search

nipy / nitransforms

a standalone fork of nipy/nibabel#656
https://nitransforms.readthedocs.io
MIT License
28 stars 17 forks source link

FIX: Accept double precision when reading ANTs' ``.mat`` affines #102

Closed oesteban closed 3 years ago

oesteban commented 3 years ago

It is a tradeoff, after all this is a 16 parameter matrix so float64 is not a big deal. However, I think I recall @hjmjohnson saying that it is critical that affines preserve the precision on some occasions at our transforms meeting in Boston two years ago.

effigies commented 3 years ago

I meant a tradeoff in the sense of losing compatibility with other tools, and if we should try to determine whether we need double and when we can fall back to float. I'm not concerned about doubling storage for affines.

oesteban commented 3 years ago

I meant a tradeoff in the sense of losing compatibility with other tools

No, perhaps the opposite. Also, this format is kind of deprecated for ANTS - at some point they will move completely to HDF5. If they haven't done so yet it is probably because of the low priority of the move.

pep8speaks commented 3 years ago

Hello @oesteban! Thanks for updating this PR.

Line 136:1: W293 blank line contains whitespace