A surface defect detection method of steel plate based on YOLOV3

At present, the steel plate surface defect detection technology based on machine vision and convolutional neural network (CNN) has achieved good results. However, these models are mostly two-stage methods, extracting features first and then classifying them, which is slow and inaccurate. Therefore, this paper proposes a single-stage surface defect detection method of steel plate based on yolov3, which can classify defects, determine the location of defects, and greatly improve the detection speed. It is of great significance to realize the automation of cold rolling production line. The experiment shows that the detection speed of this model reaches 62 fps and the accuracy reaches 73 %, which has a good prospect in industry.