archiving alignments

[martinschorb]

Hi,

what strategy do you currently advise to archive an alignment procedure when publishing a dataset.

I would like to make the aligned volume available using MoBIE/Neuroglancer but also the raw tile data. In order to enable reproducing the alignment from the raw data, what Render metadata should be archived with it?

Do you just keep the tilespecs of the aligned stack after the solve? Or would you also consider the SIFT matches relevant?

Thanks!

[trautmane]

Hi @martinschorb,

I typically keep a binary export of the mongodb data store ( using mongodump ) because that is a relatively easy way to archive/restore things without needing compute resources.

As long as the code/process you use to import tile specs and generate matches is version controlled, the minimum you'd need to keep track of is:

• the code repository versions,
• the source image locations and
• stage/scope information. To reproduce results, you just need to re-run everything.