Deep Learning for Quality Control in Woven Fabrics

Abstract

Automated defect detection in fabrics is a key challenge in textile industry
quality control. This study proposes a deep learning-based methodology to
identify defects in Batavia and Sarga fabrics. In the first stage, an autoencoder was used to filter anomalous images, thereby reducing the manual inspectionworkload. Subsequently, convolutional neural networks (DenseNet121, EfficientNetB0/B3, Xception, and VGG) were trained using data augmentation techniques and stratified cross-validation. To address class imbalance, class weighting was applied, and performance was evaluated using robust metrics such as AU-ROC, AU-PR, and G-Mean. For Batavia fabrics, DenseNet121 achieved an AU-ROC of 0.88 and an AU-PR of 0.93, demonstrating strong detection capability. For the Sarga fabrics, three different references (42402, 45433, and 43105) were considered, showing more variable performance across models and datasets. Nonetheless,
models such as ResNet101 and Xception have achieved competitive results. The results showed that the proposed system enables effective and adaptable automated inspection across different fabric types without compromising accuracy, making it a valuable tool for quality control in the textile industry.