Pith. sign in

REVIEW 6 major objections 6 minor 1 cited by

HTR-JAND: Handwritten Text Recognition with Joint Attention Network and Knowledge Distillation

T0 review · 6 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read HTR-JAND claims state-of-the-art character error rates of 1.23% on IAM, 1.02% on RIMES, and 2.02% on Bentham for handwritten text recognition.

desk verdict Reported SOTA CERs come from a pipeline with T5 post-processing and lexicon correction, not the HTR model alone, and Proxima Attention is undefined; the paper needs major revision but deserves a serious referee. read the letter →

arxiv 2412.18524 v1 pith:YB6VRHCR submitted 2024-12-24 cs.CV

Mohammed Hamdan , Abderrahmane Rahiche , Mohamed Cheriet This is my paper · ORCID
classification cs.CV
keywords handwrittentextrecognitionknowledgedistillationcombinedattentionProximaFullGatedConv2dSqueeze-and-ExcitationcurriculumlearningT5post-processing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that accurate handwritten text recognition, including on historical documents, does not require huge models. It proposes HTR-JAND, a framework that combines gated convolutions with Squeeze-and-Excitation blocks, a fused attention mechanism, knowledge distillation, curriculum learning, and T5-based post-processing. The central claim is that this pipeline achieves character error rates of 1.23% on IAM, 1.02% on RIMES, and 2.02% on Bentham, while a distilled Student model with 0.75M parameters matches much larger systems. If true, high-accuracy handwriting recognition becomes practical for resource-constrained digitization projects.

What carries the argument

The load-bearing mechanism is the Combined Attention module, which fuses Multi-Head Self-Attention with Proxima Attention, an attention variant the paper describes as using linear projections for keys and values with dynamic query updates, though its exact mathematical definition is not fully given. The other central mechanism is knowledge distillation: a Teacher-Student training scheme in which the Student is trained with a four-part loss combining CTC, cross-entropy, a KL-divergence distillation term against the Teacher's softened logits, and an auxiliary classifier loss. These mechanisms sit on a CNN backbone of FullGatedConv2d and Squeeze-and-Excitation layers feeding bidirectional LSTMs.

What would settle it

The first concrete check is to run the released implementation, or an implementation built strictly from the paper, and confirm the 1.23%, 1.02%, and 2.02% CERs; because Proxima Attention is not mathematically defined, the text alone does not permit this. A second decisive check is to ablate the Combined Attention by replacing Proxima Attention with a second standard multi-head attention head and measuring IAM CER: if it remains 1.23%, Proxima Attention is not the active ingredient; if it degrades, the reported mechanism is doing the work.

Watch

Extended reading notes

Core claim

The paper's central claim is that a compact handwritten text recognition system can reach state-of-the-art accuracy by pairing gated and Squeeze-and-Excitation convolutions with a fused attention module, knowledge distillation, curriculum learning, and neural post-processing. On the IAM, RIMES, and Bentham benchmarks, the full pipeline reports character error rates of 1.23%, 1.02%, and 2.02%, respectively. The Student model, with 0.75M parameters versus the Teacher's 1.5M, keeps competitive accuracy while cutting parameters by 48% and inference time by roughly half. The paper attributes these gains to synergy among its components rather than to any single module.

Load-bearing premise

The load-bearing premise is that the Combined Attention mechanism's 'Proxima Attention' component can be implemented as intended, yet the paper never defines it beyond saying keys and values are linear projections of the input, so if that component is not reproducible the reported error rates cannot be reproduced either.

Editorial extensions

If this is right

  • If the reported error rates hold, modern English and French handwriting recognition can move below 1.5% character error, and historical Bentham material below 2.1%, with models small enough for practical deployment.
  • The Student's 48% parameter reduction and roughly halved inference time imply that Teacher-level accuracy can be carried into resource-constrained digitization settings without a dedicated GPU at every processing node.
  • The ablation results imply that knowledge distillation, curriculum learning, ensemble learning, multi-task learning, and post-processing each contribute, so HTR systems should be designed as multi-stage pipelines rather than single architectures.
  • T5 post-processing correcting about 90% of residual errors suggests that a language-model stage can act as a general error-repair layer on top of an HTR front end.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not separately ablate Multi-Head Self-Attention from Proxima Attention, so its own evidence cannot tell which half of the Combined Attention is responsible for the gains; a reader should treat the fused module as the hypothesis to test.
  • The undefined Proxima Attention leaves open the possibility that the reported gains come from the training regimen rather than the attention mechanism, so an ablation that replaces Proxima Attention with a second standard attention head would be decisive.
  • The 23.4% CER reduction attributed to T5 post-processing suggests a testable extension: plugging the fine-tuned T5 into an unrelated HTR system and measuring whether the error correction transfers.
  • Because the Student retains competitive accuracy at 0.75M parameters, further compression via quantization or pruning is a natural extension not explored in the paper.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 6 minor

Summary. The paper proposes HTR-JAND, a handwritten text recognition framework that combines a FullGatedConv2d/SE CNN encoder, BiLSTM sequence modeling, a 'Combined Attention' mechanism (multi-head self-attention fused with 'Proxima Attention'), knowledge distillation, curriculum/synthetic/multi-task training, and a fine-tuned T5 post-processor. The authors report state-of-the-art CERs of 1.23%, 1.02%, and 2.02% on IAM, RIMES, and Bentham, with a 0.75M-parameter student model. The central claim is that the full pipeline is both more accurate and more parameter-efficient than existing HTR systems.

Significance. If properly supported, the framework would be a useful contribution to efficient HTR: it targets both accuracy and deployment cost, evaluates on several standard benchmarks, and promises released code and pretrained models. The paper also contains a reasonable high-level ablation strategy. However, the current manuscript does not establish the state-of-the-art claim because the headline results include external post-processing and an unspecified 'Proxima Attention' component, and the knowledge-distillation objective is defined inconsistently. The efficiency claim is not assessed for the complete inference pipeline. The ingredients are valuable, but the technical presentation and experimental controls need substantial revision before the claims can be accepted.

major comments (6)
  1. [§III-B4, Eqs. (16)–(18)] Proxima Attention is not defined. Eq. (17) only gives K = XW_K and V = XW_V; there is no definition of Q, no dynamic query update, and no expression for O_Proxima. Eq. (18) then uses O_Proxima in the combined output. The sentence 'Proxima Attention commuted using Eq. 16' does not specify how Proxima differs from standard multi-head attention. Since Proxima Attention is listed as one of the three key contributions and is credited in the conclusion, this gap blocks independent reimplementation and verification of the architecture.
  2. [§III-C1 and §III-D, Eqs. (23) and (28)] The knowledge-distillation loss is defined inconsistently. Eq. (23) defines Lkd = KL(softmax(zS_interp/τ), softmax(zT/τ)) · τ², while Eq. (28) defines Lkd = KL(softmax(zT/τ), softmax(zS/τ)) without the τ² factor and without the interpolation of zS. The arguments are reversed and the scaling differs. The actual training objective is therefore ambiguous, and this discrepancy affects the core distillation mechanism that the paper claims as a contribution.
  3. [§V and §VI-C, Table V] The ablation does not isolate the T5 post-processor. Section V reports that the fine-tuned T5-small model reduces CER by 23.4% on average, and the headline numbers in Table V appear after the '+LBC' column. The table has no row or column that separates the raw HTR model output from the T5-corrected output, and the text does not state whether '+LBC' includes the T5 stage or is a separate lexicon-correction step. Without this separation, the contribution of the proposed HTR architecture to the final CERs cannot be identified.
  4. [§VI-D, Table VI] The state-of-the-art comparison is not controlled. 'Ours (+LBC)' is compared with baselines that do not use an equivalent post-processing or lexicon-correction stage, so the reported advantage could come entirely from the external correction. A like-for-like comparison that reports the HTR model's raw CER, or that applies the same post-processing to the baselines, is required to support the 'state-of-the-art' claim. In addition, Table VI covers only IAM and RIMES, not Bentham, despite the abstract claiming state-of-the-art results on Bentham, and recent strong baselines are omitted.
  5. [§VI-C, Table V and §VI-F, Table VII] The claimed efficiency is not measured for the full deployed system. The abstract's 0.75M-parameter Student does not include the T5-small post-processor (60M parameters) or the ensemble/LBC stages used for the headline results. Table VII reports the Student model at CER 4.12% with 0.75M parameters, while the headline 1.23% is achieved only after the additional stages. The parameter count and inference time of the complete pipeline should be reported, and the contribution of each stage to both accuracy and cost should be separated.
  6. [§VI-C, Table V] The ablation is internally inconsistent regarding multi-task learning. The text states that Table V shows the effects of Multi-Task Learning (MTL) and reports MTL improvements (e.g., 46.17% CER improvement on Saint Gall), but the table has no MTL column. The experimental support for MTL as a component is therefore missing from the reported data.
minor comments (6)
  1. [§III-A] There is a typo: 'including including' should be 'including'.
  2. [§III-B4] 'Proxima Attention commuted using Eq. 16' should read 'computed', and the sentence should be rewritten to actually define the operation.
  3. [§VI-C] The dataset name is written as 'Rimes' in one place and 'RIMES' elsewhere; please use a consistent spelling.
  4. [References] References [14] and [35] are duplicate entries of the same paper ('Transformer for handwritten text recognition using bidirectional post-decoding').
  5. [Table II] The header 'V ocab' contains an errant space; it should be 'Vocab'.
  6. [Eq. (16)] The output projection W^O in MultiHead is not defined; please specify its dimensions or state that it is the standard multi-head output projection.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation chain found; the paper's claims are empirical and not forced by construction.

full rationale

The paper does not derive its headline CERs from a fitted parameter or from a self-referential definition. The model is evaluated on external benchmarks (IAM, RIMES, Bentham), and the ablation in Table V is a standard component-wise empirical study. The T5 post-processor described in Section V is a separately trained second-stage component; while the paper does not specify train/test splitting for the paired prediction-ground-truth data, this is a possible leakage/evaluation-fairness concern rather than an in-text equivalence by construction. The undefined Proxima Attention in Eqs. 17-18 is a reproducibility gap, not circularity: O_Proxima is simply not specified, so Eq. 18 cannot be reduced to its own input. The self-citations [15], [16] appear only in related-work and architectural-motivation contexts; they are not load-bearing for the reported results. No step of the derivation chain renames a known result or imports a uniqueness theorem from the authors' prior work. Therefore, no significant circularity is present.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The central claim rests on several unspecified hyperparameters and an undefined attention mechanism. The T5 post-processor is a separate 60M-parameter model that contributes to the final results but is not counted in the efficiency comparison.

free parameters (5)
  • alpha, beta, gamma, delta = alpha from 0.7 to 0.4, gamma from 0.2 to 0.5, delta=0.1, beta chosen to sum to 1
    Loss weights in Eq. 19 are dynamically adjusted during training, but the exact schedule is not specified.
  • temperature tau
    Used in the knowledge distillation loss (Eq. 23, Eq. 28), but no value is given.
  • r0 and rmax = r0=0.1, rmax=0.4
    Initial and maximum synthetic data ratios in the curriculum (Eq. 33).
  • epsilon in oversampling = not specified
    Small constant to prevent division by zero in Eq. 3.
  • confidence threshold theta = not specified
    Threshold for applying T5 corrections in Algorithm 5.
assumptions (4)
  • standard math Backpropagation and gradient descent work as expected for the described deep learning architectures.
    The paper assumes standard deep learning training behavior without proof.
  • domain assumption The character set unification across datasets preserves the recognition task semantics.
    Removing infrequent characters changes the label set; the paper assumes this does not harm evaluation comparability.
  • domain assumption Rendered synthetic text images are sufficiently realistic to improve real-data recognition.
    The synthetic data generation (Algorithm 2) assumes that fonts and augmentations transfer to handwriting.
  • domain assumption The T5 post-processor can correct HTR errors without introducing new errors that change the ground-truth text.
    The T5 model is trained on HTR predictions and ground truth, and the paper assumes it generalizes to test predictions.
invented entities (1)
  • Proxima Attention
    purpose: To improve sequence modeling when combined with multi-head self-attention.
    No external citation or mathematical definition is provided. The paper only states it is 'commuted using Eq. 16' with dynamic query updates, which is insufficient.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HTR-JAND: Handwritten Text Recognition with Joint Attention Network and Knowledge Distillation." pith.science (2026). https://pith.science/paper/YB6VRHCR

@misc{pith2026241218524,
  author       = {Pith},
  title        = {Pith review of: HTR-JAND: Handwritten Text Recognition with Joint Attention Network and Knowledge Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YB6VRHCR}},
  note         = {Machine review of arXiv:2412.18524}
}
read the original abstract

Despite significant advances in deep learning, current Handwritten Text Recognition (HTR) systems struggle with the inherent complexity of historical documents, including diverse writing styles, degraded text quality, and computational efficiency requirements across multiple languages and time periods. This paper introduces HTR-JAND (HTR-JAND: Handwritten Text Recognition with Joint Attention Network and Knowledge Distillation), an efficient HTR framework that combines advanced feature extraction with knowledge distillation. Our architecture incorporates three key components: (1) a CNN architecture integrating FullGatedConv2d layers with Squeeze-and-Excitation blocks for adaptive feature extraction, (2) a Combined Attention mechanism fusing Multi-Head Self-Attention with Proxima Attention for robust sequence modeling, and (3) a Knowledge Distillation framework enabling efficient model compression while preserving accuracy through curriculum-based training. The HTR-JAND framework implements a multi-stage training approach combining curriculum learning, synthetic data generation, and multi-task learning for cross-dataset knowledge transfer. We enhance recognition accuracy through context-aware T5 post-processing, particularly effective for historical documents. Comprehensive evaluations demonstrate HTR-JAND's effectiveness, achieving state-of-the-art Character Error Rates (CER) of 1.23\%, 1.02\%, and 2.02\% on IAM, RIMES, and Bentham datasets respectively. Our Student model achieves a 48\% parameter reduction (0.75M versus 1.5M parameters) while maintaining competitive performance through efficient knowledge transfer. Source code and pre-trained models are available at \href{https://github.com/DocumentRecognitionModels/HTR-JAND}{Github}.

Figures

Figures reproduced from arXiv: 2412.18524 by the authors.

Figure 1
Figure 1. Sample images from different datasets, demonstrating [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Distribution of character frequencies across the combined datasets. Note the removal of infrequent characters such as [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Proposed HTR Model Architecture: Data flow through CNN feature extraction, LSTM sequence modeling, and Combined [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Overview of our proposed knowledge distillation framework for handwritten text recognition (HTR). [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the model’s attention heatmaps for the sample predictions. The heatmaps demonstrate the character-level [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Class probabilities heatmap for character alignment in the Rimes dataset. Darker cells along the diagonal indicate [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Frame from animated attention visualization. The animation shows the model’s adaptive focus as it processes each [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. FP-THD: Full page transcription of historical documents

    cs.CV 2026-01 conditional novelty 4.0 of 10

    FP-THD combines ParseNet layout analysis with an MAE-ViT line OCR to transcribe full pages of medieval Latin while preserving archaic characters, achieving 1.78% CER on a 10-page printed test set.

Reference graph

Works this paper leans on

41 extracted references · 36 canonical work pages · cited by 1 Pith paper

  1. [1]

    Ground truth creation for handwriting recognition in historical doc- uments,

    A. Fischer, E. Inderm ¨uhle, H. Bunke, G. Viehhauser, and M. Stolz, “Ground truth creation for handwriting recognition in historical doc- uments,” in Proceedings of the 9th IAPR International Workshop on Document Analysis Systems , 2010, pp. 3–10

    work page 2010

  2. [2]

    Gated convolutional recurrent neural net- works for multilingual handwriting recognition,

    T. Bluche and R. Messina, “Gated convolutional recurrent neural net- works for multilingual handwriting recognition,” in 2017 14th IAPR International Conference on Document Analysis and Recognition (IC- DAR), vol. 1. IEEE, 2017, pp. 646–651

    work page 2017

  3. [3]

    A novel connectionist system for unconstrained hand- writing recognition,

    A. Graves, M. Liwicki, S. Fern ´andez, R. Bertolami, H. Bunke, and J. Schmidhuber, “A novel connectionist system for unconstrained hand- writing recognition,” IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, vol. 31, no. 5, pp. 855–868, 2008

    work page 2008

  4. [4]

    Dynamic and contextual information in hmm modeling for handwritten word recognition,

    A.-L. Bianne-Bernard, F. Menasri, R. A.-H. Mohamad, C. Mokbel, C. Kermorvant, and L. Likforman-Sulem, “Dynamic and contextual information in hmm modeling for handwritten word recognition,” IEEE transactions on pattern analysis and machine intelligence , vol. 33, no. 10, pp. 2066–2080, 2011

    work page 2011

  5. [5]

    Improving cnn-rnn hybrid networks for handwriting recognition,

    K. Dutta, P. Krishnan, M. Mathew, and C. Jawahar, “Improving cnn-rnn hybrid networks for handwriting recognition,” in2018 16th International Conference on Frontiers in Handwriting Recognition (ICFHR) . IEEE, 2018, pp. 80–85

    work page 2018

  6. [6]

    Markov models for offline handwriting recognition: a survey,

    T. Pl ¨otz and G. A. Fink, “Markov models for offline handwriting recognition: a survey,” International Journal on Document Analysis and Recognition (IJDAR), vol. 12, no. 4, pp. 269–298, 2009

    work page 2009

  7. [7]

    Transcription align- ment of latin manuscripts using hidden markov models,

    A. Fischer, V . Frinken, A. Forn ´es, and H. Bunke, “Transcription align- ment of latin manuscripts using hidden markov models,” in Proceedings of the 2011 Workshop on Historical Document Imaging and Processing , 2011, pp. 29–36

    work page 2011

  8. [8]

    An Efficient End-to-End Neural Model for Handwritten Text Recognition

    A. Chowdhury and L. Vig, “An efficient end-to-end neural model for handwritten text recognition,” arXiv preprint arXiv:1807.07965 , 2018

    work page Pith review arXiv 2018

Show all 41 references
  1. [9]

    Evaluating sequence- to-sequence models for handwritten text recognition,

    J. Michael, R. Labahn, T. Gr ¨uning, and J. Z¨ollner, “Evaluating sequence- to-sequence models for handwritten text recognition,” in Proc. Int. Conf. Document Analysis and Recognition (ICDAR) . IEEE, 2019, pp. 1286– 1293

    2019

  2. [10]

    Are multidimensional recurrent layers really necessary for handwritten text recognition?

    J. Puigcerver, “Are multidimensional recurrent layers really necessary for handwritten text recognition?” in 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) , vol. 1. IEEE, 2017, pp. 67–72

    2017

  3. [11]

    Enhancing handwritten text recognition with n-gram sequence decomposition and multitask learning,

    V . Tassopoulou, G. Retsinas, and P. Maragos, “Enhancing handwritten text recognition with n-gram sequence decomposition and multitask learning,” in Proc. 25th Int. Conf. Pattern Recognition (ICPR) . IEEE, 2021, pp. 10 555–10 560

    2021

  4. [12]

    Accurate, data- efficient, unconstrained text recognition with convolutional neural net- works,

    M. Yousef, K. F. Hussain, and U. S. Mohammed, “Accurate, data- efficient, unconstrained text recognition with convolutional neural net- works,” Pattern Recognition, vol. 108, p. 107482, 2020

    2020

  5. [13]

    Pay attention to what you read: Non-recurrent hand-written text-line recognition,

    L. Kang, D. Coquenet, S. R. Adam, and T. Paquet, “Pay attention to what you read: Non-recurrent hand-written text-line recognition,” in2020 25th International Conference on Pattern Recognition (ICPR) . IEEE, 2020, pp. 10 355–10 362

    2020

  6. [14]

    Transformer for handwritten text recognition using bidirectional post-decoding,

    C. Wick, J. Z ¨ollner, and T. Gr ¨uning, “Transformer for handwritten text recognition using bidirectional post-decoding,” in Proc. Int. Conf. Document Analysis and Recognition . Springer, 2021, pp. 112–126

    2021

  7. [15]

    Resnest-transformer: Joint attention segmentation-free for end-to-end handwriting paragraph recognition model,

    M. Hamdan and M. Cheriet, “Resnest-transformer: Joint attention segmentation-free for end-to-end handwriting paragraph recognition model,” Array, vol. 19, p. 100300, Sep. 2023

    2023

  8. [16]

    Refocus attention span networks for handwriting line recognition,

    M. Hamdan, H. Chaudhary, A. Bali, and M. Cheriet, “Refocus attention span networks for handwriting line recognition,” International Journal on Document Analysis and Recognition (IJDAR) , pp. 1–17, 2022

    2022

  9. [17]

    HTR-Flor: A Deep Learning System for Offline Handwritten Text Recognition,

    A. F. de Sousa Neto, B. L. D. Bezerra, A. H. Toselli, and E. B. Lima, “HTR-Flor: A Deep Learning System for Offline Handwritten Text Recognition,” in Proc. 33rd SIBGRAPI Conf. Graphics, Patterns and Images (SIBGRAPI). IEEE, 2020, pp. 07–10

    2020

  10. [18]

    Learning from multiple teacher networks,

    S. You, C. Xu, C. Xu, and D. Tao, “Learning from multiple teacher networks,” in Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2017, pp. 1285– 1294

    2017

  11. [19]

    Data augmentation for recognition of handwritten words and lines using a cnn-lstm network,

    C. Wigington, S. Stewart, B. Davis, B. Barrett, and S. Cohen, “Data augmentation for recognition of handwritten words and lines using a cnn-lstm network,” in 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) . IEEE, Nov. 2017, pp. 639–645

    2017

  12. [20]

    Connection- ist temporal classification: Labelling unsegmented sequence data with recurrent neural networks,

    A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connection- ist temporal classification: Labelling unsegmented sequence data with recurrent neural networks,” in Proceedings of the 23rd International Conference on Machine Learning , 2006, pp. 369–376

    2006

  13. [21]

    Best practices for a handwritten text recognition system,

    G. Retsinas, G. Sfikas, B. Gatos, and C. Nikou, “Best practices for a handwritten text recognition system,” arXiv preprint arXiv:2404.11339, 2024

    2024 arXiv

  14. [22]

    Improving offline handwritten text recognition with hybrid hmm/ann models,

    S. Espana-Boquera, M. J. Castro-Bleda, J. Gorbe-Moya, and F. Zamora- Martinez, “Improving offline handwritten text recognition with hybrid hmm/ann models,” IEEE transactions on pattern analysis and machine intelligence, vol. 33, no. 4, pp. 767–779, 2010

    2010

  15. [23]

    Offline continuous handwriting recognition using sequence to sequence neural networks,

    J. Sueiras, V . Ruiz, A. Sanchez, and J. F. Velez, “Offline continuous handwriting recognition using sequence to sequence neural networks,” Neurocomputing, vol. 289, pp. 119–128, 2018

    2018

  16. [24]

    Sequence- to-sequence domain adaptation network for robust text image recogni- tion,

    Y . Zhang, S. Nie, W. Liu, X. Xu, D. Zhang, and H. T. Shen, “Sequence- to-sequence domain adaptation network for robust text image recogni- tion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 2740–2749

    2019

  17. [25]

    Sequence-to-sequence contrastive learn- ing for text recognition,

    A. Aberdam, R. Litman, S. Tsiper, O. Anschel, R. Slossberg, S. Mazor, R. Manmatha, and P. Perona, “Sequence-to-sequence contrastive learn- ing for text recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 15 302–15 312

    2021

  18. [26]

    Neural machine translation by jointly learning to align and translate,

    D. Bahdanau, K. Cho, and Y . Bengio, “Neural machine translation by jointly learning to align and translate,” arXiv preprint arXiv:1409.0473, 2014

    2014 arXiv

  19. [27]

    Stacked attention networks for image question answering,

    Z. Yang, X. He, J. Gao, L. Deng, and A. Smola, “Stacked attention networks for image question answering,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 21– 29

    2016

  20. [28]

    Draw: A recurrent neural network for image generation,

    K. Gregor, I. Danihelka, A. Graves, D. Rezende, and D. Wierstra, “Draw: A recurrent neural network for image generation,” in International Conference on Machine Learning . PMLR, 2015, pp. 1462–1471

    2015

  21. [29]

    Pixelsnail: An improved autoregressive generative model,

    X. Chen, N. Mishra, M. Rohaninejad, and P. Abbeel, “Pixelsnail: An improved autoregressive generative model,” in International Conference on Machine Learning . PMLR, 2018, pp. 864–872

    2018

  22. [30]

    Long short-term memory-networks for machine reading,

    J. Cheng, L. Dong, and M. Lapata, “Long short-term memory-networks for machine reading,” arXiv preprint arXiv:1601.06733 , 2016

    2016 arXiv

  23. [31]

    A decompos- able attention model for natural language inference,

    A. P. Parikh, O. T ¨ackstr¨om, D. Das, and J. Uszkoreit, “A decompos- able attention model for natural language inference,” arXiv preprint arXiv:1606.01933, 2016

    2016 arXiv

  24. [32]

    Cnn-n-gram for handwriting word recog- nition,

    A. Poznanski and L. Wolf, “Cnn-n-gram for handwriting word recog- nition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2305–2314

    2016

  25. [33]

    Handwriting recog- nition of historical documents with few labeled data,

    E. Chammas, C. Mokbel, and L. Likforman-Sulem, “Handwriting recog- nition of historical documents with few labeled data,” in 2018 13th IAPR International Workshop on Document Analysis Systems (DAS) . IEEE, 2018, pp. 43–48

    2018

  26. [34]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in Neural Information Processing Systems , vol. 30, 2017

    2017

  27. [35]

    Transformer for handwritten text recognition using bidirectional post-decoding,

    C. Wick, J. Z ¨ollner, and T. Gr ¨uning, “Transformer for handwritten text recognition using bidirectional post-decoding,” in Document Analysis and Recognition – ICDAR 2021 . Cham, Switzerland: Springer, Sep. 2021, pp. 112–126

    2021

  28. [36]

    The iam-database: an english sentence database for offline handwriting recognition,

    U.-V . Marti and H. Bunke, “The iam-database: an english sentence database for offline handwriting recognition,” International Journal on Document Analysis and Recognition , vol. 5, no. 1, pp. 39–46, 2002

    2002

  29. [37]

    Results of the rimes evaluation campaign for handwritten mail processing,

    E. Grosicki, M. Carr ´e, J.-M. Brodin, and E. Geoffrois, “Results of the rimes evaluation campaign for handwritten mail processing,” in 2009 10th International Conference on Document Analysis and Recognition . IEEE, 2009, pp. 941–945

    2009

  30. [38]

    Building a volunteer community: results and findings from transcribe bentham,

    T. Causer and V . Wallace, “Building a volunteer community: results and findings from transcribe bentham,” Digital Humanities Quarterly, vol. 6, no. 2, 2012

    2012

  31. [39]

    Exploring the limits of transfer learning with a unified text-to-text transformer,

    C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y . Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,” Journal of Machine Learning Research, vol. 21, no. 140, pp. 1–67, 2020

    2020

  32. [40]

    Deformation- invariant networks for handwritten text recognition,

    G. Retsinas, G. Sfikas, C. Nikou, and P. Maragos, “Deformation- invariant networks for handwritten text recognition,” in Proc. IEEE Int. Conf. Image Processing (ICIP) . IEEE, 2021, pp. 949–953

    2021

  33. [41]

    Are multidimensional recurrent layers really necessary for handwritten text recognition?

    J. Puigcerver, “Are multidimensional recurrent layers really necessary for handwritten text recognition?” in Proc. 14th IAPR Int. Conf. Document Analysis and Recognition (ICDAR) , vol. 1. IEEE, 2017, pp. 67–72

    2017

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.