Maintaining quality control is of utmost importance in the manufacturing industry, particularly when it comes to small electronic devices that heavily depends on printed circuit boards (PCBs). Detection of defects throughout the production process is extremely difficult. However, the incomplete labelling of PCB defect datasets makes it harder to understand models and reduces their precision. To address this issue, this study introduced a novel ROI-based PCB defect dataset that provides comprehensive labeling for all defect classes. To evaluate the effectiveness of the proposed dataset, we employed lightweight object identification model was YOLOv7. This model was designed specifically for efficient and accurate recognition and classification tasks. The quantitative results demonstrate that, when coupled with lightweight deep learning model, with the proposed dataset outperforms existing datasets. This study makes a substantial contribution to addressing the issues related to PCB defect datasets and provides information regarding the efficiency of lightweight object identification algorithms for defect detection tasks.