A Detection Method for Yarn Ends of Cop Yarn and Warp Yarn Based on YOLOv8

In view of the complications of high employment intensity, complex targets, and low automation in the manual search for yarn ends of cop yarn and warp yarn, this paper proposes a detection method for such yarn ends based on YOLOv8. By constructing a dataset consisting of 1021 images of complete cop yarn and warp yarn and 1096 images of yarn ends, and on the basis of using YOLOv8 as the yarn - end detection network model, a dynamic convolution module is embedded in the backbone network, and a progressive feature pyramid structure is introduced into the feature fusion module to achieve automated detection. The experimental results show that the meticulousness of the algorithm in this paper is 0.894, the recall is 0.825, the $F 1$ score is 0.883, the segmentation speed reaches 98 frames per second, and the volume of components and the model size are smaller.