geetools.classification
#
Module for classification tools.
Module Contents#
Functions#
|
Get accuracy from a truth image and a classified image. |
|
Get a class raster with the following classes. |
- geetools.classification.binaryMetrics(truth, classified, scale, region=None)[source]#
Get accuracy from a truth image and a classified image.
names from: https://en.wikipedia.org/wiki/Evaluation_of_binary_classifiers
- Parameters:
truth (ee.Image) – binary truth image. Only the first band will be used
classified (ee.Image) – the classification results. Only the first band will be used
scale (int) – the scale for the analysis
region (ee.Geometry) – the region for the analysis.
- Returns:
a dictionary with the following
TPR = True Positive Rate, recall or sensitivity TNR = True Negative Rate, selectivity or specificity PPV = Positive Predictive Value or precision NPV = Negative Predictive Value FNR = False Negative Rate or miss rate FPR = False Positive Rate or fall-out FDR = False Discovery Rate FOR = False Omission Rate TS = Threat score ACC = Accuracy BA = Balanced Accuracy F1 = F1 score :rtype: ee.ConfusionMatrix
- geetools.classification.binaryRasterAccuracy(truth, classified, region=None)[source]#
Get a class raster with the following classes.
band “classes”: 0: no change detected and no real change (true negative) 1: no change detected but real change (false negative) 2: change detected but not real change (false positive) 3: change detected and real change (true positive)
band “truth”: 0: no change 1: change
source: https://en.wikipedia.org/wiki/Evaluation_of_binary_classifiers
For both truth and classified image input, it only uses the first band, therefore there is no need to specify a band.
- Parameters:
truth (ee.Image) – binary image with ground truth. Only the first band will be used
classified (ee.Image) – classified binary image. Only the first band will be used
region (ee.Geometry) – region for clipping the truth and classified images