Surface Defect Detection of Remanufactured Products by Using the Improved Yolov5

This paper presented a machine learning method to achieve accurate surface defect detection and classification of remanufactured products. An improved You Only Look Once (YOLO) network was proposed for training and testing an image detection model on a steel surface defect dataset. The results show that the proposed YOLO model has high accuracy in detecting surface defects for the remanufacturing quality detection, and the accuracy of the modified YOLO model was improved by 2.4% when compared to the original model in mAP0.5. Moreover, the improved model reasonably reduced the simulation calculation. The improved YOLO model has less operation, higher precision, and higher speed, which has practical application value in the surface defect detection of remanufactured products.