CONNECTING YOLO OBJECT DETECTION MODELS WITH FABRIC PATTERN IDENTIFICATION: A COMPREHENSIVE LITERATURE REVIEW

Abstract

Automated fabric pattern and defect detection has become essential for quality assurance due to the scale and diversity of fabric production. While the YOLO family of detectors dominates real-time vision tasks, evidence comparing the latest generations for the specific fabric challenges of fine-grained patterns, subtle defects, repetitive textures, and domain shifts remains fragmented. This review synthesizes knowledge on the use of YOLOv8, YOLOv10, and YOLOv11 for detecting patterns and defects in fabrics. The goal is to describe the advantages of each model, its ease of use, and how best to employ it. We followed a standard protocol (PRISMA) and searched for studies from the last ten years in major research databases such as Scopus, IEEE Xplore, Web of Science, and ScienceDirect. We used custom rules to determine which studies to include, then assessed their quality and summarized the findings. We analyzed reported metrics (mAP, precision/recall, F1, latency), datasets and annotation practices, augmentation, and training strategies. Among the included studies, YOLOv8 remains the most frequently adopted baseline in fabric, with strong accuracy, broad community support, and versatile variants for edge and server deployments. Emerging evidence suggests that YOLOv10 and YOLOv11 provide incremental accuracy improvements and better efficiency, especially when leveraging lightweight heads, improved label assignment, and optimized training schedules. Nano variants are generally preferred for edge devices, while medium/large variants are suitable for server-side inspection pipelines. Techniques such as multi-scale augmentation and mosaicking, copy-paste, and domain-focused data curation consistently improve recall for small and low-contrast defects. Several issues remain, such as uneven class distribution, limited public fabric datasets, unreliable latency reports, and broad application across different fabric types and lighting conditions. Ultimately, the choice of YOLOv8, YOLOv10, or YOLOv11 should be based on application limitations and desired precision, along with clear, standardized assessments and reports to support relative claims and lead to acceptance in the field.