Pith. sign in

REVIEW 4 major objections 5 minor 72 references

MegaHan97K: A Large-Scale Dataset for Mega-Category Chinese Character Recognition with over 97K Categories

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read MegaHan97K provides 97,455 Chinese character categories, six times more than any previous dataset.

desk verdict A genuinely new scale for Chinese character datasets with real benchmark findings, but the label-fidelity guarantee is thinner than the central claim requires. read the letter →

arxiv 2506.04807 v1 pith:CDNG3PTD submitted 2025-06-05 cs.CV

Yuyi Zhang , Yongxin Shi , Peirong Zhang , Yixin Zhao , Zhenhua Yang , Lianwen Jin This is my paper
classification cs.CV
keywords Chinesecharacterrecognitionmega-categoryclassificationGB18030-2022zero-shotlearningopticalfontsynthesishistoricaldocumentanalysisdataset
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

This paper introduces MegaHan97K, a dataset of 97,455 Chinese character categories assembled from handwritten, historical, and synthetic samples. The authors claim it is the largest such dataset, at least six times larger than any predecessor, and the first to cover the full GB18030-2022 standard of 87,887 categories plus additional Unicode characters. They argue that balanced per-category sampling removes the long-tail distribution problem that has made rare and variant characters difficult to study. Benchmarking shows state-of-the-art models reach 92.32% accuracy in closed-set recognition but only 79.04% in zero-shot recognition, which the authors present as evidence that mega-category recognition is a genuinely new challenge. A sympathetic reader would care because rare and archaic characters matter for digitizing historical documents, and no prior dataset allowed models to be trained or evaluated on this scale.

What carries the argument

The central object is the dataset itself, organized as a balanced category-by-subset matrix. Its construction has three load-bearing components: a lexicon built from GB18030-2022 and Unicode with IDS-based deduplication; data acquisition through volunteer handwriting, historical corpora, and FontDiffuser-generated synthetic images; and train/test splits designed for both general and zero-shot recognition. FontDiffuser, a diffusion-based font generation model, supplies the synthetic samples that drive the largest accuracy gains, while the Ideographic Description Sequence representation provides the structural prior used for deduplication and similarity analysis.

What would settle it

Recruit expert annotators to inspect a random sample of categories from the synthetic and Kangxi subsets, comparing each image with the ground-truth character and estimating the label error rate separately for characters with many strokes; if the error rate is concentrated in rare or complex characters, the coverage claim and the reported accuracies are not reliable.

Watch

Extended reading notes

Core claim

MegaHan97K is a dataset of 97,455 Chinese character categories, constructed by taking the union of the GB18030-2022 standard with Unicode 15.0.0 and removing ambiguous duplicate pairs through Ideographic Description Sequence comparison. It contains three subsets: a handwritten subset covering all categories, a historical subset built from M5HisDoc and Kangxi Dictionary images, and a synthetic subset of 35 FontDiffuser-generated samples per category. The authors report that the best closed-set model reaches 92.32% accuracy, the best zero-shot model reaches 79.04%, and that the dataset eliminates the long-tail problem that plagues earlier datasets. The paper's central claim is that this is the first dataset large enough to support mega-category Chinese character recognition, and that the performance gap between closed-set and zero-shot settings defines a new research problem.

Load-bearing premise

Every image in every category really shows the character named by its label, including the synthetic samples generated by FontDiffuser and the crawled Kangxi images; the paper checked only 10% of synthesized images and reports synthesis failures on complex characters.

Editorial extensions

If this is right

  • If the dataset is as claimed, researchers can train and evaluate Chinese-character recognizers on 97,455 classes, roughly six times the previous maximum, making mega-category OCR experimentally accessible.
  • Models trained on MegaHan97K transfer to existing handwriting and historical-document benchmarks, raising macro accuracy for rare characters on M5HisDoc, CASIA-HWDB, and CASIA-AHCDB.
  • The closed-set to zero-shot accuracy gap becomes a concrete benchmark target for methods that exploit radicals, strokes, or glyph-level priors.
  • The roughly 22-percentage-point average gain from adding synthetic data indicates that font synthesis is a viable data source, but its failures on complex and similar characters set a limit that further synthesis alone may not break.

Reading between the lines

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

  • A natural extension of the paper's logic is a continually expanding open-lexicon evaluation: as Unicode adds new codepoints, the zero-shot split can grow, and models can be reassessed without retraining on the new categories.
  • Because the authors measure confusion via IDS edit distance, the dataset could double as a glyph-similarity benchmark, letting other fine-grained recognition tasks borrow its error analysis.
  • The label-fidelity risk is asymmetric: if FontDiffuser silently corrupts rare characters, the coverage claim of 97,455 valid categories is at stake even though aggregate accuracy may move only slightly.
  • Combining MegaHan97K with text-line or page-level data, as the discussion suggests, could turn the character-level resource into a restoration tool for damaged ancient documents; that application is suggested by the paper but not implemented.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces MegaHan97K, a dataset for Chinese character recognition containing 97,455 character categories, assembled from a handwritten subset (collected from 94 volunteers), a historical subset (M5HisDoc and web-crawled Kangxi Dictionary images), and a synthetic subset (rendered by FontDiffuser). The authors claim this is the first dataset to fully cover GB18030-2022 and the largest character-class dataset to date, and they benchmark nine methods in closed-set and zero-shot settings, reporting a best closed-set accuracy of 92.32% (HierCode) and a best zero-shot accuracy of 79.04% (CCR-CLIP). They also provide cross-validation experiments with existing datasets and an analysis of synthetic-data effects.

Significance. If the data-quality claims are validated, MegaHan97K would be a valuable community resource: it is the first mega-category Chinese character dataset, it covers the full GB18030-2022 standard, and its benchmark results quantify realistic challenges (storage growth, morphologically similar characters, and zero-shot difficulties) that smaller datasets cannot reveal. The paper ships a public release, benchmarks nine diverse methods, and includes cross-dataset generalization experiments, which are useful contributions. However, the central claims depend on label fidelity across 97,455 categories, and the manuscript provides only weak evidence for that fidelity while exhibiting internal inconsistencies in category counts and balance statistics. The dataset itself is promising, but the supporting evidence needs substantial strengthening before the headline claims can be accepted.

major comments (4)
  1. [Section 3.2, Section 4.5] The label-fidelity check is insufficient relative to the central claim that all 97,455 categories are correctly labeled. Section 3.2 states that only 10% of the synthesized data was inspected, and Section 4.5 documents that FontDiffuser produces stroke misplacement and missing details for complex characters. Since the synthetic subset is used to train and test on the full lexicon, the reported accuracies (e.g., 92.32% in Table 8) presuppose that generated images actually depict their labels; similarly, the web-crawled Kangxi images and volunteer handwriting are assumed to match their labels. The paper reports no per-subset label-error rate, no stratification by character complexity (e.g., stroke count or IDS complexity), and no release of the inspection protocol. Please provide a label-fidelity audit: sample all three subsets, report error rates broken down by complexity, and either filter or relabel erroneous images or quantify the effect of label noise on the benchmark numbers.
  2. [Tables 1, 3, 4, 5, 7 and Section 3.2] The category counts are internally inconsistent. Table 1 reports Handwritten Original as 97,455 categories and Handwritten Augmented as 96,362, while Section 3.2 says the handwritten subset 'covers all 97,455 categories' and Table 5 (general test set) reports Original 96,362 and Augmented 97,455, i.e., the reverse. Table 4 lists the M5HisDoc training component with 9,369 categories but Table 5 lists 16,151 M5HisDoc test categories, and the total-category overlap structure among Original, Augmented, M5HisDoc, and Kangxi is never defined. Because the paper's first contribution is the exact number of covered categories and the claim of full GB18030-2022 coverage, these discrepancies must be reconciled and each subset's category count and overlap stated unambiguously.
  3. [Abstract, Section 3.3, Table 1, Table 7] The 'balanced samples' claim is contradicted by the Kangxi subset. Table 1 shows the Kangxi subset has 47,064 categories and 49,438 instances, i.e., about 1.05 images per category on average, and Table 7 shows the zero-shot test Kangxi portion has 24,476 categories and 26,111 instances. This is precisely the one- or two-sample long-tail regime that Section 2.1 criticizes in existing datasets. The long-tail analysis in Section 4.4 relies on macro accuracy, but a test set with many singleton categories can yield macro and top-1 accuracies that are close without the underlying category-frequency distribution being balanced. Please restrict the balance claim to the subsets that actually have multiple samples per category, or report per-subset category-frequency histograms for all subsets.
  4. [Section 4.1, Tables 8, 10, 12, 13] The dataset split is not writer-disjoint, which can inflate handwritten recognition accuracy. Section 4.1 describes splitting by character category only and does not state whether samples from the same volunteer are kept entirely in the training or test set. With 94 volunteers and roughly 900K handwritten samples, a random per-sample split will place the same writer's characters in both training and test sets. Additionally, all benchmark tables report single runs without error bars or repeated-seed statistics, so it is unclear whether the observed differences (e.g., the 31.36% gap between CCR-CLIP and SideNet-DDCM in Table 10) are significant. Please specify the writer-disjoint split, report the number of writers per split, and provide at least mean and standard deviation over multiple runs for the main methods.
minor comments (5)
  1. [Abstract, Section 3.2] The name is misspelled as 'MetaHan97K' in the abstract and as 'MegeHan97K' in Section 3.2; please use 'MegaHan97K' consistently.
  2. [Section 4.2 vs Section 4.5] Section 4.2 defines morphologically similar characters as having an IDS edit distance of no more than three, while Section 4.5 says 'less than 3'; please unify the threshold.
  3. [Figure 3 caption and Section 3.2] Figure 3 calls HandWT-O and HandWT-A the pre-processed and post-processed versions of the handwritten subset, while Section 3.2 says the original and augmented parts are sourced from different real data; please clarify whether augmented samples are transformed copies of the original samples or independently collected real handwriting.
  4. [Section 3.2] The sentence 'We extracted 10% of the synthesized data for sampling inspection' should specify whether the 10% was stratified by category and whether inspection was performed on individual images or on per-category batches.
  5. [Table 1, Table 2] The data-collection year is listed as 2024 in Table 2, while the arXiv version is dated 2025; please update the year for consistency.

Circularity Check

1 steps flagged · score 4.0 of 10

Dataset size and coverage claims are self-contained, but the synthetic-sample volume (35 per category) is selected by monitoring the benchmark test accuracy, so the reported 92.32% headline accuracy is partially self-confirmatory rather than an independent prediction.

  1. fitted input called prediction [Section 4.5 (Investigating Synthetic Data Effects and Limitations), Figure 7, and Table 8; cross-referenced from Section 3.2]
    "When the number of synthetic data samples per category is fewer than 35, accuracy exhibits an upward trajectory as additional data is incrementally introduced. However, once the number reaches approximately 35, accuracy plateaus at around 92.4%, suggesting a point of diminishing returns. ... Consequently, we select 35 samples per category for our training set to balance optimal accuracy with reasonable training time."

    The paper chooses the dataset's synthetic-sample count (35 per category) after observing accuracy on the benchmark test set (plateau at ~92.4%), then reports the accuracy obtained with that choice (HierCode, 92.32%, Table 8) as an independent benchmark finding. No separate validation set is described for this choice, so the headline accuracy is the same quantity used to tune the dataset configuration. The final number is therefore a fitted value, not a fresh prediction about the dataset.

full rationale

The central claims—97,455 categories, GB18030-2022 coverage, and balanced subsets—are derived from the lexicon construction (GB18030-2022 + Unicode 15.0.0, minus IDS-duplicates) and from the data collection pipeline, not from the benchmark results, so they are not circular. The use of FontDiffuser [42] and M5HisDoc [15] involves self-citation, but both are prior published artifacts used as tools or data sources, and no load-bearing argument reduces to an unverified self-citation. The one genuine circular element is the synthetic-data volume choice: 35 samples per category is selected by watching the test accuracy plateau, and the same test accuracy is then reported as the headline benchmark (92.32%). This makes that particular accuracy partially self-confirmatory, reflecting test-set tuning of the dataset configuration. The long-tail analysis, zero-shot splitting, and cross-dataset experiments are independent measurements and are not circular. Overall, the core dataset contribution stands independently; the circularity is limited to one benchmark finding, hence the score of 4 rather than higher.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

Central claims rest on the correctness of labels sourced from volunteers, web crawling, and FontDiffuser synthesis, plus hand-picked sample counts and thresholds. No new physical or conceptual entities are introduced.

free parameters (5)
  • synthetic samples per category = 35
    Chosen from Figure 7 as the point where accuracy plateaus near 92.4% and training time stays acceptable; this determines the 3,314,000 synthetic instances in the training set.
  • handwritten samples per category = 20 for GB18030-2000, 5 for other categories
    Set in Section 3.2 to balance volunteer effort against coverage; 5 samples per rare category is small and acknowledged as a limitation in Section 7.
  • test samples per character = 5 handwritten samples
    Stated in Section 7 as a limitation; small test samples make per-category accuracy estimates noisy.
  • IDS similarity threshold = 3
    Used in Section 4.2 to define morphologically similar characters (IDS edit distance no more than 3) for error analysis; the reported 38.34% error proportion depends on this threshold.
  • augmentation intensities = 80% blur/jitter, 30 background types
    Augmentation recipe in Section 3.2 applied to handwritten and synthetic subsets; affects the domain gap between training and the augmented test subset.
assumptions (5)
  • domain assumption GB18030-2022 and Unicode 15.0.0 codepoint lists constitute the authoritative set of Chinese characters to recognize.
    The lexicon is built from these standards in Section 3.1; if the standards or their coverage are misread, the 97,455 count and 'full support' claim fail.
  • ad hoc to paper Identical IDS strings imply the characters are ambiguous duplicates that can be merged by keeping the smaller codepoint.
    Section 3.1 removes pairs based on IDS matching; this assumes IDS uniquely identifies intended glyph identity, which is not guaranteed for variant forms.
  • ad hoc to paper FontDiffuser renders every character in the lexicon with acceptable fidelity.
    Section 3.2 uses FontDiffuser to synthesize 35 samples per category for all 97,455 categories; only 10% of outputs were inspected and Section 4.5 documents stroke-level errors, so label fidelity is an unverified load-bearing premise.
  • domain assumption Volunteers wrote the prompted character and author verification caught all or nearly all label errors.
    Section 3.2 states each sample is verified by at least one author, but no inter-annotator agreement or residual error rate is reported.
  • ad hoc to paper HandWT-Ori and HandWT-Aug are different real data, as stated in Section 3.2.
    The paper claims original and augmented parts come from different real samples; if augmentation was applied to the same samples, train/test independence for the augmented subset is compromised.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MegaHan97K: A Large-Scale Dataset for Mega-Category Chinese Character Recognition with over 97K Categories." pith.science (2026). https://pith.science/paper/CDNG3PTD

@misc{pith2026250604807,
  author       = {Pith},
  title        = {Pith review of: MegaHan97K: A Large-Scale Dataset for Mega-Category Chinese Character Recognition with over 97K Categories},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CDNG3PTD}},
  note         = {Machine review of arXiv:2506.04807}
}
read the original abstract

Foundational to the Chinese language and culture, Chinese characters encompass extraordinarily extensive and ever-expanding categories, with the latest Chinese GB18030-2022 standard containing 87,887 categories. The accurate recognition of this vast number of characters, termed mega-category recognition, presents a formidable yet crucial challenge for cultural heritage preservation and digital applications. Despite significant advances in Optical Character Recognition (OCR), mega-category recognition remains unexplored due to the absence of comprehensive datasets, with the largest existing dataset containing merely 16,151 categories. To bridge this critical gap, we introduce MegaHan97K, a mega-category, large-scale dataset covering an unprecedented 97,455 categories of Chinese characters. Our work offers three major contributions: (1) MegaHan97K is the first dataset to fully support the latest GB18030-2022 standard, providing at least six times more categories than existing datasets; (2) It effectively addresses the long-tail distribution problem by providing balanced samples across all categories through its three distinct subsets: handwritten, historical and synthetic subsets; (3) Comprehensive benchmarking experiments reveal new challenges in mega-category scenarios, including increased storage demands, morphologically similar character recognition, and zero-shot learning difficulties, while also unlocking substantial opportunities for future research. To the best of our knowledge, the MetaHan97K is likely the dataset with the largest classes not only in the field of OCR but may also in the broader domain of pattern recognition. The dataset is available at https://github.com/SCUT-DLVCLab/MegaHan97K.

Figures

Figures reproduced from arXiv: 2506.04807 by the authors.

Figure 1
Figure 1. Comparing MegaHan97K with existing Chinese character datasets. The green and blue bubbles represent handwritten and historical datasets, respectively. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The example of ambiguous Chinese characters. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The visualization of the MegaHan97K dataset. HandWT-O (Original) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The user interface of the data acquisition website. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Illustration of data processing to simulate realistic scenarios. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Comparison of challenging samples synthesized by FontDiffuser and IF-Font with ground truth (GT). [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Effects of increasing synthetic sample sizes on accuracy. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Comparison of prediction results with and without the use of MegaHan97K. “HWDB" represents predictions from the model trained only on the HWDB [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Analysis of a damaged ancient document. The left side shows a sample of a damaged ancient document, while the right side provides a detailed qualitative [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Visualization analysis of the misclassified samples from the CCR-CLIP model. The misclassified components, such as radicals or strokes, are highlighted [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

72 extracted references · 69 canonical work pages

  1. [1]

    Zhou, Y .-M

    X.-D. Zhou, Y .-M. Zhang, F. Tian, H.-A. Wang, C.- L. Liu, Minimum-risk training for semi-markov condi- tional random fields with application to handwritten Chi- nese/japanese text recognition, Pattern Recognition 47 (5) (2014) 1904–1916

    work page 2014

  2. [2]

    C. Luo, L. Jin, Z. Sun, MORAN: A multi-object recti- fied attention network for scene text recognition, Pattern Recognition 90 (2019) 109–118

    work page 2019

  3. [3]

    S. Fang, H. Xie, Y . Wang, Z. Mao, Y . Zhang, Read like humans: Autonomous, bidirectional and iterative language modeling for scene text recognition, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 7098–7107

    work page 2021

  4. [4]

    C. Wang, C. Liu, Z. Fang, Z. Han, C. Liu, X. Yin, Open set text recognition technology, Journal of Image and Graphics 28 (06) (2023) 1767–1791

    work page 2023

  5. [5]

    C. Liu, L. Jin, X. Bai, X. Li, F. Yin, Frontiers of intelligent document analysis and recognition: review and prospects, Journal of Image and Graphics 28 (08) (2023) 2223–2252

    work page 2023

  6. [6]

    Su, J.-F

    Y .-M. Su, J.-F. Wang, A novel stroke extraction method for Chinese characters using Gabor filters, Pattern Recognition 36 (3) (2003) 635–647

    work page 2003

  7. [7]

    Zhang, Y

    X.-Y . Zhang, Y . Bengio, C.-L. Liu, Online and offline hand- written Chinese character recognition: A comprehensive study and new benchmark, Pattern Recognition 61 (2017) 348–360

    work page 2017

  8. [8]

    Z.-R. Wang, J. Du, J.-M. Wang, Writer-aware CNN for parsimonious HMM-based offline handwritten Chinese text recognition, Pattern Recognition 100 (2020) 107102

    work page 2020

Show all 72 references
  1. [9]

    Huang, X

    G. Huang, X. Luo, S. Wang, T. Gu, K. Su, Hippocampus- heuristic character recognition network for zero-shot learn- ing in Chinese character recognition, Pattern Recognition 130 (2022) 108818

    2022

  2. [10]

    of Industry, I

    M. of Industry, I. Technology, GB18030-2022,https:// openstd.samr.gov.cn/bzgk/gb/newGbInfo? hcno=A1931A578FE14957104988029B0833D3 (2022)

    2022

  3. [11]

    N. S. Administration, GB18030-2000, https:// openstd.samr.gov.cn/bzgk/gb/newGbInfo? hcno=4F885660EB8B3AC463C2ED336DB3B67B (2000). 11

    2000

  4. [12]

    C.-L. Liu, F. Yin, D.-H. Wang, Q.-F. Wang, CASIA Online and Offline Chinese Handwriting Databases, in: Interna- tional Conference on Document Analysis and Recognition (ICDAR), 2011, pp. 37–41

    2011

  5. [13]

    Y . Xu, F. Yin, D.-H. Wang, X.-Y . Zhang, Z. Zhang, C.- L. Liu, CASIA-AHCDB: A Large-Scale Chinese Ancient Handwritten Characters Database, in: International Con- ference on Document Analysis and Recognition (ICDAR), 2019, pp. 793–798

    2019

  6. [14]

    Zhang, L

    H. Zhang, L. Liang, L. Jin, Scut-hccdoc: A new bench- mark dataset of handwritten chinese text in unconstrained camera-captured documents, Pattern Recognition (2020) 107559

    2020

  7. [15]

    Y . Shi, C. Liu, D. Peng, C. Jian, J. Huang, L. Jin, M5HisDoc: A Large-scale Multi-style Chinese Historical Document Analysis Benchmark, in: Advances in Neural Information Processing Systems (NeurIPS), V ol. 36, 2023, pp. 78483–78495

    2023

  8. [16]

    H. Yu, X. Wang, B. Li, X. Xue, Chinese Text Recognition with A Pre-Trained CLIP-Like Model Through Image-IDS Aligning, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 11943– 11952

    2023

  9. [17]

    Zhang, Y

    Y . Zhang, Y . Zhu, D. Peng, P. Zhang, Z. Yang, Z. Yang, C. Yao, L. Jin, Hiercode: A lightweight hierarchical code- book for zero-shot chinese text recognition, Pattern Recog- nition 158 (2025) 110963

    2025

  10. [18]

    Ao, X.-H

    X. Ao, X.-H. Li, X.-Y . Zhang, C.-L. Liu, Prototype cal- ibration with synthesized samples for zero-shot chinese character recognition, in: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 6295–6299

    2024

  11. [19]

    Z. Li, Y . Huang, D. Peng, M. He, L. Jin, SideNet: Learning Representations from Interactive Side Information for Zero- Shot Chinese Character Recognition, Pattern Recognition 148 (2024) 110208

    2024

  12. [20]

    Zhang, J

    H. Zhang, J. Guo, G. Chen, C. Li, HCL2000 - A Large- scale Handwritten Chinese Character Database for Hand- written Character Recognition, in: International Confer- ence on Document Analysis and Recognition (ICDAR), 2009, pp. 286–290

    2009

  13. [21]

    T. Su, T. Zhang, D. Guan, Corpus-based HIT-MW Database for Offline Recognition of General-purpose Chi- nese Handwritten Text, International Journal on Document Analysis and Recognition 10 (1) (2007) 27–38

    2007

  14. [22]

    L. Jin, Y . Gao, G. Liu, Y . Li, K. Ding, SCUT- COUCH2009—a comprehensive online unconstrained Chi- nese handwriting database and benchmark evaluation, In- ternational Journal on Document Analysis and Recognition 14 (2011) 53–64

    2011

  15. [23]

    H. Yang, L. Jin, W. Huang, Z. Yang, S. Lai, J. Sun, Dense and Tight Detection of Chinese Characters in Historical Documents: Datasets and a Recognition Guided Detector, IEEE Access 6 (2018) 30174–30183

    2018

  16. [24]

    W. Ma, H. Zhang, L. Jin, S. Wu, J. Wang, Y . Wang, Joint Layout Analysis, Character Detection and Recognition for Historical Document Digitization, in: International Con- ference on Frontiers in Handwriting Recognition (ICFHR), 2020, pp. 31–36

    2020

  17. [25]

    Saini, D

    R. Saini, D. Dobson, J. Morrey, M. Liwicki, F. Simistira Li- wicki, ICDAR 2019 Historical Document Reading Chal- lenge on Large Structured Chinese Family Records, in: In- ternational Conference on Document Analysis and Recog- nition (ICDAR), 2019, pp. 1499–1504

    2019

  18. [26]

    T.-L. Yuan, Z. Zhu, K. Xu, C.-J. Li, T.-J. Mu, S.-M. Hu, A large Chinese text dataset in the wild, Journal of Computer Science and Technology 34 (2019) 509–521

    2019

  19. [27]

    Cire¸ san, U

    D. Cire¸ san, U. Meier, Multi-Column Deep Neural Net- works for offline handwritten Chinese character classifi- cation, in: International Joint Conference on Neural Net- works (IJCNN), 2015, pp. 1–6

    2015

  20. [28]

    X. Xiao, L. Jin, Y . Yang, W. Yang, J. Sun, T. Chang, Build- ing fast and compact convolutional neural networks for offline handwritten Chinese character recognition, Pattern Recognition 72 (2017) 72–81

    2017

  21. [29]

    Z. Li, Q. Wu, Y . Xiao, M. Jin, H. Lu, Deep Matching Network for Handwritten Chinese Character Recognition, Pattern Recognition 107 (2020) 107471

    2020

  22. [30]

    T.-Q. Wang, F. Yin, C.-L. Liu, Radical-Based Chinese Character Recognition via Multi-Labeled Learning of Deep Residual Networks, in: International Conference on Docu- ment Analysis and Recognition (ICDAR), V ol. 01, 2017, pp. 579–584

    2017

  23. [31]

    Ao, X.-Y

    X. Ao, X.-Y . Zhang, C.-L. Liu, Cross-modal prototype learning for zero-shot handwritten character recognition, Pattern Recognition 131 (2022) 108859

    2022

  24. [32]

    C. Liu, C. Yang, X.-C. Yin, Open-Set Text Recognition via Character-Context Decoupling, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 4523–4532

    2022

  25. [33]

    C. Liu, C. Yang, H.-B. Qin, X. Zhu, C.-L. Liu, X.-C. Yin, Towards open-set text recognition via label-to-prototype learning, Pattern Recognition 134 (2023) 109109

    2023

  26. [34]

    Ao, X.-H

    X. Ao, X.-H. Li, X.-Y . Zhang, C.-L. Liu, Prototype Cali- bration with Synthesized Samples for Zero-Shot Chinese Character Recognition, in: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 6295–6299. 12

    2024

  27. [35]

    W. Wang, J. Zhang, J. Du, Z.-R. Wang, Y . Zhu, DenseRAN for Offline Handwritten Chinese Character Recognition, in: International Conference on Frontiers in Handwriting Recognition (ICFHR), 2018, pp. 104–109

    2018

  28. [36]

    T. Wang, Z. Xie, Z. Li, L. Jin, X. Chen, Radical Aggrega- tion Network for Few-Shot Offline Handwritten Chinese Character Recognition, Pattern Recognition Letters 125 (2019) 821–827

    2019

  29. [37]

    Luo, D.-H

    G.-F. Luo, D.-H. Wang, X. Du, H.-Y . Yin, X.-Y . Zhang, S. Zhu, Self-information of Radicals: A new clue for zero- shot Chinese character recognition, Pattern Recognition 140 (2023) 109598

    2023

  30. [38]

    Z. Cao, J. Lu, S. Cui, C. Zhang, Zero-Shot Handwrit- ten Chinese Character Recognition with Hierarchical De- composition Embedding, Pattern Recognition 107 (2020) 107488

    2020

  31. [39]

    J. Chen, B. Li, X. Xue, Zero-Shot Chinese Character Recognition with Stroke-Level Decomposition, in: Pro- ceedings of the International Joint Conference on Artificial Intelligence (IJCAI), 2021, pp. 615–621

    2021

  32. [40]

    X. Zu, H. Yu, B. Li, X. Xue, Chinese Character Recogni- tion with Augmented Character Profile Matching, in: Pro- ceedings of ACM International Conference on Multimedia (ACM MM), 2022, p. 6094–6102

    2022

  33. [41]

    N. S. Administration, GB18030-2005, https:// openstd.samr.gov.cn/bzgk/gb/newGbInfo? hcno=C344D8D120B341A8DD328954A9B27A99 (2005)

    2005

  34. [42]

    Z. Yang, D. Peng, Y . Kong, Y . Zhang, C. Yao, L. Jin, FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning, in: Proceedings of the AAAI Con- ference on Artificial Intelligence (AAAI), V ol. 38, 2024...

    2024

  35. [43]

    K. He, X. Zhang, S. Ren, J. Sun, Deep Residual Learn- ing for Image Recognition, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016

    2016

  36. [44]

    Zhang, B

    Y . Zhang, B. Kang, B. Hooi, S. Yan, J. Feng, Deep long- tailed learning: A survey, IEEE Transactions on Pattern Analysis and Machine Intelligence 45 (9) (2023) 10795– 10816

    2023

  37. [45]

    X. Chen, X. Ke, W. Guo, If-font: Ideographic descrip- tion sequence-following font generation, Advances in Neu- ral Information Processing Systems (NeurIPS) 37 (2025) 14177–14199

    2025

  38. [46]

    Y . T. Zheng, X. L. Li, Z. X. Yin, G. Gao, Y . Weng, Multi- feature fusion based automatic reconstruction in related to Chinese ancient manuscript fragments of dunhuang, Journal of Image and Graphics 28 (08) (2023) 2330–2342

    2023

  39. [47]

    Z. Yang, D. Peng, Y . Shi, Y . Zhang, C. Liu, L. Jin, Pre- dicting the Original Appearance of Damaged Historical Documents, Proceedings of the AAAI Conference on Arti- ficial Intelligence (2025)

    2025

  40. [48]

    Chang, Y

    J. Chang, Y . Gu, Y . Zhang, Y .-F. Wang, C. Innovation, Chinese Handwriting Imitation with Hierarchical Gener- ative Adversarial Network, in: British Machine Vision Conference (BMVC), 2018, p. 290

    2018

  41. [49]

    P. Lyu, X. Bai, C. Yao, Z. Zhu, T. Huang, W. Liu, Auto- Encoder Guided GAN for Chinese Calligraphy Synthesis, in: International Conference on Document Analysis and Recognition (ICDAR), V ol. 01, 2017, pp. 1095–1100

    2017

  42. [50]

    Tian, zi2zi: Master Chinese calligraphy with con- ditional adversarial networks, Internet] https://github

    Y . Tian, zi2zi: Master Chinese calligraphy with con- ditional adversarial networks, Internet] https://github. com/kaonashi-tyc/zi2zi 3 (2017) 2

    2017

  43. [51]

    D. Sun, T. Ren, C. Li, H. Su, J. Zhu, Learning to Write Stylized Chinese Characters by Reading a Handful of Ex- amples (2018).arXiv:1712.06424

    2018 arXiv

  44. [52]

    Zhang, Y

    Y . Zhang, Y . Zhang, W. Cai, Separating Style and Content for Generalized Style Transfer, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  45. [53]

    Y . Kong, C. Luo, W. Ma, Q. Zhu, S. Zhu, N. Yuan, L. Jin, Look Closer To Supervise Better: One-Shot Font Genera- tion via Component-Based Discriminator, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 13482–13491

    2022

  46. [54]

    S. Park, S. Chun, J. Cha, B. Lee, H. Shim, Few-shot Font Generation with Localized Style Representations and Fac- torization, in: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), V ol. 35, 2021, pp. 2393– 2402

    2021

  47. [55]

    S. Park, S. Chun, J. Cha, B. Lee, H. Shim, Multiple Heads Are Better Than One: Few-Shot Font Generation With Mul- tiple Localized Experts, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 13900–13909

    2021

  48. [56]

    C. Wang, M. Zhou, T. Ge, Y . Jiang, H. Bao, W. Xu, CF- Font: Content Fusion for Few-Shot Font Generation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 1858– 1867

    2023

  49. [57]

    Y . Xie, X. Chen, L. Sun, Y . Lu, DG-Font: Deformable Generative Networks for Unsupervised Font Generation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 5130– 5140

    2021

  50. [58]

    Y . Tang, L. Peng, Q. Xu, Y . Wang, A. Furuhata, CNN Based Transfer Learning for Historical Chinese Character Recognition, in: IAPR Workshop on Document Analysis Systems (DAS), 2016, pp. 25–29. 13

    2016

  51. [59]

    Y . Li, L. Jin, X. Zhu, T. Long, SCUT-COUCH2008: A comprehensive online unconstrained Chinese handwrit- ing dataset, in: International Conference on Frontiers in Handwriting Recognition (ICFHR), V ol. 2008, 2008, pp. 165–170

    2008

  52. [60]

    Jin, J.-X

    L.-W. Jin, J.-X. Yin, X. Gao, J.-C. Huang, Study of Sev- eral directional feature extraction methods with local elas- tic meshing technology for HCCR, in: Proceedings of the Sixth Int. Conference for Young Computer Scientist (ICYCS), 2001, pp. 232–236

    2001

  53. [61]

    F. Chang, Techniques for solving the large-scale classi- fication problem in Chinese handwriting recognition, in: Summit on Arabic and Chinese Handwriting Recognition (SACH), Springer, 2006, pp. 161–169

    2006

  54. [62]

    Pechwitz, S

    M. Pechwitz, S. S. Maddouri, V . Märgner, N. Ellouze, H. Amiri, et al., IFN/ENIT-database of handwritten Ara- bic words, in: Proceedings of the Colloque International Francophone sur lEcrit et le Document (CIFED), V ol. 2, Citeseer, 2002, pp. 127–136

    2002

  55. [63]

    N. E. B. Amara, O. Mazhoud, N. Bouzrara, N. Ellouze, ARABASE: A Relational Database for Arabic OCR Sys- tems., in: Proceedings of the International Arab Journal of Information Technology (IAJIT), V ol. 2, 2005, pp. 259– 266

    2005

  56. [64]

    Hull, A database for handwritten text recognition re- search, IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 16 (5) (1994) 550–554

    J. Hull, A database for handwritten text recognition re- search, IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 16 (5) (1994) 550–554

    1994

  57. [65]

    Marti, H

    U.-V . Marti, H. Bunke, The IAM-database: an English sentence database for offline handwriting recognition, In- ternational Journal on Document Analysis and Recognition 5 (2002) 39–46

    2002

  58. [66]

    Z. Xie, Z. Sun, L. Jin, H. Ni, T. Lyons, Learning Spatial- Semantic Context with Fully Convolutional Recurrent Net- work for Online Handwritten Chinese Text Recognition, IEEE Transactions on Pattern Analysis and Machine Intel- ligence 40 (8) (2017) 1903–1917

    2017

  59. [67]

    C.-L. Liu, S. Jaeger, M. Nakagawa, ’Online Recognition of Chinese Characters: The State-of-the-Art, IEEE Transac- tions on Pattern Analysis and Machine Intelligence 26 (2) (2004) 198–213

    2004

  60. [68]

    Q. Lin, C. Luo, L. Jin, S. Lai, Stan: A sequential transfor- mation attention-based network for scene text recognition, Pattern Recognition 111 (2021) 107692

    2021

  61. [69]

    Liu, I.-J

    C.-L. Liu, I.-J. Kim, J. H. Kim, Model-based stroke extrac- tion and matching for handwritten Chinese character recog- nition, Pattern Recognition 34 (12) (2001) 2339–2352

    2001

  62. [70]

    M. Yang, B. Yang, M. Liao, Y . Zhu, X. Bai, Class-aware mask-guided feature refinement for scene text recognition, Pattern Recognition 149 (2024) 110244

    2024

  63. [71]

    Z. Cai, M. Cao, H. Chen, K. Chen, K. Chen, X. Chen, X. Chen, Z. Chen, Z. Chen, P. Chu, X. Dong, H. Duan, Q. Fan, Z. Fei, Y . Gao, J. Ge, C. Gu, Y . Gu, T. Gui, A. Guo, Q. Guo, C. He, Y . Hu, T. Huang, T. Jiang, P. Jiao, Z. Jin, Z. Lei, J. Li, J. Li, L. Li, S. Li, W. Li, Y . Li...

    2024 arXiv

  64. [72]

    A. Yang, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Li, D. Liu, F. Huang, H. Wei, H. Lin, J. Yang, J. Tu, J. Zhang, J. Yang, J. Yang, J. Zhou, J. Lin, K. Dang, K. Lu, K. Bao, K. Yang, L. Yu, M. Li, M. Xue, P. Zhang, Q. Zhu, R. Men, R. Lin, T. Li, T. Tang, T. Xia, X. Ren, X...

    2025 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.