A Systematic Review of Textile Anomaly Detection Systems

Determining defects in textiles has become an increasing concern to maintain a standard of optimal quality. During mass production, there is a high possibility that certain fabrics are damaged and thus, identifying anomalies has become a crucial task. This paper contains a detailed overview of state-of-the-art techniques in textile defect detection, offering insights into the advancements and challenges faced by this domain. The review encompasses a thorough examination of machine learning and deep learning models like Convolution Neural Networks, Generative Adversarial Networks, autoencoders, and ensemble models, analyzing how Computer Vision has yielded promising results. Attention has also been placed while evaluating innovative algorithms like RNet, UNet, AMTFnet, and RDUNet. The purpose of this research is to guide practitioners in selecting the appropriate methodologies while recognizing anomalies, thereby fostering the development of robust systems in textile industries.