Visual Defect Detection in Printed Circuit Boards using YOLO Architectures: A Comparison on the DeepPCB Dataset

Vista sencilla de ítem
dc.creatorBlanco, Alexia
dc.creatorBenítez-González, Ivon Oristela
dc.creatorBernal-de Lázaro, José Manuel
dc.date2026-07-28
dc.date.accessioned2026-07-29T06:30:13Z
dc.descriptionThe detection of defects in Printed Circuit Boards (PCBs) remains a significant challenge in modern electronic manufacturing systems, where the need to reduce costs and improve product reliability has stimulated the development and application of new computational intelligence models. In this context, the YOLO architecture-based family has evolved through multiple versions, establishing itself as one of the leading solutions for automatic defect detection in visual inspection systems within the electronics industry. Based on this fact, the present work compared the YOLOv5 and YOLOv8 architectures for tasks focused on automatic fault detection in printed circuit boards. The methodology employed consisted of evaluating the computer vision models YOLOv5 and YOLOv8 using the DeepPCB image dataset in a GPU-enabled execution environment and comparing their precision, recall, F1-score, and mAP across different IoU thresholds. In addition, a qualitative assessment was conducted to evaluate the generalization and inference capabilities of both architectures, using images not included during the training stage. The results obtained revealed a clear contrast between the two architectures. YOLOv5m achieved the best overall balance between precision, recall, and detection stability, while YOLOv8s demonstrated superior defect localization capability under varying IoU thresholds. Overall, both computer vision models exhibited competitive performance in the DeepPCB dataset. Finally, it is concluded that YOLOv5 remains a suitable alternative for traditional visual inspection processes focused on the electronics industry due to its efficiency and scalability in quality control tasks, while YOLOv8 represents a competitive option when higher precision is prioritized in defect localization.en-US
dc.descriptionLa detección de defectos en placas de circuito impreso (PCB, por sus siglas en inglés) continúa siendo un desafío para los sistemas modernos de manufactura electrónica, donde la necesidad de reducir costos y mejorar la confiabilidad de los productos ha impulsado el desarrollo y aplicación de nuevos modelos de inteligencia computacional. En particular, la familia de arquitecturas YOLO ha evolucionado a lo largo de múltiples versiones, consolidándose como una de las principales soluciones para la detección automática de defectos en sistemas de inspección visual en la industria electrónica. A partir de esto, el presente trabajo comparó YOLOv5 y YOLOv8 para la detección automática de fallos en placas de circuito impreso. La metodología empleada consistió en evaluar ambos modelos de visión computacional mediante el conjunto de imágenes DeepPCB, bajo un entorno de ejecución con soporte GPU, comparando su precisión, sensibilidad, F1-score y mAP con distintos umbrales de solapamiento. Asimismo, se evaluó cualitativamente la capacidad de generalización e inferencia de ambas arquitecturas utilizando imágenes no disponibles durante la etapa de entrenamiento. Los resultados alcanzados mostraron un elevado contraste entre ambas arquitecturas. YOLOv5m destacó por lograr el mejor equilibrio global entre precisión, sensibilidad y estabilidad, mientras que YOLOv8s presentó una mejor capacidad de localización de defectos bajo distintos umbrales. Finalmente, se concluye que YOLOv5 continúa siendo una alternativa adecuada para los procesos tradicionales de inspección visual en la industria electrónica, al ofrecer mayor eficiencia y escalabilidad en tareas de control de calidad, mientras que YOLOv8 representa una opción competitiva cuando se prioriza una mayor precisión en la localización de defectos.es-ES
dc.formatapplication/pdf
dc.identifierhttps://revistas.itm.edu.co/index.php/tecnologicas/article/view/3665
dc.identifier10.22430/22565337.3665
dc.identifier.urihttps://hdl.handle.net/20.500.12622/8196
dc.languagespa
dc.publisherInstituto Tecnológico Metropolitano (ITM)en-US
dc.relationhttps://revistas.itm.edu.co/index.php/tecnologicas/article/view/3665/4189
dc.relation/*ref*/Q. Ling, and N. A. M. Isa, “Printed Circuit Board Defect Detection Methods Based on Image Processing, Machine Learning and Deep Learning: A Survey,” IEEE Access, vol. 11, pp. 15921-15944, Feb. 2023. http://doi.org/10.1109/ACCESS.2023.3245093
dc.relation/*ref*/Y. Zhou, M. Yuan, J. Zhang, G. Ding, and S. Qin, “Review of vision-based defect detection research and its perspectives for printed circuit board,” J. Manuf. Syst., vol. 70, pp. 557–578, Oct. 2023. https://doi.org/10.1016/j.jmsy.2023.08.019
dc.relation/*ref*/S. Alawandi, K. Mallibhat, U. Kudachi, and A. Beedanal, “PCB defects: A unified survey of trends, detection techniques, and limitations through systematic literature review,” J. Electron. Test., vol. 41, pp. 709-787, Dec. 2025. https://doi.org/10.1007/s10836-025-06198-y
dc.relation/*ref*/X. Yu, H. Sun, Y. He, J. Peng, X. Zeng, and L. Chen, “Robust classification method for printed circuit board defects based on virtual and real space cooperative diffusion model,” Eng. Appl. Artif. Intell., vol. 169, p. 114126, Apr. 2026. https://doi.org/10.1016/j.engappai.2026.114126
dc.relation/*ref*/C. Zhang, G. Shi, H. Li, M. Yang, Y. Li, and Z. Bing, “Circuit board welding defect detection based on industrial IoVT,” IEEE Internet Things J., vol. 13, no. 7, pp. 14003-14018, Apr. 2026. https://doi.org/10.1109/JIOT.2026.3651257
dc.relation/*ref*/Fineline, “Proceso de fabricación de PCB – Fineline Global,” fineline-global.com. Accessed: Mar. 21, 2026. [Online]. Available: https://www.fineline-global.com/es/knowledge-centre/pcb-manufacturing-process/
dc.relation/*ref*/PCBWay Assitant 03, “Proceso de producción y nuestros equipos,” pcbway.es. Accessed: Mar. 21, 2026. [Online]. Available: https://www.pcbway.es/pcb-service.html?step=10
dc.relation/*ref*/S. Tang, “tangsanli5201/DeepPCB,” GitHub.com. Accessed: Feb. 21, 2026. [Online]. Available: https://github.com/tangsanli5201/DeepPCB
dc.relation/*ref*/S. Lv et al., “A dataset for deep learning based detection of printed circuit board surface defect,” Sci. Data, vol. 11, p. 811, Jul. 2024. https://doi.org/10.1038/s41597-024-03656-8
dc.relation/*ref*/X. Chen, Y. Wu, X. He, and W. Ming, “A comprehensive review of deep learning-based PCB defect detection,” IEEE Access, vol. 11, pp. 139017-139038, Dec. 2023. https://doi.org/10.1109/ACCESS.2023.3339561
dc.relation/*ref*/X. Wu, Y. Ge, Q. Zhang, and D. Zhang, “PCB defect detection using deep learning methods,” in IEEE 24th International Conference on Computer Supported Cooperative Work in Design (CSCWD), Dalian, China, 2021, pp. 873-876. http://doi.org/10.1109/CSCWD49262.2021.9437846
dc.relation/*ref*/Z. He, Y. Wu, Y. Lv, and Y. He, “PCB-Faster-RCNN: An Improved Object Detection Algorithm for PCB Surface Defects,” Appl. Sci., vol. 15, no. 24, p. 12881, Dec. 2025. https://doi.org/10.3390/app152412881
dc.relation/*ref*/Z. Li, and X. Liu, “Soldering Defect Segmentation Method for PCB on Improved UNet,” Appl. Sci., vol. 14, no. 16, p. 7370, Aug. 2024. https://doi.org/10.3390/app14167370
dc.relation/*ref*/Z. Gao, Y. Li, and S. Yuan, “DECNet: An efficient improved model for printed circuit board surface defect detection,” Signal, Image Video Process., vol. 20, no. 4, p. 188, Mar. 2026. https://doi.org/10.1007/s11760-026-05175-y
dc.relation/*ref*/F. Jiang, H. Chen, S. Wei, and C. Chen, “Performance Analysis of Printed Circuit Board Defect Detection with Hybrid CNN Feature Extraction and Clustering,” Eng, vol. 7, no. 1, p. 41, Jan. 2026. https://doi.org/10.3390/eng7010041
dc.relation/*ref*/J. Shi, J. Yang, and Y. Zhang, “Research on Steel Surface Defect Detection Based on YOLOv5 with Attention Mechanism,” Electronics, vol. 11, no. 22, p. 3735, Nov. 2022. https://doi.org/10.3390/electronics11223735
dc.relation/*ref*/Z. Yang, D. Li, L. Hou, and W. Nai, “A Comprehensive Performance Evaluation of YOLO Series Algorithms in Automatic Inspection of Printed Circuit Boards,” Machines, vol. 14, no. 1, p. 94, Jan. 2026. https://doi.org/10.3390/machines14010094
dc.relation/*ref*/V. Adibhatla, H.-C. Chih, C.-C. Hsu, J. Cheng, M. F. Abbod, and J.-S. Shieh, “Defect detection in printed circuit boards using You-Only-Look-Once convolutional neural networks,” Electronics, vol. 9, no. 9, p. 1547, Sep. 2020. https://doi.org/10.3390/electronics9091547
dc.relation/*ref*/J. Xu, L. Shi, W. Yang, X. Zhu, J. Li, and Y. Wang, “YOLO-CMDS: An enhanced method for PCB defect detection,” Integration, vol. 109, p. 102732, Jul. 2026. https://doi.org/10.1016/j.vlsi.2026.102732
dc.relation/*ref*/H. Peng, W. Yang, and B. Yu, “YOLO-UMS: Multi-Scale Feature Fusion Based on YOLO Detector for PCB Surface Defect Detection,” Sensors, vol. 26, no. 2, p. 689, Jan. 2026. https://doi.org/10.3390/s26020689
dc.relation/*ref*/Z. Wang, P. Yuan, and X. Zhang, “YOLO-CD: A lightweight real-time PCB defect detection model for edge deployments,” J. Real-Time Image Proc., vol. 23, p. 46, Jan. 2026. https://doi.org/10.1007/s11554-025-01847-z
dc.relation/*ref*/V. Kumar Ancha, V. Gonuguntla, and R. Vaddi “CSPGhost-YOLO: A lightweight and robust model for real-time mixed PCB defect detection system,” Measurement, vol. 266, p. 120521, Mar. 2026. https://doi.org/10.1016/j.measurement.2026.120521
dc.relation/*ref*/X. Zuo, N. Zhao, K. Wang, and J. Hu, “MDEB-YOLO: A Lightweight Multi-Scale Attention Network for Micro-Defect Detection on Printed Circuit Boards,” Micromachines, vol. 17, no. 2, p. 192, Jan. 2026. https://doi.org/10.3390/mi17020192
dc.relation/*ref*/Y. Li et al., “FMW-YOLO: A Frequency-Enhanced and Multi-Scale Context-Aware Framework for PCB Defect Detection,” Micromachines, vol. 17, no. 5, p. 531, Apr. 2026. https://doi.org/10.3390/mi17050531
dc.relation/*ref*/S. Singh, “Leveraging YOLO object detection for accurate and efficient visual recognition,” labellerr.com. Accessed: Jun. 21, 2026. [Online]. Available: https://www.labellerr.com/blog/why-is-the-yolo-algorithm-important/
dc.relation/*ref*/G. Jocher, “Ultralytics. Box, cls, and dfl loss gain #10375,” github.com. Accessed: Jun. 21, 2026. [Online]. Available: https://github.com/ultralytics/ultralytics/issues/10375
dc.relation/*ref*/P. García Mesa, “Understanding YOLOv5 Loss: A Comprehensive Analysis,” medium.com. Accessed: Jun. 21, 2026. [Online]. Available: https://pgmesa.medium.com/understanding-yolov5-loss-a-comprehensive-analysis-ffe35e6203e0
dc.relation/*ref*/G. Naidu, T. Zuva, and E. M. Sibanda, “A review of evaluation metrics in machine learning algorithms,” Artificial Intelligence Application in Networks and Systems, R. Silhavy and P. Silhavy, Eds., Cham, Switzerland: Springer, 2023, pp. 15-25. https://doi.org/10.1007/978-3-031-35314-7_2
dc.relation/*ref*/Ultralytics, “Performance Metrics Deep Dive,” ultralytics.com/. Accessed: Jun. 21, 2026. [Online]. Available: https://docs.ultralytics.com/guides/yolo-performance-metrics/
dc.relation/*ref*/
dc.rightsCopyright (c) 2026 TecnoLógicasen-US
dc.rightshttps://creativecommons.org/licenses/by-nc-sa/4.0en-US
dc.sourceTecnoLógicas; Vol. 29 No. 66 (2026); e3665en-US
dc.sourceTecnoLógicas; Vol. 29 Núm. 66 (2026); e3665es-ES
dc.source2256-5337
dc.source0123-7799
dc.subjectdetección de defectoses-ES
dc.subjectplacas de circuito impresoes-ES
dc.subjectrepositorio DeepPCBes-ES
dc.subjectvisión artificiales-ES
dc.subjectYOLOes-ES
dc.subjectdefect detectionen-US
dc.subjectprinted circuit boardsen-US
dc.subjectdeepPCB repositoryen-US
dc.subjectmachine visionen-US
dc.subjectYOLOen-US
dc.titleVisual Defect Detection in Printed Circuit Boards using YOLO Architectures: A Comparison on the DeepPCB Dataseten-US
dc.titleInspección visual de defectos en placas de circuito impreso utilizando arquitecturas YOLO: Una comparación sobre el conjunto de datos DeepPCBes-ES
dc.typeinfo:eu-repo/semantics/article
dc.typeinfo:eu-repo/semantics/publishedVersion
dc.typeResearch Papersen-US
dc.typeArtículos de investigaciónes-ES

Archivos

Bloque original

Mostrando 1 - 1 de 1
Miniatura
Nombre:
3665_v29n66.pdf
Tamaño:
1.22 MB
Formato:
Adobe Portable Document Format
Descargar

Colecciones

Revista TecnoLógicas - Cosecha Portal de Revistas