Leather fabrics defect detection based on an improved YOLOv5s
Defect detection methods based on machine learning have been widely used in the field of fabrics,but there are still the problem of low detection accuracy and fast detection speed,which can't reach the purpose of industrial real-time detection.An improved YOLOv5s based defect detection algorithm for leather fabrics,named G3-YOLOv5,is proposed to address these issues.Ghostconv is introduced in Backbone,and a lightweight convolutional structure called GSConv is also integrated into Neck to reduce the number of parameters in the model and improve the detection efficiency.Finally,the GAM attention mechanism is added to improve the model's ability to perceive defects of different sizes,which improves the model's detection accuracy while ensuring the detection speed.According to experimental findings,the mAP value(86.2%)of the G3-YOLOv5 algorithm is 7.61%higher than that of YOLOv5s on the self-built dataset.The detection speed(FPS)on CUP is 2.68 Hz,which is 0.31 Hz faster than that of the original YOLOv5s,which basically meets the needs of detecting defects in leather fabric.
deep learningleather fabric defect detectionlightweightattention mechanism
万方数据
维普
