NewLabels#

class NewLabels[source]#

Detects labels that appear only in the test set.

Parameters
max_images_to_display_per_labelint , default: 3

maximum number of images to show from each newly found label in the test set.

max_new_labels_to_displayint , default: 3

Maximum number of new labels to display in output.

n_samplesOptional[int] , default10000

Number of samples to use for the check. If None, all samples will be used.

__init__(max_images_to_display_per_label: int = 3, max_new_labels_to_display: int = 3, n_samples: Optional[int] = 10000, **kwargs)[source]#
__new__(*args, **kwargs)#

Methods

NewLabels.add_condition(name, ...)

Add new condition function to the check.

NewLabels.add_condition_new_label_ratio_less_or_equal([...])

Add condition - Ratio of labels that appear only in the test set required to be less or equal to the threshold.

NewLabels.clean_conditions()

Remove all conditions from this check instance.

NewLabels.compute(context)

Calculate which class_id are only available in the test data set and display them.

NewLabels.conditions_decision(result)

Run conditions on given result.

NewLabels.config([include_version, ...])

Return check configuration (conditions' configuration not yet supported).

NewLabels.from_config(conf[, version_unmatch])

Return check object from a CheckConfig object.

NewLabels.from_json(conf[, version_unmatch])

Deserialize check instance from JSON string.

NewLabels.greater_is_better()

Return True if the check reduce_output is better when it is greater.

NewLabels.initialize_run(context)

Initialize run before starting updating on batches.

NewLabels.metadata([with_doc_link])

Return check metadata.

NewLabels.name()

Name of class in split camel case.

NewLabels.params([show_defaults])

Return parameters to show when printing the check.

NewLabels.reduce_output(check_result)

Reduce check result value.

NewLabels.remove_condition(index)

Remove given condition by index.

NewLabels.run(train_dataset, test_dataset[, ...])

Run check.

NewLabels.to_json([indent, include_version, ...])

Serialize check instance to JSON string.

NewLabels.update(context, batch, dataset_kind)

No additional caching required for this check.

Examples#