Fabric Defects Detection Using YOLOv8

Fabric defect detection poses a significant challenge in the textile industry due to the intricate shapes and diverse range of fabric defects. Numerous approaches have been suggested to address this issue; however, their efficacy in terms of both detection speed and accuracy has been notably limited. This paper introduces an approach for detecting fabric defects using the YOLOv8 model. The methodology involves using the Tinachi dataset, labeling the dataset with LabelImg, applying data augmentation to the dataset, and iteratively training the YOLOv8 network model. The model achieves a mean Average Precision (mAP) of 78.5% , \(box\_loss\) of 1.6209, a \(cls\_loss\) of 1.0495, and a \(dfl\_loss\) of 1.2898.