PCB Defect Detection In Manufacturing Using Deep Learning

This review paper analyzed various deep learning approaches used for PCB defect detection in manufacturing. The results discussed in the previous section confirm that deep learning models such as CNNs and YOLO-based object detection techniques significantly improve defect detection accuracy compared to traditional image processing methods. The analysis also shows that modern models can effectively detect common PCB defects such as open circuits, short circuits, mouse bites, and spurious copper. However, challenges still exist in detecting extremely small defects and handling limited datasets. Overall, deep learning provides an efficient solution for automated PCB inspection. Future research should focus on improving lightweight models, increasing dataset diversity, and developing more robust detection systems for industrial manufacturing environments.

[1]   Elsharkawy, A., Hassan, M., & Abdelrahman, T. (2025). YOLOv11-PCB: Enhanced object detection for printed circuit board inspection. Scientific Reports, 15(2), 1–14. [2]   Huang, J., Zhao, F., & Chen, L. (2025). GAN-augmented YOLOv11 for rare PCB defect detection. arXiv preprint. [3]   HoangVan, T., Nguyen, D., & Tran, H. (2025). ESRPCB: Edge-guided super-resolution for micro PCB defect detection. arXiv preprint. [4]   Ouyang, J., & Li, K. (2025). Vision-language models for PCB defect classification. ACM Transactions on Multimedia Computing, 21(1), 1–15. [5]   Pingzhen, P., Li, X., & Zhao, R. (2025). Lightweight multi-scale YOLOv8 with adaptive pruning for PCB inspection. arXiv preprint. [6]   Song, Y., Liu, H., & Wang, Z. (2025). YOLO-HDEW: Edge-enhanced multi-scale attention for small PCB defect detection. Electronics, 14(3), 1–16. [7]   Morales, D., Garcia, P., & Ortega, J. (2024). Semi-supervised learning for PCB defect detection using consistency regularization. Proceedings of the AAAI Conference on Artificial Intelligence, 4561–4568. [8]   Sharma, N., & Patel, R. (2024). Real-time PCB defect detection using YOLOv8. IEEE Sensors Journal, 24(6), 8123–8134. [9]   Singh, A., & Arora, V. (2024). Residual U-Net++ for PCB micro-crack detection. Machines, 12(4), 298. [10]               Tanaka, K., Ito, S., & Yamamoto, H. (2024). Diffusion-based synthetic PCB dataset generation. arXiv preprint.