This repository contains the following:
- code used to run the experiments on the coarse annotations of tumor areas from the TCGA-BRCA dataset.
- Code used to do inference on the held-out test set.
- The 16x downsampled coarsely annotated tumor area image files.
- The 16x downsampled finely annotated tumor area image files.
- The .p file contains the 16x scaled down polygon annotations organized in a dictionary with two main keys: 'incl_vec' and 'excl_vec'. Each key holds a dictionary where TIFF image filenames map to lists. The 'incl_vec' key represents polygon annotations for areas in the images that should be included as important, while 'excl_vec' contains annotations for areas to be excluded.