AI-Smart Agro Advisor: A Hybrid Deep Learning Based Smart Crop Disease Prediction and Recommendation System

Abstract

Crop diseases pose a major threat to agricultural productivity, particularly in rural regions where timely expert guidance and reliable internet connectivity are limited. This project presents AI-Smart Agro Advisor, a hybrid artificial intelligence–based mobile application designed for real-time crop disease detection and intelligent crop recommendation. The system employs dual-mode architecture to ensure continuous operation under both offline and online conditions. In offline mode, a MobileNetV2-based Convolutional Neural Network optimized using TensorFlow Lite performs on-device inference to identify commonly occurring crop diseases from leaf images captured using a smartphone camera. In online mode, the application integrates a cloud-based deep learning model (ResNet50) accessed through a RESTful API to enable large-scale detection of crop diseases and pests with higher accuracy. Additionally, crop suitability predictions are generated using machine learning models trained on soil parameters and seasonal data. To enhance accessibility, the system incorporates offline Tamil voice-assisted interaction implemented using on-device Text-to-Speech and Speech Recognition modules. The proposed system aims to reduce crop losses, improve farmer decision-making, and support sustainable agriculture through an efficient, scalable, and farmer-centric smart advisory solution.

Code and data availability

The paper describes a hybrid MobileNetV2/ResNet50 crop disease detection app trained on PlantVillage and Bangladesh Crop and Vegetable Disease datasets, but provides no authors' public code, model checkpoints, data deposits, or availability URLs. The mentioned datasets are generic third-party resources, not paper-archi

No evidence-backed public reproduction asset is currently recorded.