There's a lot of people that want to make a difference on the way we affect our environment.

We often want to make changes that will benefit us and the environment, but where do we start? Will our actions really make a difference?

(C)arbon can help to clear the air around these issues.

What it does

Ever wonder how much environmental impact something has? How does your computer affect the environment? Or that banana from Ecuador?

Well, (C)arbon can tell you just that. Simply install the app and point your phone's camera at just about anything and view it's Co2e score!

Now everyone can make conscious and meaningful decisions to help protect our planet!

How I built it

The Android app utilizes TensorFlow Lite for object recognition and displays Co2e data collected from a database of compiled research.

Google Cloud Storage was utilized to store model training assets and training on Docker.

Challenges I ran into

The unholy sins of outdated, poorly written documentation. Every step we took was met with an error that set us back hours. But alas, we move forward.

Accomplishments that I'm proud of

The most divine fortitude that we had to build to plow through the never ending errors and stack traces that lurked behind every successful command line instruction.

What I learned

  • Tinkering with Docker and GCP services to train TensorFlow models.
  • How TensorFlow models are trained.
  • Sometimes Google just can't help you find all the errors you get.
  • TensorFlow documentation is truly something else.

What's next for (C)arbon

Train models to:

  • recognize just about anything made by people
  • display more accurate Co2e based on an item's origination
  • Co2e produced by organic life (e.g. cats, dogs, whales, etc.)
  • potential Co2e removed by plants
  • "ranked" bounding boxes (e.g. red for high, green for low Co2e)
Share this project: