Paper Info
Title
Generative adversarial network based adaptive data augmentation for handwritten Arabic text recognition
Abstract
Training deep learning based handwritten text recognition systems needs a lot of data in terms of text images and their corresponding annotations. One way to deal with this issue is to use data augmentation techniques toincrease the amount of training data. Generative AdversarialNetworks (GANs) based data augmentation techniques are popular in literature especiallyin tasks related toimages. However, specific challenges need to be addressed in order to effectively use GANs for data augmentation in the domain of text recognition. Text data is inherently imbalanced in terms of frequency of different characters appearing in training samples and the training data as a whole. GANs trained on the imbalanced dataset leads to augmented data that do es not represent the minority characters well,. In this paper, we present an adaptive data augmentation technique using GANs that deals with the issue of class imbalance arising in text recognition problems. We show, using experimental evaluations on two publicly available datasets for handwritten Arabic text recognition, that the GANs trained using the presented technique is effective in dealing with class imbalanced problem by generating augmented data that is balanced in terms of character frequencies, The resulting text recognition systems trained on the balanced augmented data improves the text recognition accuracy as compared to the systems trained using standard techniques.
Year
DOI
Venue
2022
10.7717/peerj-cs.861
PEERJ COMPUTER SCIENCE
Keywords
DocType
Volume
Adaptive data augmentation, Deep learning neural Networks, Arabic handwriting recognition, Handwritten text generation, Generative adversarial networks, Convolutional neural networks
Journal
8
ISSN
Citations 
PageRank 
2376-5992
0
0.34
References 
Authors
0
4
Name
Order
Citations
PageRank
Mohamed Eltay100.34
Abdelmalek Zidouri200.34
Irfan Ahmad312310.13
Yousef Elarian400.68