An Improved Printing Defect Detection Method Based on YOLOv5s

Abstract

In inkjet printing, various printing defects such as missing prints and stains can occur due to uncertain factors such as industrial production environment and equipment. To ensure print quality and improve detection efficiency, this paper proposes an improved YOLOv5 method for detecting printing defects, which inserts a Coordinate Attention mechanism into the main feature extraction network of YOLOv5s to achieve detection of five types of printing defects. The experimental results show that the proposed method achieves an mAP of 91.7%, which is 0.9% higher than the baseline YOLOv5s, and can well meet the requirements of printing defect detection in industrial production.