Enhanced Detection and Classification of PCB Defects Utilizing Advanced YOLO Architectures

In rapidly growing industries, speed and accuracy in manufacturing products has become paramount. In electronics production, particularly in Printed Circuit Boards (PCBs), detecting and classifying defects is crucial to ensuring product reliability and quality. PCBs are highly sophisticated, and defects are often imperceptible to the naked eye. To address this challenge, various data augmentation techniques were applied to the dataset to enhance model training. These techniques enabled the model to learn essential features, facilitating faster defect localization and classification into different categories. This research presents an advanced methodology for PCB defect detection using the latest version of YOLO (You Only Look Once) integrated with specialized enhancements. These enhancements include the Convolutional Block Attention Module (CBAM), an attention mechanism, and an Intersection over Union (IoU) based loss function. These additions significantly improve the model’s ability to accurately detect defects by minimizing false negatives. Experimental evaluations were conducted with YOLO versions 8, 9, and 10, demonstrating that integrating CBAM and the IoU loss function yields promising results compared to traditional YOLO models. This paper provides manufacturing companies with a powerful mechanism to enhance efficiency and product quality through automated PCB defect detection techniques.