Driver-monitering-system

Driver behaviour detecting app that warns if the behaviour is detrimental.

How it works:

The app starts of with the camera running and it looks for the eyes in the video.
It then looks for any distracting behaviour shown by the driver.
If the behaviour is detected to be bad it warns the driver.
The alert runs till the driver touches the screen and if it takes too long an sos is sent to his emergency contact number

Note:

The repo is divides into 3 files:

The alert folder contains the app that triggers the alarm by flashing the screen red and also makes the flashlight of the camera to blink.
The tfmobile folder consists of the android app that detects around 9 classes of distractions such as:
- safe driving
- texting - right
- talking on the phone - right
- texting - left
- talking on the phone - left
- operating the radio
- drinking
- reaching behind
- hair and makeup
- talking to passenger
- Here is the accuracy and loss of the cnn model used to build this application:
The vigilate folder contains the app that detects if the driver is sleeping or not.

example:

Installation:

First install openCv manager for android by using this link.
Next open the folder of the app you wish to use via android studio and you are good to go.

Important note:

Th three apps are now seperate apps, if you would like to contribute plases help by combining the three into one.

Credits:

Bharath Nair,
Steve Paul

References:

tensorflow for poets
OpenCV for android
The ML model for the tfmobile app