Strip Steel Defect Detection Based on Improved YOLOv5s

Jie Wang, Qiu Fang

2023

Abstract

In response to the issues of low detection accuracy for surface defects in strip steel and difficulty in detecting small target defects in modern steel production processes, this study presents an improved algorithm based on YOLOv5s is proposed for detecting strip-steel surface defects. Firstly, an improved C3 module combining large-kernel depth separable convolution and Squeeze-and-Excitation (SE) attention mechanism is proposed, which increases the global receptive field of the network while adaptively adjusting the weight relationships among different channels in order to enhance the fusion of tiny features in the model. Secondly, A multi-scale pyramidal detection head is introduced as a means to enhance the model’s proficiency in detecting small targets. Finally, the SIoU loss function is employed to more accurately calculate the regression loss and improve the model’s detection accuracy. The results indicate that the mean average precision (mAP) of the proposed algorithm on the NEU-DET dataset reaches 80.6%, an increase of 3.4% compared to the baseline algorithm; moreover, the detection speed reaches 79.3f/s, which meets the real-time requirement of industrialised defect detection.

Paper Citation

in Harvard Style

Wang J. and Fang Q. (2023). Strip Steel Defect Detection Based on Improved YOLOv5s. In Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT; ISBN 978-989-758-677-4, SciTePress, pages 375-380. DOI: 10.5220/0012284300003807

in Bibtex Style

@conference{anit23,
author={Jie Wang and Qiu Fang},
title={Strip Steel Defect Detection Based on Improved YOLOv5s},
booktitle={Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT},
year={2023},
pages={375-380},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012284300003807},
isbn={978-989-758-677-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT
TI - Strip Steel Defect Detection Based on Improved YOLOv5s
SN - 978-989-758-677-4
AU - Wang J.
AU - Fang Q.
PY - 2023
SP - 375
EP - 380
DO - 10.5220/0012284300003807
PB - SciTePress