Real-time Driver Drowsiness Detection

This is a Python application that uses neural networks and computer vision to detect whether a driver is drowsy or not. The application utilizes non-intrusive methods to accurately determine driver state in real-time and generate effective and acceptable warnings to increase driver alertness and safety.

Prerequisites

To run this application, you need to have the following installed:

• Python
• Tensorflow
• NumPy
• matplotlib
• face_recognition
• PyObjC
• playsound

You can install NumPy, matplotlib, face_recognition, PyObjC, and playsound using pip:

pip install numpy matplotlib face_recognition PyObjC playsound

To install TensorFlow, you can follow the installation instructions on the TensorFlow website: https://www.tensorflow.org/install

Usage

To use the real-time drowsy detection webcam application:

1. Start the Jupyter Notebook server by executing the following command in a terminal:

   jupyter notebook
   

2. Once the Jupyter Notebook interface is open, navigate to drowsy_detection_application.ipynb and open it.
3. Run all the cells in the notebook.

Model Training

Eye Aspect Ratio Classification

To classify eye state based on Eye Aspect Ratio (EAR):

1. Start the Jupyter Notebook server by executing the following command in a terminal:

   jupyter notebook
   

2. Once the Jupyter Notebook interface is open, navigate to eye_aspect_ratio.ipynb and open it.
3. Run all the cells in the notebook.

Convolutional Neural Network (CNN) Model

To run the CNN model:

1. Start the Jupyter Notebook server by executing the following command in a terminal:

   jupyter notebook
   

2. Once the Jupyter Notebook interface is open, navigate to cnn_model.ipynb and open it.
3. Run all the cells in the notebook.

Transfer Learning Model

You can either run a saved model or train a new model.

Running a saved model

1. Run transfer_learning.py.
2. Type 1 at the prompt to run the saved model.

Training a new model

1. Run transfer_learning.py.
2. Type 2 at the prompt to train a new model.

Authors

This application was created by Evangeli Silva [email protected], Andrew Okoro [email protected], Jahnavi Bonagiri [email protected] and Tobin Cherian [email protected] as part of the ICSI 531 course at the University at Albany.