wsi-fixtures

A small, public-license corpus of whole-slide imaging (WSI) test files. Used by wsitools and other WSI tooling for CI integration tests.

What's included

Redistributable fixtures from the openslide-testdata corpus (CC0), derivatives generated by WSI tooling (CC0), and original work contributed under CC-BY-4.0 (see manifest.json for the authoritative list):

File	Size	Format	License
fixtures/svs/CMU-1-Small-Region.svs	1.85 MB	Aperio SVS	CC0-1.0
fixtures/svs/239551.svs	25.40 MB	Aperio SVS (multi-level, JPEG-tiled; thumbnail/label/overview)	CC-BY-4.0
fixtures/svs/590_crop_jp2k_imagescope.svs	1.39 MB	Aperio SVS (JPEG 2000 tiled)	CC-BY-4.0
fixtures/svs/590_crop_jpeg70_imagescope.svs	2.59 MB	Aperio SVS (JPEG q70 tiled)	CC-BY-4.0
fixtures/svs/590_crop_lzw_imagescope.tif	8.80 MB	TIFF (LZW tiled, ImageScope export)	CC-BY-4.0
fixtures/svs/590_crop_none_imagescope.tif	22.65 MB	TIFF (uncompressed tiled, ImageScope export)	CC-BY-4.0
fixtures/ndpi/CMU-1.ndpi	188.86 MB	Hamamatsu NDPI	CC0-1.0
fixtures/philips-tiff/Philips-4.tiff	277.51 MB	Philips TIFF	CC0-1.0
fixtures/generic-tiff/CMU-1-Small-Region.stripped.tiff	1.85 MB	Generic TIFF (stripped, derived)	CC0-1.0
fixtures/ome-tiff/CMU-1-Small-Region.ome.tiff	1.89 MB	OME-TIFF (derived)	CC0-1.0
fixtures/bif/Ventana-1.bif	216.85 MB	Ventana BIF	CC0-1.0
fixtures/cog-wsi/CMU-1-Small-Region_cog-wsi.tiff	1.89 MB	COG-WSI (derived)	CC0-1.0
fixtures/dicom/3DHISTECH-JP2K/	2.66 MB	DICOM WSM (JPEG 2000, derived)	CC0-1.0
fixtures/dicom/3DHISTECH-HTJ2K/	4.74 MB	DICOM WSM (HTJ2K, derived)	CC0-1.0
fixtures/dicom/scan_621_grundium_dicom/	340.16 MB	DICOM WSM (JPEG-baseline)	CC-BY-4.0

Each fixture has a sibling <name>.LICENSE with the verbatim license deed and <name>.PROVENANCE recording source, SHA-256, and retrieval date. Most fixtures are CC0; scan_621_grundium_dicom, the four 590_crop_* ImageScope-export crops, and 239551.svs are original work contributed under CC-BY-4.0 and require attribution (see their .LICENSE). DICOM fixtures are directories of .dcm instances (the sidecars sit next to the directory).

How CI consumers use this

Each GitHub Release hosts per-format tarballs; releases are cumulative, so the latest tag holds the whole corpus. The current release is v8:

• svs.tar (~63 MB)
• ndpi.tar (~189 MB)
• philips-tiff.tar (~278 MB)
• generic-tiff.tar (~2 MB)
• ome-tiff.tar (~2 MB)
• bif.tar (~217 MB)
• cog-wsi.tar (~2 MB)
• dicom.tar (~348 MB)

Each tarball preserves the in-repo directory structure: tar xf svs.tar extracts to svs/CMU-1-Small-Region.svs. Consumers should pull only the tarballs they need (--pattern) rather than the whole corpus.

Example GitHub Actions step:

- name: Download fixtures
  run: |
    mkdir -p sample_files
    cd sample_files
    gh release download v5 --repo wsilabs/wsi-fixtures \
      --pattern 'svs.tar' --pattern 'ndpi.tar'   # only what you need
    for t in *.tar; do tar xf "$t" && rm "$t"; done
  env:
    GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Releases are immutable. Pin a specific tag (v1, v2, ...) for reproducibility.

Adding a fixture

1. Audit licensing. Capture verbatim license text in fixtures/<format>/<name>.LICENSE and SPDX identifier in manifest.json. If the license doesn't allow redistribution, do not add the fixture.
2. Write fixtures/<format>/<name>.PROVENANCE with source URL, retrieval date, SHA-256, citation.
3. Add the binary at fixtures/<format>/<name>.
4. Update manifest.json.
5. Run scripts/verify.sh — must pass.
6. Run scripts/pack.sh — produces dist/<format>.tar.
7. Publish a NEW release tag (e.g. v2) with the rebuilt tarballs. Do not modify existing release assets — releases are immutable.

License

The repo metadata, scripts, and documentation are MIT-licensed (see LICENSE). Each fixture binary is governed by its own <name>.LICENSE file.