otx.algo.utils.xai_utils#

Utils used for XAI.

Functions

`convert_maps_to_dict_all`(saliency_map)	Convert salincy maps to dict for TargetExplainGroup.ALL.
`convert_maps_to_dict_image`(saliency_map)	Convert salincy maps to dict for TargetExplainGroup.IMAGE.
`convert_maps_to_dict_predictions`(...)	Convert salincy maps to dict for TargetExplainGroup.PREDICTIONS.
`dump_saliency_maps`(predict_result, ...[, weight])	Sumps saliency maps (raw and with overlay).
`postprocess`(saliency_map, output_size)	Postprocess single saliency map.
`process_saliency_maps`(saliency_map, ...)	Perform saliency map convertion to dict and post-processing.
`process_saliency_maps_in_pred_entity`(...)	Process saliency maps in PredEntity.
`set_crop_padded_map_flag`(explain_config, ...)	If resize with keep_ratio = True was used, set crop_padded_map flag to True.

otx.algo.utils.xai_utils.convert_maps_to_dict_all(saliency_map: list[ndarray]) → list[dict[Any, array]][source]#: Convert salincy maps to dict for TargetExplainGroup.ALL.

otx.algo.utils.xai_utils.convert_maps_to_dict_image(saliency_map: list[ndarray]) → list[dict[Any, array]][source]#: Convert salincy maps to dict for TargetExplainGroup.IMAGE.

otx.algo.utils.xai_utils.convert_maps_to_dict_predictions(saliency_map: list[ndarray], pred_labels: list | None) → list[dict[Any, array]][source]#: Convert salincy maps to dict for TargetExplainGroup.PREDICTIONS.

otx.algo.utils.xai_utils.dump_saliency_maps(predict_result: list[OTXBatchPredEntitiesSupportXAI], explain_config: ExplainConfig, datamodule: EVAL_DATALOADERS | OTXDataModule, output_dir: Path, weight: float = 0.3) → None[source]#: Sumps saliency maps (raw and with overlay).

otx.algo.utils.xai_utils.postprocess(saliency_map: ndarray, output_size: tuple[int, int] | None) → ndarray[source]#: Postprocess single saliency map.

otx.algo.utils.xai_utils.process_saliency_maps(saliency_map: list[ndarray], explain_config: ExplainConfig, pred_labels: list | None, ori_img_shapes: list, image_shape: tuple[int, int], paddings: list[tuple[int, int, int, int]]) → list[dict[str, ndarray | Tensor]][source]#: Perform saliency map convertion to dict and post-processing.

otx.algo.utils.xai_utils.process_saliency_maps_in_pred_entity(predict_result: list[MulticlassClsBatchPredEntity | MultilabelClsBatchPredEntity | HlabelClsBatchPredEntity | DetBatchPredEntity | InstanceSegBatchPredEntity], explain_config: ExplainConfig, label_info: LabelInfo | int | list[str]) → list[MulticlassClsBatchPredEntity | MultilabelClsBatchPredEntity | HlabelClsBatchPredEntity | DetBatchPredEntity | InstanceSegBatchPredEntity][source]#: Process saliency maps in PredEntity.

otx.algo.utils.xai_utils.set_crop_padded_map_flag(explain_config: ExplainConfig, datamodule: OTXDataModule) → ExplainConfig[source]#: If resize with keep_ratio = True was used, set crop_padded_map flag to True.