One of my interests is science-based worldbuilding, a big part of being creating plausible geographies based on real science. I decided to adapt my interest in the area towards real-world climate sciences.

What it does

This is an AI that takes in rainfall, temperature, and altitude data for a certain region, and trains it against satellite color data of said region to project what an area would look like from space given arbitrary rainfall/temperature/altitude data.

How we built it

I used Python Tensorflow Keras to create a triple-input convolutional neural network, taking in greyscale GeoTIFF files, and matching them against a RGB PNG output. Data was sourced from WorldClim, ETOPO1, and NASA Blue Marble.

Challenges we ran into

I initially planned on working with a partner, but that ended up producing nothing for the better part of the first day due to timezone differences.

What's next for ClimateCNN

Actually finish it lmao

Built With

Share this project: