With the continuous development of agricultural intelligent technology, rapid and accurate identification of crop diseases is crucial for agricultural production. Addressing the complexities and variations of diseases, as well as the reliance on manual identification methods, which cannot effectively adapt to diverse disease conditions, we proposed a deep learning disease detection method based on an improved YOLOv5s model. Firstly, this study focused on apple leaf disease images, collecting and annotating images of five common diseases: apple gray spot disease, spot defoliation disease, brown spot disease, mosaic disease, and rust disease. Subsequently, ASFF, CA, and CBAM modules were individually integrated into the original YOLOv5s model network to enhance the model’s disease recognition capability and detection accuracy, followed by comparative analysis through experiments. The research indicates that the improved YOLOv5s network model outperforms the original model with a recognition accuracy increase of more than 20%, effectively enhancing detection precision. Particularly, the network model with the added CA module exhibits the best performance. This study provides a feasible solution for the application of agricultural intelligent technology in disease identification, bearing positive significance for the intelligent development of agricultural production.