Insurance companies have access to mostly negative data: (crashes, tickets, and more) leaving some drivers without the opportunity to prove themselves. We wanted to develop a win-win program to reduce distracted driving by incentivizing responsible drivers with lower insurance rates while providing insurance companies with data to improve their insight accuracy.

What it does

WatchDog allows drivers to self-report their focused driving while avoiding excessive intrusion. The mobile app captures a video recording of the person behind the wheel as they drive. After the trip ends, the driver can choose whether or not they want to send this video to WatchDog where it is then processed by machine learning and assigned a specific score. Drivers can view their progress overtime and improve their score before sending in their data for discounted insurance rates.

How I built it

The mobile app was build using React Native. Our backend used Flask and nginx for networking. Image recognition was accomplished using a CNN build with PyTorch and trained with the StateFarm distracted drivers dataset.

Challenges I ran into

Getting large amounts of data to be efficiently sent to the backend.

Accomplishments that I'm proud of

The CNN was able to achieve an accuracy of 87% with limited training. If given more time to train, the accuracy could be even better.

Share this project: