Inspiration: 1997 Bond movie; Tommorow Never Dies. Remote controlled cars! The rise of self-driving cars promises to make cars safer and give drivers the time to do other things while in transit. However, what if you actually enjoy to drive yourself? How can we make car enthusiasts the world over benefit from this marvel of engineering, while not taking away the thing they love to do most.

What it does: A dashboard application for your smart car. Check input/output of cameras, control it remotely, summon it automatically.

How I built it

The car uses an NVIDIA Jetson as brain, which is actually a Linux computer running on the car. Our computer vision algorithms are implemented in OpenCV, A Flask server on the Jetson transmits image data to our web app, while receiving commands.

Challenges I ran into

We wanted to make the car fully autonomous, which was a lot more work than expected. The size of the car ruled out all pre-existing ML models. Building a robust and dependable lane detection system was a huge endeavour by itself, while we had intended it to be only a part of our system.

The cars driving system broke down, which was a bit sad because we couldn't demo the driving and control part.

Accomplishments that I'm proud of:

*Building a smart car on a cutting edge platform *Implementing usable lane detection in OpenCV, without relying on ML models *

What I learned

Go easy on the ML when developing an MVP Dont do hardware projects on hackatons because they tend to break down.

What's next for Dude, where's my car going?

Trains. I like trains.

Share this project:

Updates