SkyGuard-AI

🌩️ SkyGuard AI — National Heavy Rainfall & Inundation Early Warning Platform

Smart India Hackathon (SIH) 2026 Flagship Project
Real-time multi-source data fusion combining satellite imagery, IMD Doppler radar, 3D DEM digital twin terrain profiles, ML ensemble prediction, and autonomous UAV drone reconnaissance.

---

🌐 Live Production Links

Service	Environment	URL	Status
🚀 Vercel Production CDN	Standalone Frontend	rainsafe-ai-sih2026.vercel.app	LIVE 200 OK
🟣 Render Cloud Full-Stack	Python FastAPI Backend	rainsafe-ai-sih2026.onrender.com	LIVE 200 OK
📡 Dashboard Telemetry API	REST API Endpoint	rainsafe-ai-sih2026.onrender.com/api/dashboard	LIVE 200 OK
📚 Interactive API Docs	Swagger / OpenAPI	rainsafe-ai-sih2026.onrender.com/docs	LIVE 200 OK

---

🛠️ System Architecture & Data Pipeline

                              ┌───────────────────────────────────────────────┐
                              │  SATELLITE & OBSERVATIONAL DATA SOURCES       │
                              │  • INSAT-3DR Satellite Telemetry             │
                              │  • IMD Doppler Weather Radar (DWR)            │
                              │  • 17 Automatic Weather Stations (AWS)        │
                              │  • Open-Meteo GFS/ECMWF NWP Network          │
                              └──────────────────────┬────────────────────────┘
                                                     │
                                                     ▼
                              ┌───────────────────────────────────────────────┐
                              │     FASTAPI ASYNC TELEMETRY PIPELINE          │
                              │     (Python 3.11 / Uvicorn / Async Engine)    │
                              └──────────────────────┬────────────────────────┘
                                                     │
                                                     ▼
                              ┌───────────────────────────────────────────────┐
                              │  MACHINE LEARNING ENSEMBLE CLASSIFIER         │
                              │  • Random Forest + LightGBM + XGBoost         │
                              │  • 21 Meteorological Ingestion Features       │
                              │  • 0.904 ROC-AUC Accuracy Score @ 121ms       │
                              └──────────────────────┬────────────────────────┘
                                                     │
                                     ┌───────────────┴───────────────┐
                                     ▼                               ▼
                      ┌───────────────────────────────┐ ┌───────────────────────────────┐
                      │ SAAS ENTERPRISE UI (VERCEL)   │ │ 🚁 UAV DRONE RECON ENGINE     │
                      │ • 3D DEM Digital Twin         │ │ • Thermal IR Damage Matrix    │
                      │ • Leaflet GIS Risk Polygons   │ │ • NDMA PDF Report Exporter    │
                      │ • Conversational AI Assistant │ │ • Safe GPS Evacuation Routes  │
                      └───────────────────────────────┘ └───────────────────────────────┘

---

✨ Executive Feature Highlights

1. 🧠 Machine Learning Risk Intelligence

• 0.904 ROC-AUC Accuracy: Ensemble classification of rainfall risk (LOW, MODERATE, HIGH, CRITICAL) trained on historical monsoon data.
• Strict Mathematical Synchronization: Probability and risk categories dynamically scale with precipitation volume for 100% card consistency across all user input locations.

2. 🏔️ 3D Digital Twin DEM Elevation Profile

• Native Three.js WebGL Engine rendering 30m Digital Elevation Model cross-sections.
• Calculates slope gradients ($3.4^\circ$), river basin terrain, and dynamic water accumulation surge levels ($49.5\text{m}$).

3. 🤖 Conversational Multi-City AI Assistant

• Natural language query interface supporting prompts like "for Patna", "for Motihari", "for Ranchi".
• Performs entity extraction, fetches live satellite weather data, and triggers 1-click interactive map camera flights.

4. 🚁 Autonomous UAV Drone Reconnaissance & NDMA PDF Exporter

• Low-altitude 4K Thermal IR patrol stream operating below cloud cover at 120m altitude.
• Real-time aerial damage scan of bridges, power substations, and submerged highways with 1-click official NDMA PDF Report Export.

5. 🧭 NDRF Safe Evacuation Pathfinder

• Dynamic GPS safe route dispatch to nearby emergency relief shelters and sports complexes.
• 1-click emergency escalation to NDRF Headquarters (1078) and State Emergency Relief Cell (1070).

---

💻 Local Setup & Installation

Prerequisites

• Python 3.11+
• Node.js (Optional for syntax validation)
• Git

1. Clone Repository

git clone https://github.com/arshsri0513/SkyGuard-AI.git
cd SkyGuard-AI

2. Start Backend API Server

cd backend
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload --host 127.0.0.1 --port 8000

3. Launch Frontend Dashboard

cd frontend
python -m http.server 5500 --bind 127.0.0.1

Open http://127.0.0.1:5500 in your browser!

---

🏆 Smart India Hackathon (SIH) 2026 Presentation Protocol

1. Open Command Center and demonstrate real-time location telemetry for searched cities (Patna, Ranchi, Jaipur, Kolkata, Delhi, Bokaro).
2. Click 🤖 AI Assistant and type "for Motihari" to demonstrate natural language entity parsing and interactive map flight.
3. Switch to Inundation Risk view to showcase the 3D DEM Digital Twin Elevation Profile.
4. Click 🚁 Drone Recon and select 📋 Export Drone Reconnaissance Report to download the official printable NDMA PDF report.
5. Demonstrate 🔊 Voice Warning and 🇮🇳 हिन्दी dual-language accessibility.

---

📄 License & Attribution

Developed for Smart India Hackathon (SIH) 2026. Open source under the MIT License.
Author: Arsh Srivastava (@arshsri0513)