FireSight

Inspiration:

We have family members in California, and we've had personal experience with a wildfire in our community in Texas. Because of this, we were inspired to try to resolve this problem by creating a DL model that predicts if a wildfire will occur given certain weather conditions.

What it does:

The model has been trained on Zenodo and uses data given by the user to extrapolate and make a prediction on whether those weather conditions are suitable for a wildfire to occur.

How we built it:

We got data from a database known as Zenodo, which gave us data from California on weather conditions for wildfires.

Challenges we ran into:

We had trouble optimizing the DL model but we research regularization and hyper parameter training to further optimize the DL model. We also had trouble putting in the data for the website as it was having lots of issues with the parameters, but we were eventually able to pull through.

Accomplishments that we're proud of:

We were able to successfully create an DL model with almost 80% accuracy that is linked to a website with in-depth customizable visualizations in under 12 hours.

What we learned:

We learned how to collect data and how to filter what has been collected and put it in to an actual model. We also got extensive knowledge on making the ML model and the website and trying to get those 2 things to work in harmony.

What's next for FireSight:

We plan on training the model to be more accurate and also add more parameters to the model, so that we can get more accurate results on when a wildfire can occur. We plan on making this a worldwide app where anyone can use it and put their own data in too see whether a wildfire could happen in their area or not.

Built With

• imblearn
• keras
• matplotlib
• numpy
• pandas
• plotly
• python
• scikit-learn
• streamlit
• tensorflow
• zenodo