Improved YOLOv11 Model for Crack Detection Using AFGCAttention and VoV-GSCSP Modules

Traditional crack detection methods often struggle with low efficiency and accuracy due to manual feature extraction and computational resource limitations. To overcome these challenges, we propose an improved YOLOv11-based crack detection method that incorporates the AFGCAttention and VoVGSCSP modules, enhancing precision, robustness, and real-time performance. Experimental results demonstrate significant improvements, particularly in detecting small cracks, with notable gains in F1 score, precision (P), recall (R), and mAP metrics, especially at low confidence levels. These enhancements confirm the effectiveness of the proposed model in real-world crack detection tasks.