During the opening ceremonies, RBC brought up their challenge of data visualization. At first, we weren't sure what they meant by visualization - using image recognition or displaying data in a cool way, so we decided to do both
What it does
visiCar is a web app that lets users take a picture of a car that they are interested in. After detecting what car it is, visi tells you where you can go buy it based on your location. If you scroll further down, you'll see that we have a visual representation of what cars people are looking at. We think this data is important for car vendors, dealerships and buyers to see what is popular.
How we built it
Our app is built using Ruby on Rails. We used Google Cloud Platform to help us in detecting what car the user took a picture of. After the car model, company and year are found, we use Autodata's SOAPful API to get back the car. Then we use Autodata's RESTful API to get back where we can buy that car at a nearby location. Other services used include ChartKick to visualize data and bootstrap to make the app (a little) appealing.
Challenges we ran into
There were many challenges we ran into since it was our first time using image recognition. We didn't know if we should train our own model or use Google's model. Since we were looking to detect car models, we figured Google would do a much better job with the time that we had
Accomplishments that we're proud of
Successfully building our first Rails app at a hackathon was a big accomplishment for us. Our team is only 2 people so it's amazing how much we got done (and how little sleep we got)
What we learned
We learned a lot about making API calls in Ruby and the surprising amount of information there exists in the auto industry and pretty much every other industry.
What's next for visi-car
Maybe training our own model using tensorflow so we can have better predictions. Improving the map functionality.