Synthetic Brazilian Documents Database (SBR-Doc Database)
Ground Truth of the images for SBR-Doc Database
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
Documents Database, ICDAR 2021 Competition, Image segmentation, Signature segmentation, documents analysis
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.
|edge_segmentation_gt.7z||data||(411 MB)||2||Ground Truth for Document boundary segmentation (Task 1)|
|signature_segmentation_gt.7z||data||(221 MB)||4||Ground Truth for Handwritten signature segmentation (Task 3)|
|textfield_segmentation_gt.7z.001||data||(650 MB)||1||Ground Truth for Text segmentation by zone (Task 2)|
|textfield_segmentation_gt.7z.002||data||(631 MB)||1||Ground Truth for Text segmentation by zone (Task 2)|
No comments on this dataset yet.