An Improved YOLOv8-Based Algorithm for Concrete Crack Detection

To address the issues of inadequate feature extraction, low detection accuracy, and insufficient real-time performance in existing object detection algorithms for concrete crack detection, this paper proposes an improved YOLOv8-based concrete crack detection algorithm, SDC-YOLOv8. This algorithm builds on the YOLOv8 framework, introducing the SwinTransformer module into the C2F module to enhance the ability to learn different crack features in complex backgrounds. The DyCsample (DySample-Coordinate Attention) upsampling module is introduced to improve upsampling speed and efficiency, highlighting crack feature information. Finally, to enhance the model's performance in detecting concrete cracks, the Complete Intersection over Union (CIoU) loss function is replaced with the Wise Intersection over Union Version 3 (WIoU v3) loss function. To validate the effectiveness of the proposed model, a comparative analysis is conducted with five algorithms: YOLOv3, YOLOv5, YOLOv7, YOLOv8, and YOLOv9. It is concluded that the SDC-YOLOv8 model reduces overall parameter count, lowers model complexity, and effectively improves concrete crack detection performance.