DrowsinessDetection - WIP

This project aims to detect drowsiness by analyzing eye states (open or closed) using a deep learning model. The model is trained on grayscale eye images using transfer learning with MobileNetV2.

Datasets

1. MRL Eye Dataset Description: Contains images of eyes in grayscale. Purpose: Used for training and validating the drowsiness detection model.

Model Architecture

Base Model: MobileNetV2 (pretrained on ImageNet) Input Size: 224x224 RGB images Output: Binary classification (Open/Closed eyes) Training Epochs: 20 epochs with callbacks for checkpoint saving and learning rate adjustment.

Downloading Datasets

MRL Eye Dataset: Download from Kaggle and unzip it.

Model Training

Data Preparation: The dataset is processed and augmented using a custom data generator that converts grayscale images to RGB. Model Definition: MobileNetV2 is used with additional dense layers for binary classification.

Training:

The model is trained with early stopping and learning rate reduction strategies.

Testing the Model

Preprocessing: The eye images are resized and normalized. Prediction: The trained model predicts the eye state (Open/Closed) and annotates the results. Visualization: The results are saved and displayed.