The detection of cherry leaf diseases holds a crucial significance for maintaining the health status of cherry trees and improving their quality. This paper proposes an advanced model, YOLO-Cherry Leaf Disease (YOLO-CLD), for the efficient and accurate detection of three common cherry leaf diseases: cherry leaf brown spot, bacterial leaf shot hole, and Cherry lethal yellow. First, we built the model on the fast and efficient YOLOv7 network, and we enhanced it with an additional convolutional block attention module and a context Transformer module. The model has resulted in improved feature extraction and representation capabilities. The model is compressed using knowledge distillation to meet deployment requirements for different hardware conditions. The evaluation results on an embedded device show an average recognition accuracy of 85% with a frame rate of 18.5 frames per second (FPS). The proposed model demonstrates the potential of detecting cherry leaf diseases on the fly. The solution can be deployed on the cloud for optimal use.