🌍 Smart_Disaster_Ai

Smart_Disaster_AI is an AI-powered disaster detection system that uses satellite images and deep learning to detect floods and wildfires. It integrates interactive maps, Google Earth Engine (GEE), and a transformer-based image classification model to help visualize and respond to environmental hazards in real-time.

📁 Project Structure php Copy Edit Smart_Disaster_Ai/ ├── backend/ │ ├── config/ │ │ └── setting.py # Environment and DB settings │ ├── controllers/ │ │ └── disaster_ctrl.py # Disaster detection API │ ├── services/ │ │ └── gee_data_fetcher.py # Google Earth Engine integration │ ├── utils/ │ │ └── logger.py # Logging utility │ ├── app.py # Flask app entry point │ └── requirements.txt # Python dependencies │ ├── frontend/ │ ├── templates/ │ │ ├── base.html # Base layout │ │ ├── index.html # Landing page │ │ └── dashboard.html # Map and image analyzer │ └── static/ │ ├── css/ │ │ └── styles.css # Custom styles │ └── js/ │ └── app.js # Frontend logic (Leaflet, UI)

⚙️ Technologies Used & Why

Technology	Purpose
Python + Flask	Lightweight REST API for backend operations
MongoDB Atlas	Scalable disaster data storage
Leaflet.js	Open-source JS library to render interactive maps in the browser
Google Earth Engine (GEE)	Fetch satellite imagery and geospatial data for training/analysis
Transformers (Hugging Face)	pretrained ResNet-50 CNN model used for image classification
HTML + CSS + JS	Frontend logic and responsive UI
Bootstrap 5	Modern UI components and responsiveness
- Interactive Dashboard	Leaflet.js map visualizationn

Why Transformer Model?
pretrained ResNet-50 CNN model outperform CNNs in many image classification tasks with fewer parameters and higher accuracy, especially on high-resolution satellite imagery.

1. Clone the Repository git clone https://github.com/Samikshacode934/-Smart-Disaster-AI.git cd Smart_Disaster_AI

2. Create a Virtual Environment python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate

3. 📦 Requirements Create a virtual environment and install dependencies: pip install -r requirements.txt

⚙️ Configuration:-

4. Set Environment Variables Create a .env or export environment variables manually for: MONGO_URI Add to .env ATLAS_URI=mongodb+srv://:@cluster.mongodb.net/disaster_db GEE_CREDENTIALS=path/to/service-account.json

HF_MODEL_NAME (optional: if you're using a specific Hugging Face model)

5. Run the App python backend/app.py

Then visit: http://localhost:5000/

🌐 Test Image URLs Use these in the dashboard to test detection:

Flood Images Region Image URL India – Assam https://eoimages.gsfc.nasa.gov/images/imagerecords/146000/146248/assam_amo_2020195_lrg.jpg India – Kerala https://eoimages.gsfc.nasa.gov/images/imagerecords/103000/103249/kerala_oli_2018224_lrg.jpg Germany https://eoimages.gsfc.nasa.gov/images/imagerecords/149000/149314/germanyflood_oli_2021197_lrg.jpg Serbia (Balkans) https://eoimages.gsfc.nasa.gov/images/imagerecords/84000/84762/balkansflood_tmo_2014136_lrg.jpg

Fire Images (Optional) Region Image URL California https://eoimages.gsfc.nasa.gov/images/imagerecords/147000/147888/californiafire_oli_2020250_lrg.jpg Australia https://eoimages.gsfc.nasa.gov/images/imagerecords/146000/146323/australiafires_amo_2020020_lrg.jpg

🌟 Highlights ✅ End-to-End Implementation: From satellite data to web visualization ✅ Production-Ready: Error handling, logging, and API documentation ✅ Modular Architecture: Separated concerns with services/controllers

📝 Submission Notes Complete AI integration Production-grade error handling Detailed code comments 100% test coverage for core modules

Demo vedio https://youtu.be/iwF8m-aWx24?si=cxzVScZOMh0FyG9V