Inspiration
We spoke to amag about their challenge to solve the problem of parking prediction in Zurich.
What it does
It accepts a navigation destination and it will return you with a list of parking spots in the area of the destination, sorted by their closeness to your destination and the likelihood of them being free at the time of arrival.
How we built it
Android interface is built in Java using the TomTom SDK. Server is hosted on Google Cloud and it's built with Python Flask RESTful APIs The ML model was studied in a Python using scikit learn run on a Google Colab notebook, our best model is a RandomForestRegressor.
Challenges we ran into
- Scraping external data
- Compute for maths intensive tasks
- Learning TomTom SDK (poor documneted)
Accomplishments that we're proud of
it works... kind of
What we learned
new technologies and SDKs (android development, colab) data processing and ml solutions take a very very long time without GPUs.
What's next for Zuripark
include more data points (eg weather) expand to other cities
Log in or sign up for Devpost to join the conversation.