[Question] Is DS5020 mandatory, even for phase-from-BF? (also: hard to see error)

[tlambert03]

after chatting with @mattersoflight, I'm trying to see if I can get "Phase From BF" working... (I don't have a meadowlark at the moment). I have a comment and a question:

I hit this error, "No config group contains channels" ... but because that code is run inside of this other
try/except block here, that except clause swallows the error and simply shows "failed" in the button, it was hard to figure out what was going wrong. I had to edit the source code to see the actual error message.

... so quick comment is to fix that code so that the user can actually see the error (bonus points if you use napari notifications system so that they see the error in the GUI, rather than looking back at the terminal for print statements)

question: After commenting out that KeyError ... I was able to at least get it to say connected. Should that continue to work for me? or will i run into more issues if I don't have a meadowlark DS5020

side note: I'm on https://imagesc.zulipchat.com/ if it would be easier to chat in realtime

[talonchandler]

Thanks for reporting!

DS5020 is not needed for phase-from-BF, and you can skip the "Connect to MM" step.

For phase-from-BF I recommend acquiring a tiff, converting to zarr (with iohub convert), then using the CLI (documentation here) to reconstruct with a config file. We also have a GUI wrapper around the CLI in the "Reconstruct" tab, which can help out with iterating on a reconstruction.

Apologies for WIP docs (see also, #483, cc @mattersoflight @amitabhverma).

[tlambert03]

thanks! will try.

Ultimately, my other interest here was in trying to help replace pycromanager with pymmcore-plus... without breaking anything. So I was orienting myself to the code and dependency structure a little bit.

It would also be cool to be able to integrate some of the functionality here into pymmcore-gui (as something like a plugin, with widget, essentially) without needing to bring in napari and/or pycromanager. #goals 😄

[talonchandler]

Excellent #goals!

We currently only depend on pycromanager for calibrating the liquid crystals and acquiring backgrounds (the "Acquisition / Reconstruction" tab will be deprecated), so I expect the pycromanager -> pymmcore-plus change to be achievable.

I'll leave this issue open for further discussion, and because the hard-to-see error is still an issue.

[tlambert03]

For phase-from-BF I recommend acquiring a tiff, converting to zarr (with iohub convert), then using the CLI (documentation here) to reconstruct with a config file. We also have a GUI wrapper around the CLI in the "Reconstruct" tab, which can help out with iterating on a reconstruction.

Is the io mandatory? Do you have any functions that just take in a memory buffer (numpy array or otherwise), with associated parameters? Or is writing to disk absolutely necessary?

[talonchandler]

Do you have any functions that just take in a memory buffer (numpy array or otherwise), with associated parameters?

You bet! Here's a folder of example scripts that demonstrate all of our reconstructions with arrays in and arrays out. This script is a good place to start for phase-from-BF.