otx.api.utils.dataset_utils#

Dataset utils.

Functions

`add_saliency_maps_to_dataset_item`(...[, ...])	Add saliency maps (2D array for class-agnostic saliency map, 3D array or list or 2D arrays for class-wise saliency maps) to a single dataset item.
`contains_anomalous_images`(dataset)	Check if a dataset contains any items with the anomalous label.
`get_fully_annotated_idx`(dataset)	Find the indices of the fully annotated items in a dataset.
`get_global_subset`(dataset)	Extract a subset that contains only the global annotations.
`get_local_subset`(dataset[, ...])	Extract a subset that contains only those dataset items that have local annotations.
`non_linear_normalization`(saliency_map)	Use non-linear normalization y=x**1.5 for 2D saliency maps.
`split_local_global_dataset`(dataset)	Split a dataset into the globally and locally annotated subsets.
`split_local_global_resultset`(resultset)	Split a resultset into the globally and locally annotated resultsets.

otx.api.utils.dataset_utils.add_saliency_maps_to_dataset_item(dataset_item: DatasetItemEntity, saliency_map: List[ndarray | None] | ndarray, model: ModelEntity | None, labels: List[LabelEntity], predicted_scored_labels: List[ScoredLabel] | None = None, explain_predicted_classes: bool = True, process_saliency_maps: bool = False)[source]#: Add saliency maps (2D array for class-agnostic saliency map, 3D array or list or 2D arrays for class-wise saliency maps) to a single dataset item.

otx.api.utils.dataset_utils.contains_anomalous_images(dataset: DatasetEntity) → bool[source]#

Check if a dataset contains any items with the anomalous label.

Parameters:: dataset (DatasetEntity) – Dataset to check for anomalous items.
Returns:: True if the dataset contains anomalous items, False otherwise.
Return type:: bool

otx.api.utils.dataset_utils.get_fully_annotated_idx(dataset: DatasetEntity) → List[int][source]#

Find the indices of the fully annotated items in a dataset.

A dataset item is fully annotated if local annotations are available, or if the item has the normal label.

Parameters:: dataset (DatasetEntity) – Dataset that may contain both partially and fully annotated items.
Returns:: List of indices of the fully annotated dataset items.
Return type:: List[int]

otx.api.utils.dataset_utils.get_global_subset(dataset: DatasetEntity) → DatasetEntity[source]#

Extract a subset that contains only the global annotations.

Parameters:: dataset (DatasetEntity) – Dataset from which we want to extract the globally annotated subset.
Returns:: Output dataset with only global annotations
Return type:: DatasetEntity

otx.api.utils.dataset_utils.get_local_subset(dataset: DatasetEntity, fully_annotated_idx: List[int] | None = None, include_normal: bool = True) → DatasetEntity[source]#

Extract a subset that contains only those dataset items that have local annotations.

Parameters:

dataset (DatasetEntity) – Dataset from which we want to extract the locally annotated subset.
fully_annotated_idx (Optional[List[int]]) – The indices of the fully annotated dataset items. If not provided, the function will compute the indices before creating the subset.
include_normal (bool) – When true, global normal annotations will be included in the local dataset.

Returns:

Output dataset with only local annotations

Return type:

DatasetEntity

otx.api.utils.dataset_utils.non_linear_normalization(saliency_map: ndarray) → ndarray[source]#: Use non-linear normalization y=x**1.5 for 2D saliency maps.

otx.api.utils.dataset_utils.split_local_global_dataset(dataset: DatasetEntity) → Tuple[DatasetEntity, DatasetEntity][source]#

Split a dataset into the globally and locally annotated subsets.

otx.api.utils.dataset_utils.split_local_global_resultset(resultset: ResultSetEntity) → Tuple[ResultSetEntity, ResultSetEntity][source]#

Split a resultset into the globally and locally annotated resultsets.