An Integrated Approach for Mulberry Leaf Disease Detection, Classification and Management Based on YOLOv8 and Large Language Model

Abstract

Mulberry (Morus spp.) serves as the primary food source for silkworm (Bombyx mori) larvae in sericulture, making plant health critical for silk production. Four major foliar diseases, bacterial leaf spot (Xanthomonas campestris pv. mori), Cercospora leaf spot (Cercospora moricola), Myrothecium leaf spot (Paramyrothecium roridum) and powdery mildew (Phyllactinia guttata = P. corylea), significantly impact yield. In many sericulture‐growing areas, disease diagnosis still depends on specialist support that is not always available at the farm level. We developed an integrated system combining YOLOv8 object detection with GPT‐3.5‐powered treatment advisory. We collected 1417 field images in Murshidabad (West Bengal, India), covering four disease classes and healthy leaves; experts annotated the images. We compared five YOLOv8 variants under the same training setup, whereby YOLOv8m provided the best overall balance (recall = 1.0, F1 = 0.994, precision = 0.988). We integrated YOLOv8 outputs into GPT‐3.5‐turbo through LangChain and used a fixed, structured prompt to produce disease‐specific recommendations without model fine‐tuning. The web‐based system enables agricultural extension workers and plant health clinics to receive immediate disease identification with evidence‐based treatment recommendations, providing scalable diagnostic capacity for sericulture advisory services.

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.