Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
51
53
sq7383977_0003_0001_1.png,sq7383977_0003_0001_1.txt
sq7383977_0003_0001_2.png,sq7383977_0003_0001_2.txt
sq7383977_0003_0001_3.png,sq7383977_0003_0001_3.txt
sq7383977_0003_0001_4.png,sq7383977_0003_0001_4.txt
sq7383977_0003_0002_1.png,sq7383977_0003_0002_1.txt
sq7383977_0003_0002_2.png,sq7383977_0003_0002_2.txt
sq7383977_0003_0002_3.png,sq7383977_0003_0002_3.txt
sq7383977_0003_0002_4.png,sq7383977_0003_0002_4.txt
sq7383977_0003_0003_1.png,sq7383977_0003_0003_1.txt
sq7383977_0003_0003_2.png,sq7383977_0003_0003_2.txt
sq7383977_0003_0003_3.png,sq7383977_0003_0003_3.txt
sq7383977_0003_0003_4.png,sq7383977_0003_0003_4.txt
sq7383977_0004_0001_1.png,sq7383977_0004_0001_1.txt
sq7383977_0004_0001_2.png,sq7383977_0004_0001_2.txt
sq7383977_0004_0001_3.png,sq7383977_0004_0001_3.txt
sq7383977_0004_0001_4.png,sq7383977_0004_0001_4.txt
sq7383977_0004_0002_1.png,sq7383977_0004_0002_1.txt
sq7383977_0004_0002_2.png,sq7383977_0004_0002_2.txt
sq7383977_0004_0002_3.png,sq7383977_0004_0002_3.txt
sq7383977_0004_0002_4.png,sq7383977_0004_0002_4.txt
sq7383977_0004_0003_1.png,sq7383977_0004_0003_1.txt
sq7383977_0004_0003_2.png,sq7383977_0004_0003_2.txt
sq7383977_0004_0003_3.png,sq7383977_0004_0003_3.txt
sq7383977_0004_0003_4.png,sq7383977_0004_0003_4.txt
sq7383977_0005_0001_1.png,sq7383977_0005_0001_1.txt
sq7383977_0005_0001_2.png,sq7383977_0005_0001_2.txt
sq7383977_0005_0001_3.png,sq7383977_0005_0001_3.txt
sq7383977_0005_0001_4.png,sq7383977_0005_0001_4.txt
sq7383977_0005_0002_1.png,sq7383977_0005_0002_1.txt
sq7383977_0005_0002_2.png,sq7383977_0005_0002_2.txt
sq7383977_0005_0002_3.png,sq7383977_0005_0002_3.txt
sq7383977_0005_0002_4.png,sq7383977_0005_0002_4.txt
sq7383977_0005_0003_1.png,sq7383977_0005_0003_1.txt
sq7383977_0005_0003_2.png,sq7383977_0005_0003_2.txt
sq7383977_0005_0003_3.png,sq7383977_0005_0003_3.txt
sq7383977_0005_0003_4.png,sq7383977_0005_0003_4.txt
sq7383977_0006_0001_1.png,sq7383977_0006_0001_1.txt
sq7383977_0006_0001_2.png,sq7383977_0006_0001_2.txt
sq7383977_0006_0001_3.png,sq7383977_0006_0001_3.txt
sq7383977_0006_0001_4.png,sq7383977_0006_0001_4.txt
sq7383977_0006_0002_1.png,sq7383977_0006_0002_1.txt
sq7383977_0006_0002_2.png,sq7383977_0006_0002_2.txt
sq7383977_0006_0002_3.png,sq7383977_0006_0002_3.txt
sq7383977_0006_0002_4.png,sq7383977_0006_0002_4.txt
sq7383977_0006_0003_1.png,sq7383977_0006_0003_1.txt
sq7383977_0006_0003_2.png,sq7383977_0006_0003_2.txt
sq7383977_0006_0003_3.png,sq7383977_0006_0003_3.txt
sq7383977_0006_0003_4.png,sq7383977_0006_0003_4.txt
sq7383977_0007_0001_1.png,sq7383977_0007_0001_1.txt
sq7383977_0007_0001_2.png,sq7383977_0007_0001_2.txt
sq7383977_0007_0001_3.png,sq7383977_0007_0001_3.txt
sq7383977_0007_0001_4.png,sq7383977_0007_0001_4.txt
sq7383977_0007_0002_1.png,sq7383977_0007_0002_1.txt
sq7383977_0007_0002_2.png,sq7383977_0007_0002_2.txt
sq7383977_0007_0002_3.png,sq7383977_0007_0002_3.txt
sq7383977_0007_0002_4.png,sq7383977_0007_0002_4.txt
sq7383977_0007_0003_1.png,sq7383977_0007_0003_1.txt
sq7383977_0007_0003_2.png,sq7383977_0007_0003_2.txt
sq7383977_0007_0003_3.png,sq7383977_0007_0003_3.txt
sq7383977_0007_0003_4.png,sq7383977_0007_0003_4.txt
sq7383977_0008_0001_1.png,sq7383977_0008_0001_1.txt
sq7383977_0008_0001_2.png,sq7383977_0008_0001_2.txt
sq7383977_0008_0001_3.png,sq7383977_0008_0001_3.txt
sq7383977_0008_0001_4.png,sq7383977_0008_0001_4.txt
sq7383977_0008_0002_1.png,sq7383977_0008_0002_1.txt
sq7383977_0008_0002_2.png,sq7383977_0008_0002_2.txt
sq7383977_0008_0002_3.png,sq7383977_0008_0002_3.txt
sq7383977_0008_0002_4.png,sq7383977_0008_0002_4.txt
sq7383977_0008_0003_1.png,sq7383977_0008_0003_1.txt
sq7383977_0008_0003_2.png,sq7383977_0008_0003_2.txt
sq7383977_0008_0003_3.png,sq7383977_0008_0003_3.txt
sq7383977_0008_0003_4.png,sq7383977_0008_0003_4.txt
sq7383977_0009_0001_1.png,sq7383977_0009_0001_1.txt
sq7383977_0009_0001_2.png,sq7383977_0009_0001_2.txt
sq7383977_0009_0001_3.png,sq7383977_0009_0001_3.txt
sq7383977_0009_0001_4.png,sq7383977_0009_0001_4.txt
sq7383977_0009_0002_1.png,sq7383977_0009_0002_1.txt
sq7383977_0009_0002_2.png,sq7383977_0009_0002_2.txt
sq7383977_0009_0002_3.png,sq7383977_0009_0002_3.txt
sq7383977_0009_0002_4.png,sq7383977_0009_0002_4.txt
sq7383977_0009_0003_1.png,sq7383977_0009_0003_1.txt
sq7383977_0009_0003_2.png,sq7383977_0009_0003_2.txt
sq7383977_0009_0003_3.png,sq7383977_0009_0003_3.txt
sq7383977_0009_0003_4.png,sq7383977_0009_0003_4.txt
sq7383977_0010_0001_1.png,sq7383977_0010_0001_1.txt
sq7383977_0010_0001_2.png,sq7383977_0010_0001_2.txt
sq7383977_0010_0001_3.png,sq7383977_0010_0001_3.txt
sq7383977_0010_0001_4.png,sq7383977_0010_0001_4.txt
sq7383977_0010_0002_1.png,sq7383977_0010_0002_1.txt
sq7383977_0010_0002_2.png,sq7383977_0010_0002_2.txt
sq7383977_0010_0002_3.png,sq7383977_0010_0002_3.txt
sq7383977_0010_0002_4.png,sq7383977_0010_0002_4.txt
sq7383977_0010_0003_1.png,sq7383977_0010_0003_1.txt
sq7383977_0010_0003_2.png,sq7383977_0010_0003_2.txt
sq7383977_0010_0003_3.png,sq7383977_0010_0003_3.txt
sq7383977_0010_0003_4.png,sq7383977_0010_0003_4.txt
sq7383977_0011_0001_1.png,sq7383977_0011_0001_1.txt
sq7383977_0011_0001_2.png,sq7383977_0011_0001_2.txt
sq7383977_0011_0001_3.png,sq7383977_0011_0001_3.txt
sq7383977_0011_0001_4.png,sq7383977_0011_0001_4.txt
End of preview. Expand in Data Studio

OSSQ scanned-corrected staff crops

The portable, packaged form of ossq_scanned_corrected: staff-crop images and their token labels, derived from the OpenScore String Quartets scanned corpus with manual corrections applied.

This is the payload half of github.com/jhlusko/ossq-omr-data. That repository holds the build definition — sources.lock.json, fetch/verify scripts, and the packaging tool that produced this tree. Neither is complete without the other.

Why this exists

Stage C cross-staff training (training/transformer/train_staff_context.py) batches by system, and system_batch_loader.group_by_system can only group OSSQ-stemmed entries (<score>_<page>_<system>_<part>). That makes this the only corpus Stage C can consume. Before this upload it existed on exactly one machine, which is how a previous instance's trained weights were already lost once.

Contents

ossq-scanned-corrected-v1.tar.gz unpacks to ossq_scanned_corrected_portable/:

train valid
rows 34,510 3,911
systems 8,631 978
systems usable by Stage C (--min-staves 2) 8,630 978

Real systems are 4 staves almost everywhere — 8,621 of 8,631 train systems — because this is a string-quartet corpus.

Each row is <image>,<tokens>, with a .notation.json sidecar alongside (schemaVersion, annotatedSymbols, notation).

Portability

Packaged with ossq-omr-data/tools/package_dataset.py, which materialises symlinks, rewrites index paths relative to the dataset root, and then re-reads its own output to fail on a dangling link or an unresolved reference.

  • 38,421 index rows rewritten from the build machine's absolute layout (/workspace/b0/phase7bar/...) to relative paths
  • 0 symlinks
  • verify() passes: no symlinks, no absolute paths, every reference resolves

Verified independently of the packaging tool: re-running group_by_system over the packaged indexes reproduces the pre-packaging system counts exactly, with no unresolved paths.

Integrity

archive sha256 c1342d31999ef7dc36324837fbef01a221030b24e24aa66582a4bf9b29603703
archive bytes 794,305,790
unpacked files 115,265
unpacked bytes 1,276,265,760 (1.19 GB)
tree_digest 7b33bbc4adfd950cd5fa294dd92cf5aa954fc3f81ae0ab34144b6c664cfb7467

MANIFEST.json carries the file count, byte total and tree_digest for the unpacked tree; check it after extracting rather than trusting the download alone.

Rights and access boundary

Private hosting is not a licence grant. The images derive from the OpenScore String Quartets scanned corpus and its own upstream sources; consult those notices before redistributing. Do not switch this repository public without a rights review — the same boundary lieder-omr-private-all-sources documents for its own sources.

Downloads last month
43