<p>In the process of crop growth, real-time detection and treatment of leaf diseases is extremely important. To improve detection accuracy and reduce the waste of resources, a you only look once version 8 nano (YOLOv8n) crop leaf disease detection model is proposed. First, we use the wise intersection over union (WIoU) loss function to intelligently focus on the model anchor boxes, accelerating model convergence. Next, to accommodate the size of the target being inspected, the detection heads are replaced to improve model accuracy. Then, we add the multi-head self-attention (MHSA) mechanism to the backbone network, enriching the feature extraction capability of the model for images. At last, we introduce deformable convolutional networks version 2 (DCNv2) in cross stage partial feature fusion (C2f) to improve the model’s ability to accurately localize input image targets. The experimental results show that the YOLOv8n_leaf disease model accuracy is improved by 1.3%, recall rate is improved by 0.4%, mean average precision at IoU threshold 0.5 (<i>mAP</i>0.5) reaches 83.7%, and mean average precision at IoU threshold 0.95 (<i>mAP</i>0.95) reaches 69.7%. The improved YOLOv8 model can provide technical support for crop leaf disease detection.</p>