Textile and colour defect detection using deep learning methods

Abstract

Recent advances in deep learning (DL) have significantly enhanced the detection of textile and colour defects. This review focuses specifically on the application of DL-based methods for defect detection in textile and coloration processes, with an emphasis on object detection and related computer vision (CV) tasks. The first section systematically categorises existing DL approaches including convolutional neural networks (CNNs), generative adversarial networks (GANs), and transformer-based models—and examines their implementation in tasks such as surface defect localisation, colour inconsistency identification, and anomaly detection. Core algorithms are outlined alongside their underlying principles, practical challenges, and emerging solutions. The second section further compares DL approaches with traditional methods such as CV and expert systems (ESs) for diagnosing defects in coloration processes. This work offers a structured framework that integrates both model architecture taxonomy and methodological comparison, providing deeper technical insight than prior surveys. By highlighting the trade-offs between DL, CV, and ES methods, and identifying future research opportunities, this review serves as a reference for designing cost-effective, high-performance defect detection systems in textile manufacturing.