histolab

You can extract training tiles by loading a whole-slide image into a Slide object and applying a tiling strategy like RandomTiler, GridTiler, or ScoreTiler to automatically detect tissue and extract actionable tiles for deep learning datasets.

Digital pathology dataset preparation supports RandomTiler for exploratory sampling, GridTiler for systematic coverage, and ScoreTiler for quality-ranked extraction, allowing flexible training data curation from whole-slide images.

Tissue mask detection in whole-slide image processing filters out non-tissue regions using pluggable mask pipelines, ensuring that only relevant tissue areas are processed during tile extraction and dataset curation.

Yes, whole-slide image processing supports tissue characterization and analysis across both H&E and IHC slides, enabling balanced training dataset creation and exploratory tile sampling for various histology workflows.

End-to-end whole-slide image processing is automated through pluggable architecture classes like Slide, TissueMask, and tiling strategies, handling tissue detection, tile extraction, and dataset curation without manual intervention.

No external dependencies are required to run whole-slide image tiling, as the processing architecture operates independently with its own Slide, TissueMask, and tiling classes to generate actionable tiles.