Each year, over 50 million people are affected by car accidents, and over 1.35 million people have lost their lives, and many more millions have been injured. A study has also shown that 90% of countries lose 3% of their GDP due to road accidents. If we look closer into the statistics of road accidents we notice that 35% of accidents on roads are due to drowsy drivers. Along with this, 93% of fatalities are due to people driving on unsafe roads.
What it does
This webapp makes roads safer by using machine learning. This webapp detects car crash hotspots, and is trained on data used for 350,000 other large counties/cities. The webapp detects if the road you want to go on is safe or not. Along with this the webapp detects if you are drowsy by tracking eye movement. It then plays a loud alarm sound so you will automatically be alerted.
How we built it
I have built SafeNavigate using openCV and 4 machine learning models --> Logarithmic Regression(94%), Decision Tree (96% accuracy), Random Forest Classifier (97% accuracy), KNN (95% accuracy). I tested all these models and chose the Random Forest Classifier due to its accuracy. I have used python, cockroach DB for the backend, html, css, and js for frontend
Challenges we ran into
One Challenge I ran into was designing the random Forest classifier. It was my first time using it, and althought it was a fun experience it was a challenging one
Accomplishments that we're proud of
I am proud of the following things
- designing a fully functional app
- using not one not two but 4 ml models
- utilizing openCV and math to detect eyes and blinking
- training on a large dataset without using google AutoML vision or any other tool that could help
- first time using a cockroachdb database
What's next for SaveNavigate
I would like to partner with someone in my community to export this app in our county. From there I would like to pitch this to my senator who will be willing to try this out and see a change in results.