Driver Monitoring System

An AI-powered real-time driver monitoring system that detects unsafe driving behaviors to enhance road safety.

Quick Links

• Live Demo: https://driver-monitoring.vercel.app/
• Repository: https://github.com/AlbatrossC/Driver-Monitoring-System

---

Project Overview

The Driver Monitoring System (DMS) is an AI-based solution designed to detect unsafe driving behaviors in real-time. Using computer vision and deep learning, the system identifies critical safety violations including:

• Drowsiness detection
• Phone usage while driving
• Eating and drinking
• Smoking
• Seatbelt violations
• Driver distraction

The system is optimized to run directly in the browser using lightweight object detection models, making it accessible and efficient.

---

Models Used

We developed three distinct models for comprehensive driver behavior analysis:

YOLO Models (Object Detection)

Model	Description	Status
YOLOv8n	Fast and lightweight object detection	✅ Deployed in browser
YOLO11n	Improved accuracy with better class separation	✅ Deployed in browser

CNN Model (Image Classification)

Model	Description	Status
Custom CNN	Deep learning model for behavior classification	⚠️ Local deployment only

Live Demo Note

The browser-based demo runs only the YOLO models due to performance constraints. The CNN model is too large and memory-intensive for browser deployment.

To run the CNN model locally:

git clone https://github.com/AlbatrossC/Driver-Monitoring-System
cd Driver-Monitoring-System
python app.py

---

Data Sources

Dataset 1: Soham Jadhav - Roboflow Dataset

Link: Driver Monitoring System Dataset

Classes (8):

• 0: Distracted
• 1: Drinking
• 2: Drowsy
• 3: Eating
• 4: PhoneUse
• 5: SafeDriving
• 6: Seatbelt
• 7: Smoking

Dataset 2: Chaitanya kulkarni - Roboflow Universe Dataset

Link: Abnormal Driver Behaviour Dataset

Classes:

• Cigarette
• Drinking
• Eating
• Phone
• Seatbelt

Dataset 3: Divyanshu Mishra - Kaggle Dataset

Link: Multi-Class Driver Behavior Image Dataset

Classes:

• other_activities
• safe_driving
• talking_phone
• texting_phone
• turning

---

Data Processing & Training

All datasets underwent rigorous preprocessing:

• Data cleaning and structuring
• Data augmentation for improved model generalization
• Class balancing to handle imbalanced datasets
• YOLO annotation format conversion
• Separate training pipelines for each model

Training Notebooks

All training notebooks are located in the training_notebooks/ folder:

training_notebooks/
├── 1_soham_yolo11.ipynb      # YOLO11 model training
├── 2_chaitanya_yolo8.ipynb   # YOLOv8 model training
└── 3_divyanshu_cnn.ipynb     # CNN model training

To Rerun the Notebooks:

1. Download the datasets manually from the links provided above
2. Update dataset paths in each notebook
3. Execute cells sequentially

---

Project Structure

Driver-Monitoring-System/
│
├── data/                    # Raw and processed datasets
├── demo-images/             # Sample output images
├── driver_monitoring/       # Core ML logic and utilities
├── driver_monitor_sys/      # Experimental modules
├── models/                  # Trained YOLO & CNN models
├── public/                  # Public frontend assets
├── resources/               # Supporting resources
├── runs/                    # YOLO auto-generated outputs
├── static/                  # CSS, JS, images
├── templates/               # HTML templates (Flask)
├── testing_notebooks/       # Model testing and evaluation
├── training_notebooks/      # Model training notebooks
│
├── app.py                   # Main Flask application
├── convert.py               # Model conversion utilities
├── requirements.txt         # Python dependencies
├── vercel.json              # Deployment configuration
└── README.md                # Project documentation

---

Installation & Setup

Prerequisites

• Python 3.8 or higher
• pip package manager
• Git

Installation Steps

1. Clone the repository:

   git clone https://github.com/AlbatrossC/Driver-Monitoring-System
   cd Driver-Monitoring-System

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   python app.py

4. Access the application:

   Open your browser and navigate to http://localhost:5000

---

Usage

Browser-Based Demo

1. Visit the live demo
2. Upload an image
3. The system will automatically detect and highlight unsafe behaviors

Local Deployment (All Models)

Run the Flask application locally to access all three models including the CNN classifier:

python app.py

---

Team & Contributions

This project was developed as part of the Advanced Course on Green Skills and Artificial Intelligence under the Skills4Future Program.

Team Members

Name	Contribution
Chaitanya kulkarni	YOLOv8 Model Training & Dataset Preparation
Soham Jadhav	YOLO11 Model Training & Data Annotation
Divyanshu Mishra	CNN Model Training & Image Classification
Anurag Pawar	Backend Logic & System Integration

Program Details

• Program: Advanced Course on Green Skills and Artificial Intelligence
• Organized by: Edunet Foundation, AICTE, Shell India Markets Pvt. Ltd.
• Mentor: Professor Sarthak Narnor

---

Model Performance

Model	Accuracy	Speed	Deployment
YOLOv8n	High	Fast	✅ Browser
YOLO11n	Very High	Fast	✅ Browser
Custom CNN	Low	Moderate	🖥️ Local Only

---

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Acknowledgments

Special thanks to:

• Edunet Foundation, AICTE, and Shell India Markets Pvt. Ltd. for organizing the Skills4Future Program
• Professor Sarthak Narnor for mentorship and guidance
• The open-source community for providing datasets and tools