Motivation

Driver drowsiness is a leading cause of accidents, with over 1 in 5 fatal crashes involving drowsy driving. This project aims to leverage machine learning to build a robust model capable of generalizing to new instances. We explore Convolutional Neural Network (CNN) techniques to develop a model that detects drowsiness in drivers, thereby aiming to reducing accidents caused by driver fatigue and improving public safety.

Project Structure and Contents

This repository contains the implementation of various CNN models for driver drowsiness detection. Key components include:

• DriverDrowsiness/: Main directory containing all project files.
• Presentation.pdf: Project presentation highlighting objectives and key findings.
• DriverDrowsiness/207_Driver_Drowsiness_CNN.ipynb: Jupyter notebook detailing model development.
• DriverDrowsiness/data/: Directory for dataset storage and preprocessing outputs.
• DriverDrowsiness/images/: Images referenced in the notebook and for documentation.
• DriverDrowsiness/models/: Saved model checkpoints.
• DriverDrowsiness/plots/: Visualizations generated during model training and evaluation.
• DriverDrowsiness/results/: Output files capturing model performance metrics.
• DriverDrowsiness/src_bkp/: Backup files and additional resources.
• Annotations_sample_data_Drowsy.pdf and Annotations_sample_data_NonDrowsy.pdf: Sample annotations for drowsy and non-drowsy driver images.

Presentation Highlights

For a deeper understanding of the project objectives, methodology, and results, please refer to Presentation.pdf. Highlights include:

• Statistics on drowsy driving accidents in the United States.
• Objectives of the CNN model exploration.
• Insights into model performance and generalization capabilities.

Notebook Details

The primary notebook 207_Driver_Drowsiness_CNN.ipynb covers:

• Dataset exploration and preprocessing steps.
• Implementation and evaluation of various CNN architectures.
• Analysis of experimental results and model performance metrics.

Getting Started

To replicate the experiment or further explore the models:

1. Clone this repository to your local machine.
2. Refer to 207_Driver_Drowsiness_CNN.ipynb for detailed code and methodology.
3. Utilize the provided datasets in DriverDrowsiness/data/ for training and evaluation.

Conclusion

This project underscores our commitment to using advanced AI techniques to address real-world challenges like driver safety. We invite you to explore the code, datasets, and results to gain insights into our approach and findings.