/classification-testing

Primary LanguagePython

classification-testing

A scratch repo for classification of regions in lungmap images.

=======

Purpose

The purpose of this repository is to start with the assumption that the immunofluorescence confocal images have adequately had instance segmentation applied, meaning that a contour of an anatomical structure is available. This repo tests difference classification techniques for labeling the segmented regions and also segmenting cells within these larger anatomical entities and ideally tie these to an ontology which can tie the color of the cell to the type of cell segmented.

Usage

To run the entire pipeline:

  1. Run data_generator.py: This will produce the training and test data in a model_data directory.
  2. Run xception_transfer.py: This will train the Xception pipeline, and can take quite a while.
  3. Run test.py: This will evaluate the test image for accuracy, saving the results in results.csv in the model_data directory.