kaishi.image.filters.similar
¶
Class definition for filtering similar images in a dataset.
Module Contents¶
-
class
kaishi.image.filters.similar.
FilterSimilar
¶ Bases:
kaishi.core.pipeline_component.PipelineComponent
Filter near duplicate files, detected via perceptual hashing (using the imagehash library).
-
__call__
(self, dataset)¶ Perform filter operation on a specified dataset.
- Parameters
dataset (
kaishi.image.dataset.ImageDataset
) – dataset to perform operation on
-
configure
(self, perceptual_hash_threshold=3)¶ Configure the filter with a perceptual hash threshold.
- Parameters
perceptual_hash_threshold (int or float) – threshold for determining whether or not images are similar (> are deemed not similar)
-