Edge Implementation of Lightweight YOLO Models for Real-Time Mango Leaf Disease Detection on RISC-V Powered Device

Abstract

An important part of Bangladesh’s income is agriculture, and the mango is both a cultural and an economic symbol. Rajshahi and Chapai Nawabganj, which are known as the “mango hubs of Bangladesh,” help make the country the seventh-largest mango producer in the world, and they ship a lot of mangoes. Mango leaf diseases like Anthracnose, Die Back, Gall Midge, and Sooty Mould have a big effect on Bangladesh’s farming output, which is hard for farmers who use old, time-consuming ways to find diseases. This study shows a new edge-based system that uses light YOLO models (YOLOv5, YOLOv8, and YOLOv11) to find diseases on mango leaves in real time on RISC-V-powered devices. The suggested method aims to offer an effective, easy-to-reach, and low-cost answer for places with limited resources. The method involves teaching YOLO models on a balanced collection of 1,920 pictures that show four types of disease. The models are then quantized to INT8 so they can be used on devices with less power. Using performance measures such as mean Average Precision (mAP), precision, recall, and inference speed, YOLOv8 (small) is found to be the most reliable model, combining high accuracy with low computational needs. The study emphasizes the positive effects on people and the environment of the suggested system, such as lowering the need for pesticides, improving food security, and encouraging environmentally friendly farming methods. Ethical issues like justice, ease, and openness are taken into account to make sure that all farmers can adopt the technology equally. The results show that it is possible to combine advanced AI models with traditional farming, bringing together modern technology and traditional farming methods. In the future, researchers will focus on adding more types of datasets, making models more general, and using green energy sources to make the world more sustainable. This new way of doing things shows how edge computing could change the way diseases are managed in farmland, making farming groups stronger and more productive.