OME-NGFF community call: 2025-02-21

Please paste this into the Zoom chat as new people join:

Code of conduct

The OME community is open to everybody and built upon mutual respect. Please take the time to review the code of conduct below.

https://github.com/ome/.github/blob/master/CODE_OF_CONDUCT.md

Preliminary agenda (not for notes)

| discourse thread | zulip thread |

• Updates (30 minutes) - slides
  • Zarr v3
  • OME-Zarr RFCs
  • OME-Zarr 1.0
  • Community (calls, etc.)
• Open discussion (30 minutes)
  • … Please list items here before the meeting 📝 …

OME-NGFF community call (17:00 CET; 21 Feb 2025)

"User registration" Session 2

Session 2 Notes

• Erick: https://www.bioimagingnorthamerica.org/events/ome-2025-community-meeting/

• Mark: 1.0 prerequisites

  • Zarr v3 complaints
    • "Under construction" needs to go away. Yup. See https://github.com/zarr-developers/zarr-specs/pull/330
    • Cross testing has posed some issues
      • e.g. ts writes v2 such zstd isn't readable by other implementations like zarr-python
      • Norman: Yes! definitely put on the list 1.0
    • zstd becoming a prominent codec (zarr-python v3 default)
      • how do we enumerate codecs in OME-Zarr
      • Norman: see https://github.com/zarr-developers/zarr-extensions
  • Ken: if "more reliable then extensions"?
    • Norman: there are different goals. mine:
      • every version has been breaking so far
      • want to get to the point where we can not break for a long time
      • KH:
        Image Not Showing Possible Reasons

        • The image file may be corrupted
        • The server hosting the image is unavailable
        • The image path is incorrect
        • The image format is not supported

        Learn More →
        for something more stable.
    • Dan:
      • 1. sample implementation that's really robust will go a long way to make it feel serious.
        • An "implementation" should codify all the MUSTs, SHOULDs, MAYs etc. from the spec..
      • 2. metadata: official way of dealing with more metadata, even if it is just one field that says, "put it here"
        • or a statement that "you're on your own"
      • 3. thinking a lot about collections, distinction between zarr groups and sources from different filesystems. (perhaps completely required in 1.0)
    • David:
      • reference implementation in https://ome-zarr-models-py.readthedocs.io
        • v0.4 done; working on v0.5. Love that to be a community project.
        • outcome of the last hackathon
    • Norman:
      • discussed collections at the hackathon. working on stories to motivate
      • working towards a RFC. definitely part of 1.0
      • get in touch on zulip, etc.
      • likely to get a WG/meeting series started soon
    • Luca: regarding transforms
      • with Wouter, contributing an implementation to ome-zarr-models-py
      • working other than displacements… (currently without pydantic)
      • please say if development needs prioritizing say the word. (April? Yes!)
    • Abbas
      • as an end user, very interested in transforms. (webknossos, etc.)
      • need transforms for stitching
        • Luca (in chat): https://github.com/multiview-stitcher/multiview-stitcher
          • Joel (in chat): That library is super cool, we've had a lot of fun using it
      • don't want to duplicate the large datasets.
      • top priority
    • John
      • thanks Dan + others for reviewing RFC transforms
      • currently working on what parts are necessary vs optional for implementation
      • currently working on complete set of examples
    • Josh
      • Displacement field might be something in its own page so fully optional
    • Joel
      • Community calls are helpful for people to catch up on what's going into RFCs

• Meetings

  • Erick: need to change how people view this meeting - wasn't attending because viewed it as heavily technical spec meeting
  • Virginie taking over community-facing meeting series

• Misc

  • Abbas: infrastructure based on zarr
    • Working on new parallelization for chunk-wise compute, but having trouble getting in touch with dask developers
      • Davis: DON'T set Dask chunk size to be same as Zarr chunk size. Overhead of 1 millisecond for every task.
        • Dask chunks should not be as small as Zarr chunks
      • Davis: dask forums aren't very lively, but ping people directly (e.g. Davis) on github repo
        • zulip is also a good option for these kinds of questions
  • Will: what's the Python reference implementation for working with arrays? is everyone using different things?
    • Josh: likely a dozen different tools…
      • 2023 hackathon outcomes: https://forum.image.sc/t/outcomes-of-the-next-generation-bioimage-analysis-workflows-hackathon/88733
    • David: next steps need to be limited scope - started with validating metadata, before moving on to writing metadata
      • would be ok with adding support for reading/writing data to ome-zarr-models-py
        • https://ome-zarr-models-py.readthedocs.io/en/stable/
  • Ken: Java for getting users to adopt… what's the latest on java development for ome-zarr?
    • John: progress on Zurich hackathon last year for reading/writing zarr 3 from java. can expect Fiji support for that "soon"
      • smaller community working on java than python, and Fiji maintenance is a big task