A segmentation-free approach for printed Devanagari script recognition

Karayil, Tushar; Ul-Hasan, Adnan; Breuel, Thomas M.

doi:10.1109/icdar.2015.7333901

Cited by 15 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning : The popularity of deep learning models is monotonically increasing day by day. For OCR systems, CNN, multilayer perceptron, bi‐directional ANNs, deep belief ANN, LSTM, few‐shot Siamese networks, and variations of LSTM are quite in practice 5,24,42,43 . Using Urdu Printed Text Images (UPTI) dataset, multi‐dimensional LSTM and CNN are trained 3 and 98.12% accuracy is achieved.…”

Section: Related Workmentioning

confidence: 99%

“…For OCR systems, CNN, multilayer perceptron, bi-directional ANNs, deep belief ANN, LSTM, few-shot Siamese networks, and variations of LSTM are quite in practice. 5,24,42,43 Using Urdu Printed Text Images (UPTI) dataset, multi-dimensional LSTM and CNN are trained 3 and 98.12% accuracy is achieved. Using zooning features, 2D-LSTM is trained and 93.39% accuracy is achieved.…”

Section: Classification Approaches For Ocrmentioning

confidence: 99%

See 1 more Smart Citation

Meta‐feature based few‐shot Siamese learning for Urdu optical character recognition

Naseer

Zafar

2022

Computational Intelligence

View full text Add to dashboard Cite

Standard convolution neural network (CNN) achieves high level of accuracy for the recognition of characters in different languages. However, like other deep neural networks, training of CNN requires a substantial amount of data. Lack of sufficient training data invokes dataset bias, during learning process, which leads to a decay in the performance of CNN. The limitation of training data can be addressed by using few-shot learners. In this research, CNN-based few-shot Siamese learner is trained on meta-features, extracted from Urdu text images using a novel graph-based normal to tangent line (GNTL) technique, for Urdu optical character recognition (OCR) across different font sizes. The learner is trained on three corpora (datasets) including one benchmark corpus "Centre for Language Engineering Text Images" and two other corpora, that is, "Urdu Thickness Graphs" (UTG) and "Urdu OCR Font 16 to 36" (UOF) which are developed and released in this research. 80% of data is used for training while 20% of data is used for testing. To create UTG corpus, the proposed novel feature extraction technique GNTL is used and a meta-features-based corpus is developed in form of thickness graphs. The third corpus UOF is based on five different font sizes, that is, 16, 20, 26, 30, and 36. The performance of few-shot Siamese learner is compared with a standard CNN, trained on the same three corpora. Meta-feature based few-shot Siamese learner achieves a promising recognition accuracy and outperforms

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Classification Approaches For Ocrmentioning

confidence: 99%

Meta‐feature based few‐shot Siamese learning for Urdu optical character recognition

Naseer

Zafar

2022

Computational Intelligence

View full text Add to dashboard Cite

show abstract

“…For the purpose of applying segmentation free approach to Devnagri scripts, long short-term memory (LSTM) networks are suitable due to their good context-aware processing. LSTM networks are kind of Recurrent Neural Networks and wisely used in unsegmented character recognition 21 . Premaratne & Bigun 22 proposed a novel approach for printed character recognition using linear symmetry, where features are directly used to recognise characters with standard alphabets without performing actual segmentation.…”

Section: Segmentation Free Approachmentioning

confidence: 99%

Analysis of Segmentation Methods for Brahmi Script

Singh¹,

Kushwaha²

2019

DESIDOC Jl. Lib. Info. Technol.

View full text Add to dashboard Cite

Segmentation is an important step for developing any optical character recognition (OCR) system, which has to be redesigned for each script having, non-uniform nature/property. It is used to decompose the image into its sub-units, which act as a basis for character recognition. Brahmi is a non-cursive ancient script, in which characters are not attached to each other and have some spacing between them. This study analyses various segmentation methods for different scripts to develop the best suitable segmentation method for Brahmi. MATLAB software was used for segmentation purpose in the experiment. The sample data belongs to Brahmi script-based ‘Rumandei inscription’. In this paper, we discuss a segmentation methodology for distinct components, namely text lines, words and characters of Rumandei inscription, written in Brahmi script. For segmenting distinct components of inscription different approach were used like horizontal projection profile, vertical projection profile and Relative minima approach. This is fundamental research on an inscription based on Brahmi script, which acts as a foundation for developing a segmentation module of an OCR solution/system of similar scripts in future. Information search and retrieval is an important activity of a library. So, to ensure this support for digitised documents written in ancient script, their character recognition is mandatory through the OCR system.

show abstract

“…Segmentation based (when ligatures are divided into characters) and segmentation free (ligature based), both approaches are used for OCR systems for these languages. OCRopus, based on LSTM, is also trained on raw pixel values of Devanagri text images and 91% accuracy is achieved [11].…”

Section: Related Workmentioning

confidence: 99%