5 based on attention mechanism", Proc. SPIE 12246, 2nd International Conference on Signal Image Processing and Communication (ICSIPC 2022), 122460S (9 October 2022); https://doi.org/10.1117/12.2643639

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