Q about numpy files created by evaluation method

[paul-reiners]

This is a question, not an issue.

I understand that the evaluation method in evaluate.py saves a number of numpy files:

• dice. npy
• hausdorff.npy
• hausdorff_95.npy

and so on. I understand that these are two-dimensional arrays where the rows correspond to segmentation files and the columns correspond to the region numbers for anatomical regions.

How can I determine the row names (i.e., the segmentation files) and the column names (i.e., the region IDs)? I want to load these numpy files and do analysis using them by creating perhaps Pandas data frames or something similar.

Are the implicit row and column headers constant across each of these files (for a given call to evaluation)?

I'm guessing these are defined by the values of label_list and path_segs, but I wanted to be sure.

[BBillot]

It's actually the opposite, rows correspond to segmented structures (given in the order of label-list), and columns to the images that are segmented (given in the order when sorting the files contained in the provided seg_dir).
Sorry I understand this was not well explained in the docstring of evaluate.evaluation

[paul-reiners]

When you say "label-list", what do you mean exactly? Is this a file created somewhere by SynthSeg? Or a variable created somewhere by SynthSeg? Where, exactly?