Skip to content

cellpheno/nis: add test tile + models#2136

Open
Chrisa142857 wants to merge 1 commit into
nf-core:modulesfrom
Chrisa142857:cellpheno-nis
Open

cellpheno/nis: add test tile + models#2136
Chrisa142857 wants to merge 1 commit into
nf-core:modulesfrom
Chrisa142857:cellpheno-nis

Conversation

@Chrisa142857

Copy link
Copy Markdown
Member

Test data for the cellpheno/nis module (nf-core/modules#12179).

  • data/imaging/segmentation/cellpheno_nis/cellpheno_nis_tile.zip — a real P4 partial-brain crop: 30 z-slices, 384×384 (2D OME-TIFF), ~4 MB.
  • data/imaging/segmentation/cellpheno_nis/cellpheno_nis_models.zip — the TorchScript NIS models (cuda:0 set), ~64 MB.

Both are zipped so the module test extracts them via the UNZIP subworkflow (dir inputs), following the numorph/3dunet pattern. Verified: NIS segments 351 nuclei on this crop.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this file is too big, we usually aim for as small as possible, target <1MB. Any chance you can shrink this and the second file?

@Chrisa142857

Copy link
Copy Markdown
Member Author

Thanks @mashehu! Addressed:

  • Shrank the tile to a smaller real P4 crop (16 slices, 224×224) — now ~700 KB (<1 MB).
  • Dropped the models zip entirely: the TorchScript weights (~64 MB) are large and available via gdown from the source G-drive, so they don't belong in test-datasets — the module test fetches them at runtime instead of bundling them here.

So this PR now adds a single ~700 KB file. NIS still segments 221 nuclei on the crop.

@mashehu mashehu left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

awesome and thanks for cleaning up the git history!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants