Research on Real-Time Detection of Fabric Defects Using an Improved YOLOv8n Method

In response to the slow speed and low accuracy of textile defect detection during industrial production processes, this paper, after studying a vast array of textile defect types, proposes a textile defect detection algorithm based on an improved YOLOv8. Initially, methods such as rotation and shifting were employed for data augmentation to enhance the dataset's universality and the model's generalizability. Subsequently, a lightweight network, Starnet, which features fewer model parameters and lower computational complexity, was introduced, making it more suitable for devices with limited computing capabilities. This paper verifies through experiments that the model's accuracy in detecting fabric defects on the dataset reached 0.98% over the original YOLOv8 model. The parameter count decreased by 9.0%, and GFLOPs decreased from 8.2 to 7.2, an increase of 12.19%, thereby proving the effectiveness of the model.