PCB Bare Board Defect Detection Based on Improved YOLOv7-Tiny

A PCB bare board defect detection model based on an improved YOLOv7-tiny is proposed in order to simplify computation, increase detection accuracy, and minimize the number of model parameters. The coordinate attention mechanism is introduced with the purpose of helping the network locate defects more precisely, and the SimCSPSPPF module is introduced to further improve the network's feature extraction powers. The RepGhostN et network is used as a new feature extraction network to boost the extraction of the features of images, reduce the computational complexity, capture richer semantic information, as well as improve the accuracy of detection. According to the results from the experiment, the number of parameters and computing effort in the modified model are correspondingly down from the original model by 33.3 % and 43.2 %, respectively, but instead the mAP (mean Average Precision) rises by 2.95%. The enhanced YOLOv7-tiny makes it possible to be implemented in edge devices by increasing detection accuracy and decreasing processing and parameter counts.