Pith. sign in

REVIEW 5 major objections 6 minor 43 references

SignBart -- New approach with the skeleton sequence for Isolated Sign language Recognition

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Isolated sign language recognition can be made accurate and lightweight by encoding a skeleton's x and y coordinates separately in a BART-style encoder-decoder, then reconnecting the streams with cross-attention; the paper reports 96.04%…

desk verdict Interesting architecture idea, but the headline accuracy claim collapses against the paper's own tables; worth a look if revised honestly. read the letter →

arxiv 2506.21592 v1 pith:PTVXJYFD submitted 2025-06-18 cs.CL cs.CV

Tinh Nguyen , Minh Khue Phan Tran This is my paper · ORCID
classification cs.CLcs.CV
keywords signlanguagerecognitionskeletonsequencescoordinatetheorymodelcomplexityencoder-decodercross-attentionLSA-64WLASL
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

SignBart sets out to break the usual trade-off between accuracy and model size in isolated sign language recognition. Instead of feeding each skeleton keypoint as an inseparable (x, y) pair, the model encodes the x coordinates in a bidirectional encoder and the y coordinates in a causal decoder, then uses cross-attention so the y stream can query the encoded x stream. The paper reports 96.04% top-1 accuracy on the 64-word LSA-64 dataset with 749,888 parameters; on WLASL-1000 and WLASL-2000 it reports 81.45% and 68.00%, above NLA-SLR's 75.72% and 58.31%; and on the 2731-gloss ASL-Citizen set it reports 75.22% Rec@1, above I3D and ST-GCN. Ablations attribute most of the gain to the linear coordinate projection, per-part bounding-box normalization, and combining body with both hands. The authors note the method has not been tested on continuous sign recognition and that its three attention mechanisms may add computational cost on keypoint-heavy data.

What carries the argument

The load-bearing object is the coordinate-split BART-style encoder-decoder. BART is a transformer sequence-to-sequence architecture whose encoder reads bidirectionally and whose decoder reads causally. Here an input skeleton $I \in \mathbb{R}^{T \times K \times 2}$ is split into $x_{\text{coord}}$ and $y_{\text{coord}}$, each projected to the model dimension by its own linear layer. The encoder runs bidirectional self-attention on the x stream; the decoder runs self-causal attention on the y stream with a mask $M(i,j)=1$ when $i \ge j$, then cross-attention in which queries come from the y attention map while keys and values come from the encoder's x attention map. A second load-bearing piece is preprocessing: the body, left hand, and right hand are each normalized inside their own bounding box with a 5% margin, and the ablations show that the projection step, the three-box normalization, and the combination of all three skeleton components each add large accuracy gains.

What would settle it

A reader could settle it by training SignBart with the exact keypoint extraction and normalization pipeline on the official LSA-64 split used by SPOTER and on NLA-SLR's WLASL splits; if the 96.04% figure cannot be reproduced, or if the margins over ST-GCN and NLA-SLR reverse under those splits, the central comparison fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the x and y coordinates of skeleton keypoints carry separable information, and that encoding them separately instead of as fused (x, y) pairs makes a small transformer more accurate. The architecture gives x to the encoder's bidirectional self-attention and y to the decoder's self-causal attention; cross-attention then lets each y-based query attend to x-based keys and values, so the independence never becomes a disconnection. The authors report that this scheme reaches 96.04% top-1 accuracy on LSA-64 with 749,888 parameters, higher than the ST-GCN and 3DGCN baselines listed for that dataset, though below SPOTER's reported 100%, and that accuracy degrades less than previous methods as WLASL grows from 100 to 2000 glosses. On the 2731-gloss ASL-Citizen version the model reaches 75.22% Rec@1, which is reported as better than I3D and ST-GCN from the dataset's original paper.

Load-bearing premise

The load-bearing assumption is that every comparison uses the same train/test splits as the cited baselines, because the paper never specifies the split protocol for LSA-64, WLASL, or ASL-Citizen; if SignBart was evaluated under an easier split, the headline accuracy and generalization claims lose their basis.

Editorial extensions

If this is right

  • On LSA-64, SignBart reports 96.04% top-1 accuracy with 749,888 parameters, which exceeds the ST-GCN (92.81%) and 3DGCN (94.84%) baselines in the same table while using far fewer parameters; SPOTER still reports the highest number at 100% with 5,918,848 parameters.
  • On WLASL, the model's top-1 accuracy stays between 68.00% and 81.45% as the vocabulary grows from 100 to 2000 glosses, and its 81.45% on WLASL-1000 and 68.00% on WLASL-2000 exceed NLA-SLR's 75.72% and 58.31% despite NLA-SLR using both RGB and skeleton inputs.
  • On ASL-Citizen-2731, SignBart's 75.22% Rec@1 exceeds the I3D (63.10%) and ST-GCN (59.52%) baselines from the original dataset paper.
  • The ablations show that removing the coordinate projection drops LSA-64 accuracy from 96.04% to 62.08%, that normalization adds 13.54 points over no normalization, and that the three-box normalization outperforms one-box and two-box variants.
  • The component ablation shows body-only keypoints reach 86.97%, right-hand-only 70.20%, and left-hand-only 23.02%, so the combined body-plus-two-hands input is essential to the reported 96.04%.

Reading between the lines

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

  • Because the paper never states the train/test split protocol, the cross-dataset comparisons should be read as provisional until SignBart is run on the official SPOTER and NLA-SLR splits; this is an editorial caution, not a claim in the paper.
  • The coordinate-split design is a general mechanism that could be tried on other skeleton-based tasks such as action recognition or gesture control; if the mechanism is what carries the gain, separating x and y before cross-attention should help wherever horizontal and vertical motion carry different meaning, but the paper does not test this.
  • The large right-hand versus left-hand accuracy gap (70.20% versus 23.02%) suggests the model may be exploiting signer handedness rather than learning symmetric language content; flipping hand keypoints as augmentation would be a direct test, and the paper does not report it.
  • The WLASL pattern (below NLA-SLR on 100 and 300 glosses, above on 1000 and 2000 glosses) suggests the model's advantage is scalability rather than small-vocabulary discrimination; a matched-split comparison would be needed to confirm, which the paper omits.
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

5 major / 6 minor

Summary. The paper proposes SignBart, a skeleton-based encoder-decoder transformer for isolated sign language recognition. The encoder processes x-coordinates and the decoder processes y-coordinates, with cross-attention linking the two streams; inputs are MediaPipe keypoints normalized per body part, and the model is evaluated on LSA-64, WLASL, and ASL-Citizen. The abstract claims 96.04% accuracy on LSA-64 with 749,888 parameters, 'significantly outperforming previous models with over one million parameters,' and superior generalization on WLASL and ASL-Citizen. Ablation studies examine the effects of coordinate projection, normalization, and skeleton components.

Significance. The proposed architecture is an interesting lightweight alternative in skeleton-based ISLR: separating x and y coordinate streams and reconnecting them through cross-attention is a non-standard design, and the paper reports results on three datasets together with ablations that isolate useful components. If the claims were fully supported, the parameter-efficiency result would be a useful contribution. However, the central comparative claims are contradicted by the paper's own tables, and the evaluation protocol is not specified, so the significance cannot currently be assessed at face value. The paper also provides no code or data release, which limits reproducibility and verification.

major comments (5)
  1. [Abstract; §4.3; Tables 2 and 4] The claim that SignBart 'significantly outperforms previous models with over one million parameters' is contradicted by the reported numbers. In Table 4, SPOTER (100%, 5,918,848 parameters), HWGATE (98.59%, 10,758,354), and SL-GCN (98.13%, 4,872,306) all exceed SignBart's 96.04% on LSA-64. In Table 2, NLA-SLR outperforms SignBart on WLASL-100 (93.08% vs. 78.00%) and WLASL-300 (87.33% vs. 78.50%). The abstract and conclusion therefore need to be reworded to claim competitive accuracy at a lower parameter count rather than superior accuracy; alternatively, the empirical comparisons must be redone under a common protocol.
  2. [§4.2–§4.3; Tables 2–6] No train/validation/test split protocol is stated for any dataset. LSA-64, WLASL, and ASL-Citizen all have multiple signers and multiple class subsets, so the comparisons in Tables 2–6 are only meaningful if SignBart uses the same splits as the cited baselines. The paper must specify the exact split (for example, signer-independent vs. random), the subsetting procedure, and the number of runs, and should report mean and standard deviation over multiple seeds.
  3. [§4.3; Table 2] The text states that the model achieves a 5.73% increase on WLASL-300 and a 9.69% increase on WLASL-2000. Table 2 shows the WLASL-300 accuracy is 78.50% for SignBart versus 87.33% for NLA-SLR, a decrease rather than an increase; the 5.73% improvement corresponds to WLASL-1000 (81.45% vs. 75.72%). This numerical error weakens the 'superior generalization' claim and should be corrected.
  4. [§4.3; Table 6] The ASL-Citizen comparison is incomplete: SignBart's Rec@5 is reported as '-', so the comparison against I3D and ST-GCN on Rec@5 cannot be evaluated. Additionally, the statement that there has been no benchmark comparison on ASL-Citizen should be checked against the evaluation protocol used in the original ASL-Citizen paper before asserting that this is the first such comparison.
  5. [§3.1; §3.2] The conceptual motivation is presented as a 'coordinate theory' in which x and y are independent but interdependent, but the cited reference [34] concerns motor coordination and does not support this specific claim. The separation of x and y streams should be framed explicitly as a design hypothesis with references or preliminary evidence, rather than as an established theory.
minor comments (6)
  1. [§3.2] The sentence 'With separate encoding, the encoder will encode the x coordinates while the decoder will encode the y coordinates' appears twice in the same paragraph; one occurrence should be removed.
  2. [Tables 4, 7, 8] The LSA-64 dataset is cited as [9] in the table captions, but [9] is a modified LSTM paper; the correct reference for LSA-64 is [35].
  3. [§3.3] The paper states that only 6 of the 33 MediaPipe body keypoints are used, but does not specify which 6 keypoints are retained; this is needed for reproducibility.
  4. [§3.2; Title] The model is called 'BART-based,' but it does not use BART's denoising pretraining objective; using the name 'BART' without qualification may mislead readers about the relationship to the original BART model.
  5. [Tables 7–9] Ablation results are reported as single accuracy numbers without error bars, confidence intervals, or significance tests, so the relative importance of the ablated components is not statistically established.
  6. [General] No code, training configuration, or data availability statement is provided; given the number of unspecified preprocessing details, this substantially limits reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SignBart's claims are empirical model evaluations, not derivations that reduce to their inputs.

full rationale

The paper's central claims are empirical: a supervised encoder-decoder is trained on labeled skeleton sequences and evaluated on held-out splits of LSA-64, WLASL, and ASL-Citizen. The architecture choices (encoding x in the encoder, y in the decoder, cross-attention) are design decisions, not outputs derived from the evaluation numbers. The 'coordinate theory' citation [34] is used only as motivation for the design, not as a load-bearing derivation, and the paper does not invoke any uniqueness theorem or machine-checked self-citation. The ablation studies are standard empirical tests of preprocessing and input components, not fitted parameters renamed as predictions. The reviewer-observed inconsistency between the abstract's 'significantly outperforming previous models with over one million parameters' and the paper's own Tables 2 and 4, where SPOTER, HWGATE, and SL-GCN report higher LSA-64 accuracy, is a correctness/claim-support issue, not circularity. No equation is shown to equal another by construction, and no fitted value is presented as an independent prediction. Accordingly, the circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The model's accuracy depends on a set of hand-chosen preprocessing and architecture choices whose values are not fully reported. The most important is the unspecified dataset split, which determines whether the headline numbers are comparable to the baselines. The 'coordinate theory' is motivated by a citation that does not support it. The paper introduces no new entities; all inputs are Mediapipe keypoints.

free parameters (4)
  • d_model and ffdim per dataset = not reported
    Adjusted per dataset 'to optimize efficiency and prevent overfitting' (Section 4.1); values are not listed, so the architecture cannot be reproduced and the choice is fitted to validation performance.
  • number of body keypoints = 6
    Only 6 of 33 Mediapipe body keypoints are kept (Section 3.3), a hand choice that affects input and is not justified by analysis.
  • bounding box margin = 5%
    Margin added to bounding boxes during normalization (Section 3.4), chosen by hand; no sensitivity analysis.
  • number of layers and attention heads = 2 blocks, 16 heads
    Fixed architecture choice without ablation or sensitivity study.
assumptions (4)
  • domain assumption Mediapipe provides accurate 2D skeleton keypoints for all frames
    The whole pipeline depends on pose extraction; missing keypoints are set to 0 (Section 3.3), which may inject artifacts.
  • domain assumption The 6 selected body keypoints plus 21 hand keypoints per hand capture enough sign information
    Ablation shows body+hands gives the highest accuracy, but the drop from using all 33 body keypoints is not tested.
  • ad hoc to paper Treating x and y as independent input streams with cross-attention preserves the spatial relationship of keypoints
    This is the paper's design thesis (Section 3.1), motivated by a citation to a motor coordination paper [34] rather than by a demonstrated property of skeleton sign data.
  • domain assumption The used data splits match those of the cited baselines
    Not stated in the paper; comparisons in Tables 2-6 are only valid if splits are identical.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SignBart -- New approach with the skeleton sequence for Isolated Sign language Recognition." pith.science (2026). https://pith.science/paper/PTVXJYFD

@misc{pith2026250621592,
  author       = {Pith},
  title        = {Pith review of: SignBart -- New approach with the skeleton sequence for Isolated Sign language Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PTVXJYFD}},
  note         = {Machine review of arXiv:2506.21592}
}
read the original abstract

Sign language recognition is crucial for individuals with hearing impairments to break communication barriers. However, previous approaches have had to choose between efficiency and accuracy. Such as RNNs, LSTMs, and GCNs, had problems with vanishing gradients and high computational costs. Despite improving performance, transformer-based methods were not commonly used. This study presents a new novel SLR approach that overcomes the challenge of independently extracting meaningful information from the x and y coordinates of skeleton sequences, which traditional models often treat as inseparable. By utilizing an encoder-decoder of BART architecture, the model independently encodes the x and y coordinates, while Cross-Attention ensures their interrelation is maintained. With only 749,888 parameters, the model achieves 96.04% accuracy on the LSA-64 dataset, significantly outperforming previous models with over one million parameters. The model also demonstrates excellent performance and generalization across WLASL and ASL-Citizen datasets. Ablation studies underscore the importance of coordinate projection, normalization, and using multiple skeleton components for boosting model efficacy. This study offers a reliable and effective approach for sign language recognition, with strong potential for enhancing accessibility tools for the deaf and hard of hearing.

Figures

Figures reproduced from arXiv: 2506.21592 by the authors.

Figure 1
Figure 1. Model architecture. With the input skeleton data, the Encoder encodes the x coordinate, and the Decoder encodes the y coordinate and query information from the encoded x coordinate. Before encoding, both the x and y coordinates will go through a mapping process via Projection. Self Attention After mapping the coordinates to dmodel via Projection and adding positional information through Positional Encoding, the x-co… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 31 canonical work pages

  1. [34]

    Optimal feedback control as a theory of motor coordination,

    E. Todorov and M. I. Jordan, “Optimal feedback control as a theory of motor coordination,” Nature neuroscience, vol. 5, pp. 1226–1235, 2002

    work page 2002

  2. [1]

    Structure of sign language,

    A. Othman, “Structure of sign language,” in Sign Language Processing: From Gesture to Meaning.Cham:SpringerNatureSwitzerland,2024,pp.17– 40, isbn: 978-3-031-68763-1. doi: 10.1007/978-3-031-68763-1_2 . [On- line]. Available:https://doi.org/10.1007/978-3-031-68763-1_2

    work page doi:10.1007/978-3-031-68763-1_2 2024

  3. [2]

    Evaluation of manual and non-manual components for sign language recognition,

    M. Mukushev, A. Sabyrov, A. Imashev, K. Koishibay, V. Kimmelman, and A. Sandygulova, “Evaluation of manual and non-manual components for sign language recognition,” inProceedings of The 12th Language Re- sources and Evaluation Conference, European Language Resources Asso- ciation (ELRA), 2020

    work page 2020

  4. [3]

    Sign language recognition: A com- prehensive review of traditional and deep learning approaches, datasets, and challenges,

    T. Tao, Y. Zhao, T. Liu, and J. Zhu, “Sign language recognition: A com- prehensive review of traditional and deep learning approaches, datasets, and challenges,”IEEE Access, vol. PP, pp. 1–1, Jan. 2024.doi: 10.1109/ ACCESS.2024.3398806

    work page arXiv 2024

  5. [4]

    W. H. Organization, Deafness and hearing loss, https://www.who.int/ news- room/fact- sheets/detail/deafness- and- hearing- loss, [On- line; accessed: 10-Feb-2025], 2021. SignBart 13

    work page 2025

  6. [5]

    A comprehen- sive overview and comparative analysis on deep learning models: Cnn, rnn, lstm, gru,

    F. M. Shiri, T. Perumal, N. Mustapha, and R. Mohamed, “A comprehen- sive overview and comparative analysis on deep learning models: Cnn, rnn, lstm, gru,” arXiv preprint arXiv:2305.17473, 2023

    arXiv 2023

  7. [6]

    Isolated sign language recognition using hidden markov models,

    K. Grobel and M. Assan, “Isolated sign language recognition using hidden markov models,” in1997 IEEE International conference on systems, man, and cybernetics. Computational cybernetics and simulation, IEEE, vol. 1, 1997, pp. 162–167

    work page 1997

  8. [7]

    Understanding vision-based continuous sign languagerecognition,

    N. Aloysius and M. Geetha, “Understanding vision-based continuous sign languagerecognition,” Multimedia Tools and Applications,vol.79,pp.22177– 22209, 2020

    work page 2020

Show all 43 references
  1. [8]

    Continuous chinese sign language recognition with cnn-lstm,

    S. Yang and Q. Zhu, “Continuous chinese sign language recognition with cnn-lstm,” in Ninth international conference on digital image processing (ICDIP 2017), SPIE, vol. 10420, 2017, pp. 83–89

    2017

  2. [9]

    A modified lstm model for continuous sign language recognition using leap motion,

    A.Mittal,P.Kumar,P.P.Roy,R.Balasubramanian,andB.B.Chaudhuri, “A modified lstm model for continuous sign language recognition using leap motion,” IEEE Sensors Journal, vol. 19, pp. 7056–7063, 2019

    2019

  3. [10]

    Mediapipe’s landmarks with rnn for dynamic sign language recognition,

    G. H. Samaan, A. R. Wadie, A. K. Attia,et al., “Mediapipe’s landmarks with rnn for dynamic sign language recognition,” Electronics, vol. 11, p. 3228, 2022

    2022

  4. [11]

    Rnn-transducer based chinese sign language recognition,

    L. Gao, H. Li, Z. Liu, Z. Liu, L. Wan, and W. Feng, “Rnn-transducer based chinese sign language recognition,”Neurocomputing, vol. 434, pp. 45–54, 2021

    2021

  5. [12]

    Signbert: Pre-training of hand-model-aware representation for sign language recognition,

    H. Hu, W. Zhao, W. Zhou, Y. Wang, and H. Li, “Signbert: Pre-training of hand-model-aware representation for sign language recognition,” inPro- ceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 11087–11096

    2021

  6. [13]

    Sign pose-based transformer for word-level sign language recognition,

    M. Boháček and M. Hrúz, “Sign pose-based transformer for word-level sign language recognition,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, 2022, pp. 182–191

    2022

  7. [14]

    Jiang, B

    S. Jiang, B. Sun, L. Wang, Y. Bai, K. Li, and Y. Fu, Skeleton aware multi-modal sign language recognition, 2021. arXiv:2103.08833 [cs.CV]. [Online]. Available:https://arxiv.org/abs/2103.08833

    2021 arXiv

  8. [15]

    Spatial-temporal graph convolutional networks for sign language recognition,

    C. C. de Amorim, D. Macêdo, and C. Zanchettin, “Spatial-temporal graph convolutional networks for sign language recognition,” in International Conference on Artificial Neural Networks, Springer, 2019, pp. 646–657

    2019

  9. [16]

    Computation of 3d skeletons using a generalized delaunay triangulation technique,

    J. M. Reddy and G. M. Turkiyyah, “Computation of 3d skeletons using a generalized delaunay triangulation technique,”Computer-Aided Design, vol. 27, pp. 677–694, 1995, issn: 0010-4485. doi: https : / / doi . org / 10 . 1016 / 0010 - 4485(94 ) 00025 - 9. [Online]. Available: htt...

    1995

  10. [17]

    Skelemotion: A new representation of skeleton joint sequences based on motion information for 3d action recognition,

    C. Caetano, J. Sena, F. Brémond, J. A. Dos Santos, and W. R. Schwartz, “Skelemotion: A new representation of skeleton joint sequences based on motion information for 3d action recognition,” in2019 16th IEEE Inter- national Conference on Advanced Video and Signal Based Surveill...

    2019

  11. [18]

    A survey on graph neural networks and its applications in various do- mains,

    T. R. Murgod, P. S. Reddy, S. Gaddam, S. M. Sundaram, and C. Anitha, “A survey on graph neural networks and its applications in various do- mains,” SN Comput. Sci., vol. 6, 2024.doi: 10.1007/s42979-024-03543-

    2024 doi

  12. [19]

    Available:https://doi.org/10.1007/s42979-024-03543-4

    [Online]. Available:https://doi.org/10.1007/s42979-024-03543-4

    doi

  13. [20]

    A survey on skeleton-basedactivityrecognitionusinggraphconvolutionalnetworks(gcn),

    M. Fanuel, X. Yuan, H. N. Kim, L. Qingge, and K. Roy, “A survey on skeleton-basedactivityrecognitionusinggraphconvolutionalnetworks(gcn),” in2021 12th International Symposium on Image and Signal Processing and Analysis (ISPA), IEEE, 2021, pp. 177–182

    2021

  14. [21]

    Lewis, Y

    M. Lewis, Y. Liu, N. Goyal, et al., Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehen- sion, 2019. arXiv: 1910 . 13461 [cs.CL]. [Online]. Available: https : / / arxiv.org/abs/1910.13461

    2019 arXiv

  15. [22]

    Sift based approach on bangla sign language recognition,

    F. Yasir, P. Prasad, A. Alsadoon, and A. Elchouemi, “Sift based approach on bangla sign language recognition,” in 2015 IEEE 8th International Workshop on Computational Intelligence and Applications (IWCIA), 2015, pp. 35–39. doi: 10.1109/IWCIA.2015.7449458

    2015

  16. [23]

    Learning sign language by watching tv (using weakly aligned subtitles),

    P. Buehler, A. Zisserman, and M. Everingham, “Learning sign language by watching tv (using weakly aligned subtitles),” in2009 IEEE Conference on Computer Vision and Pattern Recognition, IEEE, 2009, pp. 2961–2968

    2009

  17. [24]

    A com- prehensive review of dimensionality reduction techniques for feature selec- tion and feature extraction,

    R. Zebari, A. Abdulazeez, D. Zeebaree, D. Zebari, and J. Saeed, “A com- prehensive review of dimensionality reduction techniques for feature selec- tion and feature extraction,”Journal of Applied Science and Technology Trends, vol. 1, pp. 56–70, 2020

    2020

  18. [25]

    Sign language recognition: A deep survey,

    R. Rastgoo, K. Kiani, and S. Escalera, “Sign language recognition: A deep survey,” Expert Systems with Applications, vol. 164, p. 113794, 2021

    2021

  19. [26]

    Deep learning for sign language recognition: Current techniques, benchmarks, and open issues,

    M. Al-Qurishi, T. Khalid, and R. Souissi, “Deep learning for sign language recognition: Current techniques, benchmarks, and open issues,”IEEE Ac- cess, vol. 9, pp. 126917–126951, 2021

    2021

  20. [27]

    Action recognition in videos using pre-trained 2d convolutional neural networks,

    J.-H. Kim and C. Won, “Action recognition in videos using pre-trained 2d convolutional neural networks,”IEEE Access, vol. 8, pp. 60179–60188,

  21. [28]

    Desai, L

    A. Desai, L. Berger, F. O. Minakov, et al., Asl citizen: A community- sourced dataset for advancing isolated sign language recognition, 2023. arXiv: 2304.05934 [cs.CV] . [Online]. Available: https://arxiv.org/ abs/2304.05934

    2023 arXiv

  22. [29]

    A review of convolutional-neural-network- based action recognition,

    G. Yao, T. Lei, and J. Zhong, “A review of convolutional-neural-network- based action recognition,”Pattern Recognition Letters, vol. 118, pp. 14–22, 2019

    2019

  23. [30]

    A comprehensive study on deep learning-based methods for sign language recognition,

    N. Adaloglou, T. Chatzis, I. Papastratis,et al., “A comprehensive study on deep learning-based methods for sign language recognition,”IEEE trans- actions on multimedia, vol. 24, pp. 1750–1762, 2021. SignBart 15

    2021

  24. [31]

    Self-supervised video transformers for isolated sign language recognition,

    M.Sandoval-Castaneda,Y.Li,D.Brentari,K.Livescu,andG.Shakhnarovich, “Self-supervised video transformers for isolated sign language recognition,” arXiv preprint arXiv:2309.02450, 2023

    2023 arXiv

  25. [32]

    Graph convolutional net- works: A comprehensive review,

    S. Zhang, H. Tong, J. Xu, and R. Maciejewski, “Graph convolutional net- works: A comprehensive review,”Computational Social Networks, vol. 6, pp. 1–23, 2019

    2019

  26. [33]

    Spatial temporal graph convolutional net- works for skeleton-based action recognition,

    S. Yan, Y. Xiong, and D. Lin, “Spatial temporal graph convolutional net- works for skeleton-based action recognition,” inProceedings of the AAAI conference on artificial intelligence, vol. 32, 2018

    2018

  27. [35]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar,et al., Attention is all you need, 2023. arXiv: 1706.03762 [cs.CL] . [Online]. Available: https://arxiv.org/ abs/1706.03762

    2023 arXiv

  28. [36]

    D. Li, C. R. Opazo, X. Yu, and H. Li,Word-level deep sign language recog- nition from video: A new large-scale dataset and methods comparison,2020. arXiv: 1910.11006 [cs.CV] . [Online]. Available: https://arxiv.org/ abs/1910.11006

    2020 arXiv

  29. [37]

    Lsa64: A dataset of argentinian sign language,

    F.Ronchetti,F.Quiroga,C.Estrebou,L.Lanzarini,andA.Rosete,“Lsa64: A dataset of argentinian sign language,” XX II Congreso Argentino de Ciencias de la Computación (CACIC), 2016

    2016

  30. [38]

    Hand pose guided 3d pooling for word-level sign language recognition,

    A. A. Hosain, P. S. Santhalingam, P. Pathak, H. Rangwala, and J. Kosecka, “Hand pose guided 3d pooling for word-level sign language recognition,” in Proceedings of the IEEE/CVF winter conference on applications of com- puter vision, 2021, pp. 3429–3439

    2021

  31. [39]

    R. Zuo, F. Wei, and B. Mak,Natural language-assisted sign language recog- nition, 2023. arXiv: 2303.12080 [cs.CV] . [Online]. Available:https:// arxiv.org/abs/2303.12080

    2023 arXiv

  32. [40]

    Patra, A

    S. Patra, A. Maitra, M. Tiwari, et al., Hierarchical windowed graph at- tention network and a large scale dataset for isolated indian sign language recognition, 2024. arXiv:2407.14224 [cs.CV]. [Online]. Available:https: //arxiv.org/abs/2407.14224

    2024 arXiv

  33. [41]

    arXiv:2302.05075 [cs.CV]

    W.Zhao,H.Hu,W.Zhou,J.Shi,andH.Li, Best: Bert pre-training for sign language recognition with coupling tokenization, 2023. arXiv:2302.05075 [cs.CV]. [Online]. Available:https://arxiv.org/abs/2302.05075

    2023 arXiv

  34. [43]

    Spatialattention- based 3d graph convolutional neural network for sign language recogni- tion,

    M.Al-Hammadi,M.A.Bencherif,M.Alsulaiman, et al.,“Spatialattention- based 3d graph convolutional neural network for sign language recogni- tion,” Sensors, vol. 22, 2022,issn: 1424-8220. doi: 10.3390/s22124558. [Online]. Available:https://www.mdpi.com/1424-8220/22/12/4558

    2022 doi

  35. [2020]

    doi: 10.1109/ACCESS.2020.2983427

    2020

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.