Sericulture industry has a profound historical and cultural heritage, and is also a low-carbon, green and sustainable livelihood industry. Mulberry is an important cash crop, and its leaves are the main raw materials for silkworm rearing. However, in the planting process of mulberry, the occurrence of pests and diseases has seriously affected the yield and quality of mulberry leaves, and the Sangnong lacks sufficient professional knowledge and skills, which is often difficult to find and correctly identify the corresponding diseases in time, so that appropriate remedial measures can not be taken in time. Therefore, based on the PaddleDetection framework and the PP-YOLOE model, this paper improved the model's performance on the task of mulberry leaf disease recognition by improving the model structure and training methods, with a view to providing scientific management guidance for mulberry leaf disease. In the experiment, we compared the model with the YOLOv3 model. The experimental results showed that the model designed in this paper achieved good results in the task of mulberry leaf disease recognition, with mAP reaching 96.7%, which was significantly better than the YOLOv3 model. The research in this paper provides a method for the effective identification and management of mulberry pests and diseases, which is helpful to improve the yield and quality of mulberry leaves and ensure the sustainable development of silk industry.