Research on Tea Disease Detection System Based on Improved YOLOv8

Aiming at the problems of low efficiency and poor accuracy of traditional tea pest and disease detection, this study constructs an improved detection system based on YOLOv8.By integrating FasterBlock, EMA multi-scale attention mechanism and BiFPN module, we combine migration learning and multi-scale training to improve the detection accuracy and robustness. The performance of the improved model is excellent, with mAP50 reaching 75.0% and mAP50-95 51.6%, with a small number of parameters and low computational complexity. Finally, the detection results are fed into DeepSeek to generate prevention and control recommendations. The improved model has excellent performance, and the detection results are connected to the DeepSeek large model knowledge base to automatically analyse and generate control recommendations, which provides powerful support for the management of tea plantations and promotes the intelligent development of the tea industry.