Fabric Defect Detection Using Yolov11 And Yolov12: A Comparative Study

Abstract

The textile industry is a major contributor to global environmental degradation, generating significant waste due to defective products and inefficient quality control processes. Traditional manual fabric inspection—labor-intensive and error-prone—often results in high rates of material rejection, exacerbating resource depletion and pollution. Advances in computer vision offer a pathway to mitigate these impacts through automated defect detection. This study evaluates the efficacy of two state-of-the-art deep learning models, YOLOv11 and YOLOv12, for real-time fabric defect detection, leveraging the Fabric Detection dataset. By replacing manual inspection with AI-driven systems, we demonstrate potential reductions in textile waste and energy use. Our results show that YOLOv12 outperforms YOLOv11 in accuracy (↑12%) and speed (↑18%), enabling faster, more reliable quality control. These findings highlight the role of AI in promoting sustainable manufacturing, aligning with circular economy principles by minimizing resource waste and optimizing production efficiency.