WCSE

Surface Defect Detection of Components Based on SE-YOLOv5

Abstract—With the continuous development of modern industry, all kinds of equipment have higher and higher requirements for components, and minor faults of components may cause great impact on equipment. Therefore, it is very important to detect component defects under modern industrialization conditions. The contrast and defect characteristics of industrial components are different, and the traditional visual detection method has poor effect and low accuracy, which cannot meet the requirements of modern industrial detection. Aiming at the deficiency of traditional visual detection for small defect target detection, a fault diagnosis method for surface defects of industrial components based on SE-YOLOv5 is proposed in this paper. YOLOv5 is very suitable for visual defect detection because of its fast calculation speed and good expression of bearing defect characteristics. However, its accuracy of defect detection is not high enough. Therefore, SE attention module is embedded on the basis of YOLOv5 and the new model SE-YOLOV5 is obtained, which not only greatly improves the calculation speed, but also improves the accuracy of visual defect detection. Moreover, higher accuracy of defect diagnosis is obtained.

Index Terms—SE-YOLOV5 defect detection SE attention module

Zheng Lehui, Sun Junjie, Niu Rui
Graduate School, Engineering University of PAP, Xi’an 710086, CHINA
Huang Ying
Graduate School, Information and Communication, Engineering University of PAP, Xi’an 710086, CHINA

[Download]

Cite: Zheng Lehui, Sun Junjie, Niu Rui, Huang Ying, "Surface Defect Detection of Components Based on SE-YOLOv5, " Proceedings of 2022 the 12th International Workshop on Computer Science and Engineering (WCSE 2022), pp. 191-197, June 24-27, 2022.