bzamith / IJCNN2023PantanalFireDetection

Pantanal Fire Detection project, as published in IJCNN 2023

Pantanal Fire Detection

About

• Authors: Bruna Zamith Santos (programmer), Ricardo Cerri, Marcelo Narciso, Balbina Soriano, Diego Furtado
• This is the codebase for the paper "A New Time Series Framework for Forest Fire Risk Forecasting and Classification" - IJCNN 2023.

Install

Using virtualenv (PREFERRED)

virtualenv -p /usr/bin/python3 env
source env/bin/activate
pip3 install -r requirements.txt
deactivate

Local

sudo python3 setup.py clean --all install

Data for Fitting the Model

There are 2 required data files:

• One with climatic data. Must contain the columns Year, Month, Day, T (temperature), P (precipitation), UR (relative humidity) and V (wind speed).
• One with hotspot detection. Must contain the column Date, representation the dates where a hotspot was identified.

To define which data file will be used, you must:

1. Place the data file in folder under /datasets that represents this data source. Either /datasets/hotspot_data or /datasets/climatic_data
2. Define the file names in /config/general_settings.py

Data for Predictions

There is 1 required data file:

• One with climatic data, with at least X days, in which X is equal to OBSERVATION_WINDOW in /config/forecast_settings.py

To define which data file will be used, you must:

1. Place the data file in folder under /datasets/prediction_data
2. Define the file name in /config/general_settings.py

Settings

You can provide custom settings for all files in /config folder

Run

chmod +x run.sh

# Don't forget to activate the virtualenv, if you are using one!
# First, build the code:
./run.sh build

# Then, train the different models:
./run.sh fit

# Finally, predict the risk rate:
./run.sh predict

# To run everything at once, just run:
./run.sh all