In view of the low accuracy of the existing defect detection methods for stamped parts,the principle and method of deep learning were analyzed,and the VGG13 network was used as the reference model to improve by adding CBAM modules after the feature extraction layer,and five kinds of network model based on VGG13 and CBAM were proposed.The network model (VGG13-CBAM) combined with the attention mechanism module was experimentally studied on the stamped parts defect dataset collected by a manufacturing workshop in Wuhan with the improved new models and the original VGG13 model.The dataset was divided into training set,validation set,and test set according to 6∶2∶2,and data enhancement was used to further expand the training set,the generalization performance of the model was increased,and the improvement effects before and after data enhancement were compared.The experimental results show that the effect is significantly improved on the improved VGG13-CBAM03 network and VGG13-CBAM04 network,and the test sets accuracy rate is increased from 79.65% to 81.55% and 81.40%,respectively;after expanding the training set by using data enhancement,the accuracy of the test sets reach 84.25% and 84.15%,respectively,which effectively improves the accuracy of stamped parts defect detection.