pith. sign in

arxiv: 2606.19139 · v1 · pith:SCLSHI2Snew · submitted 2026-06-17 · 💻 cs.CV · cs.CL

Urdu Katib Handwritten Dataset: A Historical Document Dataset for Offline Urdu Handwritten Text Recognition with CRNN-Based Baseline Evaluation

Ramza Basharat , Muhammad Usman Ali This is my paper

Pith reviewed 2026-06-26 21:12 UTC · model grok-4.3

classification 💻 cs.CV cs.CL
keywords Urdu handwritten text recognitionhistorical documentsNastalique scriptCRNN modelsdataset curationKatib handwritingoffline HTRCNN-BGRU-CTC
0
0 comments X

The pith

The Urdu Katib Handwritten Dataset supplies the first offline benchmark of historical Nastalique Urdu lines for CRNN-based recognition.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces the Urdu Katib Handwritten Dataset, the first real offline collection of Urdu text lines written by historical Katibs in the Nastalique calligraphic style with flat nibs. The work responds to the scarcity of benchmark data that has slowed progress on Urdu Handwritten Text Recognition by providing authentic samples that capture varied writing styles. Multiple CRNN hybrid architectures are tested on the new data, and the CNN-BGRU-CTC model records the lowest character and word error rates among them. The dataset is released to help researchers build recognition systems that can aid preservation of Urdu handwritten literature.

Core claim

The authors create the UKHD as the first offline Urdu handwritten text lines dataset from historical Katib materials, covering diverse flat nib variations in Nastalique calligraphy, and show that among CRNN models the CNN-BGRU-CTC architecture achieves the most robust performance with low CER and WER.

What carries the argument

The UKHD dataset of historical Nastalique calligraphic text lines, used to benchmark CRNN hybrid models for Urdu handwriting recognition.

If this is right

  • Recognition models can now be trained and tested on authentic historical Urdu samples rather than synthetic or scarce data.
  • The CNN-BGRU-CTC model supplies a reproducible baseline architecture for subsequent Urdu text recognition work.
  • Automated systems developed from this resource can support digitization and long-term preservation of Urdu literary materials.
  • Curation methods used for UKHD can guide creation of comparable datasets for other cursive historical scripts.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Supplementing UKHD with contemporary Urdu handwriting samples would likely be needed before claiming coverage of all writing styles.
  • Replacing the recurrent layers in the best-performing model with attention or transformer components could be tested directly on the released data.
  • The same historical-document sourcing strategy could accelerate benchmark creation for other low-resource cursive scripts facing similar data shortages.

Load-bearing premise

The historical Katib materials encompass a diverse range of flat nib writing variations in the Nastalique calligraphic style that are sufficient to support development of robust general-purpose Urdu handwritten text recognition systems.

What would settle it

A new collection of Urdu handwriting samples drawn from non-Katib sources or different historical periods on which models trained only on UKHD produce markedly higher CER and WER would indicate the dataset does not yet support general-purpose recognition.

Figures

Figures reproduced from arXiv: 2606.19139 by Muhammad Usman Ali, Ramza Basharat.

Figure 1
Figure 1. Figure 1: Urdu character set has 38 basic characters including 10 non-joiner characters, 27 joiner characters, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (a) Nastalique Style: Diagonal Behavior (consumes less space), (b) Naskh Style: Horizontal Behavior (consumes more space); often used for writing Urdu but generally used for writing Arabic [22]. In both (a) and (b), the arrows indicate the writing direction, whereas the red horizontal line shows the baseline3 . 2An Optical Character Recognition (OCR) system transforms printed text from images into machine-… view at source ↗
Figure 3
Figure 3. Figure 3: (a) Cursive Nature: The given phrase is made up of five words, each word has one or more ligatures i.e. 1st word has two ligatures (2L) —first ligature is formed by joining four characters together while second ligature has only one character. In the given example, the arrow above each word points to the details of ligatures present in the corresponding word i.e. individual characters are written in blue c… view at source ↗
Figure 4
Figure 4. Figure 4: (a) Group of characters having similar primary strokes [28]. (b) There are three primary types of diacritics: “Dot/Nuqta”, “Chota Toay” in superscript, and “Aerabs”. Nuqta and small toay are compulsory diacritics. The remaining diacritics are known as Aerabs, which are optional and used for removing any ambiguity in pronunciation [22, 30, 35, 45]. (c) List of characters accompanied by dots. Dots can range … view at source ↗
Figure 5
Figure 5. Figure 5: PUTL Subset Labels Length Histogram —Minimum Length is 3, whereas Maximum Length is 90 [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: MUTL Subset Labels Length Histogram —Minimum Length is 2, whereas Maximum Length is 103 [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: UKHD Generation Process 4.2 Preprocessing Page images were first renamed using correct page numbers for consistency, then the following preprocessing steps were applied to improve image quality and reduce computational complexity. 4.2.1 Grayscale Conversion. During text recognition, the structure of the text is important. There￾fore, the color information of the acquired RGB images was eliminated by conver… view at source ↗
Figure 8
Figure 8. Figure 8: Sample Images from Source Books used in UKHD/ Urdu Katib Handwriting Samples [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: (a) Samples of Acquired RGB Images (b) After Grayscale Conversion 3816 3498 3438 3385 3341 3373 3513 3497 3357 3263 3236 3498 3511 3521 3519 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 255 254 254 254 254 255 251 254 255 255 255 255 255 255 255 254 254 254 254 254 200 87 162 254 254 254 255 254 254 254 253 254 253 253 254 90 170 137 253 253 253 254 253 254 254 253 253 253 253 235 139 119 233 244 138 253 253 253 253… view at source ↗
Figure 10
Figure 10. Figure 10: Horizontal Projection Profile (HPP) —Column Vector Hx1 is the HPP of the Image HxW (It converts a [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: (a) Noise Free Grayscale Skewed Image, (b) After Applying Sobel Filter: It detected horizontal and vertical edges, and eliminated the extra information such as the gradual changes in intensities which are not associated with significant edges. (c) After Performing Image Inversion/Negation: It enhanced the visibility of the text. It is transformed skewed image which has been subsequently rotated at differe… view at source ↗
Figure 12
Figure 12. Figure 12: Horizontal Projection Profiles (HPPs) of the Rotated Images; transformed skewed image given in [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: UKHD Generation Application Interface – Preprocessing Phase [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: (a) Preprocessed Image, (b) After Applying Otsu’s Threshold 4.3.2 HPP Calculation and Estimating the Potential Regions for Line Boundaries. The horizontal projection of the resultant binary image was then computed i.e. the sum of white pixel values along each row of the binary image, as shown in [PITH_FULL_IMAGE:figures/full_fig_p015_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Horizontal Projection Profile (HPP) of the Resultant Binary Image [PITH_FULL_IMAGE:figures/full_fig_p015_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: The red line in HPP of the binary image is the optimal threshold for line boundaries. The regions [PITH_FULL_IMAGE:figures/full_fig_p016_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: (a) Invalid Estimated Regions of Line Boundaries (zoom the image and see these regions are causing over-segmentation), (b) Detected Line Boundaries The estimated regions for line boundaries were then further refined by excluding those regions that did not actually correspond to a line boundary. This exclusion was necessary because some of these regions are caused by diacritics, as illustrated in Figure 17… view at source ↗
Figure 18
Figure 18. Figure 18: (a) Auto-segmented Text lines, (b) After Manual Adjustments If you observe the auto-segmented text lines in Figure 18a which are automatically segmented using above the horizontal projection profile-based method, then you will see that they are not in an optimal form. There are some instances of incorrect segmentation, so these were further adjusted manually. These manual adjustments were carried out with… view at source ↗
Figure 19
Figure 19. Figure 19: The text after manual corrections is the final transcription of the text line image, that [PITH_FULL_IMAGE:figures/full_fig_p018_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: UKHR Model Architecture – This is essentially the CNN-BGRU-CTC hybrid model architecture, [PITH_FULL_IMAGE:figures/full_fig_p019_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Visualization of feature maps produced by Conv2D layers in the CNN-BGRU-CTC model. The initial [PITH_FULL_IMAGE:figures/full_fig_p021_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: The CNN component takes a grayscale text line image as input and returns a feature sequence. The [PITH_FULL_IMAGE:figures/full_fig_p022_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: (a) Input sequence length is 95. It is the input of the model (RNNs part input) which consists of a sequence of 95 time steps, where each time step represents a segment of an image. (b) Raw output sequence length is 95. It represents the predictions made by the model (RNNs part output) at each time step of the input sequence. In actual it predicts a probability distribution over all possible characters, i… view at source ↗
Figure 24
Figure 24. Figure 24: Frequency of Each Class i.e. Character and Symbol, in Experimental Dataset ( [PITH_FULL_IMAGE:figures/full_fig_p024_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: (a) Input Text Line Image: Text is written from right-to-left, (b) After Horizontal Flipping: Reverses the character order in the image (making text left-to-right), (c) After Resizing: After experimenting with different sizes, following standard height and width dimensions were chosen because they maintained the image quality without losing the textual information; height=32 pixels, width=384 pixels, padd… view at source ↗
Figure 26
Figure 26. Figure 26: (a) Input Text Line Image, (b) Corresponding Label/ Image Transcription: This is the original text which is a sequence of characters, (c) Preprocessed Label: Each character in the label is replaced by its numeric identifier which creates a sequence of numbers that represents the original text; actual sequence length is 69, after padding it becomes 90, (d) Encoding Detail: Showing the unique numeric identi… view at source ↗
Figure 27
Figure 27. Figure 27: Types of Recognition Errors: Substitution error occurs when a character is incorrectly re￾placed/misspelled in the predicted text. Insertion error occurs when a character is erroneously included, whereas Deletion errors occur when a character is skipped/omitted in the predicted text [PITH_FULL_IMAGE:figures/full_fig_p027_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: UKHR Model Output – In Most Cases, It Recognized the Text with 100% Accuracy [PITH_FULL_IMAGE:figures/full_fig_p031_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: Comparison Between the Output of the Cloud Vision API and the UKHR Model [PITH_FULL_IMAGE:figures/full_fig_p032_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: UKHR Model Output — Failure Case Samples (Recognition Errors are Highlighted with Rectangular [PITH_FULL_IMAGE:figures/full_fig_p032_30.png] view at source ↗
read the original abstract

Automatic Handwritten Text Recognition (HTR) is inherently a challenging task, and its complexity is further increased when dealing with cursive scripts. Although significant efforts have been made on various cursive scripts, research regarding Urdu Handwritten Text Recognition (UHTR) has been relatively limited. This lag of research is primarily due to the unique challenges posed by its script, and the scarcity and unavailability of benchmark datasets. Therefore, to advance research in UHTR, this study presents a specialized real dataset called the Urdu Katib Handwritten Dataset (UKHD). To the best of our knowledge, this is the first offline Urdu handwritten text lines dataset specifically curated from the materials written by Katibs in historical times. It encompasses a diverse range of flat nib writing variations in the Nastalique calligraphic style. Additionally, the effectiveness of different CRNN-based hybrid models has been evaluated to identify the optimal architecture for Urdu Katib Handwriting Recognition (UKHR). Among the analyzed models, the CNN-BGRU-CTC model showed more robust performance, with low Character Error Rate (CER) and Word Error Rate (WER). This research work aims to support and encourage the research community in developing a robust recognition system for preserving Urdu handwritten literature.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces the Urdu Katib Handwritten Dataset (UKHD), presented as the first offline handwritten text-line dataset curated from historical Katib materials in Nastalique calligraphic style. It evaluates multiple CRNN-based hybrid architectures for Urdu handwritten text recognition and concludes that the CNN-BGRU-CTC model exhibits the most robust performance, characterized by low CER and WER.

Significance. The curation of a specialized historical dataset from Katib sources addresses a documented scarcity of real-world benchmarks for Urdu HTR. If released with complete statistics, splits, and reproducible baselines, the resource could enable targeted progress on Nastalique-style recognition and support preservation of Urdu literature. The empirical model comparison supplies initial architecture rankings, though the absence of quantitative results limits evaluation of claimed robustness.

major comments (2)
  1. [Dataset description] Dataset description section: no total number of text lines, number of source documents or writers, train/test split ratios, or statistics on writing variation diversity are supplied. These details are required to substantiate that UKHD constitutes a usable benchmark for the claimed research advancement.
  2. [Results] Results section: the claim that CNN-BGRU-CTC achieves 'low CER and WER' and 'more robust performance' is unsupported because exact numerical values, comparisons against the other evaluated CRNN variants, and any error bars or statistical tests are omitted. This prevents verification of the model ranking.
minor comments (2)
  1. The abstract and introduction do not define the threshold for 'low' CER/WER relative to existing Urdu or Nastalique HTR literature, reducing interpretability of the performance claim.
  2. [Model evaluation] Hyperparameter settings, layer counts, and training protocols for the CRNN variants are not specified, hindering reproducibility of the reported baseline.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments. We address each major comment below and will revise the manuscript to incorporate the requested information.

read point-by-point responses

  1. Referee: [Dataset description] Dataset description section: no total number of text lines, number of source documents or writers, train/test split ratios, or statistics on writing variation diversity are supplied. These details are required to substantiate that UKHD constitutes a usable benchmark for the claimed research advancement.

    Authors: We agree that these quantitative details are necessary to fully establish UKHD as a usable benchmark. The revised manuscript will add the total number of text lines, number of source documents and writers, explicit train/test split ratios, and statistics on writing variation diversity to the Dataset description section. revision: yes

  2. Referee: [Results] Results section: the claim that CNN-BGRU-CTC achieves 'low CER and WER' and 'more robust performance' is unsupported because exact numerical values, comparisons against the other evaluated CRNN variants, and any error bars or statistical tests are omitted. This prevents verification of the model ranking.

    Authors: We acknowledge that the Results section currently lacks the specific numerical values and direct comparisons needed to substantiate the performance claims. In the revision we will include the exact CER and WER figures for all CRNN variants evaluated, a clear ranking table, and any available error bars or statistical tests to support the statement that CNN-BGRU-CTC exhibits the most robust performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a dataset release plus empirical baseline evaluation on CRNN variants. No equations, derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described contribution. The central claim (new historical Urdu text-line resource with reproducible model rankings) rests on the released splits and measured CER/WER values, which are externally falsifiable and do not reduce to the paper's own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical dataset creation and model benchmarking paper. No free parameters, mathematical axioms, or invented entities are introduced.

pith-pipeline@v0.9.1-grok · 5756 in / 1182 out tokens · 24979 ms · 2026-06-26T21:12:00.092926+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 7 canonical work pages · 4 internal anchors

  1. [1]

    Riaz Ahmad, M Zeshan Afzal, S Faisal Rashid, Marcus Liwicki, Thomas Breuel, and Andreas Dengel. 2016. Kpti: Katib’s pashto text imagebase and deep learning benchmark. In2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR). IEEE, 453–458

    2016

  2. [2]

    Saad Bin Ahmed, Saeeda Naz, Salahuddin Swati, and Muhammad Imran Razzak. 2019. Handwritten Urdu character recognition using one-dimensional BLSTM classifier.Neural Computing and Applications31 (2019), 1143–1151

    2019

  3. [3]

    Sana Al-azzawi, Elisa Barney, and Marcus Liwicki. 2026. Cross-Language Learning within Arabic Script for Low- Resource HTR. arXiv:2605.02089 [cs.CV] https://arxiv.org/abs/2605.02089

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  4. [4]

    Saad Albawi, Tareq Abed Mohammed, and Saad Al-Zawi. 2017. Understanding of a convolutional neural network. In 2017 international conference on engineering and technology (ICET). Ieee, 1–6

    2017

  5. [5]

    Anjum Ali, Mohmood Ahmad, Nasir Rafiq, Javed Akber, Usman Ahmad, and Shahwar Akmal. 2004. Language independent optical character recognition for hand written text. In8th International Multitopic Conference, 2004. Proceedings of INMIC 2004.IEEE, 79–84

    2004

  6. [6]

    Tayaba Anjum and Arifah Azhar. 2025. A Survey on Urdu Handwritten Text Recognition: State of the Art, Challenges, and Future Directions.Journal of Computing and Artificial Intelligence3, 1 (2025)

    2025

  7. [7]

    Tayaba Anjum and Nazar Khan. 2020. An attention based method for offline handwritten Urdu text recognition. In 2020 17th International conference on frontiers in handwriting recognition (ICFHR). IEEE, 169–174

    2020

  8. [8]

    Saad Bin Ahmed, Saeeda Naz, Salahuddin Swati, Imran Razzak, Arif Iqbal Umar, and A Ali Khan. 2017. UCOM offline dataset-an Urdu handwritten dataset generation. (2017)

    2017

  9. [9]

    Lei Chen and Shaobin Li. 2018. Improvement research and application of text recognition algorithm based on CRNN. InProceedings of the 2018 international conference on signal processing and machine learning. 166–170

    2018

  10. [10]

    Liren Chen, Ruijie Yan, Liangrui Peng, Akio Furuhata, and Xiaoqing Ding. 2017. Multi-layer recurrent neural network based offline Arabic handwriting recognition. In2017 1st international workshop on Arabic script analysis and recognition (ASAR). IEEE, 6–10

    2017

  11. [11]

    Kyunghyun Cho, Bart Van Merriënboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078(2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  12. [12]

    Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling.arXiv preprint arXiv:1412.3555(2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  13. [13]

    MUHAMMAD FAHAD, MALIK MUHAMMAD SAAD MISSEN, MUJTABA HUSNAIN, ALI SAMAD, DALER ALI, and ASAD ALI. 2023. MULTI-ASPECT URDU HANDWRITING DATA COLLECTION.Tianjin Daxue Xuebao (Ziran Kexue yu Gongcheng Jishu Ban)/Journal of Tianjin University Science and Technology(2023)

    2023

  14. [14]

    Takwa Ben Aïcha Gader and Afef Kacem Echi. 2022. Attention-based deep learning model for Arabic handwritten text recognition.Machine Graphics and Vision31, 1/4 (2022), 49–73

    2022

  15. [15]

    Aejaz Farooq Ganai and Farida Khursheed. 2022. A novel holistic unconstrained handwritten urdu recognition system using convolutional neural networks.International Journal on Document Analysis and Recognition (IJDAR)25, 4 (2022), 351–371

    2022

  16. [16]

    Aejaz Farooq Ganai and Farida Khursheed. 2023. Computationally Efficient Holistic Approach for Handwritten Urdu Recognition using LRCN Model.International Journal of Intelligent Systems and Applications in Engineering11, 4s (2023), 536–551

    2023

  17. [17]

    Aejaz Farooq Ganai and Farida Khursheed. 2023. Computationally efficient recognition of unconstrained handwritten Urdu script using BERT with vision transformers.Neural Computing and Applications35 (2023), 1–17

    2023

  18. [18]

    Alex Graves, Santiago Fernández, Faustino Gomez, and Jürgen Schmidhuber. 2006. Connectionist temporal classifica- tion: labelling unsegmented sequence data with recurrent neural networks. InProceedings of the 23rd international conference on Machine learning. 369–376

    2006

  19. [19]

    Shahbaz Hassan, Ayesha Irfan, Ali Mirza, and Imran Siddiqi. 2019. Cursive handwritten text recognition using bi-directional LSTMs: a case study on Urdu handwriting. In2019 International conference on deep learning and machine learning in emerging applications (Deep-ML). IEEE, 67–72

    2019

  20. [20]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation9, 8 (1997), 1735–1780

    1997

  21. [21]

    S Afaq Husain, Asma Sajjad, and Fareeha Anwar. 2007. Online Urdu Character Recognition System.. InMV A. 98–101. A Study on Urdu Katib Handwritten Dataset Generation and CRNN-Based Baseline Evaluation 35

    2007

  22. [22]

    Sarmad Hussain. 2003. Complexity of Asian writing systems: a case study of Nafees Nasta’leeq for urdu. InProceedings of the 12th AMIC Annual Conference on e-Worlds: Governments, Business and Civil Society, Asian Media Information Center, Singapore

    2003

  23. [23]

    Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. InInternational conference on machine learning. pmlr, 448–456

    2015

  24. [24]

    Iqbal Cyber Library

    Iqbal Academy Pakistan. 2023. "Iqbal Cyber Library". [Online]. Available: https://iqbalcyberlibrary.net. Accessed: 2023

    2023

  25. [25]

    Sobia Tariq Javed and Sarmad Hussain. 2009. Improving Nastalique specific pre-recognition process for Urdu OCR. In 2009 IEEE 13th International Multitopic Conference. IEEE, 1–6

    2009

  26. [26]

    Sobia T Javed, Sarmad Hussain, Ameera Maqbool, Samia Asloob, Sehrish Jamil, and Huma Moin. 2010. Segmentation free nastalique urdu ocr.World Academy of Science, Engineering and Technology46 (2010), 456–461

    2010

  27. [27]

    Yuxiang Jiang, Haiwei Dong, and Abdulmotaleb El Saddik. 2018. Baidu Meizu deep learning competition: Arithmetic operation recognition using end-to-end learning OCR technologies.IEEE Access6 (2018), 60128–60136

    2018

  28. [28]

    Muhammad Kashif. 2021. Urdu Handwritten Text Recognition Using ResNet18.arXiv preprint arXiv:2103.05105(2021)

    work page arXiv 2021

  29. [29]

    Khalil Khan, Rehan Ullah, Nasir Ahmad Khan, and Khwaja Naveed. 2012. Urdu character recognition using principal component analysis.International Journal of Computer Applications60, 11 (2012)

    2012

  30. [30]

    Naila Habib Khan and Awais Adnan. 2018. Urdu optical character recognition systems: Present contributions and future directions.IEEE Access6 (2018), 46019–46046

    2018

  31. [31]

    Herleen Kour and Naveen Kumar Gondhi. 2020. Machine Learning approaches for Nastaliq style Urdu handwritten recognition: A survey. In2020 6th International Conference on Advanced Computing and Communication Systems (ICACCS). IEEE, 50–54

    2020

  32. [32]

    Gurpreet Singh Lehal and Ankur Rana. 2013. Recognition of nastalique urdu ligatures. InProceedings of the 4th International Workshop on Multilingual OCR. 1–5

    2013

  33. [33]

    LB Mahanta and Alpana Deka. 2013. Skew and slant angles of handwritten signature.International Journal of Innovative Research in Computer and Communication Engineering1, 9 (2013), 2030–2034

    2013

  34. [34]

    Ahmed Muaz. 2010. Urdu optical character recognition system MS thesis.Diss. National University of Computer & Emerging Sciences(2010)

    2010

  35. [35]

    Omar Mukhtar, Srirangaraj Setlur, and Venu Govindaraju. 2010. Experiments on urdu text recognition.Guide to OCR for Indic scripts: Document recognition and retrieval(2010), 163–171

    2010

  36. [36]

    Saeeda Naz, Khizar Hayat, Muhammad Imran Razzak, Muhammad Waqas Anwar, Sajjad A Madani, and Samee U Khan. 2014. The optical character recognition of Urdu-like cursive scripts.Pattern Recognition47, 3 (2014), 1229–1248

    2014

  37. [37]

    Saeeda Naz, Arif I Umar, Riaz Ahmad, Saad B Ahmed, Syed H Shirazi, Imran Siddiqi, and Muhammad I Razzak. 2016. Offline cursive Urdu-Nastaliq script recognition using multidimensional recurrent neural networks.Neurocomputing 177 (2016), 228–241

    2016

  38. [38]

    Saeeda Naz, Arif Iqbal Umar, Riaz Ahmed, Muhammad Imran Razzak, Sheikh Faisal Rashid, and Faisal Shafait. 2016. Urdu Nasta’liq text recognition using implicit segmentation based on multi-dimensional long short term memory neural networks.SpringerPlus5 (2016), 1–16

    2016

  39. [39]

    Keiron O’Shea and Ryan Nash. 2015. An introduction to convolutional neural networks.arXiv preprint arXiv:1511.08458 (2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  40. [40]

    U Pal and Anirban Sarkar. 2003. Recognition of printed Urdu script. InSeventh International Conference on Document Analysis and Recognition, 2003. Proceedings., Vol. 3. Citeseer, 1183–1183

    2003

  41. [41]

    Dhuha Rashid and Naveen Kumar Gondhi. 2022. Scrutinization of Urdu handwritten text recognition with machine learning approach. InInternational Conference on Emerging Technologies in Computer Engineering. Springer, 383–394

    2022

  42. [42]

    Nauman Riaz, Haziq Arbab, Arooba Maqsood, Khuzaeymah Nasir, Adnan Ul-Hasan, and Faisal Shafait. 2022. Conv- transformer architecture for unconstrained off-line Urdu handwriting recognition.International Journal on Document Analysis and Recognition (IJDAR)25, 4 (2022), 373–384

    2022

  43. [43]

    Malik Waqas Sagheer, Chun Lei He, Nicola Nobile, and Ching Y Suen. 2010. Holistic Urdu handwritten word recognition using support vector machine. In2010 20th international conference on pattern recognition. IEEE, 1900–1903

    2010

  44. [44]

    Danish Altaf Satti. 2013. Offline Urdu Nastaliq OCR for printed text using analytical approach.MS thesis report(2013), 141

    2013

  45. [45]

    Danish Altaf Satti and Khalid Saleem. 2012. Complexities and implementation challenges in offline urdu Nastaliq OCR. InProceedings of the Conference on Language & Technology. 85–91

    2012

  46. [46]

    Adnan Hussain Shah, Muhammad Majid Mahmood Bagram, Muhammad Munwar Iqbal, and Farooq Ali. 2021. Urdu Handwritten Words Recognition Using Machine Learning.Technical Journal26, 02 (2021), 81–88

    2021

  47. [47]

    Nabeel Shahzad, Brandon Paulson, and Tracy Hammond. 2009. Urdu Qaeda: recognition system for isolated urdu characters. InProceedings of the IUI Workshop on Sketch Recognition, Sanibel Island, Florida. 36 Ramza Basharat et al

    2009

  48. [48]

    Mohammad Daniyal Shaiq, Musa Dildar Ahmed Cheema, and Ali Kamal. 2022. Transformer based Urdu Handwritten Text Optical Character Reader.arXiv preprint arXiv:2206.04575(2022)

    work page arXiv 2022

  49. [49]

    Baoguang Shi, Xiang Bai, and Cong Yao. 2016. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition.IEEE transactions on pattern analysis and machine intelligence 39, 11 (2016), 2298–2304

    2016

  50. [50]

    Ralf C Staudemeyer and Eric Rothstein Morris. 2019. Understanding LSTM–a tutorial into long short-term memory recurrent neural networks.arXiv preprint arXiv:1909.09586(2019)

    work page arXiv 2019

  51. [51]

    Guofeng Tong, Yong Li, Huashuai Gao, Huairong Chen, Hao Wang, and Xiang Yang. 2020. MA-CRNN: a multi-scale attention CRNN for Chinese text line recognition in natural scenes.International Journal on Document Analysis and Recognition (IJDAR)23 (2020), 103–114

    2020

  52. [52]

    Adnan Ul-Hasan, Saad Bin Ahmed, Faisal Rashid, Faisal Shafait, and Thomas M Breuel. 2013. Offline printed Urdu Nastaleeq script recognition with bidirectional LSTM networks. In2013 12th international conference on document analysis and recognition. IEEE, 1061–1065

    2013

  53. [53]

    Noor ul Sehr Zia, Muhammad Ferjad Naeem, Syed Muhammad Kumail Raza, Muhammad Mubasher Khan, Adnan Ul-Hasan, and Faisal Shafait. 2022. A convolutional recursive deep architecture for unconstrained Urdu handwriting recognition.Neural Computing and Applications(2022), 1–14

    2022