Campfire Burn Severity Prediction

This repo contains the code (Jupyter Notebook) for a research project I was involed in which studied the use ofconvolutional neural networks to predict the burn severity of affect human structures from the 2018 Campfire using post-wildfire aerial imagery.

Jupyter Notebook

The Jupyter Notebook contains code for building and creating the dataset (images and labels), as well as train/test split, model building, training, and test evaluation. The notebook has headers for each of these sections. Some extras are included at the bottom of the notebook.

Sample dataset

The sample dataset provides a CSV as well as one example of each image for each class.

Conda environment

Conda can be used to install the necessary packages to run the code in this project. This can be done by running the command:

conda env create -f environment.yml -n campfire

And then activating the environment:

source activate campfire

Since n-dimensional arrays are not currently supported in Imbalanced-Learn, I created a forked version which can be installed by running the command:

pip install git+https://github.com/gustaver/imbalanced-learn.git