Faculty of Electronic Engineering, Niš, Serbia

Faculty of Electronic Engineering, Niš, Serbia

The paper is available at the following link: PDF

Abstract:

Printed circuit board (PCB) form the foundation of almost every electronic system, and its defects can lead to many problems in functionality. Manual inspection of those defects is time consuming and prone to human error, since not all of them can be spotted immediately. This paper represents an efficient solution for automated PCB defects detection using a pre-trained Faster R- CNN deep-learning model. The model was trained on a Kaggle dataset, with YOLO annotated images for six most common defects. Training was finished after 12 epochs, and model had great achievements of around 90% of accuracy.

Keywords:

Printed circuit board, defect detection, Deep Learning, Faster R-CNN.

REFERENCES:

[1] Neha Kardam, "Introduction to Printed Circuit Board Desing and Layout using Ki-CAD", Lake Washington Institute of Technology, 2024.
[2] Kenji Suzuki, Artificial Neural Networks - Methodological Advances and Biomedical Applications, InTech, 2011.
[3] Available at: https://www.ibm.com/think/topics/deep- learning, accessed on: 22/09/2025.
[4] Available at: https://www.geeksforgeeks.org/deep- learning/what-is-forward-propagation-in-neural-networks/, accessed on: 01/10/2025.
[5] Available at: https://www.geeksforgeeks.org/machine- learning/activation-functions-neural-networks/, accessed on: 09/10/2025.
[6] Sebastian Raschka & Vahid Mirjalili," Python Machine Learning Third Edition", Packt Publishing, 2019.
[7] Sebastian Raschka & Vahid Mirjalili," Python Machine Learning Third Edition", Packt Publishing, 2019.
[8] Predrag M. Petković, Miljana Milić, Dejan Mirković, " VHDL i VHDL-AMS podrška projektovanju elektronskih kola i sistema", Univerzitet u Nišu Elektronski fakultet, 2009.
[9] Vančo B. Litovski, Osnovi testiranja elektronskih kola, Univerzitet u Nišu Elektronski fakultet, 2009.