Lightweight Deep Learning-Based Plant Leaf Disease Assisted Diagnosis System

Abstract

Plant diseases pose a significant threat to agricultural productivity and global food security, making accurate and timely diagnosis essential. Although deep learning has shown promising performance in plant disease recognition, most existing methods focus on single-task classification and lack interpretability and deployability. To address these limitations, this paper proposes a lightweight deep learning-based plant leaf disease assisted diagnosis system. The proposed framework integrates EfficientNet-B3 for disease classification and Small U-Net for leaf and lesion segmentation, enabling precise localization and quantitative analysis. Disease severity is further estimated based on the lesion-to-leaf area ratio, providing interpretable results for end users. Experimental results demonstrate that the proposed method achieves high accuracy and robust performance while maintaining low computational cost. The system also supports mobile-friendly deployment and generates practical recommendations, forming a complete closed-loop diagnostic pipeline suitable for real-world agricultural applications.

Plant phenotyping relevance

葉の病変セグメンテーションと病変面積比による病害重症度推定を開発しており、植物状態の画像ベース表現型取得が中心です。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.