Application of Neural Networks to Testing Printed Circuit Boards Using Data from a X-ray 3D Microtomograph

A method for defect recognition in printed circuit boards using neural networks is discussed. An analysis of various neural network architectures is performed to identify the most effective one. An approach to data filtering simulating the operation of a microtomograph using convolutional autoencoders is also presented. The quality of the proposed approaches was evaluated using the mean Average Precision (mAP) metric for YOLOv8 and Faster R-CNN models.