Vehicular Facial Recognition

Inspiration

Leaving my car key in the car and had to embarrassingly call AAA.

What it does

Using facial recognition to unlock a car door and with future implementation start a car.

How we built it

We used openCV library to build our machine learning software to register and verify the user's face. We set up the raspberry pi, connected the hardware component and our code to put together the project. We installed 2 buttons, a servo motor, and a display. One button registers the users face, and the other verifies an user's face. The motor simulates unlocking a car door and triggering the ignition. Displays outputs the result after comparing the scanned face against the registered face.

Challenges we ran into

The code was unable to differentiate Mike and Ryan, they're both Asians with somewhat similar facial feature. Finding a LCD package that works on a Raspberry pi 5, memba forge was tricky to add libraries to. The current supply to the servo motor was not constant, causing jittering before it was supposed to move.

Accomplishments that we're proud of

Being able differentiate similar looking people. Fixing the power issues by adding a transistor. Finding libraries that are compatible with the Raspberry Pi. Working around the display to make it compatible with the Raspberry Pi.

What we learned

Newest hardware does not mean it is the best option, as Pi 5 is not compatible with a lot of our components, we had to work our way around it. Computer vision library can take a while to install. For a facial recognition to be accurate enough, we need sufficient amount of training data, face distance and low comparison threshold.