SIH26084 — WeatherNow AI Nowcasting System

AI-Powered Weather Nowcasting and Early Warning System integrating the SIH V3 PyTorch Deep-Convection ConvLSTM Model with a high-performance FastAPI backend and a modern React/Vite dashboard frontend.

---

📂 Repository Architecture

weather nowcast/
├── backend/                # Python FastAPI backend serving SIH V3 Nowcast model
│   ├── models/             # PyTorch model weights (sih_v3_best.pth)
│   ├── routes/             # Endpoints (/api/health, /api/model-info, /api/predict, /api/hazards)
│   ├── utils/              # Normalization & preprocessing pipeline
│   ├── main.py             # FastAPI service entry point
│   ├── model_loader.py     # SIHV3Nowcast architecture & checkpoint loader
│   ├── inference.py        # Inference pipeline (tensor prep, forward pass, sigmoid)
│   ├── schemas.py          # Pydantic validation schemas
│   ├── config.py           # Model constants, fixed mean/std arrays, settings
│   ├── requirements.txt    # Python dependencies
│   ├── .env.example        # Backend environment template
│   └── README.md           # Backend documentation
│
├── frontend/               # React 19 + TypeScript + Vite frontend application
│   ├── src/                # UI source code, components, pages, services, types
│   ├── public/             # Favicon (SVG) and static assets
│   ├── package.json        # Frontend package config ("weathernow-ai")
│   ├── index.html          # HTML entry point & metadata
│   ├── vite.config.ts      # Vite configuration
│   ├── .env.example        # Frontend environment template
│   └── README.md           # Frontend documentation
│
└── README.md               # Root project documentation

---

🚀 Quick Start Guide

1. Run the Python FastAPI Backend

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn main:app --reload --host 0.0.0.0 --port 8000

• API Docs & Swagger UI: http://localhost:8000/docs
• Health Status: http://localhost:8000/api/health
• Model Info: http://localhost:8000/api/model-info

2. Run the React Frontend

cd frontend
npm install
npm run dev

• Dashboard UI: http://localhost:5173

---

🔬 Scientific Model Semantics (SIH V3 Nowcast)

• Model Architecture: ConvLSTM with Encoder-Decoder & 4 Horizon Output Heads (SIHV3Nowcast).
• Input Dimensions: (1, 6, 8, 128, 128) — 6 temporal frames (10-min cadence), 8 meteorological channels, 128x128 spatial resolution.
• Strict Channel Order: ["B13", "t2m", "d2m", "u10", "v10", "cape", "cin", "tp"].
• Target Meaning: P(future B13 < 235 K) — Cold-Cloud / Deep-Convection Proxy Probability.
• Important Note: Output represents deep convective proxy probabilities. It is not direct lightning, hail, or cloudburst detection. All derived hazard scores are explicitly marked as proxies.
• Horizons: 30, 60, 90, and 120 minutes.