🚗 NoNapDrive

Real-Time Driver Drowsiness Detection using AI

NoNapDrive is an intelligent real-time driver monitoring system that detects drowsiness and microsleep using computer vision and deep learning.
It analyzes eye behavior through a webcam and triggers severity-based alerts to help prevent fatigue-related road accidents.

---

✨ Key Features

• 🎥 Live webcam-based monitoring
• 👁️ Eye Aspect Ratio (EAR) calculation using MediaPipe Face Mesh
• 🧠 LSTM-based deep learning model (3-class classification)
• 🟢 Alert | 🟡 Drowsy | 🔴 Critical (Microsleep)
• ⏱️ Temporal validation to reduce false positives
• 🔊 Escalating audio alerts based on severity
• ⚙️ Fully configurable using YAML
• 💻 Software-only solution (no additional hardware required)

---

🧩 System Workflow

1. Webcam captures real-time video frames
2. Face landmarks are detected using MediaPipe
3. Eye Aspect Ratio (EAR) is computed per frame
4. Sequential EAR data is passed to an LSTM model
5. Driver state is classified into 3 levels
6. Corresponding alert is triggered

---

🛠️ Tech Stack

Component	Technology
Programming Language	Python 3.10
Deep Learning	TensorFlow (tf.keras)
Computer Vision	MediaPipe, OpenCV
Web Interface	Streamlit
Utilities	NumPy, PyYAML

---

📁 Project Structure

NoNapDrive/
│
├── app/
│ ├── pycache/
│ ├── alert.py # Audio alert handling logic
│ ├── app.py # Streamlit application entry point
│ ├── features.py # EAR feature extraction
│ ├── model.py # LSTM model loading & inference
│ └── state.py # Driver state management
│
├── assets/
│ ├── alarm.wav # Critical alert sound
│ └── beep.wav # Mild warning sound
│
├── models/
│ └── drowsiness_lstm_3class_tf.keras # Trained LSTM model
│
├── notebooks/
│ └── drowsiness_model.ipynb # Model training notebook
│
├── config.yaml # Thresholds & runtime configuration
├── requirements.txt # Python dependencies
├── README.md # Project documentation
└── .gitignore

---

▶️ Getting Started

1️⃣ Create Virtual Environment (Python 3.10 recommended)

py -3.10 -m venv venv310
venv310\Scripts\activate

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Train the Model (One-Time)

jupyter notebook

Run:

notebooks/drowsiness_model.ipynb

After training, place the model file inside:

models/drowsiness_lstm_3class_tf.keras

4️⃣ Run the Application

streamlit run app/app.py

Open in browser:

http://localhost:8501

---

🚨 Alert Logic

🟢 refers Alert -> Normal monitoring 🟡 refers Drowsy -> Mild warning beep 🔴 refers Critical -> Loud alarm after time confirmation

⚙️ Configuration

All thresholds and runtime parameters can be tuned using:

config.yaml

This allows behavior changes without modifying code.

---

📌 Use Cases

1. Driver safety systems
2. Long-distance driving assistance
3. Academic research in computer vision
4. Fatigue detection systems

👤 Author

Anubhab Pradhan BE – Artificial Intelligence & Data Science CMR Institute of Technology, Bangalore

📜 License

This project is intended for academic and educational use only.