AlertDrive: Driver Drowsiness Detection

A comparative analysis project for detecting driver drowsiness from video using multiple machine learning models.

Project Overview

This project implements and compares multiple approaches for driver drowsiness detection from video data:

• Baseline Models: Existing models from research papers and open-source repositories
• Custom Model: A newly implemented model for comparison
• Comparative Analysis: Performance evaluation across different datasets and metrics

Project Structure

AlertDrive/
├── data/                   # Dataset storage
│   ├── raw/               # Original video datasets
│   ├── processed/         # Preprocessed data
│   └── interim/           # Intermediate processing
├── src/                   # Source code
│   ├── models/           
│   │   ├── baseline/     # Existing model implementations
│   │   └── custom/       # Custom model implementation
│   ├── preprocessing/    # Data preprocessing
│   ├── evaluation/       # Evaluation metrics
│   ├── visualization/    # Plotting utilities
│   └── utils/            # Helper functions
├── notebooks/            # Jupyter notebooks
│   ├── exploratory/     # Data exploration
│   └── experiments/     # Model experiments
├── trained_models/      # Saved model weights
├── results/             # Experiment results
│   ├── metrics/        # Performance metrics
│   ├── figures/        # Visualizations
│   ├── reports/        # Analysis reports
│   └── comparisons/    # Comparative analysis
├── configs/            # Configuration files
├── tests/              # Unit tests
├── docs/               # Documentation
└── logs/               # Training logs

Setup Instructions

1. Create Virtual Environment

python -m venv venv

2. Activate Virtual Environment

Windows:

venv\Scripts\activate

Linux/Mac:

source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

Key Features

• Reproducibility: Properly implement and evaluate existing models
• Dataset Diversity: Test on multiple datasets with varying characteristics
• Comparative Analysis: Consistent evaluation metrics (accuracy, F1, precision, recall)
• Comprehensive Results: Detailed performance comparison and insights

Evaluation Metrics

• Accuracy
• Confusion Matrix
• ROC-AUC

Contributors

Sanjana Afreen Adrita Ahsan Nishat Tasneem Oishee Nafisa Tasnim Rodoshi Jarin Subha Sneha