🌱 AgroPredict

Smart Agriculture Web and Mobile application that provides intelligent crop recommendations, fertilizer suggestions, and plant disease classification using Machine Learning and Deep Learning.

---

📖 Table of Contents

• About the Project
• Key Features
• Concepts & Methodologies
• Tech Stack
• Project Structure
• Setup & Installation
• Running the Application
• Datasets
• Machine Learning Models

---

🌾 About the Project

AgroPredict is an end-to-end full-stack AI platform designed to help farmers and agriculture enthusiasts make data-driven decisions. The system integrates machine learning to provide accurate crop recommendations based on soil testing and real-time weather data. Additionally, it offers a deep learning-powered visual disease classification system for plant leaves and smart fertilizer selection based on soil NPK thresholds.

The project is natively available as a Flask Web Application and a fully functional cross-platform Mobile App built with React Native (Expo).

---

✨ Key Features

1. Crop Recommendation 🌱

   • Suggests the most suitable crop to plant based on Soil variables (Nitrogen, Phosphorous, Potassium, pH) and environmental metrics (Rainfall).
   • Dynamically fetches realtime Temperature and Humidity for the provided city using the OpenWeatherMap API.

2. Fertilizer Recommendation 🧪

   • Employs rule-based heuristics comparing target nutrient levels corresponding to crops and current soil NPK ratios.
   • Prescribes actionable advice to raise or lower specific soil nutrient levels.

3. Plant Disease Classification 🦠

   • Users can upload an image of a plant leaf.
   • The application leverages a finely-tuned ResNet9 PyTorch Deep Learning model to classify the image against 38 different plant disease classes.

4. Multi-Platform & Accessibility 📱

   • Functional Web Interface using Flask, HTML, CSS.
   • Cross-Platform React Native App enabling mobile functionality.
   • Multi-language Support (i18n) with English and Hindi translations.
   • Text-to-Speech (SpeechButton) integration for better accessibility.

---

🧠 Concepts & Methodologies

Core Concept	Description
Classification (ML)	The project utilizes a Random Forest Classifier mapped over an extensively trained agricultural dataset to predict the optimal crop category.
Computer Vision (DL)	A convolutional neural network architecture (ResNet9) maps pixel data of crop leaves into predicted categorical distributions for disease profiling.
API Integrations	Incorporates the OpenWeatherMap API to compute live weather markers mapping out temperature and humidity constraints in recommendation queries.
Mobile Accessibility	Built with Expo SDK & React Native utilizing custom hooks, React Navigation (Tabs/Stack), native device sensors, gestures, and audio accessibility.
Data Imputation	Notebooks cover robust pandas operations for joining raw CP data, fertilizer, and crop metrics into clean pre-processed sets for training pipelines.

---

🛠 Tech Stack

🚀 Backend & Machine Learning

• Python 3.x
• Flask (Web Application & API serving)
• PyTorch & Torchvision (ResNet9 Plant Disease Model Training and Inference)
• Scikit-Learn & XGBoost (Random Forest for Crop Forecasting)
• Pandas & NumPy (Data Processing)
• Jupyter Notebook (Mathematical EDA and Data Modelling pipelines)

🎨 Frontend Web

• HTML5 & CSS3
• JavaScript
• Jinja2 (Flask Templating)

📱 Mobile Dashboard (Cross-Platform)

• React Native & Expo SDK 52
• Expo Router / React Navigation
• i18next (Multilingual Context - Hindi/English)
• Axios (API interfacing)

---

📂 Project Structure

Agropredict/
│
├── app/                       # Flask Web Backend and API Routes
│   ├── app.py                 # Core Flask application and endpoints
│   ├── config.py              # Environment and API configurations
│   ├── static/                # CSS Styles and JS scripts (cities.js)
│   ├── templates/             # HTML Templates (Jinja2)
│   └── utils/                 # ML processing scripts (disease.py, fertilizer.py)
│
├── MobileApp/                 # React Native / Expo Mobile App codebase
│   ├── app/                   # Expo Router layout & structural screens
│   ├── components/            # UI components (Collapsible, HapticTab, SpeechButton)
│   ├── constants/             # Theming (Colors) and i18n locales (en, hi)
│   ├── hooks/                 # Custom React Hooks
│   ├── App config files       # package.json, app.json, tsconfig.json
│
├── model-building/            # Core Jupyter Notebooks detailing algorithm choices
│   ├── Crop_Recommendation_Model.ipynb
│   └── plant-disease-classification.ipynb
│
├── models/                    # Exported and Serialized ML weights
│   ├── plant_disease_model.pth # PyTorch Weights
│   └── RandomForest.pkl       # Sklearn classifier (inside app/models logically)
│
├── data-prep/                 # Pre-processing configurations
├── data-processed/            # Cleaned final datasets ready for regression algorithms
├── Data-raw/                  # Open-source agriculture raw datasets
└── requirements.txt           # Python application dependencies

---

⚙️ Setup & Installation

1. Prerequisites

Ensure you have the following installed on your local machine:

• Python 3.8+
• Node.js 18+ & npm (For Mobile App)
• OpenWeatherMap API Key.

2. Backend / Web App Deployment

1. Clone the repository and navigate to the project directory.
2. Create a virtual environment (Optional but Recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install the Python dependencies:

   pip install -r requirements.txt

4. Configuration: Open app/config.py and set your OpenWeather API Key:

   weather_api_key = "YOUR_API_KEY_HERE"

3. Mobile App Deployment

1. Navigate into the MobileApp/ directory:

   cd MobileApp

2. Install npm packages:

   npm install

3. You may need to edit MobileApp/app/config.js or .env files if they possess absolute API URL properties mapping natively to your local python backend server IP network.

---

💻 Running the Application

Running the Web Server Backend

To spin up the FLASK server, run:

cd app
python app.py

The web API will successfully run locally at http://127.0.0.1:5000.

Running the Mobile Client App

To run the React Native Expo app:

cd MobileApp
npm start

• Press a to open on an Android Emulator.
• Press w to open on Web.
• Alternatively, download the Expo Go application on your mobile device and scan the generated QR code.

---

📊 Datasets

The raw data sources reside in the Data-raw/ folder, spanning combinations of historical agricultural records encompassing crop parameters and fertilizer dependencies. The data-prep/ pipeline joins these datasets ensuring cohesive data frames generated to ./data-processed/ to be uniformly fed into the scikit-learn algorithms.

---

🤖 Machine Learning Models

1. Crop Recommendation (Random Forest)

An ensemble learning algorithm. A multitude of decision trees are generated during training to account for non-linear soil permutations accurately outputting an individual nominal crop classification to minimize entropy.

2. Plant Disease Recognition (ResNet-9)

Utilizes Transfer Learning on a ResNet9 infrastructure (Residual Neural Networks), tackling vanishing gradients and ensuring deep layered training. Sourced out of the PyTorch ecosystem (torch & torchvision), it accepts transposed normalized RGB images mapping inferences correctly targeting categories like 'Tomato___Bacterial_spot' or 'Healthy'.