<p>The production of hot rolled strip steel plates is affected by various uncertainties, resulting in numerous defects on the steel plate surface, such as scratches, cracks, and inclusions, which significantly impact the surface quality and performance of the steel plates. Considering the high rates of missed and false detections by traditional manual visual inspections, along with the limited accuracy of existing defect recognition algorithms due to insufficient defect sample data, a novel intelligent identification method is proposed for the recognition of steel plate surface defects based on deep convolutional generative adversarial network (DCGAN). Aiming at the problem of insufficient samples of actual industrial surface defect images, the proposed method generates steel plate surface image data with a similar distribution to real data based on DCGAN, which will expand the original sample data. Moreover, batch normalization is used to process the data, and an appropriate activation function is selected to better combine the original GAN and the deep convolutional neural network (DCNN) model in structure, thereby effectively improving the quality of the generated image data. Finally, the extended dataset is used to train the DCNN, and the recognition accuracy of the trained model on the test set is 96.16%. To the high error rate of manual inspection and the difficulty of creating and labeling datasets, the experimental results on actual industrial data demonstrate that the proposed method has good engineering application for the recognition of steel plate surface defects.</p>