Enhancing Fabric Defect Detection With Attention Mechanisms and Optimized YOLOv8 Framework

Fabric defect detection is a critical task in the textile industry, requiring high precision and recall to ensure effective quality control. This study presents an enhanced YOLOv8-based framework that integrates novel attention mechanisms and advanced architectural modules to improve detection accuracy and robustness. The framework incorporates the SimAM attention mechanism within the SPPF module and adopts an optimized Dilation-wise Residual (DWR) structure in the backbone. Comprehensive ablation studies and comparisons with state-of-the-art methods validate the effectiveness of the proposed approach. The enhanced model achieves a mAP50-95 of 74.3%, outperforming the baseline by 4.7 percentage points, with marked improvements in detecting challenging defect categories. While the framework demonstrates significant advancements, limitations in dataset diversity and computational efficiency are acknowledged. Future work will focus on resource optimization, dataset augmentation, and extending the framework’s applicability to other domains.