Synthetic Brazilian Documents Database (SBR-Doc Database)

Ground Truth

Ground Truth of the images for SBR-Doc Database

2021-08-22 (v. 1)

Contact author

Celso A M Lopes Junior

Universidade de Pernambuco

camlj@ecomp.poli.br

+55 81 992469364


This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.

Keywords

Documents Database, ICDAR 2021 Competition, Image segmentation, Signature segmentation, documents analysis

Description

Ground Truth (GT) images are divided into three groups:

 

  • Task 1- GT for Document boundary segmentation (15,000 images);
  • Task 2- GT for Text segmentation by zone (15,000 images);
  • Task 3- GT for Handwritten signature segmentation (15,000 images).

Because of this, it is possible to train and test your model with one or all of the tasks you need.

FileTypeSizeDownloadsDescription
edge_segmentation_gt.7zdata(411 MB)0Ground Truth for Document boundary segmentation (Task 1)
signature_segmentation_gt.7zdata(221 MB)0Ground Truth for Handwritten signature segmentation (Task 3)
textfield_segmentation_gt.7z.001data(650 MB)0Ground Truth for Text segmentation by zone (Task 2)
textfield_segmentation_gt.7z.002data(631 MB)0Ground Truth for Text segmentation by zone (Task 2)

Comments

No comments on this dataset yet.

Add your comment

In order to comment on a dataset you need to be logged on
Register Now!

Valoration

In order to rate this dataset you need to be logged on
Register Now!