Defect Detection in the Textiles Industry using a Multi-Task Transfer Learning Approach

There are a number of drawbacks of using the manual process for the detection of fabric defects while making the defect-free fabric itself. The list of issues includes the additional manpower required, the time-consuming process to detect the defects, the lack of perfection, focus degradation in the sense of tiredness, and many more that would distract from achieving efficiency. There are many techniques such as CNN, SVM, Random forest, GMM, deep networks, and transfer learning to be used in this domain. Among these, a novel category technique, such as multi-task transfer learning, can be applied, which not only discovers the defects but also provides the type of fabric and pattern-related features. When the defect detection system is framed, the results are compared against the recent approaches, such as CNN, SVM, and MTTL. The efficiency and accuracy of MTTL are found to be better than existing approaches, is demonstrated in the visual representations. The outcome of the textile industry is to make profits and make customers satisfied with the best fabric quality without defects. The losses to be experienced by the textile industry would be overcome by this designed model and it adds income to the organization.