Deep Learning - Driven Fabric Fault Detection Using YOLOv9 in Textile Industrial Sittings
TL;DR
The results confirm the model’s scalability and practical applicability for deployment in high-speed manufacturing environments, contributing to intelligent textile inspection systems that enhance productivity, reduce economic losses, and support sustainable industrial practices.
Abstract
Automated fabric inspection is vital for maintaining quality in modern textile manufacturing, where manual inspection remains slow, inconsistent, and prone to human error. This study presents a deep learning–driven approach for fabric fault detection using YOLOv9, evaluated under real industrial conditions with datasets collected from Chenab Textiles. The dataset encompasses seven defect categories across plain, regularly printed, and randomly printed fabrics. The YOLOv9 framework achieved a mAP@0.5 of 86.3%, a precision of 0.832, and a recall of 0.847, demonstrating robust performance in detecting high-variance defect classes. Comparative experiments with MobileNetV3-SSD highlight YOLOv9’s superior accuracy and inference efficiency. The results confirm the model’s scalability and practical applicability for deployment in high-speed manufacturing environments, contributing to intelligent textile inspection systems that enhance productivity, reduce economic losses, and support sustainable industrial practices.