Eggplant Leaf Diseases Detection and Counting System on Agricultural Robot Based on YOLOv8

Abstract

Eggplants play an important role in Indonesia's economic and food sectors. Production increases yearly, but diseases such as earworm, flea beetle, leaf spot, and leafhopper cause significant losses. Additionally, manual disease detection methods are time-consuming and prone to errors. This research develops an automated system using YOLOv8 and the Telegram Bot system to detect and count eggplant leaf diseases. The YOLOv8 algorithm can detect objects in videos or images quickly and accurately in real-time. The model uses the OpenVINO format for faster inference compared to PyTorch. The research results show that the YOLOv8 model achieved a mAP50 of 0.606. This system effectively detects, tracks, and counts eggplant leaf diseases and can send PDF files via email through the Telegram Bot. Accuracy evaluation shows an accuracy rate of 98% from the confusion matrix, with video testing showing the highest accuracy variation at 97.05% and the lowest at 23.07%. These results indicate that the automated eggplant leaf disease detection system has great potential to help agriculture prevent eggplant diseases more efficiently and accurately, supporting increased production and quality of harvests