Faster Metallic Surface Defect Detection Using Deep Learning with Channel Shuffling

作     者：Siddiqui Muhammad Yasir Hyunsik Ahn 

作者机构：Department of Robot System EngineeringTongmyong UniversityBusan48520Korea School of Artificial IntelligenceTongmyong UniversityBusan48520Korea 

出 版 物：《Computers, Materials & Continua》 (计算机、材料和连续体（英文）)

年 卷 期：2023年第75卷第4期

页      面：1847-1861页

核心收录：

学科分类：0710[理学-生物学] 12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Tongmyong University Innovated University Busan Metropolitan City 

主　　题：Defect detection deep learning convolution neural network object detection YOLOv5 shuffleNetv2 

摘      要：Deep learning has been constantly improving in recent years,and a significant number of researchers have devoted themselves to the research of defect detection *** and recognition of small and complex targets is still a problem that needs to be *** authors of this research would like to present an improved defect detection model for detecting small and complex defect targets in steel *** steel strip production,mechanical forces and environmental factors cause surface defects of the steel ***,the detection of such defects is key to the production of high-quality ***,surface defects of the steel strip cause great economic losses to the high-tech *** far,few studies have explored methods of identifying the defects,and most of the currently available algorithms are not sufficiently ***,this study presents an improved real-time metallic surface defect detection model based on You Only Look Once(YOLOv5)specially designed for small *** the smaller features of the target,the conventional part is replaced with a depthwise convolution and channel shuffle *** assigning weights to Feature Pyramid Networks(FPN)output features and fusing them,increases feature propagation and the network’s characterization *** experimental results reveal that the improved proposed model outperforms other comparable models in terms of accuracy and detection *** precision of the proposed model achieved by mAP@0.5 is 77.5%on the Northeastern University,Dataset(NEU-DET)and 70.18%on the GC10-DET datasets.