<p>The structural integrity of printed circuit boards (PCB) plays a critical role in ensuring the performance and reliability of electronic devices. Traditional defect detection methods often struggle to identify minute defects, particularly in complex industrial environments. This paper proposes a novel PCB defect detection framework based on the You Only Look Once architecture. Key innovations include the introduction of a dual-branch backbone network that replaces conventional architectures and enhances multi-scale feature extraction while reducing the number of parameters. The method also incorporates a deformable attention mechanism that optimizes information flow between the backbone and feature fusion layers. In addition, an asymmetric detection head based on depthwise separable convolutions is designed to improve detection performance and reduce computational overhead. Finally, the Wise Intersection over Union version 3 loss function is integrated to improve the accuracy of bounding box regression. Experimental results demonstrate that the proposed method achieves a detection accuracy of 86.6%, representing a 3.2% improvement over the baseline model, while reducing the number of model parameters by 30%. These findings indicate that the proposed method offers an efficient and practical solution for PCB defect detection in complex industrial settings.</p>