CoolFace
Datasetpublic

wellesley-easel/randall-segmentation-masks

Randall Manuscript Segmentation Masks This dataset contains all hand-annotated segmentation masks for manuscripts indexed in Lilian Randall's Images in the Margins of Gothic Manuscripts (1966). It contains 482 training masks and 154 test masks across 87 manuscript groups. 7 categories of decorative features are labeled. Text, blank page, and other non-decorative regions are grouped together as background. Fields name, uuid, labels, segments_metadata, and… See the full description on the dataset page: https://huggingface.co/datasets/wellesley-easel/randall-segmentation-masks.

sourceHugging Faceotherupdated 1mo agoView on Hugging Face
0likes21downloads
Dataset Card

Randall Manuscript Segmentation Masks

This dataset contains all hand-annotated segmentation masks for manuscripts indexed in Lilian Randall's Images in the Margins of Gothic Manuscripts (1966). It contains 482 training masks and 154 test masks across 87 manuscript groups.

7 categories of decorative features are labeled. Text, blank page, and other non-decorative regions are grouped together as background.

Fields

  • —name, uuid, labels, segments_metadata, and segments_annotations are preserved unchanged from the pinned sources.
  • —labels is a single-channel semantic segmentation mask containing class IDs 0 through 7.
  • —source_image_credit identifies the holding institution, manuscript or the most precise non-guessed dataset identifier, and folio/page when available.
  • —pixel_values is intentionally not included. No replacement source image, image URL, or local image path is provided.
LabelCategory
"0""background"
"1""line filler"
"2""marginal images"
"3""painted initials"
"4""miniatures"
"5""borders"
"6""pen flourishing"
"7""pen flourished texts"

Source-image attribution and rights

The source pages represented here come from institutions with different image reuse policies. source_image_credit records provenance for the omitted image; it is not a license and does not mean attribution alone authorizes reuse. Some source images are CC0 or marked public domain, while others are conditional, permission-based, or unresolved. Generic upstream identifiers are retained as such rather than being presented as verified shelfmarks.

Users who obtain source images separately must review the current item-level rights and institutional terms for their intended use.