hemit_test_no_overlap_no_empty_ids.txt lists the 292 HEMIT test sample IDs
used for the paper evaluation. The list contains identifiers only; it does not
contain HEMIT image data.
The original 1024x1024 tiles were extracted from WSIs at a 512-pixel stride.
To remove overlap, we retained tiles whose _patch_<row>_<column> indices are
both even. The six IDs in hemit_test_empty_excluded_ids.txt were subsequently
excluded as empty tiles.
To regenerate the released selection from a file containing all original test filenames, run:
python scripts/generate_hemit_test_split.py \
--all-test-ids /path/to/all_hemit_test_ids.txt \
--output /path/to/hemit_test_no_overlap_no_empty_ids.txtThe HEMIT dataset itself is governed by its original access and redistribution terms; users must obtain it through the dataset provider.