Agni-Chakshu (Jharkhand Forest Fire Intelligence System)

Agni-Chakshu is a high-resolution Geospatial AI system designed to predict and simulate forest fire spread across Jharkhand, India.

Overview

The system fuses multi-source satellite and atmospheric data to generate 12-hour predictive risk maps and real-time fire spread simulations.

Technical Architecture

• Model: Custom U-Net Deep Learning architecture optimized for Mac M-series (MPS).
• Data Pipeline: Fuses COP90 (Elevation), Bhuvan LULC (Fuel), NASA FIRMS (Fire History), OSM (Human Activity), and NetCDF (Weather).
• Simulation: Cellular Automata (CA) engine for temporal fire spread forecasting.
• Frontend: Streamlit dashboard with interactive Folium maps and animated spread visualizations.

Project Structure

ForestFire/
├── data/
│   ├── raw/                 # Original satellite/GIS data
│   ├── processed/           # AI-ready feature stacks
├── src/                     # Core Python engines
│   ├── model.py            # U-Net Architecture
│   ├── preprocess.py       # GIS Data Fusion
│   ├── simulation.py       # Fire Spread Engine
│   └── utils.py            # Visualization & GIS Tools
├── web/                     # Dashboard & API
│   ├── app.py              # Streamlit Interface
│   └── api_server.py       # FastAPI Backend
└── main.py                  # Pipeline Orchestrator

Setup & Usage

1. Install Dependencies:

   pip install -r requirements.txt

2. Run Pipeline:

   python main.py

3. Launch Dashboard:

   streamlit run web/app.py

CI/CD

The project includes GitHub Actions workflows for:

• Automated Jupyter Notebook testing.
• Automated package building and publishing validation.
• GIS system dependency management on Ubuntu runners.

---

Developed for Forest Fire Intelligence in Jharkhand.