Farm Sage Multilingual Agricultural Q&A Assistant And Plant Leaf Disease Detection

Abstract

This work proposes an AI-powered agricultural decision-support system to overcome the limitations of traditional advisory mechanisms that often fail either in their diagnostic acumen or contextual awareness. In this respect, the proposed approach integrates a hybrid convolutional neural network-Swin Transformer architecture for plant disease detection with a multilingual conversational module powered by Google Gemini. The hybrid visual framework merges convolutional feature extraction with shifted-window selfattention, thus accurately identifying subtle and globally distributed disease patterns within natural farm images. The conversational module interprets farmer queries across regional languages, assuring reliable intent understanding and domainspecific response generation. Real-time meteorological data from the OpenWeatherMap API have been integrated to align diagnoses with localized environmental risks, rendering more actionable recommendations. The empirical investigation shows that the system has outperformed traditional CNN-based architectures on accuracy, precision, recall, and robustness metrics. Collectively, the framework offers a scalable, contextaware, and multilingual solution for enhanced crop health management under precision agriculture.

Plant phenotyping relevance

植物の自然画像から病害パターンを検出・分類する画像解析手法が中核であり、植物病害状態のフェノタイピングに該当する。対話型農業支援や気象情報も含むが、病害検出モデルの技術評価が明示されている。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.