Abstract: A fast classification method based on transfer learning is proposed for detecting defects in magnetic tile. This method pre-trains the deep convolutional neural network VGG16 on the large image dataset ImageNet and then uses the transfer learning method to transfer the model to the study of magnetic tile defect classification. First, freeze the parameters of the first few layers of the model.Secondly, the magnetic tile defect dataset is used for training and adjustment of fully connected layer.Finally, test the classification effect of the model on the test set. Experimental results show that the recognition accuracy of the six types of magnetic tile defects reached 98.69%, which is significantly higher than the recognition accuracy of manual classification and traditional machine vision classification methods. The superiority of this method provides a reliable basis for practical applications in industrial production.