YOLOv5-GXT: A New PCB Surface Defect Detection Algorithm Based on YOLOv5

Printed Circuit Board (PCB) surface defect detection is the most important step in the PCB manufacturing process. How to accurately and efficiently implement PCB quality inspection is still a challenging task in the field of automatic detection. This paper proposes an improved PCB surface defect detection algorithm YOLOvS-GXT based on YOLOvS. The main structure of the network includes GHostv2 Block in the backbone network to achieve feature map reduction and long-distance information capture, reduce redundant computing, and more efficient feature extraction; C3XTR was used to replace C3 before the output of the backbone network, to explore the global information induction with self-attention, to speed up the training of the model, and Mish has used to improve the performance of the model. A large number of experiments on the PKU-Market-PCB dataset show that the YOLOvS-GXT model achieves 98.4% map (Mean Average Precision) and 97% F1-Score, and has faster training speed and better detection performance than the YOLOvS model.