search

dirac-institute / kbmod

KBMOD (Kernel-Based Moving Object Detection)
BSD 2-Clause "Simplified" License
46 stars 14 forks source link

Refactor the image-level meta data in WorkUnit #739

Closed jeremykubica closed 1 week ago

jeremykubica commented 2 weeks ago

As a step toward adding more provenance data in the workflow (specifically traveling through the WorkUnit), I regrouped a bunch of the data about constituent images into a single astropy Table. I think this makes it a little easier to track and reason about, but I could also make a case that this adds unneeded complexity. So let me know if you think this is worth it.

DinoBektesevic commented 2 weeks ago

Hah, funny, sorry to double dip but I believe these two PRs might be related: https://github.com/dirac-institute/kbmod/pull/738

If we're already collating into BinTables from Headers - de-duplicating things like time stamps, per-image WCS et. al. via imagecollection HDU could be helpful here.

jeremykubica commented 2 weeks ago

Hah, funny, sorry to double dip but I believe these two PRs might be related: #738

If we're already collating into BinTables from Headers - de-duplicating things like time stamps, per-image WCS et. al. via imagecollection HDU could be helpful here.

I agree that there is more cleanup we can do, especially with de-duplicating things. I was didn't want to make any changes to the serialized data in this PR though, since I don't want to introduce any incompatibilities with WorkUnits currently living on disk. I think we can do that in a future round.

I would need to talk through how to use the imagecollection HDU correctly.