train_test_validation#

Module containing the train test validation check in the vision package.

Classes

HeatmapComparison

Check if the average image brightness (or bbox location if applicable) is similar between train and test set.

ImageDatasetDrift

Calculate drift between the entire train and test datasets (based on image properties) using a trained model.

ImagePropertyDrift

Calculate drift between train dataset and test dataset per image property, using statistical measures.

NewLabels

Detects labels that appear only in the test set.

PropertyLabelCorrelationChange

Return the Predictive Power Score of image properties, in order to estimate their ability to predict the label.

TrainTestLabelDrift

The TrainTestLabelDrift check is deprecated and will be removed in the 0.14 version.

LabelDrift

Calculate label drift between train dataset and test dataset, using statistical measures.