Defect detection in the textile industry using the YOLOv8 model

The textile industry plays a fundamental role in the global economy, being responsible for the production of a wide variety of products. Defect detection in this context is a critical aspect as it presents a series of defects that include objects of reduced dimensions and inspection is often carried out manually. This article improves quality control processes in the textile industry by applying the YOLOv8 model to the public AITEX dataset for automated defect detection, representing a technological advance when compared to the YOLOv5 model, both developed by ultralytics. Furthermore, issues such as the architecture and variants of the YOLOv8 model, limitations, parameter optimization and training strategies, as well as model evaluation are discussed. Promising results are presented, reaching an mAP of 85.11%, indicating that YOLOv8 can effectively be used in the context of textile defect detection.