GDCP-YOLO: Enhancing Steel Surface Defect Detection Using Lightweight Machine Learning Approach

• DOI: 10.3390/electronics13071388
• Publication Date: 2024-04-08T07:11:33Z
• Authors: Zhaohui Yuan; Hao Ning; Xiangyang Tang; Zhengzhe Yang
• Keywords: Unmanned surface vehicle

Surface imperfections in steel materials potentially degrade quality and performance, thereby escalating the risk of accidents engineering applications. Manual inspection, while traditional, is laborious lacks consistency. However, recent advancements machine learning computer vision have paved way for automated defect detection, yielding superior accuracy efficiency. This paper introduces an innovative deep model, GDCP-YOLO, devised multi-category detection. We enhance reference YOLOv8n architecture by incorporating adaptive receptive fields via DCNV2 module channel attention C2f. These integrations aim to concentrate on valuable features minimize parameters. incorporate efficient Faster Block employ Ghost convolutions generate more feature maps with reduced computation. modifications streamline extraction, curtail redundant information processing, boost detection speed. Comparative trials NEU-DET dataset underscore state-of-the-art performance GDCP-YOLO. Ablation studies generalization experiments reveal consistent across a variety types. The optimized lightweight facilitates real-time inspection without sacrificing accuracy, offering invaluable insights further techniques surface identification manufacturing sectors.