Geoparky | Devpost

Add car functionality - stores parking location and tells the number of danger zones around you
Home Map View
Profile Page for a particular user
Streamlit Danger zone visualization (Bangalore specific due to data limitation)
Redis Cloud analytics

Inspiration

Imagine a world where parking frustrations are a thing of the past. No more wandering aimlessly through vast parking lots, no more desperate searches on crowded city streets. We envision a future where finding your parked car is as simple as a few taps on your smartphone. Our project is about more than just convenience; it's about reclaiming your time and reducing stress. Whether you're in a bustling metropolis or a quiet suburban neighborhood, our solution will be your trusted companion, ensuring you always know where you park.

What it does

The application has 3 features - add car, delete car, and find car. Add car presents you with the option to save your current parking location in the application. Find car provides you a google maps route between your current location and parked location. Delete car removes your parked location from the application once you have removed it from parking. There is also a global map view to see the parking density in your locality as well as all over the world. A profile page is also provided to ensure that you can see your last parked locations. Add car can also tell the number of danger zones within a 1km radius of your parked location (Bangalore-specific due to data limitation).

How we built it

The application frontend and db logic are built using react and supabase - an open-source Firebase alternative. The danger spot visualization is done using Streamlit and Python. We used sci-kit learn to train a K-means clustering schema and visualize the centers using Streamlit. We also stored the danger points on Redis Geospatial and exploited its LRU cache to retrieve the nearest danger zones quickly and efficiently.

Challenges we ran into

We ran into multiple challenges in deploying the Streamlit application as we were deploying it from a private repo as opposed to a public one. Nested files were not accessible and dependencies had to be configured manually. Furthermore, we also ran into multiple issues with the node-redis client which was solved by switching the code to FastAPI and redis-python.