patch_predictor¶

Defines the PatchPredictor engine for patch-level inference in digital pathology.

This module implements the PatchPredictor class, which extends the EngineABC base class to support patch-based and whole slide image (WSI) inference using deep learning models from TIAToolbox. It provides utilities for model initialization, post-processing, and output management, including support for multiple output formats.

Classes:
  • PatchPredictor:

    Engine for performing patch-level predictions.

  • PredictorRunParams:

    TypedDict for configuring runtime parameters.

Example

>>> images = [np.ndarray, np.ndarray]
>>> predictor = PatchPredictor(model="resnet18-kather100k")
>>> output = predictor.run(images, patch_mode=True)

Classes

PatchPredictor

Patch-level prediction engine for digital histology images.

PredictorRunParams

Parameters for configuring the PatchPredictor.run() method.