Image classification of industrial parts based on image generation

The application of computer vision in industrial parts image classification has become more and more widespread. In order to solve the problem of accurately classifying industrial precision parts in industrial automation, a classification method based on image generation technology is proposed. First, different types of part image data are collected, the image data is expanded through personalized generation models, and then classifier filtering is used to construct a new data training set. This paper proposes a method for data enhancement and expansion of small sample data sets, which can effectively increase the number of training samples in the classifier model and improve the quality of model training. The research results show that compared with the method that does not use image generation data as a training set, the overall accuracy rate increases from 51.46% to 59.06%, and the part with the lowest accuracy rate increases from 38.75% to 44.375%. When classifying small samples of industrial precision parts images, the method of using the generative model proposed in this paper to expand image data can significantly improve the recognition accuracy of the classifier model.