Park Better

Inspiration

Frequently when scurry about town (in our whips), we've noticed how atrocious Ottawa drivers are at parking. We wanted to devise a solution to publicly shame those park-ers so they will change their ways

What it does

Our app uses computer vision and AI to evaluate a users "parking job" based on our proprietary criteria to score between 0-100 based on photo input. Then, the score is uploaded and averaged amongst all of the users other parking jobs. You can also view a leaderboard of other users parking scores.

How we built it

We used Auth0 for our authentication paired with NeonDB for hosting our user data. We then used a DigitalOcean bucket for storing pictures.

Other technologies used Yolo Model - Car detection; GDino - Boundary detection for parking spaces; OpenCV - Parking line detection;

Stack: FastAPI React React Router TailwindCSS

Challenges we ran into

Computer vision within this scope was a new challenge. Learned a lot, and came in unaware of how to solve our designated problem. It was more difficult than we expected.

Filtering output from blackbox pretrained models made it difficult to control the system as we had little control of how the model actually worked.

Accomplishments that we're proud of

Having something to actually submit lol

What we learned

How to use cameras within a react app, how to connect Auth0 to a custom db, computer vision, learned a lot more about models and messing with parameters

What's next for Park Better

Public repo release after we rotate all the keys (I accidentally pushed the .env to a public repo... oops)

Built With

• auth0
• digitalocean
• fastapi
• gdino
• neondb
• opencv
• react
• tailwind
• yolo