YOLO-BGS Optimizes Textile Production Processes: Enhancing YOLOv8n with Bi-Directional Feature Pyramid Network and Global and Shuffle Attention Mechanisms for Efficient Fabric Defect Detection

Timely detection of fabric defects is crucial for improving fabric quality and reducing production losses for companies. Traditional methods for detecting fabric defects face several challenges, including low detection efficiency, poor accuracy, and limited types of detectable defects. To address these issues, this paper chose the YOLOv8n model for continuous iteration enhancement in order to improve its detection performance. First, multiscale feature fusion was realized by the Bi-directional Feature Pyramid Network (BiFPN). Second, the Shuffle Attention Mechanism (SA) is introduced to optimize feature classification. Finally, the Global Attention Mechanism (GAM) was used to improve global detection accuracy. Empirical findings demonstrated the improved model's efficacy, attaining a test set mean average precision (mAP) value of 96.6%, which is an improvement of 3.6% compared to the original YOLOv8n. This validates that YOLO-BGS excels in detecting textile defects. It effectively locates these defects, minimizes resource waste, and fosters sustainable production practices.