Pothole Detection

Inspiration

My car is equipped with ultrasonic sensors, cameras, and software that's built by engineers smarter than I've ever imagined but it still can't avoid the most obvious potholes on the road in Seattle so I built one myself to see how challenging it would be.

What it does

The project takes a video file (which could be easily replaced with a live video feed) and highlights the potholes that it detects in the frame

How we built it

I used tensorflow, a dataset from Kaggle, cv2 to access the video feeds, and a lot of debugging and errors to get through everything and get an MVP running.

Challenges we ran into

Tensorflow isn't actually that stable on newer Python versions so I had to downgrade to version 3.11 instead of using version 3.12.

Accomplishments that we're proud of

I'm honestly just super happy that I was able to make this within the span of an entire day. I didn't think it was possible between the time spent at the career fair and dinner but I got through it.

What we learned

Tensorflow is finicky at best and could be hard to work on. The error descriptions in Tensorflow aren't the easiest to understand so a lot of googling was required on my end.