Digital Platform for Crop Health and Agricultural Services

Abstract

Modern precision agriculture requires the incorporation of high-accuracy diagnostic instruments to guarantee food security for inexperienced practitioners. This paper introduces an AI-driven agricultural web architecture that connects deep learning-based diagnostics with real-world farm management. The main contribution is a Convolutional Neural Network (CNN) framework that can automatically find diseases in five common crops: Capsicum annuum, Vitis vinifera, Zea mays, Solanum tuberosum, and Solanum lycopersicum. The proposed model reached a final training accuracy of 98.30% and a validation accuracy of 90.12% over 10 epochs by using a sequential architecture with optimized convolutional layers and data augmentation. The platform has a localized marketplace, a government scheme eligibility engine, and a Crop Journal for long-term record-keeping to make it useful in the real world. Results demonstrate that this unified ecosystem provides a transparent and accessible framework for data-informed agricultural management, effectively lowering the technical barrier for new farmers.

Plant phenotyping relevance

CNNによる作物病害の自動検出が中心的な技術貢献であり、植物の病害状態を画像ベースで推定するため、農業サービス部分を含んでも植物フェノタイピング手法として採用する。

Code and data availability

The paper describes a custom CNN trained on the PlantVillage dataset, but contains no public code, model checkpoint, or dataset deposit for this paper's own analysis. PlantVillage is a cited external benchmark, not a paper-specific asset, and no availability statement or authors' URL for code/models appears in any sup

No evidence-backed public reproduction asset is currently recorded.