Dataset.autotag_items method to export the individual items and their Autotag scores from an existing Autotag of an indexed Dataset.
import nucleus client = nucleus.NucleusClient("YOUR_SCALE_API_KEY") dataset = client.get_dataset("YOUR_DATASET_ID") dataset.autotag_items("YOUR_AUTOTAG_NAME", for_scores_greater_than=-0.5)
You can optionally supply a lower bound on the Autotag score to increase the precision of the exported items; default is 0. All Autotag scores are normalized to the range [-1, 1].
This will return a set of items tagged by the Autotag represented by their reference ID with their Autotag score. The dataset item can then be fetched using the
Dataset.autotag_training_items method to export the user-selected positives used to train the Autotag (from refinement).
You can export items for both image and object level Autotags.
Updated about 1 year ago