Design Compact YOLO based Network for Small Target Detection on Infrared Image

Shuang Liu; Zhicheng Liu; Yuhai Li; Wancheng Liu; Chenyang Ge; Longjun Liu

doi:10.1109/CAC57257.2022.10054751

Design Compact YOLO based Network for Small Target Detection on Infrared Image

Shuang Liu
, Zhicheng Liu
, Yuhai Li
, Wancheng Liu
, Chenyang Ge
, Longjun Liu

College of Artificial Intelligence and Robotics

Xi'an Jiaotong University
Science and Technology on Electro-Optical Information Security Control Laboratory

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

6 Scopus citations

Abstract

most YOLO object detection neural networks prefer to focus on traditional RGB image, but previous studies rarely consider special YOLO network with compact architecture for infrared image. In this paper, we analyze original YOLO network architecture, and we propose a compact YOLO based network by using different blocks from previous work for small target detection on infrared image. We use small target detection layer, GhostConv convolution, Focus structure, Focal EIOU Loss and soft NMS modules to improve original YOLO network structure, which improves the accuracy and speed of target detection for infrared image. The experimental results show that the accuracy of target detection of original model can be effectively improved by adding the small target detection layer, the Focus structure and the BI-FPN structure. By replacing the GhostConv convolution, the calculation speed is significantly improved. At last, after the model is quantized and deployed with NCNN framework, the object detection speed can reached to 35ms per infrared image.

Original language	English
Title of host publication	Proceedings - 2022 Chinese Automation Congress, CAC 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	4991-4996
Number of pages	6
ISBN (Electronic)	9781665465335
DOIs	https://doi.org/10.1109/CAC57257.2022.10054751
State	Published - 2022
Event	2022 Chinese Automation Congress, CAC 2022 - Xiamen, China Duration: 25 Nov 2022 → 27 Nov 2022

Publication series

Name	Proceedings - 2022 Chinese Automation Congress, CAC 2022
Volume	2022-January

Conference

Conference	2022 Chinese Automation Congress, CAC 2022
Country/Territory	China
City	Xiamen
Period	25/11/22 → 27/11/22

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 7 Affordable and Clean Energy

Keywords

DNN architecture
Infrared image
Object detection
YOLO based neural network

Access to Document

10.1109/CAC57257.2022.10054751

Cite this

Liu, S., Liu, Z., Li, Y., Liu, W., Ge, C., & Liu, L. (2022). Design Compact YOLO based Network for Small Target Detection on Infrared Image. In Proceedings - 2022 Chinese Automation Congress, CAC 2022 (pp. 4991-4996). (Proceedings - 2022 Chinese Automation Congress, CAC 2022; Vol. 2022-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CAC57257.2022.10054751

@inproceedings{1fcf9383fda044fea6c92fe89b782d7f,

title = "Design Compact YOLO based Network for Small Target Detection on Infrared Image",

abstract = "most YOLO object detection neural networks prefer to focus on traditional RGB image, but previous studies rarely consider special YOLO network with compact architecture for infrared image. In this paper, we analyze original YOLO network architecture, and we propose a compact YOLO based network by using different blocks from previous work for small target detection on infrared image. We use small target detection layer, GhostConv convolution, Focus structure, Focal EIOU Loss and soft NMS modules to improve original YOLO network structure, which improves the accuracy and speed of target detection for infrared image. The experimental results show that the accuracy of target detection of original model can be effectively improved by adding the small target detection layer, the Focus structure and the BI-FPN structure. By replacing the GhostConv convolution, the calculation speed is significantly improved. At last, after the model is quantized and deployed with NCNN framework, the object detection speed can reached to 35ms per infrared image.",

keywords = "DNN architecture, Infrared image, Object detection, YOLO based neural network",

author = "Shuang Liu and Zhicheng Liu and Yuhai Li and Wancheng Liu and Chenyang Ge and Longjun Liu",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 Chinese Automation Congress, CAC 2022 ; Conference date: 25-11-2022 Through 27-11-2022",

year = "2022",

doi = "10.1109/CAC57257.2022.10054751",

language = "英语",

series = "Proceedings - 2022 Chinese Automation Congress, CAC 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "4991--4996",

booktitle = "Proceedings - 2022 Chinese Automation Congress, CAC 2022",

}

Liu, S, Liu, Z, Li, Y, Liu, W, Ge, C & Liu, L 2022, Design Compact YOLO based Network for Small Target Detection on Infrared Image. in Proceedings - 2022 Chinese Automation Congress, CAC 2022. Proceedings - 2022 Chinese Automation Congress, CAC 2022, vol. 2022-January, Institute of Electrical and Electronics Engineers Inc., pp. 4991-4996, 2022 Chinese Automation Congress, CAC 2022, Xiamen, China, 25/11/22. https://doi.org/10.1109/CAC57257.2022.10054751

Design Compact YOLO based Network for Small Target Detection on Infrared Image. / Liu, Shuang; Liu, Zhicheng; Li, Yuhai et al.
Proceedings - 2022 Chinese Automation Congress, CAC 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 4991-4996 (Proceedings - 2022 Chinese Automation Congress, CAC 2022; Vol. 2022-January).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Design Compact YOLO based Network for Small Target Detection on Infrared Image

AU - Liu, Shuang

AU - Liu, Zhicheng

AU - Li, Yuhai

AU - Liu, Wancheng

AU - Ge, Chenyang

AU - Liu, Longjun

PY - 2022

Y1 - 2022

N2 - most YOLO object detection neural networks prefer to focus on traditional RGB image, but previous studies rarely consider special YOLO network with compact architecture for infrared image. In this paper, we analyze original YOLO network architecture, and we propose a compact YOLO based network by using different blocks from previous work for small target detection on infrared image. We use small target detection layer, GhostConv convolution, Focus structure, Focal EIOU Loss and soft NMS modules to improve original YOLO network structure, which improves the accuracy and speed of target detection for infrared image. The experimental results show that the accuracy of target detection of original model can be effectively improved by adding the small target detection layer, the Focus structure and the BI-FPN structure. By replacing the GhostConv convolution, the calculation speed is significantly improved. At last, after the model is quantized and deployed with NCNN framework, the object detection speed can reached to 35ms per infrared image.

AB - most YOLO object detection neural networks prefer to focus on traditional RGB image, but previous studies rarely consider special YOLO network with compact architecture for infrared image. In this paper, we analyze original YOLO network architecture, and we propose a compact YOLO based network by using different blocks from previous work for small target detection on infrared image. We use small target detection layer, GhostConv convolution, Focus structure, Focal EIOU Loss and soft NMS modules to improve original YOLO network structure, which improves the accuracy and speed of target detection for infrared image. The experimental results show that the accuracy of target detection of original model can be effectively improved by adding the small target detection layer, the Focus structure and the BI-FPN structure. By replacing the GhostConv convolution, the calculation speed is significantly improved. At last, after the model is quantized and deployed with NCNN framework, the object detection speed can reached to 35ms per infrared image.

KW - DNN architecture

KW - Infrared image

KW - Object detection

KW - YOLO based neural network

UR - https://www.scopus.com/pages/publications/85151120680

U2 - 10.1109/CAC57257.2022.10054751

DO - 10.1109/CAC57257.2022.10054751

M3 - 会议稿件

AN - SCOPUS:85151120680

T3 - Proceedings - 2022 Chinese Automation Congress, CAC 2022

SP - 4991

EP - 4996

BT - Proceedings - 2022 Chinese Automation Congress, CAC 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 Chinese Automation Congress, CAC 2022

Y2 - 25 November 2022 through 27 November 2022

ER -

Design Compact YOLO based Network for Small Target Detection on Infrared Image

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this