Nuclei number of the Kumar dataset.

[windygooo]

The number of nuclei in the downloaded Kumar dataset is 16966, which is different from the reported 21623 in the paper.

cell_count = 0
for root, _, files in os.walk('raw_data/kumar'):
    for file in files:
        if file.endswith('mat'):
            data = scipy.io.loadmat(os.path.join(root, file))
            cell_count += data['inst_map'].max()
print(cell_count)

[simongraham]

Have you ensured that you are counting across train, validation and test sets?