A Deep Learning based
Arabic Script Recognition System: Benchmark on KHAT
Riaz Ahmad1, Saeeda Naz2, Muhammad
Afzal3, Sheikh Rashid4, Marcus Liwicki5, and
Andreas Dengel6
1Shaheed
Banazir Bhutto University, Sheringal, Pakistan
2Computer
Science Department, GGPGC No.1 Abbottabad, Pakistan
3Mindgarage, University
of Kaiserslautern, Germany
4Al Khwarizmi Institute of Computer
Science, UET Lahore, Pakistan
5Department of Computer
Science, Luleå University of Technology, Luleå
6German Research Center for
Artificial Intelligence (DFKI) in Kaiserslautern, Germany
Abstract: This paper presents a deep learning benchmark on a complex dataset known
as KFUPM Handwritten Arabic TexT (KHATT). The KHATT data-set consists of
complex patterns of handwritten Arabic text-lines. This paper contributes
mainly in three aspects i.e., (1) pre-processing, (2) deep learning based
approach, and (3) data-augmentation. The pre-processing step includes pruning
of white extra spaces plus de-skewing the skewed text-lines. We deploy a deep
learning approach based on Multi-Dimensional Long Short-Term Memory (MDLSTM)
networks and Connectionist Temporal Classification (CTC). The MDLSTM has the
advantage of scanning the Arabic text-lines in all directions (horizontal and
vertical) to cover dots, diacritics, strokes and fine inflammation. The
data-augmentation with a deep learning approach proves to achieve better and
promising improvement in results by gaining 80.02% Character Recognition (CR)
over 75.08% as baseline.
Keywords: Handwritten Arabic text recognition, deep learning, data augmentation.