Beth Cimini

+1 to this being useful!

Aha! I'm not crazy. @shntnu ran the below in 2018 (found it in Slack, though it seems it may have been kicked off by an in-person conversation because the discussion...

So I downloaded one plate from the last round of JUMP pilots* (`s3://cellpainting-gallery/cpg0001-cellpainting-protocol/source_4/workspace/backend/2021_03_04_Stain5_CondA_PE_Standard/BR00120527/BR00120527.sqlite`) and ran the code below to load just the columns I wanted (since the whole SQLite is...

Maybe in #2129 when we moved to skimage watershed?

Nope, because at that point [centrosome was already using skimage](https://github.com/CellProfiler/centrosome/blob/8c5d136cc3de98fe9d469053a31cf62f2194a3f7/centrosome/watershed.py) . No obvious things come up when searching the skimage issues either. I'm stumped. Happy, suspicious, and stumped.

Per our conversation today - I "approve" in the sense that I think this was done correctly, but I think this module needs an overhaul (or at least, an evaluation...

Ok, so based on now going through this module more carefully (thank you @callum-jpg for making the visualizations, they were a big help!), I'm definitely sure we want to rip...

Thanks @callum-jpg ! >However, it's important to note that there is something extra going on behind the scenes in regional maxima. As you can see in the above image, some...

Cool, one more thing, then I think there's enough data to propose a way forward - can you generate the same image you just did, with no second mahotas grouping...

(maybe a lightly smoothed version of the intensity image, that is after all what we do in IDPrimary)