Datasets:

---

nvidia

/

OpenH-RF

Like 17

Follow

NVIDIA 68.7k

Tidying the corpus documents ahead of the release. Nothing about the reconstructions changes except where noted.

Pipelines

• num_patches removed (25 files) — zea patches the beamforming grid itself now, at a fixed patch size.
• TUe-Cardiac written in block style (4 files) — same values, just readable.
• twente-cavitation now names its beamformer. It was falling back to delay-and-sum while the submission uses minimum variance, so this one does change the image.
• waterloo-largeartery gains the parameters: dynamic_range block it was missing; tumunich uses log compression instead of an unresolvable power_compress.

Missing configs added — oslo/parameters.yaml, which maps each acquisition to one of the pipeline_*.yaml files already here, and unc-openpros/pipeline.yaml, which was absent entirely.

Python files removed — the corpus ships data cards and pipelines; the code lives in the companion repo. That clears twente-microbubblesim/{migrate_custom_names,utils}.py and tue-cardiac/reconstruct.py.

Line endings normalized to LF (14 files) and a text=auto eol=lf rule added, so a Windows-authored file no longer reads as wholly modified. It sits above the LFS rules, which keep -text on the binaries.

Cross-link to the companion repo added to the dataset card.

Worth reviewing commit by commit: the line-ending commit rewrites whole files, so the combined diff hides the real changes.

@tristan-deep why are some READMEs completely modified? Did you modify spacing or is there something else going on? I cannot tell...

LGTM