ROA (Road Operational Assistant)

Leveraging machine learning and computer vision balanced with algorithmic precision to bring the safety features of self driving vehicles to everyone on the road.

Comment

Inspiration

Our mission statement was to leverage artificial intelligence, computer vision, and innovative road threat detection to make traveling safer. We wanted to make self-driving car technology and machine learning available to everyone on the road, ultimately saving lives.

What it does

Our project analyzes historical accident data and patterns, traffic density, weather, and more to alert and warn the driver of potential risks ahead of time. ROA detects if the road you are driving on and the conditions of the road are safe for driving on; it prescribes driving suggestions and alerts drivers of unsafe conditions

How we built it

We compiled 14 million data points and trained a computer vision binary classification algorithm based on thousands (only used partial data for timing purposes) of pictures of safe and unsafe roads or areas. We created an algorithm that focuses on the road itself to balance the unpredictability of the model with the advantages of structured analysis.

Challenges we ran into

We also struggled to deal with balancing time constraints and training an effective and viable machine learning algorithm. We worked around this problem by training on only 10,000 data points of our homemade 14 million row-long dataset. We ran into challenges coming up with a way to demo our project since we couldn't get in a car. We solved this problem by presenting our project in parts so that we could demonstrate each section of our software's effectiveness. Along the way, we also struggled to find data sets of non-accident zones. We combatted this issue by utilizing random sampling to identify what a normal road looks like.

Accomplishments that we're proud of

We created a 14 million-row-long dataset of images to use in training our computer vision binary-classification model. We are extremely proud of our front-end continuity use, which is toggled if the website is accessed on a laptop. We are also proud of the algorithms we used to look at the weather, location, wind speed, precipitation, ice conditions, temperature, speed, and speed limit in an attempt to balance out the unpredictable nature of machine learning.

What's next for ROA (Road Operational Assistant)

We aim to make this into a business that saves lives and helps individuals map according to live safety hazards.

Share this project:

Updates