YOLO-IMF: An Improved YOLOv8 Algorithm for Surface Defect Detection in Industrial Manufacturing Field

Surface quality is an effective metric to evaluate the quality of industrial products. The traditional automatic detection methods such as eddy current detection method, infrared detection method and magnetic flux leakage method are limited to particular environments and cannot achieve satisfied accuracy. Deep learning based vision detection such as YOLO algorithm is a new promising method. However, due to complex real industrial environment, the direct use of existing methods still has some limitations. To solve the challenge, in this paper, we propose an improved YOLOv8 algorithm - YOLO-IMF to address the issue of surface defect recognition on aluminum plates. By replacing the CIOU loss function with the EIOU loss function, we can better measure the similarity between small targets and targets with irregular shapes, thereby enhancing the effectiveness of bounding box regression. Experimental results demonstrate an obvious improvement in defect detection, with a mean precision increasing from 98.1% to 99.3%. Moreover, the detection performance outperforms YOLOv5m and Faster R-CNN algorithms.