Training method and system of steel surface defect detection model

The invention relates to the field of computer vision, and provides a training method and system for a steel surface defect detection model, and the method comprises the steps: designing a data enhancement network comprising an image data generation module and an image data identification module to carry out the confrontation game, so as to obtain more real and representative synthetic image data; according to the method, the manufactured steel surface defect image data set is richer and more comprehensive, the generalization ability of a steel surface defect detection model is greatly improved, the learning rate is adaptively adjusted through the adaptive moment estimation optimization module to optimize model training, the learning rate selection difficulty is avoided, and the learning efficiency is improved. The problems that the convergence speed is low and the local optimal solution is difficult to jump out are solved, and the performance of the steel surface defect detection model is greatly improved. 本