YOLOv10-Powered Detection of Tea Leaf Diseases: Enhancing Crop Quality through AI
Abstract
Rapid and sensitive diagnosis of tea leaf diseases is crucial to ensure the yield and quality of tea leaves. Classical manual diagnosis is time-consuming, subjective, and a source of human errors, introducing incoherencies and time awaiting in the management of the disease. This research presents a machine learning-based method for detecting and diagnosing tea leaf diseases via multiple advanced models. We trained the models YOLOv8n, YOLOv8x, YOLOv9c, YOLOv10s, and YOLOv10n on 5180 annotated tea leaf images with five typical disease types and healthy leaves. From these models, the best-performing was YOLOv10 with a precision of 96.0%, a recall of 97.4%, and a mean Average Precision (mAP@50) of 98.9%. YOLOv10n achieved the best performance across all models, with up to 0.9% higher mAP50 and F1 score, while demonstrating competitive success in recall and precision. In contrast, YOLOv9 and YOLOv10s obtain slightly lower mAPs of 98.3%, and 97.9%, respectively. These results have shown that YOLOv10n has an excellent accuracy and processing speed performance, indicating that it can be a reliable, automatic method for detecting tea leaf disease. This method significantly cuts down the labor base for expert diagnosis, avoids potential financial losses, and can extend the current approach to the large tea plantations, providing efficient and reliable management of the disease outbreak.