🌱 AgriAidAI

AI-powered crop disease detection and agricultural intelligence platform for modern farmers.

AgriAidAI combines a fine-tuned MobileNetV2 computer vision model with a FastAPI backend and a React frontend to deliver instant crop disease diagnosis, expert treatment recommendations, live weather intelligence, and a persistent scan history — all in one platform.

[image ]

✨ Features

Feature	Description
🔬 AI Disease Detection	MobileNetV2 model trained on the PlantVillage dataset — 38 disease classes across 16 crops
📋 Structured Results	Per-scan: disease name, confidence score, symptoms, treatment plan, preventive tips
🌤️ Live Weather	Real-time weather via Open-Meteo using GPS or IP geolocation — no API key required
📈 Prediction History	All scans persisted in-memory; auto-refreshes on the dashboard
📊 Admin Statistics	Real prediction counts, average confidence, 7-day activity chart
📄 Export to PDF	Print any scan result as a PDF directly from the browser
🔐 Auth (lightweight)	Email + fixed password login; username derived from email; localStorage session
🌍 Multilingual	English + Hindi support via LanguageContext
📱 Responsive	Works across desktop, tablet, and mobile

---

🖥️ Tech Stack

Backend

Layer	Technology
Framework	FastAPI
ML Runtime	TensorFlow / Keras
Model	MobileNetV2 (fine-tuned, PlantVillage dataset)
Weather	Open-Meteo API + Nominatim reverse geocoding
HTTP client	httpx
Server	Uvicorn with ASGI lifespan

Frontend

Layer	Technology
Framework	React 18 + TypeScript
Styling	Tailwind CSS (minimal design system)
UI Components	shadcn/ui
Charts	Recharts
Routing	React Router v6
Data Fetching	Native fetch + useCallback hooks

---

🗂️ Project Structure

agri-aid-master/
├── agriaidai_backend/          # FastAPI backend
│   ├── main.py                 # App entry point, routes, lifespan
│   ├── models/
│   │   ├── prediction_models.py    # Pydantic schemas
│   │   └── train_plant_disease.py  # Model training script
│   └── services/
│       ├── prediction_service.py   # MobileNetV2 inference engine
│       ├── history_service.py      # In-memory prediction history
│       ├── weather_service.py      # Open-Meteo live weather
│       └── admin_service.py        # Real admin stats from history
│
├── agri-aid-bright/            # React frontend
│   └── src/
│       ├── pages/
│       │   ├── Index.tsx       # Home + login (merged)
│       │   ├── Dashboard.tsx   # Live dashboard with stats & charts
│       │   ├── Predict.tsx     # Split-screen: upload + results
│       │   ├── Admin.tsx       # Admin stats dashboard
│       │   └── About.tsx       # Platform info & tech details
│       ├── components/
│       │   └── Navigation.tsx  # Sticky nav with auth + logout
│       └── contexts/
│           └── LanguageContext.tsx
│
└── models/                     # Trained model artifacts
    ├── plant_disease_model.h5
    └── class_labels.npy

---

🚀 Getting Started

Prerequisites

• Python 3.10+
• Node.js 18+
• A trained model file at models/plant_disease_model.h5 (see Training)

1. Backend Setup

cd agri-aid-master

# Create and activate a virtual environment
python -m venv .venv
.venv\Scripts\activate          # Windows
# source .venv/bin/activate     # macOS/Linux

# Install dependencies
pip install fastapi uvicorn tensorflow httpx numpy pillow python-multipart

# Start the backend
python -m agriaidai_backend.main

Backend runs at: http://localhost:8000 Interactive API docs: http://localhost:8000/docs

Environment variables (optional — set to suppress TF noise):

TF_ENABLE_ONEDNN_OPTS=0
TF_CPP_MIN_LOG_LEVEL=2

2. Frontend Setup

cd agri-aid-bright

npm install
npm run dev

Frontend runs at: http://localhost:8081

---

🔑 Login

The app uses a lightweight auth system with localStorage.

Field	Value
Email	Any valid email (yourname@example.com)
Password	password123#

Your username in the navigation is automatically derived from the email prefix (e.g. john@gmail.com → displays as john).

---

🌿 Supported Crops & Diseases

The model classifies 38 disease classes across 16 crops:

Crop	Example Diseases
Tomato	Early Blight, Late Blight, Leaf Mold, Mosaic Virus
Corn	Common Rust, Northern Leaf Blight, Gray Leaf Spot
Potato	Early Blight, Late Blight
Apple	Apple Scab, Black Rot, Cedar Apple Rust
Grape	Black Rot, Esca, Leaf Blight
Pepper	Bacterial Spot
Strawberry	Leaf Scorch
Peach, Cherry, Squash, Soybean, Orange, Blueberry, Raspberry	Various

All crops also include a Healthy class.

---

🧠 Model Training

To retrain or improve the model:

# Place PlantVillage dataset inside:
# agriaidai_backend/models/dataset/PlantVillage/

python agriaidai_backend/models/train_plant_disease.py

Training configuration (in train_plant_disease.py):

• Architecture: MobileNetV2 (ImageNet pre-trained)
• Input size: 224 × 224 RGB
• Phase 1: 20 epochs — top classifier only
• Phase 2: 20 epochs — fine-tune top 30 layers
• Augmentation: flip, rotation, zoom, brightness, contrast
• Callbacks: EarlyStopping, ReduceLROnPlateau, ModelCheckpoint

Artifacts saved to models/:

• plant_disease_model.h5 — Keras model weights
• class_labels.npy — ordered class name array

---

🌐 API Reference

Method	Endpoint	Description
POST	/predict	Multipart: crop (str) + file (image) → disease prediction
GET	/weather	?city= or ?lat=&lon= → live weather object
GET	/history	Returns last N predictions
POST	/history	Save a prediction entry
GET	/admin/stats	Returns real prediction counts and 7-day chart data

Full schema available at http://localhost:8000/docs

---

📍 Weather & Location

Weather is fetched automatically using a 3-tier location strategy:

1. Browser GPS — asks for location permission; uses real coordinates
2. IP Geolocation (ipapi.co) — no permission needed; returns city from IP
3. Fallback — Coimbatore (default if both above fail)

The backend uses Open-Meteo (free, no API key) and Nominatim for reverse geocoding.

---

🗺️ Roadmap

• Persistent database (SQLite / PostgreSQL) for prediction history
• Real user authentication with JWT
• Crop calendar & seasonal advisories
• Batch image analysis
• Push notifications for disease alerts
• Offline PWA support
• OpenWeatherMap integration for extended forecast

---

📄 License

MIT License — open for personal and commercial use. Contributions welcome.

---

🙏 Acknowledgements

• PlantVillage Dataset — disease image training data
• Open-Meteo — free, no-key weather API
• Nominatim / OpenStreetMap — reverse geocoding
• shadcn/ui — React component system
• TensorFlow / Keras — ML framework