AnnoPage Dataset

AnnoPage Dataset contains pages of mostly historical documents and annotations of non-textual objects, such as images, diagrams, symbols, initials, etc. The dataset contains data mostly from czech written documents downloaded from Digital Library, but it also contains annotations for pages from other already published datasets which are listed below together with links. The provided ZIP archive includes a text file referencing the pages from these datasets. The complete list of the annotated object categories is also listed below. The dataset contains a split into training and testing subsets, the creation of the validation subset is up to the users.

The ZIP archive of the provided dataset is organized as follows:

images/ - directory with published 5690 images divided into train and test subdirectories

labels/ - directory with annotations in 7550 text files divided into train and test subdirectories

README.md - README file with the same information as displayed on this website

classes.txt - text file with object categories

dataset.yaml - YAML file with dataset definition

images.txt - text file with references to the 1860 annotated images from other datasets

This dataset was crated within Orbis Pictus – book revival for cultural and creative sectors project.

Extra datasets

illuhisdoc

PRImA Layout Analysis Dataset

PRImA RDCL2019

PRImA Europeana Newspapers

READ-ICDAR2019-cBAD-dataset

TexBiG

Object categories

Chemical formula and equation

Symbol, logo, coat of arms

Exlibris

Photograph

Geometric drawing

Graph

Initial

Caricature and comics

Map

Mathematical expression and equation

Musical notation

Image

Other book decor

Other technical drawing

Decorative inscription

Floor plan

Barcode and QR code

Stamp

Advertisement

Handwritten note

Diagram

Signet

Table

Vignette

Frieze