DriveSafe: AI Driving Assistant

Revolutionizing Driving Safety with Deep Learning and OpenCV Implementations

Comment

Inspiration

After observing the community around us, we realized that there are greater than 6 million car accidents per year, and close to 23 percent of them end in fatalities. Car accidents are the second leading cause of death for teens, 87 percent of which involve distracted drivers. We created DriveSafe to tackle this problem and make an effort to reduce these numbers.

What it does

DriveSafe is an automated computer vision driving security assistant. It uses Deep Learning and Computer Vision to identify if and when a driver is distracted or feels sleepy, and accordingly alerts the driver and/or passengers.

How it works

The DriveSafe automated driving assistant was built through a 5-step cumulative process:

  1. A stable, face-monitoring video stream was set up with the use of a webcam

  2. The stream is fed to the laptop, which uses deep-learning facial recognition algorithms to detect the presence of the face in the stream

  3. Once a face is detected, the algorithm performs eye localization and draws eye-shaped bounds around the driver’s eyes

  4. An eye aspect ratio is determined and updated constantly based on the height of the top eyelid to the bottom eyelid. If the height difference ever crosses a certain threshold (reduces) for a given amount of time, the driver is determined to be distracted or sleepy

  5. Upon identification, a sound is played constantly until the driver is predicted to be focused

How We built it

DriveSafe was built using OpenCV, Python, deep learning facial recognition/object localization algorithms. The website was developed using HTML, CSS, and javascript.

Challenges I ran into

Implementing the face detection with eye detection made it so that there was a lot of lag between frames, leading to less consistent eye detection. We optimized the face detection to have it run a minimal amount of times, and the lag was fixed. A lot of the issues lay with consistency - we had to keep experimenting with parameters to make sure that it would work at least 90% of the time.

What's next for DriveSafe: AI Driving Assistant

Currently, DriveSafe can only be implemented with a webcam and computer--we only had a day to create it. We want to make it more convenient and make it a mobile application that can link with a webcam. In addition to functionality with webcams and Bluetooth webcams, the mobile application will utilize GPS libraries to determine if the driver is exceeding speed limits. This product will essentially eliminate common risks that drivers encounter.

Share this project:

Updates