Abstract: Surface defect detection is a key part of product quality inspection, and with the rapid development of deep learning technology in recent years, the surface defect detection technology of metal materials has been greatly improved. This paper compares and analyzes the surface defect detection methods of metal materials based on deep learning in recent years, and discusses the status of research and application effects in recent years from three aspects: supervised methods, unsupervised methods, and weakly supervised methods. Finally, the key problems and solutions in the detection of surface defects in metal materials are systematically summarized. The further development of surface defect detection is considered and foreseen in the light of industrial needs.