REAL-TIME EGGPLANT DISEASE DETECTION USING YOLO ALGORITHM AND TELEGRAM NOTIFICATIONS

Abstract

The urgent need for enhancing agricultural production has been a critical focus due to agriculture's significance in providing food, supporting public health, and fostering the well-being of farmers. Among the crops cultivated year-round in Indonesia, eggplant stands out as a vital staple. However, persistent pest infestations, particularly by the Step nursery Pilchna sp., have hindered the full realization of eggplant's potential. Ideal eggplant growth conditions call for sandy loam soil with a pH range of 6.5-7, and a temperature range of 22-30˚C, making it suitable for cultivation during the dry season. Nevertheless, eggplant plants in Indonesia are under constant attack by polyphagous pests like the Beetle, which damages the lower epidermis of leaves, leaving them vulnerable to various pathogens from the Solanaceae family, causing diseases such as leaf spot, root neck rot, fruit rot, anthracnose, and bacterial wilt. To combat these issues effectively, an automatic classification system capable of recognizing leaves, fruit, and stems on eggplant plants based on recorded pest categories is essential.

This research focuses on the implementation of a high-performance automatic classification system for pest-disease detection in eggplant plants. The system, executed on devices with graphical processing units, achieves real-time video stream processing with less than 25 ms latency per second and a significantly improved Mean Average Precision (MAP). The system's accuracy value for identification and calculation stands at an impressive 92.85%, with an inference time of 11.88 seconds for detecting one plant and 25.29 seconds for sending notifications. Further studies have shown an average classification accuracy of 99.25%, with an average bounding box detection of 74.57% and an average classification and detection speed of 0.911 seconds per image.

With this automated pest-disease detection system, eggplant farmers can promptly identify and address infestations, leading to faster and more accurate intervention measures, thereby ensuring better yields and reduced crop losses. The proposed system heralds a significant step towards sustainable and efficient agricultural practices, contributing to food security and economic prosperity.