SCA-MobiPlant: smartphone-deployed multistage attention fusion model for accurate field detection of chili leaf curl complex.

Abstract

Background Field-scale assessment of chili leaf curl complex presents a significant diagnostic challenge, as both chili leaf curl virus (ChiLCV) and mite infestations produce visually overlapping symptoms difficult to distinguish by untrained personnel. This diagnostic confusion frequently leads to inappropriate application of either insecticides or acaricides, resulting in economic losses and environmental concerns. To address this issue, we propose SCA-MobiPlant, an improved MobileNetV3-Small model integrated with a novel multistage Squeeze-and-Excitation Coordinate Attention (SCA) fusion mechanism, designed for accurate differentiation of these apparently similar symptoms and precise field assessment of the disease. Results The proposed model effectively focuses on subtle diagnostic features including leaf texture, petiole elongation, and irregular curling patterns to achieve reliable classification. The multistage SCA fusion module demonstrated superior performance, achieving 99.64% accuracy, 99.61% precision, 99.64% recall, and 99.62% F1-score through K = 5 cross-validation, outperforming other attention modules such as the Convolutional Block Attention Module (CBAM) and Coordinate Attention (CA). Gradient-Weighted Class Activation Mapping (Grad-CAM) provided visual interpretability of the model's decision-making process. Comparative evaluation against state-of-the-art architectures, including EfficientNetB0, ResNet50, VGG19 and YOLO advanced series, confirmed the computational efficiency of the proposed model for mobile deployment. Conclusions The final system, termed SCA-MobiPlant, has been successfully implemented on smartphones, along with a Disease Incidence (DI) calculation module, enabling rapid and accurate field assessment of the disease. This facilitates appropriate intervention strategies while minimizing unnecessary pesticide use. The study highlights the potential of lightweight, attention-enhanced models for real-world plant disease diagnostics, particularly in resource-constrained agricultural settings.

Plant phenotyping relevance

植物病徴画像から葉の症状を分類し、病害発生率を算出するスマートフォン実装手法の開発・比較評価が研究の中心であり、植物の病害状態を直接推定しているため。

Code and data availability

植物フェノタイピング解析を再現する公開資産であることを、入力本文と直接リンクから確認できなかったため保留しました。

No evidence-backed public reproduction asset is currently recorded.