You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am trying to create N5 datasets for stitching/alignment with BigStitcher, N5-Spark, etc. Currently building the dataset via Fiji in a headless format, but it is exceedingly slow. My experience is that operating with n5-spark is substantially faster, so I am looking for ways to leverage this library.
Do you anticipate adding the ability to create multichannel/multi-timepoint N5 datasets with n5-spark anytime in the future? Would you recommend something else?
In the long-run, I'm hoping to register the time series via iterative closest point as well... As always, thank you for your help.
Best,
Kevin
The text was updated successfully, but these errors were encountered:
I am trying to create N5 datasets for stitching/alignment with BigStitcher, N5-Spark, etc. Currently building the dataset via Fiji in a headless format, but it is exceedingly slow. My experience is that operating with n5-spark is substantially faster, so I am looking for ways to leverage this library.
Do you anticipate adding the ability to create multichannel/multi-timepoint N5 datasets with n5-spark anytime in the future? Would you recommend something else?
In the long-run, I'm hoping to register the time series via iterative closest point as well... As always, thank you for your help.
Best,
Kevin
The text was updated successfully, but these errors were encountered: