I was carpooling from Waterloo to Toronto, then my friend accelerated to 160km/h right towards a truck. Me: "sh**********t". Friend wakes up: "oh, hey what's up". Me: "That was close"
So I want to prevent traffic accidents resulted from careless drowsy driving like this. One of the way to do it is to use technology to detect drowsy driving, and uses human nature (love of free/reduced stuff) to prevent it.
What it does
This is an Android application that tracks users' eyes for eye lids movement to determine if the driver is drowsy and offers recommendations(coupons) for good economical deals(such as coffee) at nearby stores. Thus save drivers' lives, give then rewards for being responsible drivers and help local business grow.
How I built it
Android Framework OpenCV to scrape the users' face and locate eyes TensorFlow to adapt the users' face JobService to manage deals popup Google Map for the background Navigation
Challenges I ran into
I was thinking about make an all in one app that is like good, but this Hackathon is really short: 12 hrs of pure work time. So I managed to delegate tasks such as opening map/navigation to Google, which is actually the google's recommended way of forming implicit intent
Accomplishments that I'm proud of
I have always wanted to try something with Vision and machine learning, however, my previous attempts has been all failures as I either somehow messes up the data sets or the trained data pool. This time, things did not go wrong.
What I learned
Horizontal System design, where: instead of having an almighty app that does it all, have it does a core function, and delegate other tasks to other apps that does it the best. A key is to make the transition as smooth as possible.
What's next for Seyet
Create a network layer to process locations through an api: because we will have more vendors than just one business.