🚦 Smart Traffic Management Dashboard

An AI-powered real-time traffic monitoring system that detects emergency vehicles and traffic accidents using YOLO object detection models with automated email alerts.

✨ Features

• 🚑 Emergency Vehicle Detection - Real-time identification of ambulances, fire trucks, and police vehicles
• 🚨 Accident Detection - Automatic detection of traffic accidents with visual alerts
• 📧 Email Alerts - Automated email notifications with accident images and GPS location coordinates
• 🎥 Real-time Video Processing - Live video analysis with side-by-side detection results
• 🌍 GPS Integration - Browser-based geolocation capture for accident incidents
• ⚙️ Configurable Confidence - Adjustable detection confidence threshold (0.1 - 1.0)
• 📱 Web Dashboard - Interactive Streamlit web interface

📋 Prerequisites

• Python 3.8 or higher
• CUDA-capable GPU (recommended for real-time processing)
• Webcam or video files for input

🚀 Installation

1. Clone or extract the project

   cd "Intelligent Traffic Incident Detection"

2. Install dependencies

   pip install -r requirements.txt

3. Verify CUDA installation (optional but recommended)

   python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

⚙️ Configuration

Environment Variables

Set the following environment variables for email alerts functionality:

# Windows PowerShell
$env:SENDER_EMAIL = "your_email@gmail.com"
$env:SENDER_PASSWORD = "your_app_password"
$env:RECEIVER_EMAIL = "alert_recipient@gmail.com"

# Windows CMD
set SENDER_EMAIL=your_email@gmail.com
set SENDER_PASSWORD=your_app_password
set RECEIVER_EMAIL=alert_recipient@gmail.com

# Linux/Mac
export SENDER_EMAIL="your_email@gmail.com"
export SENDER_PASSWORD="your_app_password"
export RECEIVER_EMAIL="alert_recipient@gmail.com"

Note: For Gmail, use an App Password instead of your regular password.

📖 Usage

Run Web Dashboard

streamlit run app.py

The dashboard will open at http://localhost:8501

Dashboard Features:

• Upload traffic video files (mp4, avi, mov, webm)
• Adjust confidence threshold with slider
• View real-time detection results
• Monitor email alert status
• See accident detection alerts

Run Standalone Notebook

jupyter notebook accident_detection.ipynb

Processes video/webcam feed and saves output to output_video/ folder.

📁 Project Structure

├── app.py                          # Main Streamlit web application
├── accident_detection.ipynb        # Standalone Jupyter notebook
├── requirements.txt                # Python dependencies
├── best.pt                         # Accident detection model
├── emergency.pt                    # Emergency vehicle detection model
├── alarm.wav                       # Audio alert sound
│
├── Model Results and Details/
│   ├── data.yaml                   # Dataset configuration
│   ├── train.py                    # Model training script
│   ├── yolo12m.pt                  # YOLOv12 medium pretrained model
│   ├── yolo11n.pt                  # YOLOv11 nano pretrained model
│   │
│   ├── Accident model results/     # Accident model training results
│   │   ├── args.yaml
│   │   ├── results.csv
│   │   └── [confusion matrices, PR curves, etc.]
│   │
│   └── Vehicle Classification/     # Emergency vehicle model results
│       ├── args.yaml
│       ├── results.csv
│       ├── weights/
│       │   ├── best.pt
│       │   └── last.pt
│       └── [training visualizations]
│
├── Testing Videos/                 # Sample video files
│   ├── rear_collision_10.mp4
│   ├── stock-footage-a-uk-ambulance...webm
│   └── WhatsApp Video 2025-12-05...mp4
│
└── output_video/                   # Generated video outputs

🤖 Models

Emergency Vehicle Detection Model

• File: emergency.pt
• Framework: YOLOv12
• Classes: Ambulance, Fire Truck, Police, Vehicle
• Dataset: Indian Emergency Vehicles (Roboflow)
• License: CC BY 4.0

Accident Detection Model

• File: best.pt
• Framework: YOLO
• Purpose: Real-time accident/collision detection
• Training Data: Custom accident dataset

🔧 Training (Advanced)

To retrain models with new data:

python Model\ Results\ and\ Details\train.py

Training Configuration:

• Epochs: 120
• Batch Size: 8 (RTX 3050 compatible)
• Learning Rate: 0.001
• Optimizer: AdamW
• Image Size: 640x640

📊 Model Performance

Both models include comprehensive evaluation metrics:

• Confusion matrices
• Precision-Recall (PR) curves
• F1, Precision, and Recall curves
• Training/validation visualizations
• Results CSV files

🐛 Troubleshooting

Streamlit Won't Start

• Issue: Exit code 1
• Solution: Set environment variables for email credentials

  $env:SENDER_EMAIL = "email@gmail.com"
  $env:SENDER_PASSWORD = "password"
  $env:RECEIVER_EMAIL = "recipient@gmail.com"

GPU Not Detected

• Ensure NVIDIA CUDA Toolkit is installed
• Update NVIDIA drivers
• Verify PyTorch installation: python -c "import torch; print(torch.cuda.is_available())"

Model Not Found

• Ensure best.pt and emergency.pt are in the project root directory
• Check file paths are correct

Email Alerts Not Working

• Verify environment variables are set correctly
• Use App Password for Gmail (not regular password)
• Ensure 5-minute cooldown between alerts
• Check SENDER_EMAIL and RECEIVER_EMAIL are valid

📦 Dependencies

See requirements.txt for complete list:

• streamlit - Web UI framework
• opencv-python - Video processing
• ultralytics - YOLO models
• pygame - Audio alerts
• torch - Deep learning framework
• numpy - Numerical operations
• Pillow - Image processing

📝 Notes

• Email alerts have a 5-minute cooldown to prevent spam
• Confidence threshold default: 0.3 (adjustable in sidebar)
• GPS location requires browser permission
• Accident frame screenshots are automatically saved and deleted after email

🎯 Future Enhancements

• Real-time traffic flow analysis
• Traffic congestion prediction
• Multi-camera support
• Database logging for incident history
• Mobile app integration
• SMS alerts
• Integration with traffic management systems

📧 Support

For issues or questions, refer to:

• Ultralytics YOLOv8 Documentation
• Streamlit Documentation
• Roboflow Dataset

📄 License

Dataset License: CC BY 4.0 Project: For educational and commercial use.

---

Version: 1.0
Last Updated: December 2025