At present, most yarn factories still rely on manual inspection, which has a high missed detection rate and consumes a long time. There are many types of yarn defects, such as paper tube stains, deformation, dirty yarn, broken yarn, thrown yarn, fluff, and two-color yarn. Manual inspection is not conducive to actual quality requirements, and automatic optical inspection (AOI) is also difficult to detect when facing non-fixed defects, and the false detection rate is high, and manual re-inspection is still required. In order to allocate labor costs to more efficient work, yarn inspection should seek higher efficiency inspection solutions.Using SolVision's Segmentation technology, feature extraction is performed on various defects on paper tubes and yarns, and AI model training is performed to enable AI to learn to identify defect features and quickly and accurately find various defects. It can effectively improve the detection rate, finished product yield and reduce the quality inspection burden. With the increase in the number of learning samples, the ability of AI to identify defects can be continuously optimized, and the learning results can also be quickly introduced into various production lines.