Fabric Defect Detection and Classification Using Machine Learning

Authors- Assistant Professor Dr. Pankaj Malik, Parth Akhand, Vinayak Pandiya, Krapansh Dubey, Shashwat Kanungo

Abstract-In the textile industry, identifying and classifying fabric defects is critical for maintaining product quality and minimizing waste. Traditional methods of fabric inspection, typically involving manual inspection, are labor-intensive, subjective, and prone to error. This study explores the application of machine learning techniques to automate the detection and classification of fabric defects, aiming to improve accuracy, efficiency, and scalability in textile quality control. We focus on common fabric defects, including holes, stains, misweaves, and pattern irregularities, and evaluate several machine learning models, including Convolutional Neural Networks (CNNs) and Transfer Learning architectures such as ResNet and VGG, for their performance in defect detection. A labeled dataset of fabric images with various defect types was used to train and evaluate the models, with data augmentation techniques applied to enhance robustness. Key performance metrics, including accuracy, precision, recall, and F1-score, were used to assess model efficacy. Results indicate that CNN-based models, particularly those leveraging Transfer Learning, achieve high accuracy in detecting and classifying fabric defects, significantly outperforming traditional approaches. The findings underscore the potential of machine learning to transform textile quality management by enabling real-time, automated defect detection on production lines. This research demonstrates that machine learning-driven systems can improve the speed and consistency of fabric inspection processes, providing valuable insights for developing automated quality control solutions in the textile industry. Future work may explore real-time implementation and advanced architectures for further optimization.

DOI: /10.61463/ijset.vol.12.issue5.301