Intelligent Plant Disease Detection & IoT Soil Health Monitoring System
Access the production dashboard here:
🚀 https://cropguard-ai-1-ys7p.onrender.com
CropGuard AI is a full-stack agricultural intelligence platform that combines deep learning-based plant disease detection with real-time IoT soil monitoring and external weather data to deliver actionable farming insights.
The system fuses three data sources — an AI image classifier, live Arduino sensor readings, and Open-Meteo weather forecasts — through a rule-based Fusion Engine that produces crop-specific risk scores, treatment plans, and irrigation recommendations.
Recently upgraded from a Streamlit prototype to a fully unified, production-ready Flask Web Application, CropGuard AI now features a responsive, dynamic UI with direct hardware connectivity right from your browser.
| Feature | Description |
|---|---|
| 🔬 AI Disease Detection | Client-side image cropping + server-side MobileNetV2 classification |
| 🔌 Web Serial Connect | Browser-level hardware connection filters & reads Arduino USB directly |
| 📡 Live IoT Monitoring | Robust Python background publisher auto-detects ports & heals from unplug/replug events |
| 🌦️ Dynamic Weather | Uses Browser Geolocation GPS to pull hyper-local real-time weather & AQI dynamically |
| 🧬 Fusion Engine | Cross-references disease diagnosis with 7-day sensor trends + weather to generate risk scores |
| 📊 Interactive Dashboard | Custom HTML/JS/CSS with real-time Chart.js graphs, gauge cards & threshold alerts |
| 📄 Report Generation | Export sensor data as CSV or professional PDF reports from the dashboard |
| 🗃️ Auto Data Retention | Background thread purges records older than 7 days and auto-vacuums the database |
- Plug and Play: Connect your Arduino Uno (with the flashed firmware) to your system via USB.
- Manual Connection: If the dashboard does not automatically display sensor readings, use the "🔌 Connect USB Sensor" button located in the bottom-right corner of the interface. This uses the Web Serial API to bridge your hardware directly to the browser.
- Stabilization Period: After connecting, reconnecting, or refreshing, please wait for approximately 1 minute for the system to stabilize, establish a consistent data stream, and update the "Online/Offline" status indicators.
- Navigate to the Microscope icon (Disease Detection) tab.
- Upload or drag-and-drop a photo of a plant leaf.
- Use the interactive cropper to focus on the affected area and click "Run AI Diagnosis".
- View the combined report which includes the AI's confidence score and the Fusion Engine's environmental risk analysis.
- Check the Live Soil Monitor for real-time graphs of temperature, humidity, and soil moisture.
- The system automatically stores your data for up to 7 days. Older data is purged automatically to maintain system performance.
- Export your findings using the "Export CSV" or "Download PDF Report" buttons in the History tab.
The Arduino Uno R3 acts as the edge node, polling data from DHT11 (Air) and HL-69 (Soil) sensors every 2 seconds. This data is packaged as a JSON string and broadcasted over the Serial USB interface.
- Backend Reader: A local Python script (
serial_reader.py) can run in the background, auto-detecting COM/tty ports and pushing data to the cloud API. - Web Serial Bridge: For a seamless experience without local setup, the frontend (
app.js) uses the Web Serial API to read the USB port directly from the browser and sync it to the database.
The Flask Backend serves as the central orchestrator:
- SQLite Database: Stores all historical readings, disease predictions, and user preferences.
- Retention Worker: A background thread monitors the database and purges records older than 168 hours (7 days) to ensure high performance and low storage overhead.
The core logic resides in fusion_engine.py. It evaluates the Cumulative Environmental Stress:
- Disease Model: MobileNetV2 (TensorFlow) classifies the leaf pathogen.
- Soil Context: Fetches 24-hour averages of soil moisture and humidity to check if conditions favor fungal growth.
- Weather Context: Integrates live precipitation and temperature forecasts from Open-Meteo.
- Outcome: Generates a unified Risk Score (0-100) and a prioritized action list.
CropGuard-AI/
│
├── run_all.py # 🚀 Master launcher — starts services
├── requirements.txt # Python dependencies
├── backend/
│ ├── server.py # Central server (DB, Retention, API Launcher)
│ ├── dashboard_api.py # Flask REST Endpoints & Routing
│ ├── fusion_engine.py # AI + Environmental logic
│ ├── database.py # SQLite abstractions
│ └── serial_reader.py # USB-to-Cloud bridge script
│
├── frontend/
│ ├── static/
│ │ ├── css/ # Glassmorphism design system
│ │ └── js/app.js # SPA logic, Charts, Web Serial API
│ └── templates/ # Jinja2 HTML templates
│
├── iot/
│ └── arduino_sensor.ino # Arduino C++ firmware
│
├── soil_data.db # 🗃️ SQLite database (auto-created)
└── plant_disease_model_new.keras # 🧠 Trained AI Model
| Component | Model | Pin | Purpose |
|---|---|---|---|
| Microcontroller | Arduino Uno R3 | — | Central controller |
| Temp/Humidity Sensor | DHT11 | D2 | Air temperature & humidity |
| Soil Moisture Sensor | HL-69 (Analog) | A0 | Soil moisture level (0–100%) |
| Alert LED | Standard LED | D8 | Visual alert indicator |
| Alert Buzzer | Passive Buzzer | D9 | Audible alert |
- Python 3.10+ with
pip - Arduino IDE (to flash the firmware)
- Arduino Uno R3 with sensors connected
- USB Cable connecting Arduino to your machine
- Google Chrome or Edge (required for Web Serial API features)
git clone <your-repo-url>
cd plant_disease_iot_v3
python -m venv .venv
source .venv/bin/activate # Linux/macOS
# .venv\Scripts\activate # Windows
pip install -r requirements.txt- Open
iot/arduino_sensor.inoin Arduino IDE - Install the DHT sensor library (by Adafruit) via Library Manager
- Select Arduino Uno board and the correct serial port
- Click Upload
python run_all.pyThis single command starts:
mqtt_subscriber.py— The MQTT listener, SQLite DB manager, and the embedded Flask API server (Port 5000).mqtt_publisher.py— The background hardware scanner that bridges Arduino Serial to the cloud.
Open your browser (Chrome/Edge recommended) to: https://cropguard-ai-1-ys7p.onrender.com
All REST endpoints are served by the Flask API on port 5000.
| Method | Endpoint | Description |
|---|---|---|
GET |
/api/sensor |
Latest sensor reading + connection status |
POST |
/api/sensor/upload |
Upload dynamic Web Serial client sensor readings to SQLite |
POST |
/api/predict |
Run MobileNetV2 inference on an uploaded image crop |
GET |
/api/history?hours=N |
Sensor history (max 168h / 7 days) |
GET |
/api/weather?lat=X&lon=Y |
Local weather, hourly/daily forecasts |
GET |
/api/insights?lat=X&lon=Y |
Rule-based crop insights with health score |
GET |
/api/thresholds |
Current alert thresholds |
POST |
/api/thresholds |
Update alert thresholds |
GET |
/api/calibration |
Current sensor calibration offsets |
POST |
/api/calibration |
Update calibration offsets |
GET |
/api/export/csv?hours=N |
Download sensor data as CSV |
GET |
/api/export/pdf?hours=N |
Download full PDF report |
| Layer | Technology |
|---|---|
| AI Model | TensorFlow/Keras (MobileNetV2), OpenCV |
| Backend Web Framework | Flask, Jinja2, Flask-CORS |
| Frontend Utilities | Chart.js, Cropper.js, Web Serial API, Geolocation API |
| IoT Hardware | Arduino Uno R3, DHT11, HL-69 Soil Sensor |
| Communication | MQTT (paho-mqtt) via HiveMQ public broker |
| Serial I/O | PySerial (9600 baud) |
| Database | SQLite3 |
| Weather | Open-Meteo API (free, no key) |
| Styling | Custom Vanilla CSS (Glassmorphism, CSS Variables) |
This project is for educational and research purposes.
Built with 💚 for smarter agriculture