Socks have a variety of defect forms, including snags, wrinkles, and tears. The shape, size and position of these defects are not fixed. Traditional AOI is suitable for the inspection of whole pieces of cloth, but it has difficulty in detecting defects that are not fixed, and it is easy to make false detections. Therefore, manual re-inspection is still required.By collecting images of sock defects such as snags and wrinkles, and using SolVision's Segmentation technology to complete the training of the AI model, it is possible to quickly and accurately find defects, classify different defects and remove defective products. This can help to control product quality and improve production efficiency. By classifying and analyzing defects, it is also possible to optimize the overall manufacturing process.