DFP-YOLO: An Efficient Algorithm for Detecting Steel Surface Defects.

Saved in:

Bibliographic Details
Title:	DFP-YOLO: An Efficient Algorithm for Detecting Steel Surface Defects.
Authors:	Jiawei Chai¹ 2192520875@qq.com, Ziwei Zhou² 381431970@qq.com
Source:	IAENG International Journal of Computer Science. Jun2025, Vol. 52 Issue 6, p17-54. 1763p.
Subjects:	Surface plates, Surface defects, Iron & steel plates, Feature extraction, Boosting algorithms
Abstract:	To enhance the accuracy of steel plate surface defect detection and minimize the incidence of misdetection and leakage, this paper proposes a DFP-YOLO algorithm based on YOLOv8n to achieve efficient detection of steel plate surface defects. Firstly, the C2f module of the backbone network and Neck layer is substituted by the DWR_DRB module to strengthen the ability of capturing defects at various scales and enhance the efficiency of model feature extraction. Secondly, the Feature Pyramid Share Convolution module is devised to extract multi-scale features through convolutional layers with different dilation rates, integrating local details with global contextual information for a better comprehension of complex scenes. Finally, a Powerful-IoU loss function is utilized to control the scale size of the auxiliary boundary to accelerate the detection speed and improve the model accuracy. The experimental results demonstrate that the proposed algorithm in this paper boosts the mean accuracy (mAP) of the steel plate surface defect detection task by 3.4% compared to the original YOLOv8n model, increases the accuracy by 8.8%, and raises the inference speed of the model by 49 frames per second when conducting DFP-YOLO detection on the dataset NEU-DET. Meanwhile, the generalization and robustness of the model are verified through tests on the industrial steel plate surface defects dataset GC10-DET and the larger public benchmark dataset PASCAL VOC 2012. [ABSTRACT FROM AUTHOR]
	Copyright of IAENG International Journal of Computer Science is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	To enhance the accuracy of steel plate surface defect detection and minimize the incidence of misdetection and leakage, this paper proposes a DFP-YOLO algorithm based on YOLOv8n to achieve efficient detection of steel plate surface defects. Firstly, the C2f module of the backbone network and Neck layer is substituted by the DWR_DRB module to strengthen the ability of capturing defects at various scales and enhance the efficiency of model feature extraction. Secondly, the Feature Pyramid Share Convolution module is devised to extract multi-scale features through convolutional layers with different dilation rates, integrating local details with global contextual information for a better comprehension of complex scenes. Finally, a Powerful-IoU loss function is utilized to control the scale size of the auxiliary boundary to accelerate the detection speed and improve the model accuracy. The experimental results demonstrate that the proposed algorithm in this paper boosts the mean accuracy (mAP) of the steel plate surface defect detection task by 3.4% compared to the original YOLOv8n model, increases the accuracy by 8.8%, and raises the inference speed of the model by 49 frames per second when conducting DFP-YOLO detection on the dataset NEU-DET. Meanwhile, the generalization and robustness of the model are verified through tests on the industrial steel plate surface defects dataset GC10-DET and the larger public benchmark dataset PASCAL VOC 2012. [ABSTRACT FROM AUTHOR]
ISSN:	1819656X