razzak-ucomofflinedataset-2017.pdf (415.42 kB)
UCOM offline dataset-an Urdu handwritten dataset generation
journal contribution
posted on 2017-01-01, 00:00 authored by S Bin Ahmed, S Naz, S Swati, Imran RazzakImran Razzak, A I Umar, A Ali Khan© 2017, Zarka Private University. All rights reserved. A benchmark database for character recognition is an essential part for efficient and robust development. Unfortunately, there is no comprehensive handwritten dataset for Urdu language that would be used to compare the state of the art techniques in the field of optical character recognition. In this paper, we present a new and publically available dataset comprising 600 pages of handwritten Urdu text written in Nasta’liq style in conjunction with detailed ground truth for the evaluation of handwritten Urdu character recognition. This dataset contains text lines written in Nasta’liq style by limited individuals on A4 size paper. The acquired data on page was scanned and text lines were segmented. UCOM database covers all Urdu characters and ligatures with different variation in addition to Urdu numeric data. We have considered that ligature consists of up to five characters in this dataset. The UCOM dataset can be used for handwritten character recogntition as well as writer identification. We proposed and evaluated the strength of Recurrent Neural Networks (RNN) on UCOM offline database sample text line.
History
Journal
The international Arab journal of information technologyVolume
14Issue
2Pagination
239 - 245Publisher
Zarqa Private UniversityLocation
Zarqa, JordanLink to full text
ISSN
1683-3198eISSN
2309-4524Language
engPublication classification
C1.1 Refereed article in a scholarly journalUsage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC