Occurrences of many road accidents because of minor issues like sleepiness and fatigue. There were many cases of accidents that took place in remote areas and the person was not able to get help sooner. We want to address this problem with our model.

What it does

Our model detects the face of the driver and alerts the person if he/she closes eyes for more than 3 seconds. In addition, we designed an alert system that sends a notification along with their current location to emergency contacts if the person doesn't wake up for more than 10 seconds.

How we built it

We used OpenCV and dlib libraries to detect the facial features of the driver from the video stream. We took every frame of the video as an input image. We used 'Eye Aspect Ratio(EAR)' to determine the condition of the driver i.e, whether he is awake or sleepy. If the person closes eyes for more than 3 seconds we trigger a thread that starts an alarming sound. If the person doesn't wake for more than 10 seconds we used Twilio client to send a personal message along with location to his emergency contacts.

Challenges we ran into

We are new to image processing and we had to do research and learn how each library works and then implement it. We tried to develop an ios app for sending realtime location to contacts in case the person doesn't wake up for a long time.

Accomplishments that we're proud of

We are proud to be able to finish most of our application which we didn't expect to do so.

What we learned

We learned new concepts like video processing and face detection and also ios app development.

What's next for Drowsiness detection

In the event of failure to wake the driver, we thought we can at least save others traveling near this vehicle by alerting them. We can integrate this model with the vehicle and make all light flicker as an indicator of danger. Currently, we are able to send location through IP which is not accurate. Using an app, we can send a precise location and users can set their own alarm tone and configure emergency contacts.

We added a demo video in this link

Built With

Share this project: