Underwater concrete crack detection of dams via CycleGAN-based data enhancement and optimized multi-scale YOLO11

Key Points

• Crack detection performance improves significantly with data enhancement strategies.
• The CycleGAN model enhances datasets by generating realistic images, augmenting training resources effectively.
• Optimized multi-scale YOLO11 architecture achieves higher accuracy in identifying structural issues than traditional methods.
• Results emphasize the potential for enhanced monitoring of underwater structures, which may prevent failures.