摘要
为了解决传统算法需要人工提特征、检测准确率低以及成本高的问题,设计了一种基于注意力机制的YOLOv5模型用于织物瑕疵检测。该方法通过引入注意力机制,提高了模型对织物图像中瑕疵区域的关注度,从而提高了检测性能。通过试验验证了该方法的有效性。结果表明,在织物瑕疵检测中,相较传统YOLOv5模型,该检测方法的结果准确率提高了1.2%,检测效率可达82.8%。
In order to solve the problem that traditional algorithms require manual feature lifting,low detection accuracy and high cost,a YOLOv5 model based on the attention mechanism was designed for fabric defect detection.The method improves the model's attention to the defective region in the fabric image by introducing the attention mechanism(CBAM model),which improves the detection performance.The effectiveness of the method was verified through experiments.The results showed that detection efficiency of the proposed method was up to 82.8%and a 1.2%improvement in fabric defect detection relative to the traditional YOLOv5 model.
作者
李文泽
LI Wenze(College of Science and Technology,Tianjin University of Finance and Economics,Tianjin 300222,China)
出处
《纺织检测与标准》
2023年第6期16-19,35,共5页
Textile Testing and Standard
关键词
瑕疵检测
注意力机制
深度学习
损失函数
数据增强
defect detection
attention mechanism
deep learning:loss function
data enhancement