Steel Defect Detection Based on YOLO

Steel defect detection plays a critical role in the quality control and maintenance of manufacturing processes, especially in the steel industry. This project presents a Steel Defect Detection system using the YOLOv12 (You Only Look Once) algorithm, which leverages deep learning for efficient and accurate defect classification. The system is trained to detect various defects such as crazing, inclusion, patches, pitted surfaces, rolled-in scale, and scratches in steel plates. The YOLOv12 model is employed for its ability to perform real-time object detection and classification with high accuracy, enabling the automatic identification of defects in steel surfaces. The system achieves notable performance metrics, including precision, recall, and F1-scores across different defect classes. For example, the "patches" class shows an impressive F1-score of 0.932, while other classes like "crazing" and "scratches" display significant detection capabilities as well. This project utilizes Python-based libraries and frameworks to train and evaluate the model, optimizing it for the specific needs of steel defect detection. The results demonstrate the potential of deep learning techniques, particularly YOLOv12, in enhancing industrial processes by automating defect detection, reducing human error, and improving the overall efficiency of quality control in steel production. The system contributes to advancing the automation of defect detection in the steel industry, ensuring better product quality and consistency in production lines.

Keywords: Steel Defect Detection, YOLOv12, Deep Learning, Object Detection, Industrial Quality Control, Crazing, Inclusion, Patches, Pitted Surface, Rolled-in Scale, Scratches, Machine Learning, Precision, Recall, F1-Score, Python.