MapBiomas Accuracy Source Code
This repository organizes the MapBiomas accuracy codes.
Requisites:
-
Python 3.7 or above
-
scikit-learn library for Python
-
R 3.6.2 o above
-
R Studio 1.1.2.5 or above
-
tidyverse , ggplot2 , stringr , reshape2 , shiny , htmlwidgets , highcharter and dplyr packages for R
Recommendations:
-
For Windows, install Miniconda - Python 3.7 or above and add it to the system variable PATH like:
- PATH = C:\ProgramData\Miniconda3; C:\ProgramData\Miniconda3\Library\bin; C:\ProgramData\Miniconda3\Scripts;
-
Install scikit-learn library using conda install -c intel scikit-learn on system terminal (windows prompt)
How to use
1. Export reference and classification matrix from Google Earth Engine (GEE)
Copy and paste the code src/estimates_codes/1_export_gee_input.js on your GEE code editor and click on Run. After that, click on Task and run every task.
2. Getting Accuracy Assessment information with Python
Open your system terminal and run the following line like python3 src/estimates_codes/2_accuracy_estimates.py <INPUT_DIR> <OUTPUT_DIR>. Enter the directory address of the exported files as INPUT_DIR and the output files as OUTPUT_DIR. Sit in a comfortable chair, grab a book, a coffee and wait ... this process can take a while.
When the process ends, run the code 3_format_toShinny_for_Unix.sh (For Windows: 3_format_toShinny_for_Windows.bat) inside the output directory. This code prepares the file names and directory hierarchy for the Shiny reading format.
One last change needs to be made. Download the Sublime code editor, drag and drop all .CSV files from the lv1, lv2 and lv3 output directories, then press Ctrl + Shift + F and replace all the "NA" for NaN (without quotes).
3. Visualize accuracy information using R/Shiny
To visualize the accuracy information that you have process you need to:
- Copy the directories lv1, lv2 and lv3 to the directory src/visualization_codes/integracao/
- Open the src/visualization_codes/App.R code with RStudio
- Click on Run App (at the top center of the screen)