🔥 RtoG: Wildfire Fast Detection Server (Red to Green)

This repository powers the RtoG (Red to Green) server — an AI-driven system for fast and accurate wildfire detection. It uses YOLOv8 for real-time object detection and integrates Gemini Vision API for verification, enabling early fire detection from video footage like CCTV or drone sources.

From red (wildfire) to green (safe forest) — detect fast, act faster.

---

🚀 Project Structure (Mono Repository)

This project has been refactored into a monorepo with independent virtual environments:

RtoG/
├── server/           # Web server (FastAPI)
│   ├── venv/             # App server virtual environment
│   └── requirements.txt  # App server dependencies
│
├── preprocessing/        # Image preprocessing tools
│   ├── venv/             # Preprocessing virtual environment
│   └── requirements.txt  # Preprocessing dependencies
│
├── model-training/       # Model training code
│   ├── venv/             # Model training virtual environment
│   └── requirements.txt  # Model training dependencies
│
├── Other project files...

💻 Virtual Environment Setup

Each project uses an independent virtual environment to prevent dependency conflicts:

server

cd server
python3 -m venv venv
source venv/bin/activate  # Linux/Mac
pip install -r requirements.txt

preprocessing

cd preprocessing
python3 -m venv venv
source venv/bin/activate  # Linux/Mac
pip install -r requirements.txt

model-training

cd model-training
python3 -m venv venv
source venv/bin/activate  # Linux/Mac
pip install -r requirements.txt

🧠 Model Class Mapping

ID	Class Label
0	흑색연기 (Black Smoke)
1	백색/회색연기 (White/Grey Smoke)
2	화염 (Flame)
3	구름 (Cloud)
4	안개/연무 (Fog/Mist)
5	굴뚝연기 (Chimney Smoke)