Intelligent Surface Defect Detection of Hot Rolled Strip Steel Using YOLOv8 Framework

Hot-rolled strip steel is a common metal product widely used in intelligent industries such as construction, machinery manufacturing, and automotive production. With the development of industrial modernization, the quality inspection of hot-rolled strip steel is of great significance. Previously, deep learning-based computer vision technology has proven to be highly effective in defect detection. In this paper, we propose a method for detecting surface defects in hot-rolled strip steel using the You Only Look Once v8m (YOLOv8m) framework. We validate the performance of the YOLOv8m model using the open-source NEU Surface Defect Dataset from Northeastern University. Furthermore, we train the dataset using the Single Shot MultiBox Detector (SSD) and Faster R-CNN for comparison. The results in terms of accuracy and lightweight design indicate that YOLOv8m outperforms the others and is better suited for practical applications. Additionally, due to the lightweight design of YOLOv8m, our method can run in real-time on embedded devices, enabling automated detection on production lines. The surface defect detection method based on YOLOv8 holds significant application potential, marking an important step for the manufacturing industry in quality inspection and supporting intelligent production in the era of Industry 4.0.