DriverSense

Inspiration

We were inspired by using some form of AI model to do photo recognition to make a game similar to the 2008 show 'Hole in the Wall' but after hearing one of the tracks we were able to adapt the idea to DriverSense, a program that would guide and teach better and safer driving practices.

What it does

It is comprised of two main programs that take a video feed from a camera. One of these uses an AI model to detect if a mobile phone is visible - which would indicate unsafe driving - and prompts the user to put their phone down after a short amount of time of it being visible. The other program performs facial recognition and identifies the current position of the user's face. If it is detected that the user's face is pointing away from the camera - essentially looking away from the road - then they are alerted to focus on what is ahead of them.

How we built it

Matthew used some face tracking Good phone detection idk Oh yeah

Challenges we ran into

While trying to create a program to detect mobile phones, we attempted to create our model and even trained one ourselves to do this. However, despite the good dataset used for this, the model produced by this method was very inaccurate at identifying mobile phones in pictures. This meant we had to search for other methods of finding phones in images, which is why we turned to yolo11.

Accomplishments that we're proud of

Even though it may not have contributed to the final result, creating and training an AI model is still an impressive feat we are all proud of.

What we learned

To search for pre-existing models instead of attempting to train our own especially in time sensitive situations as a large amount of time was spent setting up the model and then training it. For time limited events like Hackathon, pre-existing models should be used due to their high accuracy and built upon with other features.

What's next for DriverSense?

If DriverSense were to be developed further, it could potentially be improved to include tracking for other objects that shouldn't be seen while driving, such as drinks to ensure users are detected being dangerous. Moreover, more facial tracking could be implemented to detect dangerous driving, such as searching for two eyes to be visible at all times to ensure the user does not close them and potentially fall asleep while driving.