README.md

Accuracy Assessment Tools

The simplest way to run this is to open the Jupyter Notebook AccuracyAssessmentTools.ipynb in Google Colab and follow the instructions in the notebook. To open the notebook in Colab, open the file in github, then click the open in colab button at the top of the file.

Requires pandas and numpy.

Input data is expected to be in a table with a column containing the mapped values for each point and a column containing the reference value for each point. Each row of the table should be a separate point.

Additional data such as mapped proportions or strata proportions should be given as a dictionary whose keys match the labels used in the columns e.g. if your class labels are 0, 1, 2, ... then the keys of the given dictionary should also be 0, 1, 2, ...

Running each file will print test values from the corresponding paper to verify that the math is being done properly, e.g.

python naive_acc_assessment.py

Usage Example

After running pip install . from the root directory:

import pandas as pd
from acc_assessment.olofsson import Olofsson

data = pd.read_csv("/path/to/file/containing/assessment/points.csv")

mapped_areas = {"forest": 200000, "deforestation": 1000}

assessment = Olofsson(
    data, "name of map value col", "name of ref value col",
    mapped_areas)

print(assessment.overall_accuracy())
print(assessment.users_accuracy("forest"))