Pith. sign in

REVIEW 3 major objections 4 minor 36 references

Signs of the Past, Patterns of the Present: On the Automatic Classification of Old Babylonian Cuneiform Signs

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A standard image classifier trained on depth renderings of Old Babylonian clay tablets identifies cuneiform signs with 87.1% top-1 accuracy after fine-tuning, and training across three proveniences lifts out-of-distribution accuracy to…

desk verdict First Old Babylonian cuneiform sign classification benchmark with genuinely useful imaging and transfer experiments, but the headline accuracy is crop-level and likely inflated by a per-crop split that leaks tablet-level cues. read the letter →

arxiv 2507.13959 v1 pith:NRMMENJZ submitted 2025-07-18 cs.LG

Eli Verwimp , Gustav Ryberg Smidt , Hendrik Hameeuw , Katrien De Graef This is my paper · ORCID
classification cs.LG
keywords cuneiformsignclassificationOldBabylonianResNet502D+visualizationhandwrittentextrecognitiontransferlearningculturalheritagepalaeography
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 establishes the first automatic classification benchmark for handwritten Old Babylonian cuneiform signs, using a standard ResNet50 network on crops from 2D+ visualizations of tablets from Nippur, Dūr-Abiešuḫ, and Sippar. It reports $87.1\% \pm 0.3$ top-1 accuracy after fine-tuning on the target provenience, with top-5 accuracy of $96.5\%$. The central claim is that sign shape varies enough across cities and imaging conditions to dominate the difficulty, and that training on several proveniences nearly closes the gap to unseen sites. The authors argue the practical path forward is acquiring depth-based visualizations and more geographically varied data rather than more of the same.

What carries the argument

The load-bearing object is a ResNet50 image classifier trained on square crops of individual signs cut from 2D+ visualizations, where the crucial input type is the SketchB visualization: a rotation-invariant sketch derived from the surface normal map. The argument is carried by controlled training-set ablations over visualizations (twelve render types), proveniences (single versus combined), and a fine-tuning stage that removes augmentations and lowers the learning rate to match the test distribution. A secondary mechanism is the TSNE projection of the 2048-dimensional penultimate-layer features, used to show that clusters correspond to script variants documented in sign lists.

What would settle it

Take a random set of crops from the test data, have two trained Assyriologists independently assign sign classes without seeing the corpus labels, and compare each human's agreement with the model's accuracy against the same reference labels. If human-human agreement is no higher than human-model agreement, the $87.1\%$ figure is partly an artifact of ambiguous labels rather than of visual recognition.

Watch

Extended reading notes

Core claim

The authors claim that a standard convolutional network can classify Old Babylonian cuneiform signs cut from dome-captured visualizations, provided the rendering exposes depth: non-photorealistic sketches and normal maps beat any single lighting angle by 4–5 percentage points. The model reaches $87.1\% \pm 0.3$ top-1 on Nippur after fine-tuning, and the base model ranges $77.2\%$ to $85.6\%$ across Dūr-Abiešuḫ, Nippur, and Sippar. Training on all three proveniences raises accuracy on the held-out Marad site to $93.4\%$ of in-distribution performance, whereas single-provenience training drops to $51.4\%$–$79.0\%$ of in-distribution levels. Qualitative TSNE analysis shows that the model's feature space mirrors palaeographic variants that Assyriologists recognize, including the known ANSZE/GIRI3 overlap. The paper concludes that this is a viable assistive tool for reading tablets, with top-5 performance strong enough for a suggestion tool, while top-1 remains limited without context.

Load-bearing premise

The reported accuracies assume that each cropped sign has exactly one correct class, namely the contextual reading assigned by a human annotator in the Cuneur software; if those readings are inconsistent or the Unicode-based sign list splits or merges signs differently from actual palaeography, every accuracy number inherits that bias.

Editorial extensions

If this is right

  • A top-5 accuracy of $96.5\%$ means a human-in-the-loop tool can offer five likely signs per crop, which is already useful for reading unfamiliar proveniences.
  • Including multiple proveniences in training is the single most effective step for generalizing to unseen sites, raising out-of-distribution accuracy to $93.4\%$ of in-distribution performance.
  • Fine-tuning the base model on the target provenience's own data improves results for every city tested, making specialization a cheap final step after broad training.
  • Acquisition standards should prioritize depth-revealing visualizations (normal maps, sketches) and variety of provenience over raw image quantity.
  • Signs on the curved left and right edges of tablets are the hardest to classify under directional light, so even illumination or depth-based renderings are preferable for those zones.

Reading between the lines

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

  • The authors do not test legacy flatbed scans or photographs; a direct extension would be to measure how much accuracy drops on such images and then apply domain adaptation from SketchB renderings to legacy data.
  • The embedding space that TSNE visualizes could be turned into a quantitative palaeographic instrument, for example dating undated tablets by nearest-neighbor positions among dated exemplars—something the paper only gestures at.
  • The out-of-distribution claim rests on a single unseen site, Marad; a stronger stress test would hold out an entire region or a chronologically distinct corpus, which could lower the $93.4\%$ figure.
  • Because labels are contextual readings, the accuracy ceiling may be set by inter-annotator agreement; measuring that agreement would place the reported scores in perspective.
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

3 major / 4 minor

Summary. The paper trains ResNet50 models to classify individual cuneiform signs cropped from 2D+ visualizations of Old Babylonian tablets from three proveniences (Nippur, Dūr-Abiešuḫ, Sippar), evaluates a held-out Marad set for out-of-distribution performance, compares twelve visualization types, studies the effect of fine-tuning, and uses TSNE plots for qualitative palaeographic analysis. The headline results are a top-1 accuracy of 87.1% ± 0.3 and top-5 accuracy of 96.5% ± 0.1 on the Nippur test set after fine-tuning, with base-model accuracies of 77.2–85.6% top-1 across the three in-distribution proveniences and an OOD performance of 93.4% of the in-distribution performance when training on all three.

Significance. If the evaluation is trustworthy, this is a useful first benchmark for automated Old Babylonian sign classification. The paper's strengths include repeated runs with reported standard deviations, a systematic comparison of lighting angles and depth visualizations, an external OOD test set (Marad) completely withheld from training, and a publicly available code repository and suggestion tool. The results also speak directly to data acquisition standards for cuneiform tablets, which is a practical contribution to digital Assyriology.

major comments (3)
  1. [§3.4] The 80/20 split is performed per sign category at the level of cropped signs, not at the level of tablets or tablet sides. Because crops are square bounding boxes that deliberately include partial neighboring signs (§3.3), crops from the same tablet side can appear in both training and test partitions, sharing lighting, surface texture, scribal hand, and often overlapping context. A ResNet50 can exploit such tablet-specific cues, so the reported accuracies, especially the fine-tuned Nippur row in Table 1 (87.1% ± 0.3), may be optimistic. The standard deviations only reflect random re-splitting of crops, not the choice of which tablets or tablet sides are held out. Please report results with a tablet-level or tablet-side-level split (for example, leave-one-tablet-out or grouped split), or at minimum quantify the fraction of test crops that share a tablet side with training crops.
  2. [§3.2.2, §3.1.1] The class labels are contextual readings assigned by human annotators using the Nuolenna sign list. As the paper acknowledges, many signs have multiple readings and visually similar signs from different classes can overlap, as the ANSZE/GIRI3 confusion in §4.3.2 shows. This means the reported accuracy conflates visual classification with the annotators' context-based reading decisions. The paper should report inter-annotator agreement on a subset of signs, or at least explicitly discuss how label noise and class overlap affect the interpretation of the accuracy numbers. Without this, the absolute accuracies are difficult to interpret as purely palaeographic classification performance.
  3. [§4.3.3, Table 1] The fine-tuning procedure is not fully specified. The text states that the base model is 'fine-tuned using either all data or only the data of one specific provenience,' but it does not explicitly state that the held-out 20% test crops are excluded from the fine-tuning set. If the test partition is included in fine-tuning, the results in Table 1 would be invalid; if not, this should be stated clearly. Please clarify the exact data split used for each fine-tuning experiment, including whether the test set is the same 20% crop-level partition described in §3.4.
minor comments (4)
  1. [§2] In the related work section, 'Cobanaglu' should be 'Cobanoglu' to match reference [7].
  2. [§4.3.1] The phrase 'do to their combined state' should be 'due to their combined state.'
  3. [References] Reference [8] contains a typo in the title: 'Old Babylonian Peiod' should be 'Old Babylonian Period.'
  4. [§3.2.2] The data availability statement says 'The data and scripts published with this paper can be found on our Zenodo/Github page (link),' but the link is a placeholder. Please provide the actual DOI or URL for the dataset and scripts.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the accuracy and transfer claims are measured on held-out test crops and a fully withheld Marad set, not derived from fitted parameters or from self-citations.

full rationale

The paper's quantitative claims are supervised-learning evaluations. The label space is defined externally by the Nuolenna Unicode sign list (Section 3.2.2), and the per-class 80/20 split (Section 3.4) creates held-out crops; the Marad set is kept outside training at all times (Section 4.3.3). The reported top-1/top-5 numbers are therefore empirical measurements on examples not used for the reported fit, not quantities forced by construction. The self-citations (e.g., Hameeuw et al. [11] for visualization naming, Rattenborg et al. [22] for corpus spread, Willems et al. [32] for light-dome acquisition) are provenance and methodology references, not load-bearing arguments that the results are correct by definition; no uniqueness theorem or ansatz is imported from the authors' prior work. TSNE-based variant analysis is explicitly post hoc and the authors caution about TSNE artifacts (Section 3.4). The main nontrivial risk is the crop-level rather than tablet-level split (Section 3.4), which could let tablet-specific surface/lighting cues appear in both train and test and inflate in-distribution numbers; this is an evaluation-design concern, not circularity. Likewise, the placeholder '(link)' for the promised Zenodo/Github resources and the deferral of annotation-data evaluation to 'Smidt forthcoming' are completeness issues, not circular steps.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

There are no fitted free parameters in the traditional sense, but the paper relies on labeling choices and a text-level assumption that cropped signs are independent samples. The 20-instance threshold is a pragmatic choice, not a fitted constant, and the authors are transparent about it.

assumptions (3)
  • domain assumption The Nuolenna Unicode sign list provides a consistent and correct mapping from visual signs to class labels.
    Section 3.2.2 states that the sign list was used to compile classes. The entire accuracy metric depends on this labeling being internally consistent.
  • domain assumption Random rotation by 0-360 degrees is a valid augmentation because sign orientation carries no class information in the training distribution.
    Section 3.3 applies random rotation and argues that semantics-changing rotations are resolved by visible surrounding signs. This is plausible but only partially verified by the reported accuracies.
  • domain assumption Each cropped sign is an independent sample despite multiple crops coming from the same tablet or same annotator.
    Section 3.4 randomly splits 80/20 per sign category without grouping by tablet, so crops from the same tablet can appear in both train and test, inflating apparent accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Signs of the Past, Patterns of the Present: On the Automatic Classification of Old Babylonian Cuneiform Signs." pith.science (2026). https://pith.science/paper/NRMMENJZ

@misc{pith2026250713959,
  author       = {Pith},
  title        = {Pith review of: Signs of the Past, Patterns of the Present: On the Automatic Classification of Old Babylonian Cuneiform Signs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NRMMENJZ}},
  note         = {Machine review of arXiv:2507.13959}
}
read the original abstract

The work in this paper describes the training and evaluation of machine learning (ML) techniques for the classification of cuneiform signs. There is a lot of variability in cuneiform signs, depending on where they come from, for what and by whom they were written, but also how they were digitized. This variability makes it unlikely that an ML model trained on one dataset will perform successfully on another dataset. This contribution studies how such differences impact that performance. Based on our results and insights, we aim to influence future data acquisition standards and provide a solid foundation for future cuneiform sign classification tasks. The ML model has been trained and tested on handwritten Old Babylonian (c. 2000-1600 B.C.E.) documentary texts inscribed on clay tablets originating from three Mesopotamian cities (Nippur, D\=ur-Abie\v{s}uh and Sippar). The presented and analysed model is ResNet50, which achieves a top-1 score of 87.1% and a top-5 score of 96.5% for signs with at least 20 instances. As these automatic classification results are the first on Old Babylonian texts, there are currently no comparable results.

Figures

Figures reproduced from arXiv: 2507.13959 by the authors.

Figure 1
Figure 1. Examples of three sign categories that are included in the data used in this paper, together with the predictions of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the data that is used in this paper. Left (a): map from Google Earth of southern Iraq, with the locations [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Results for models trained on different visualisations of the tablets. Each result is the average top-1 accuracy, i.e. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The location of a sign on a tablet has a significant influence on the classification accuracy. To study this we divide [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Top-1 and top-5 accuracy of different sign categories for a model trained on the SketchB visualisations, grouped by [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: TSNE visualization of the representations of a model trained on the SketchB visualizations of the three datasets [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Training results for different dataset combinations, split by provenience on the SketchB visualizations. Dark bars [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Screenshot from the first version of the Cune-AI-form tool. The user draws the (red) rectangle on the tablet on a sign [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 34 canonical work pages

  1. [1]

    Anderson and Marc Levoy

    Sean E. Anderson and Marc Levoy. 2002. Unwrapping and visualizing cuneiform tablets. IEEE Computer Graphics and Applications 22, 6 (2002), 82–88

    work page 2002

  2. [2]

    Marine Béranger. 2023. Dur-Abi-ešuh and the Abandonment of Nippur During the Late Old Babylonian Period: A Historical Survey. Journal of Cuneiform Studies 75 (2023), 27–47. doi:10.1086/725217

    work page doi:10.1086/725217 2023

  3. [3]

    Bartosz Bogacz and Hubert Mara. 2022. Digital Assyriology—Advances in Visual Cuneiform Analysis. Journal on Computing and Cultural Heritage 15, 2, Article 38 (may 2022), 22 pages. doi:10.1145/3491239

    work page doi:10.1145/3491239 2022

  4. [4]

    Kai Brandenbusch, Eugen Rusakov, and Gernot A. Fink. 2021. Context aware generation of cuneiform signs. In Document Analysis and Recognition - International Conference on Document Analysis and Recognition , Josep Lladós, Daniel Lopresti, and Seiichi Uchida (Eds.). Springer International Publishing, Cham, Switzerland, 65–79

    work page 2021

  5. [5]

    Dominique Charpin. 2010. Writing, Law, and Kingship in Old Babylonian Mesopotamia . University of Chicago Press, Chicago and London

    work page 2010

  6. [6]

    Dominique Charpin, Dietz Otto Edzard, Marten Stol, Pascal Attinger, Walther Sallaberger, and Markus Wäfler. 2004. Mesopotamien - Die altbabylonische Zeit. Orbis Biblicus et Orientalis, Vol. 160. Academic Press and Vandenhoeck & Ruprecht, Fribourg and Göttingen

    work page 2004

  7. [7]

    Yunus Cobanoglu, Luis Sáenz, Ilya Khait, and Enrique Jiménez. 2024. Sign detection for cuneiform tablets. it-Information Technology 66, 1 (2024), 28–38

    work page 2024

  8. [8]

    Rients de Boer. 2013. Marad in the Early Old Babylonian Peiod: Its Kings, Chronology, and Isin’s Influence. Journal of Cuneiform Studies 65 (2013), 73–90. https://www.jstor.org/stable/10.5615/jcunestud.65.2013.0073

    arXiv 2013

Show all 36 references
  1. [9]

    Tobias Dencker, Pablo Klinkisch, Stefan M Maul, and Björn Ommer. 2020. Deep learning of cuneiform sign detection with weak supervision using transliteration alignment. Plos one 15, 12 (2020), e0243039

    2020

  2. [10]

    Ilya Gershevitch. 1979. The Alloglottography of Old Persian. Transactions of the Philological Society 77 (1979), 114–190

    1979

  3. [11]

    Hendrik Hameeuw, Katrien De Graef, Gustav Ryberg Smidt, Anne Goddeeris, Timo Homburg, and Krishna Kumar Thirukokaranam Chandrasekar. 2024. Preparing multi-layered visualisations of Old Babylonian cuneiform tablets for a machine learning OCR training model towards automated sig...

    2024 doi

  4. [12]

    Hendrik Hameeuw and Geert Willems. 2011. New visualization techniques for cuneiform texts and sealings. Akkadica 132, 2 (2011), 163–178

    2011

  5. [13]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . IEEE Computer Society, Los Alamitos, CA, USA, 770–778

    2016

  6. [14]

    Timo Homburg, Robert Zwick, Hubert Mara, and Kai-Christian Bruhn. 2022. Annotated 3D-Models of Cuneiform Tablets. Journal of Open Archaeology Data (JOAD) 10, 92 (2022), 4 pages. doi:10.5334/joad.92

    2022 doi

  7. [15]

    John Huehnergard. 2011. A Grammar of Akkadian (3rd ed.). Eisenbrauns, Winona Lake, Indiana

    2011

  8. [16]

    Tommi Jauhiainen, Heidi Jauhiainen, Tero Alstola, and Krister Lindén. 2019. Language and Dialect Identification of Cuneiform Texts. In Proceedings of the Sixth Workshop on NLP for Similar Languages, Varieties and Dialects , Marcos Zampieri, Preslav Nakov, Shervin Malmasi, Niko...

    2019 doi

  9. [17]

    René Labat. 1976. Manuel D’Épigraphie Akkadienne (5th ed.). Librairie Orientaliste Paul Geuthner, Paris

    1976

  10. [18]

    Karel Van Lerberghe and Gabriella Voet. 2009. A Late Old Babylonian Temple Archive from Dūr-Abiešuḫ . Cornell University Studies in Assyriology and Sumerology, Vol. 8. CDL Press, Bethesda, Maryland

    2009

  11. [19]

    Piotr Michalowski. 2007. The Lives of the Sumerian Language (2nd print ed.). Oriental Institute Seminars, Vol. 2. The Oriental Institute of the University of Chicago, Chicago, Illinois, Chapter 10, 163–188

    2007

  12. [20]

    Rachel Mikulinsky, Morris Alper, Shai Gordin, Enrique Jiménez, Yoram Cohen, and Hadar Averbuch-Elor. 2025. ProtoSnap: Prototype Alignment for Cuneiform Signs. arXiv: 2502.00129 [cs.CV] https://arxiv.org/abs/2502.00129

    2025 arXiv

  13. [21]

    Joseph Nockels, Paul Gooding, and Melissa Terras. 2024. The implications of handwritten text recognition for accessing the past at scale. Journal of Documentation 80, 7 (2024), 148–167

    2024

  14. [22]

    Rune Rattenborg, Gustav Ryberg Smidt, Carolin Johansson, Nils Melin-Kronsell, and Seraina Nett. 2023. The Archaeological Distribution of the Cuneiform Corpus. A Provisional Quantitative and Geospatial Survey. Altorientalische Forschungen 50, 2 (2023), 178–205. doi:doi:10.1515/...

    2023 doi

  15. [23]

    Christopher Rest, Denis Fisseler, Frank Weichert, Turna Somel, and Gerfrid GW Müller. 2022. Illumination-based augmentation for cuneiform deep neural sign classification. Journal on Computing and Cultural Heritage (JOCCH) 15, 3 (2022), 1–20

    2022

  16. [24]

    Christian Reul, Dennis Christ, Alexander Hartelt, Nico Balbach, Maximilian Wehner, Uwe Springmann, Christoph Wick, Christine Grundig, Andreas Büttner, and Frank Puppe. 2019. OCR4all—An open-source tool providing a (semi-) automatic OCR workflow for historical printings. Applie...

    2019

  17. [25]

    Eugen Rusakov, Turna Somel, Gernot A Fink, and Gerfrid GW Müller. 2020. Towards query-by-eXpression retrieval of cuneiform signs. In 2020 17th International Conference on Frontiers in Handwriting Recognition (ICFHR) . IEEE, Dortmund, 43–48. Sign of the Past, Patterns of the Pr...

    2020

  18. [26]

    Walther Sallaberger. 2004. Das Ende des Sumerischen: Tod und Nachleben einer altmesopotamischen Sprache . Münchner Forschungen zur historischen Sprachwissenschaft, Vol. 2. Hempen Verlag, Bremen, 108–140

    2004

  19. [27]

    Thea Sommerschield, Yannis Assael, John Pavlopoulos, Vanessa Stefanak, Andrew Senior, Chris Dyer, John Bodel, Jonathan Prag, Ion Androutsopoulos, and Nando De Freitas. 2023. Machine learning for ancient languages: A survey. Computational Linguistics 49, 3 (2023), 703–747

    2023

  20. [28]

    Ernst Stötzner, Timo Homburg, and Hubert Mara. 2023. CNN based Cuneiform Sign Detection Learned from Annotated 3D Renderings and Mapped Photographs with Illumination Augmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision . IEEE Computer Societ...

    2023

  21. [29]

    Michael P. Streck. 2010. Großes Fach Altorientalistik: Der Umfang des keilschriftlichen Textkorpus. Mitteilungen der Deutschen Orient-Gesellschaft zu Berlin 142 (2010), 35–58

    2010

  22. [30]

    Laurens Van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of Machine Learning Research 9, 11 (2008), 2579–2605

    2008

  23. [31]

    Martin Wattenberg, Fernanda Viégas, and Ian Johnson. 2016. How to use t-SNE effectively. Distill 1, 10 (2016), e2

    2016

  24. [32]

    Geert Willems, Frank Verbiest, Wim Moreau, Hendrik Hameeuw, Karel Van Lerberghe, and Luc Van Gool. 2005. Easy and cost-effective cuneiform digitizing. In The 6th International Symposium on Virtual Reality, Archaeology and Cultural Heritage (V AST 2005). Eurographics Associatio...

    2005

  25. [33]

    Williams, Grace Su, Sandra R

    Edward C. Williams, Grace Su, Sandra R. Schloen, Miller C. Prosser, Susanne Paulus, and Sanjay Krishnan. 2025. DeepScribe: Localization and Classification of Elamite Cuneiform Signs Via Deep Learning. J. Comput. Cult. Herit. (mar 2025), 31 pages. doi:10.1145/3716850 Just Accepted

    2025 doi

  26. [34]

    Christopher Woods. 2007. Bilingualism, Scribal Learning, and the Death of Sumerian (2nd print ed.). Oriental Institute Seminars, Vol. 2. The Oriental Institute of the University of Chicago, Chicago, Illinois, Chapter 6, 95–124

    2007

  27. [35]

    Saining Xie, Ross Girshick, Piotr Dollár, Zhuowen Tu, and Kaiming He. 2017. Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition . IEEE Computer Society, Los Alamitos, CA, USA, 1492–1500

    2017

  28. [36]

    Vasiliy Yugay, Kartik Paliwal, Yunus Cobanoglu, Luis Sáenz, Ekaterine Gogokhia, Shai Gordin, and Enrique Jiménez. 2024. Stylistic classification of cuneiform signs using convolutional neural networks. IT-Information Technology 66, 1 (2024), 15–27. A Cune-AI-form Tool As part o...

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.