Efficient and accurate crack detection can provide a basis for assessing the structural safety of tunnels. Aiming at the shortcomings of traditional crack detection methods, which are complex and weak in generalization ability, an improved algorithm YOLOv5-CT for tunnel lining crack detection is proposed.Considering the slender morphology of the cracks, the network introduces the Transformer module to improve the crack detection effect.The strong long-range dependency capture ability of the Transformer module enables the proposed detection model to fully learn the contextual information of the crack region. In addition, the network integrates the convolutional attention mechanism CBAM in Neck.The experiment shows that the AP50 of YOLOv5-CT can reach 85.2%, which exceeds the YOLOv5 model by 8.9%. It is better than other one-stage object detection networks in terms of accuracy, and the inference speed reaches 161.3fps under pixel conditions, which meets real-time detection of tunnel lining cracks.