Detection of garment manufacturing defects using CFPNet and deep beliefnetwork: an image-based approach

The demand for high-quality items and the quickly shifting economic landscape increase the importance of ready-made garment manufacturers in providing the correct quality product. It is difficult work in the textile industry since the efficacy and efficiency of automatic flaw identification determine the quality and cost of every textile surface. In the past, the textile industry used manual human efforts to find flaws in the manufacturing of clothing. The main downsides of the manual garment fault identification technique include lack of concentration, human tiredness, and time requirements. Applications based on digital image processing and computer vision can overcome the aforementioned restrictions and shortcomings. In this article, we use intelligent algorithms like Channel-wise Feature Pyramid Network (CFPNet) based on deep learning-based techniques with Deep Belief Network (DBN) to monitor the quality and predict any occurrences of manufacturing problems in clothing. The suggested algorithm is mostly utilised in the textile industry to find flaws in clothing while estimating client needs based on the environment and the economy to react quickly and meet business objectives. The performance evaluation was used to determine the 12 kinds of garment faults, which included holes, excessive margins, stains, cracks, inappropriate stitch balancing, needle breaks, ink stains, torn clothing, drop stitches, soil content, and broken clothing. The suggested model obtains a 95.85% stain defect detection rate, a 97.33% defect-free garment recognition rate, and a 97.16% hole defect recognition rate.