german-brazilian-newspapers-dataset_1

The GBN Dataset consists German-Brazilian historical newspapers, along with their digital and binarized images and ground truth files.

https://github.com/lemmbraalemao-dpb/german-brazilian-newspapers-dataset_1

Science Score: 26.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
  • Academic publication links
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (4.6%) to scientific vocabulary

Keywords

ground-truth newspaper ocr page-xml training
Last synced: 6 months ago · JSON representation

Repository

The GBN Dataset consists German-Brazilian historical newspapers, along with their digital and binarized images and ground truth files.

Basic Info
Statistics
  • Stars: 0
  • Watchers: 4
  • Forks: 0
  • Open Issues: 0
  • Releases: 59
Topics
ground-truth newspaper ocr page-xml training
Created over 2 years ago · Last pushed 10 months ago
Metadata Files
Readme License Citation

README.md

German-Brazilian-Newspapers-Dataset Part 1

The GBN Dataset consists from four different German-Brazilian historical newspapers, along with their digital and binarized images and ground truth files. The GBN-Dataset Corpus consists of two parts. The parts are designated numerically. German-Brazilian-Newspapers-Dataset_1 = Part 1, German-Brazilian-Newspapers-Dataset_2 = Part 2. It is a dataset created for evaluation of layout analysis methods.

Metadata

Language:
deu
Format:
Page-XML
Time:
1912-1940
GT Type:
data_structure
License:
PublicDomainMark 1.0
Project:
German-Brazilian Newspapers (GBN)
Project-URL:
https://web.inf.ufpr.br/vri/databases/gbn/

Sources

The volume of transcriptions:

TxtRegion ImgRegion LineDrawRegion GraphRegion TabRegion ChartRegion SepRegion MathRegion ChemRegion MusicRegion AdRegion NoiseRegion UnknownRegion CustomRegion TextLine Page
1113 18 0 71 0 0 176 0 0 0 0 0 0 0 0 68

List of transcriptions

document TxtRegion ImgRegion LineDrawRegion GraphRegion TabRegion ChartRegion SepRegion MathRegion ChemRegion MusicRegion AdRegion NoiseRegion UnknownRegion CustomRegion TextLine Page
DerJugendfreund 283 2 16 28 15
DerLandwirt 274 14 24 65 17
DerGemeindebote 226 12 49 19
EvLuthKirchenblatt 330 2 19 34 17

About

The GBN Dataset (German-Brazilian historical Newspapers Dataset)

The GBN Dataset (Part 1 and Part 2)[1] consists of 102 page images from eight different German-Brazilian historical newspapers, along with their binarized images and ground files. It is a dataset created for evaluation of layout analysis methods.

Complex layouts, poor preservation conditions of the paper, old spelling, different font sizes and the presence of multiple languages in the same page are some of the challenges of historical newspaper layout analysis that are present in the GBN Dataset. There are also pages with a slightly transparent paper and some blurs, which makes the dataset more challenging.

The dataset samples belong to the dbp digital collection from the dokumente.br initiative which has as one of the goals to digitally reconstruct the complete collection of these newspapers.

The pages of eight different newspaper titles were scanned at 600 dpi and are available in PNG format. The following table shows the composition and distribution of these images from the dataset.

Newspaper Title Publishing date Script Language Image size (in pixels)
Der Gemeindebote 1935-1938 Fraktur German 3850 x 5480
Der Jugendfreund 1911-1917 Fraktur German from 2650 x 3950 to 5320 x 8050
Der Pioner 1888-1891 Fraktur / Latin Portuguese / German 7100 x 10590
Der Landwirt 1934-1940 Fraktur German 4250 x 6020
Evangelissch-Lutherisches Kirchenblatt 1916-1919 Fraktur German 2590 x 3690
Kolonie Zeitung 1863-1889 Fraktur / Latin Portuguese / German

5470 x 7010

6700 x 8400

7050 x 9300

Gemeindeblatt 1938 Fraktur German 3850 x 5870
Heimatbote 1936 Fraktur German 3850 x 5480

The GBN Dataset is composed of:

  • grayscale images of the pages;
  • ground truth files for each of them (XML PAGE[2]).
    The XML PAGE files were produced with the Aletheia software and contain the geometric representation of different classes (e.g. text, image, graphic or separator regions) of each region of the page.

How to obtain the origin Dataset

The GBN dataset is released for academic research only and is free to researchers from educational or research institutes for non-commercial purposes.

Training GBN v1.0 Dataset: training-GBNv1.zip

References

  1. Araujo, A. B. Anlise de layout de pgina em jornais histricos germano-brasileiros, Federal University of Paran, Curitiba, 2019. https://acervodigital.ufpr.br/handle/1884/63706
  2. Pletschacher, S. and Antonacopoulos, A. The PAGE (Page Analysis and Ground-Truth Elements) Format Framework, in IEEE 20th International Conference on Pattern Recognition (ICPR), 2010, pp. 257260.

Contact

Please contact Alessandra Belzia Araujo (alessandra.belezia@ufpr.br) with questions or comments.

GitHub Events

Total
Last Year