CropGuard-AI

A pocket doctor for Pakistani farmers, diagnosing plant diseases and aiding in farming practices.

Comment

🌾 CropGuard: AI-Powered Crop Disease Dignosis

Event: DevLaunch Hackathon 2026 (WPBrigade × iCodeGuru)

CropGuard is a multi-modal AI-powered system designed to assist traditional farmers by detecting crop diseases from leaf images and providing spoken Urdu guidance. The platform leverages state-of-the-art computer vision, natural language processing, and speech synthesis to make advanced agricultural intelligence accessible to non-technical users.

The system allows farmers to simply upload a crop image. The AI then identifies the crop and disease, explains the issue in English, translates the result into Urdu, and finally converts it into natural Urdu voice output, enabling farmers to interact with the system without needing literacy or technical skills.

🎯 Project Overview

CropGuard addresses a real-world problem in agriculture:
farmers lack instant access to expert-level crop disease diagnosis.

By using modern AI models, CropGuard provides:

  • Crop name detection
  • Disease identification
  • Confidence score
  • Disease description
  • Recommended pesticide
  • Urdu text guidance
  • Urdu voice assistance (hands-free)

This project demonstrates an end-to-end multi-modal AI system combining:

  • Vision AI
  • NLP translation
  • Speech synthesis

Built as a hackathon demo, but architected to be scalable and production-ready.

🎥 Demo Video (YouTube)

Watch the complete demo of CropGuard – AI-Powered Crop Disease Dignosis:

CropGuard AI Demo Video
▶️ Click to Play

🏗️ System Architecture

Three-Tier Architecture

Layer Description
Presentation Layer Streamlit web interface for farmers
Application Layer FastAPI backend for request handling
AI Layer Gemini Flash 3 (Vision) + Gemini NLP + Google Voice API

AI Processing Flow

graph TD
    A[Crop Image Upload] --> B[Gemini Flash 3 Vision Mode]
    B --> C[ Disease Diagnosis]
    C --> D[Gemini Translation API]
    D --> E[Urdu Text]
    E --> F[Google Voice API]
    F --> G[Urdu Voice Output for Farmer]

This makes CropGuard a fully multi-modal AI pipeline:
Image → Text → Translation → Speech

🚀 Technology Stack

Backend

Component Technology
API Framework FastAPI
AI Vision Model Google Gemini Flash 3
Image Processing Pillow
Translation Gemini Text API
Voice Google Voice API
Environment Python 3.11
API Docs Swagger (OpenAPI)

Frontend

Component Technology
UI Streamlit
User Interaction Image upload + audio output
Target Users Traditional farmers

🧠 AI Model & Intelligence Pipeline

Gemini Flash 3 (Vision AI)

Gemini Flash 3 is a lightweight and high-performance multimodal model capable of understanding images and generating structured textual outputs.

In CropGuard, Gemini Flash 3 is used to:

  • Analyze crop leaf images
  • Identify crop type
  • Detect disease patterns
  • Generate human-readable diagnosis

Language & Voice Intelligence

After vision inference:

  1. Output is generated in English
  2. Gemini Translation API converts English → Urdu
  3. Google Voice API converts Urdu text → natural Urdu speech

This enables:

  • Voice-based interaction
  • Accessibility for illiterate farmers
  • Real-world usability in rural Pakistan

📊 Supported Crops

CropGuard is designed to work with multiple crop types, including but not limited to:

  • Wheat
  • Rice
  • Cotton
  • Tomato
  • Potato
  • Sugarcane
  • Maize
  • Vegetables and fruits

The system is model-agnostic and can scale to more crops without retraining.

🔬 Testing & Evaluation

Since this project is based on real-time AI inference (no custom training), evaluation was performed through:

  • Manual testing on multiple crop images
  • Cross-verification with online agricultural sources
  • Live testing during hackathon demo

The system demonstrated:

  • High confidence predictions for common crops
  • Accurate disease identification for visible symptoms
  • Consistent translation and voice generation

This validates CropGuard as a practical real-world AI assistant, not just a theoretical model.

🚀 Deployment Architecture

Current Status

  • Hackathon Demo Deployment
  • Runs locally with full functionality
  • Scalable microservice design

Scalable by Design

The system can be deployed to:

  • Cloud platforms (Render, AWS, GCP, Azure)
  • Mobile apps (Flutter / React Native)
  • IoT devices (field cameras)

📂 Project Structure


CropGuard-AI/
├── backend/
│   ├── app/
│   │   ├── api/
│   │   │   ├── routes.py
│   │   │   └── schemas.py
│   │   ├── services/
│   │   │   ├── gemini_client.py
│   │   │   ├── image_processing.py
│   │   │   └── voice_service.py
│   │   ├── main.py    # FastAPI app + endpoints
|   |   └── .gitignore 
│
├── frontend/
│   └── app.py
├── .env
├── requirements.txt
└── README.md

🔌 API Endpoint

POST /analyze

Input: Multipart image file
Output: JSON response

{
  "crop_name": "Wheat",
  "disease_name": "Leaf Rust",
  "confidence": 0.95,
  "description": "Fungal disease affecting wheat leaves.",
  "recommended_pesticide": "Propiconazole",
  "urdu_message": "یہ گندم کی بیماری ہے اور اس کے علاج کے لیے مخصوص دوا استعمال کریں۔"
}

⚙️ Local Installation & Usage

1️⃣ Clone Repository

git clone https://github.com/YOUR_USERNAME/cropguard-backend.git
cd cropguard-backend

2️⃣ Create Virtual Environment

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

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Environment Variables

Create a .env file in the root directory:

PORT=8000 GEMINI_API_KEY=GOOGLE VOICE API + GEMINI API GEMINI_MODEL=gemini-3-flash-preview

5️⃣ Run Backend

uvicorn app.main:app --reload 🔌 API Access Base URL: http://127.0.0.1:8000

**Swagger Docs: **http://127.0.0.1:8000/docs

🖥️ Streamlit Frontend

Run the Streamlit application:

streamlit run app.py The frontend allows farmers to:

Upload crop images

View disease predictions

Receive Urdu guidance

Listen to voice instructions

🔐 Security Notes

.env file is ignored via .gitignore

API keys are never committed to the repository

No sensitive data is stored

All inference is stateless and secure

🎯 Key Achievements

Multi-modal AI system (Vision + NLP + Voice)

Real-world agricultural use case

Built for non-technical users

Urdu voice accessibility 🇵🇰

Scalable microservice architecture

Production-ready design

Hackathon-grade innovation

🔮 Future Enhancements

  • Mobile app (Flutter)

  • Offline disease detection

  • Multi-language farmer support

  • Crop advisory dashboard

  • IoT camera integration

  • Government agriculture data APIs

  • Model fine-tuning on local Pakistani crops

👥 Team

Name Role GitHub
Hanif Ullah Team Lead, Frontend Developer https://github.com/hanifullah313
Moneka Meghwar Backend Developer, API Integration https://github.com/mmoneka11
Kashmala Saddiqui Documentation Lead,API Integrate Support https://github.com/kashmalaasif

📜 License

This project is licensed under the MIT License.

🤝 Contributing

We welcome contributions! Please open issues or pull requests.

⭐ Support this project: If you find this project useful, please star the repo! GitHub stars

✨ Built with ❤️ for smart agriculture

Built With

Share this project:

Updates