A collection of Ground Truth data for handwritten and printed text recognition for South Asian scripts provided by FID4SA - Specialized Information Service South Asia. Interested researchers can download the data archived here and use it as training data for their own text recognition models. At the same time, researchers working on text recognition for South Asian scripts are invited to use this archive to make their own ground truth data available and to contribute to the creation of a ground truth data archive at a central site.
Featured Dataverses

In order to use this feature you must have at least one published dataverse.

Publish Dataverse

Are you sure you want to publish your dataverse? Once you do so it must remain published.

Publish Dataverse

This dataverse cannot be published because the dataverse it is in has not been published.

Delete Dataverse

Are you sure you want to delete your dataverse? You cannot undelete this dataverse.

Advanced Search

1 to 4 of 4 Results
Mar 21, 2023
Derrick, Tom; British Library, 2023, "Ground Truth transcriptions for training OCR of historical Bengali printed texts – Recognition of Early Indian Printed Documents competition - updated with improved XML coordinates", https://doi.org/10.11588/data/AIQSXL, heiDATA, V1
This dataset comprises 81 digitised images (TIFF files) drawn from a selection of early printed Bengali books (1713-1914) digitised through the Two Centuries of Indian Print project (https://www.bl.uk/projects/two-centuries-of-indian-print). Also contained are ground truth transc...
Feb 24, 2023
Tübingen University Library, 2023, "Ground Truth data for printed Malayalam", https://doi.org/10.11588/data/L2KRZO, heiDATA, V1
Ground Truth (GT) data (JPG, PAGE and ALTO XML files) which can be used to train OCR models that recognize printed text in Malayalam script. The training material is gathered from 19th and 20th centuries prints. The GT data was trained in Transkribus with the HTR+ and the PyLaia...
Dec 8, 2022
O'Neill, Alexander, 2022, "Ground Truth Model for Pracalit for Sanskrit and Newar MSS 16th to 19th C.", https://doi.org/10.11588/data/WI9184, heiDATA, V1
Ground truth data for a an OCR model. Will be continually updated. Originally trained on Transkribus with a PyLaia model created from ground truth data based on transcripts into Pracalit Unicode of four Nepalese manuscripts. The manuscripts used to create this model are Staatsbib...
Oct 26, 2022
Merkel-Hilf, Nicole, 2022, "Ground Truth data for printed Devanagari", https://doi.org/10.11588/data/EGOKEI, heiDATA, V1
Ground truth (GT) data (jpg and alto xml files) for an OCR model that recognizes printed text in Devanagari script. The GT data was trained on Transkribus with the HTR+ engine. The training was performed on appr. 220 pages with appr. 27,000 words. The validation set was 10% of th...
Add Data

Sign up or log in to create a dataverse or add a dataset.

Share Dataverse

Share this dataverse on your favorite social media networks.

Link Dataverse
Reset Modifications

Are you sure you want to reset the selected metadata fields? If you do this, any customizations (hidden, required, optional) you have done will no longer appear.