AI-Driven Predictive Plant Disease Diagnosis and Treatment Recommendation System using IoT

Abstract

In precision farming, plant diseases destroy 20-40% of crops worldwide every year, threatening global food security. Our project delivers a working prototype that combines IoT sensors with deep learning to catch diseases early and prevent losses. The system uses CNN models to analyze leaf images and LSTM to predict how diseases will spread over time. Real-time IoT data (soil moisture, temperature, humidity) feeds a hybrid CNN-LSTM model that spots trouble 7-14 days ahead and calculates exact pesticide/nutrient doses needed. Tested on PlantVillage dataset (54,000+ images, 14 crops), it hits 97.2% disease detection accuracy and 95.8% prediction accuracy – beating standalone CNN (94.1%) and LSTM (89.5%). F1-scores stay above 0.96. This shifts farming from "react when plants die" to "predict and prevent". Farmers cut losses by 30%, use fewer chemicals, and grow more sustainably. Smallholder farmers win big – affordable, scalable, climate-proof agriculture. Future plans: drone integration + federated learning.1][2].

Code and data availability

The paper describes a CNN-LSTM IoT plant disease system using the PlantVillage dataset and a custom Tamil Nadu image set, but provides no public deposit, URL, or availability statement for its code, trained models, custom dataset, or sensor data. PlantVillage is a generic third-party dataset, not a paper-specific asset

No evidence-backed public reproduction asset is currently recorded.