Research on Metal Surface Defect Detection Method Based on Deep Learning

In deep learning single-stage object detection algorithms, the YOLO series has emerged as a prominent focus in the object detection domain due to its streamlined architecture, high detection efficiency, and excellent accuracy. The YOLOv8 variant, in particular, demonstrates superior overall performance in terms of real-time processing and precision, making it an ideal algorithm for defect detection tasks. To address the challenges in metal surface defect detection, we developed CDA-YOLOv8, an enhanced architecture based on the YOLOv8s framework. The proposed enhancements encompass three critical components: the feature extraction network, the feature fusion module, and the detection head network. Through this series of optimizations, the model's effectiveness in detecting metal surface defects is enhanced, thereby providing technical support for the application of deep learning in practical industrial production scenarios.