C2f-Enhanced YOLOv5 for Lightweight Concrete Surface Crack Detection

In recent years, there have been significant breakthroughs in the field of detection using deep learning technology. What used to be challenging defects for traditional visual methods can now be addressed with the help of deep learning techniques. This paper employs the YOLOv5 network architecture to achieve rapid and precise detection of concrete surface cracks. Additionally, it integrates the C2f module to overcome issues like gradient vanishing and information bottlenecks, which can limit the performance of the traditional YOLOv5 network. The C2f module enhances feature propagation and utilization by introducing strategies such as feature reuse, cross-stage partial connections, and attention mechanisms, thereby improving feature representation and information flow. Various training techniques are also applied to enhance training speed and detection accuracy, including weighted residual connections (WRC), cross-stage partial connections (SCP), cross mini-batch normalization (CmBN), self-adaptive training (SAT), and mish activation function. As a result, the system achieves a detection accuracy of 96.91% on a concrete crack detection dataset.