LADaS: Layout Analysis Dataset with Segmonto

Dataset Details

LADaS, created by the ALMANaCH team-project at Inria, continued in partnership with other researchers, is a multidocuments diachronic layout analysis dataset. This dataset includes:

• Monographs from the Bibliothèque Nationale de France (17th century - today);
• PhD Thesis, in various fields (not only STEM, 20th-21st century);
• Selling Catalogs (for manuscripts and art pieces), in various fields (18th-20th century);
• Noisy digitization (with fingers for example, 20th-21st century);
• Academic papers (mostly Humanities and Social Sciences) (19th-21st century);
• Magazines about technologies and video games, from 1920s to 2010;
• Misc stuff found here and there.

The data are in YoloV8 bbox format (center_x center_y width height).

The script in document is mostly Latin script, and language is mostly French with some representation of the main western academic languages.

Annotations

Label Annotation have been conducted using the SegmOnto vocabulary. An annotation guide is available here.

Uses

Direct Use

• Layout Analysis

Dataset Structure

You'll find 8 fields in the parquet:

• image_path is a string identifier for the file
• year is the year of production. It can be null.
• dating-certainty is used to specify when a date is automatically provided based on a date range and a century range. They will usually be centuries, such as 1600.
• set represents the subset of the data, there are a few, including theatre for plays, monographies for monographs, theses for PhD thesis, etc.
• image contains the image
• width and height are quite clear...
• objects is a dictionary with entries:
  • bbox which represents a series of bbox using the COCO / YOLO format (tuple of relative center_x, center_y, width, height)
  • category which contains a plain text representation of the annotation class of each bbox. There are 41 classes, further described on the annotation guide.
  • objects can contain an empty bbox list and an empty category list: it means the image does not contain any objects.

Annotations

Annotation process

The annotation process is described in the dataset paper.

Who are the annotators?

• Clérice, Thibault
• Janès, Juliette
• Scheithauer, Hugo
• Bénière, Sarah
• Bougrelle, Roxane
• [Anonymous place holder until a paper is published]

Citation

BibTeX:

@misc{Clerice_Layout_Analysis_Dataset,
  author = {Clérice, Thibault and Janès, Juliette and Scheithauer, Hugo and Bénière, Sarah and Bougrelle, Roxane and Romary, Laurent and Sagot, Benoit},
  title = {{Layout Analysis Dataset with SegmOnto (LADaS)}},
  url = {https://github.com/DEFI-COLaF/LADaS}
}

Dataset Card Contact

Thibault Clérice or Juliette Janes (first.last@inria.fr)