search

AllenInstitute / ophys_nway_matching

Other
4 stars 0 forks source link

PSB-147 Fix issues #73

Closed aamster closed 1 year ago

aamster commented 1 year ago
aamster commented 1 year ago

@morriscb the plots look identical. There is 1 change though to the structural similarity numbers displayed on the plots. In the version of skimage installed in this build (0.21) the api forces you to pass in data_range argument to the structural_similarity function. In the version of skimage installed in the currently deployed nway matching (0.18) it guesses the data range, and there's a note about it usually being wrong, which is why it forces you to pass it in the newer api https://scikit-image.org/docs/stable/api/skimage.metrics.html#skimage.metrics.structural_similarity . The data_range is 1.0 because the images have been normalized to [0,1], so I think the numbers in this PR are correct whereas before they were wrong.

morriscb commented 1 year ago

Cool. Looks good. After/Before merging this, you should ping the Ophys processing mailing list to give them a heads up regarding the "slight" change to the comparison plots. I'll slack you the mailing list.