We were heavily focused on the machine learning aspect and realized that we lacked any datasets which could be used to train a model. So we tried to figure out what kind of activity which might impact insurance rates that we could also collect data for right from the equipment which we had.
What it does
Insurity takes a video feed from a person driving and evaluates it for risky behavior.
How we built it
We used Node.js, Express, and Amazon's Rekognition API to evaluate facial expressions and personal behaviors.
Challenges we ran into
This was our third idea. We had to abandon two major other ideas because the data did not seem to exist for the purposes of machine learning.