Defect Detection and Identification in PCBs Using Single Stage Object Detection Models

In industrial electronics, detection of defects in PCBs is an essential part in ensuring quality and reliability of printed circuit boards. Conventional methods of identifying such defects are sometimes ineffective, as they may not detect the defects efficiently owing to human error. Machine learning (ML) methods can be a better alternative to these traditional methods. Convolutional neural networks (CNNs) are well known for detecting abnormal patterns and irregularities in datasets efficiently. By training ML models using large datasets of PCB image samples, the models can effectively recognize the presence of such defects. The defects may include mouse bites, short circuits, spur defects and so on that may sometimes go unnoticed by the human eye. The implementation of ML models for this purpose not only improvises on accuracy, but also reduces the time taken. This paper focuses on a solution to detecting PCB defects using ML models.