A driver drowsiness detection system built using a CNN model

The model is trained using the Transfer Learning technique in the domain of Deep Learning. Transfer learning is a machine learning method where a model developed for a task is reused as the starting point for a model on a second task (i.e to retrain a pre-trained model). There are various classifiers in Transfer learning namely MobileNet, GoogleNet, ResNet, VGG16 etc. In our project, we have used the MobileNet model. MobileNet is a light-weighted model having fewer parameters as compared to others.

The dataset that we have used for the project is a MRL Eye Dataset. Download from here

The dataset contains eye images of 37 candidates. So for our project we created two separate folders namely Train_Dataset and Test_Dataset. In each of these folders , we created two more folders with the names Closed_Eyes and Open_Eyes. We kept 8000 images of closed and open eyes from the actual dataset into Train_Dataset and 2500 images in Test_Dataset.

• numpy
• pandas
• matplotlib
• tensorflow
• opencv
• flask

We have created a website as well for running the project using flask. If you want to run the model on your system then run the test1.py file. Firstly, to start with the project, the user first needs to go to the website and should click on Start button. Once the button is clicked, the webcam gets turned on & it starts detecting the face and eyes. After the detection of the eyes, it splits the eyes into left & right and it is then fed to the model. The model then checks whether the eyes are open or not. If the eyes are open, the status is shown as Active & if the eyes are closed, the status is shown as Sleep Alert and beeps for the user to become aware.