YOLOv4-MN3 for PCB Surface Defect Detection

Surface defect detection for printed circuit board (PCB) is indispensable managing PCB production quality. However, automatic of surface defects still a challenging task because, even within the same category defect, present great differences in morphology and pattern. Although many computer vision-based detectors have been established to handle these problems, current struggle achieve high accuracy, fast speed low memory consumption simultaneously. To address those issues, we propose cost-effective deep learning (DL)-based detector based on cutting-edge YOLOv4 detect quickly efficiently. The improved upon with respect its backbone network activation function neck/prediction network. evaluated customized dataset, collected from factory. experimental results show that achieved performance, scoring 98.64% mean average precision (mAP) at 56.98 frames per second (FPS), outperforming other compared SOTA detectors. Furthermore, reduced parameter space 63.96 M 39.59 number multiply-accumulate operations (Madds) 59.75 G 26.15 G.