Effective Approaches of Deep Learning Models for Automated Fabric Defect Classification

Accurate fabric defect classification is a critical component process of textile quality control. A robust blemish sorting system is instrumental in ensuring that the final product adheres to consumer strict standards. Recent advancements in technology have facilitated the development of automated defect classification systems leveraging Deep Learning (DL) methods. This research aims to construct an automated fabric defect arrangement model utilizing DL architectures, specifically ResNet152 and MobileNetV3, within the context of textile printing. Comparative analysis of these two models indicates comparable performance, with ResNet152 slightly surpassing MobileNetV3 in terms of testing accuracy (99.72%) and computational time (0.21 seconds).