Pith. sign in

REVIEW 3 major objections 5 minor 50 references

Training Strategies for Isolated Sign Language Recognition

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A model-agnostic training pipeline for isolated sign language recognition lifts accuracy on every tested dataset and architecture, reaching state-of-the-art results on WLASL and Slovo.

desk verdict A useful, model-agnostic training recipe for ISLR with consistent gains, but the SOTA claims are undermined by test-set-based flip selection and unmatched pre-training. read the letter →

arxiv 2412.11553 v2 pith:JX4CNRPI submitted 2024-12-16 cs.CV

Karina Kvanchiani , Roman Kraynov , Elizaveta Petrova , Petr Surovcev , Aleksandr Nagaev , Alexander Kapitanov This is my paper · ORCID
classification cs.CV
keywords isolatedsignlanguagerecognitiontrainingpipelinevideoaugmentationimageboundaryregressionIoU-balancedcross-entropySlovoExtdatasetWLASLandbenchmarks
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 claims that a large share of the remaining error in isolated sign language recognition—classifying one sign from a short RGB video—can be removed by changing how the model is trained, not what it is. The proposed recipe applies video augmentations that simulate different signing speeds and shifted sign boundaries, image augmentations that reproduce poor recording quality, and two loss-side additions that teach the model where in the clip the sign actually happens. Across the WLASL, AUTSL, Slovo, and newly assembled SlovoExt datasets, and across transformer and CNN architectures, the pipeline improves top-1 accuracy in every configuration tested, with the largest gain being +10.12 points on Slovo. On WLASL and Slovo the paper reports state-of-the-art results. The practical significance is that a sign recognition system can become substantially more robust without new hardware, new modalities, or a new architecture.

What carries the argument

The load-bearing mechanism is the combination of a sign boundary regression head and an IoU-balanced cross-entropy loss, together with the augmentations that feed them. The regression head is a fully connected branch parallel to the classification head that outputs the predicted first and last sign frames; Huber loss trains it. The IoU-balanced loss multiplies each class score by the fraction of the sampled window that overlaps the true sign interval (intersection length divided by window size), down-weighting clips where the sign is partially outside the window. The video augmentations (speed up, slow down, random frame add/drop, random boundary shift) and image augmentations (compression, noise, sharpness, flip, color jitter, CutMix, MixUp) are what expose the model to the real-world variability this loss-based supervision exploits.

What would settle it

Collect a test set of signs whose meaning changes under mirroring (for example, signs that use left versus right hand) and train the same pipeline with and without horizontal flip: the paper's account predicts mirroring should hurt those classes, and if per-class accuracy is unchanged or improves, the meaning-preservation assumption is not what drives the result.

Watch

Extended reading notes

Core claim

The central claim is that sign boundary awareness, injected at training time through an auxiliary regression head and an IoU-scaled classification loss, is the missing ingredient on top of augmentation. The model receives a 32-frame window from an untrimmed or trimmed video; a small regression branch predicts the start and end frames of the sign, trained with Huber loss, while the classification score is multiplied by the intersection-over-union between the sampled window and the true sign interval, so frames that miss the sign contribute less. The authors argue that this makes the backbone attend to the frames that carry the sign and that the effect is additive with the augmentations: every component contributes in the ablation, and the full pipeline reaches 62.89% top-1 on WLASL and 78.21% mean accuracy on Slovo, which the paper reports as surpassing prior state of the art.

Load-bearing premise

The load-bearing premise is that every augmentation leaves the sign's meaning intact: mirroring a video can turn one sign into a different sign, and duplicating frames can distort hand shapes, but the paper assumes these changes are harmless and picks the mirroring policy by comparing test-set accuracies.

Editorial extensions

If this is right

  • Teams can apply the same training recipe to a new ISLR model without changing its architecture, since the gains appear for both transformers and a CNN.
  • Datasets with untrimmed videos and "no event" frames stand to gain most, as Slovo shows the largest improvement (+10.12 top-1 points); the boundary regression head is designed to exploit exactly that structure.
  • Because the pipeline is RGB-only, it preserves real-time deployment constraints and can be combined with later pose or depth modalities if available.
  • The released SlovoExt dataset and pre-trained models give Russian Sign Language researchers a larger, signer-diverse training resource and a new benchmark for isolated signs.

Reading between the lines

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

  • Editorial inference: the same two losses could be dropped into a continuous sign language recognition model, where segment boundaries are not given; the gain may be larger there because localization is a core difficulty.
  • Editorial inference: the ablation numbers suggest the losses add less than the augmentations (0.52% and 0.05% vs up to 4% for image augmentations), so the headline mechanism might be data diversity rather than boundary awareness; a controlled study holding augmentations fixed while toggling losses could test this.
  • Editorial inference: the policy of choosing horizontal flip by comparing test-set accuracy risks overfitting the benchmark; a linguistically verified list of non-mirrored signs would be a more transportable decision rule.
  • Editorial inference: the 0.66% attributed to slow-down frame duplication could be re-interpreted as robustness to duplicated handshapes rather than to slower signing; a minimal-pair test on handshape-sensitive signs would disambiguate.
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 / 5 minor

Summary. The paper proposes a training pipeline for isolated sign language recognition (ISLR) consisting of video-level augmentations (speed up/down, random add/drop, boundary shift), image-level augmentations (quality degradations, horizontal flip, color jitter, CutMix/MixUp), an auxiliary sign-boundary regression head trained with Huber loss, and an IoU-balanced cross-entropy loss. The pipeline is evaluated by fine-tuning MViTv2-S, MViTv2-B, and I3D on WLASL, AUTSL, Slovo, and a newly introduced SlovoExt dataset, comparing against a basic fine-tuning baseline. The paper reports consistent accuracy gains across datasets and architectures, ablation studies isolating each component, and state-of-the-art results on WLASL and Slovo.

Significance. If the results are validated, the pipeline is a useful and model-agnostic contribution to ISLR: it improves accuracy without changing the model architecture, and the ablation study provides actionable evidence about which components matter. The paper also releases code and pretrained models, provides detailed hyperparameters in the supplementary material, and documents careful ethical and community involvement in dataset construction. However, the strongest empirical claims are currently weakened by two issues: the horizontal-flip policy is selected using the reported accuracy itself rather than an independent validation split, and the state-of-the-art comparisons use additional pre-training on the authors' own SlovoExt corpus that the cited baselines did not receive. These issues affect the central claim of consistent, unbiased gains, so the paper needs revision before the claims can be accepted at face value.

major comments (3)
  1. [Section 6.1 and Section 5.1] The decision to apply horizontal flip to AUTSL but not to WLASL is made by comparing the accuracy of two runs per dataset and selecting the better configuration. The paper does not state that this comparison was performed on a held-out validation split; if it was performed on the test split, the final configurations reported in Table 2 are partially selected to maximize the reported test accuracy, which introduces optimistic bias into the claimed gains on WLASL and AUTSL. The problem is especially serious for AUTSL, where the reported advantage of flipping is only 0.11%, a difference that is likely within run-to-run noise; no multiple-seed or variance estimates are reported anywhere in the paper. The authors should either fix the flip policy using a validation split, report results for both configurations, or provide repeated-seed means and confidence intervals.
  2. [Section 5.3] The state-of-the-art claims on WLASL and Slovo are not apples-to-apples comparisons. The WLASL SOTA result is obtained with MViTv2-S pre-trained with MaskFeat and then further pre-trained on SlovoExt, a corpus introduced in this paper; the cited NLA-SLR baseline was not trained on SlovoExt. Similarly, the Slovo SOTA result is reported on a benchmark built from the authors' own Slovo data, with models pre-trained on SlovoExt. Additional in-domain pre-training is a substantial advantage that is not controlled for in the comparison. The authors should compare against baselines that receive the same pretraining data, remove the extra pre-training from the SOTA runs, or clearly label these results as 'with in-domain pre-training' and refrain from claiming a direct SOTA improvement over methods that did not use that data.
  3. [Section 6 and Table 3] The ablation study is conducted only on SlovoExt with MViTv2-S pretrained with MaskFeat. The central claim is that the pipeline yields consistent, model-agnostic improvements across datasets and architectures, but the component-level evidence comes from a single architecture and a single dataset. The concern is compounded by the fact that some pipeline components are dataset-specific (the boundary-shift interval differs for WLASL/AUTSL, and the flip policy is chosen per dataset). The authors should either repeat the ablation on at least one other dataset or architecture, or explicitly limit the component-level conclusions to SlovoExt and discuss the risk that component contributions vary across datasets.
minor comments (5)
  1. [Section 3.3, Eq. (1)] The formula for the IoU score does not specify what happens when the sampled window does not overlap the sign at all; in that case the numerator is negative, and the text gives no clipping or normalization rule. Please define the value of the score for non-overlapping windows and clarify whether the score is clipped to [0, 1].
  2. [Section 4.1 and Section 8] The contribution section says SlovoExt is released, while the Limitations and the dataset description say it 'will be available as part of a larger RSL dataset in future work.' Please state clearly the current availability status of SlovoExt, the exact release conditions, and whether the evaluation on Slovo uses the original Slovo test split or a new split defined by the authors.
  3. [Algorithm 1 and Section 5.2] Algorithm 1's notation 'y.extend([videok]^ylen_{k=32})' is unclear; please rewrite the extension step in standard pseudocode. Also, the text in Section 5.2 says one random video augmentation is applied per batch, but the ablation table and Figure 3 suggest each augmentation is applied with a given probability; please clarify how the per-batch selection interacts with the probabilities in Table 6.
  4. [Table 6] The table lists blacklisted labels for the flip augmentation but does not state which dataset those labels belong to. The text says non-mirrored signs are excluded for Slovo/SlovoExt, yet the table gives only a list of IDs without indicating the dataset or the corresponding sign names; please make this explicit.
  5. [Section 3.1] The 'slow down' augmentation duplicates frames, and the paper does not discuss whether duplicated frames can change the perceived speed or articulation in a way that alters sign identity. A short discussion or a simple sanity check would address this concern, especially because the ablation attributes a 0.66% gain to this operation.

Circularity Check

1 steps flagged · score 4.0 of 10

Flip-augmentation policy is selected on the reported top-1 metrics, making WLASL/AUTSL results partly fitted; the core pipeline evaluation is otherwise independent.

  1. fitted input called prediction [Section 6.1 (Image Augmentations Necessity), applied via Section 5.1 and Table 2]
    "Additionally, since we are uncertain which signs are non-mirrored in WLASL and AUTSL datasets, we search for the optimal option via two experiments for each: one with flip augmentation and one without for all signs. On WLASL, the metric with a flip is lower by 1.3%, indicating that this dataset probably contains numerous non-mirrored signs. On AUTSL, conversely, the metric with flip was higher by 0.11%. Therefore, in the main experiments, we apply flip augmentation to the AUTSL dataset and refrain from using it with the WLASL dataset."

    The flip on/off hyperparameter is selected by comparing the same top-1 benchmark metric that Table 2 later reports as the pipeline's result, and Section 5.1 confirms the choice was carried into the main results: 'the WLASL signs are not being flipped, and the AUTSL signs are flipped with no exceptions because Section 6 shows the efficacy of such decisions.' No validation split is identified for this search, so the reported WLASL/AUTSL accuracies are the maximum over two test-set configurations rather than an independent prediction. The AUTSL flip advantage (0.11%) is within plausible run-to-run noise, and no variance or multiple-seed estimates are reported.

full rationale

The paper's central contribution is an empirical training pipeline (video/image augmentations, boundary regression head, IoU-scaled CE loss) evaluated against a 'basic approach' baseline. That comparison is not circular: the basic approach is defined as the pipeline without the proposed components, the Slovo/SlovoExt gains in Table 2 are independent test-set measurements, and the losses and augmentations are standard supervised techniques rather than quantities derived from the target metric. The IoU score in Eq. (1) uses ground-truth sign boundaries as supervision; it is a training loss, not a self-defined evaluation metric. The Slovo and SlovoExt assets come from the authors, but the Slovo benchmark is public and the SlovoExt evaluation is a separately reported dataset, so these are self-produced data rather than a self-citation chain that forces the conclusion. The one genuine selection loop is the flip-augmentation policy for WLASL and AUTSL in Section 6.1: the authors choose flip on/off by comparing top-1 accuracy on each benchmark and then report the selected configuration as the pipeline's result, without stating that the comparison was on a validation split. This makes the reported WLASL/AUTSL numbers partially selected rather than predicted; the effect is small (AUTSL flip advantage is 0.11%) and does not affect the main ablation, so the overall circularity score is moderate rather than severe.

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

The pipeline's effectiveness rests on a collection of hand-chosen augmentation hyperparameters and domain assumptions about sign semantics and boundary annotation quality. The central empirical comparison is against the authors' own baseline, and the strongest SOTA results add the authors' own SlovoExt pre-training, so the external grounding of the headline claims is weaker than the internal consistency.

free parameters (4)
  • Image augmentation probabilities and magnitudes = Table 6 values (color jitter p=0.5, noise p=0.5, sharpness p=0.35, flip p=0.5, random erasing p=0.25, compression…
    Selected by hand as 'expertly selected' (Section 5.2); ablations show this block contributes up to 4.06% accuracy.
  • Video augmentation rates = drop ratio 0.1 (p=0.5), add ratio 0.3 (p=0.25), speed up factor 2 (p=0.25), slow down factor 2 (p=0.25), boundary…
    Hand-chosen in Section 5.2 and Table 6; ablations show each contributes roughly 0.3-0.7%.
  • Per-dataset flip decision = AUTSL: flip all signs; WLASL: no flip; Slovo/SlovoExt: flip except labels 739, 635, 148, 636
    Selected by comparing test-set accuracies in Section 6.1, a test-set-based choice.
  • CutMix and MixUp alphas = CutMix alpha=0.8, MixUp alpha=1
    Taken as-is from prior work and applied only to transformer models (Section 5.2).
assumptions (4)
  • domain assumption Sign boundary annotations are accurate enough to serve as regression targets and IoU weights.
    Section 3.3 trains the regression head with Huber loss against these boundaries; Section 6.3 shows the head adds 0.52%.
  • domain assumption The augmentations (frame duplication, speed change, boundary shift, flip) preserve the semantic meaning of signs.
    Section 3.1 and 3.2 introduce these; the paper itself flags uncertainty about non-mirrored signs for WLASL and AUTSL.
  • ad hoc to paper Pre-training on SlovoExt transfers positively to WLASL.
    Used for the WLASL SOTA result (Section 5.3); no controlled comparison against pre-training on another large SL dataset is provided.
  • ad hoc to paper The overlap score in Eq. (1) is an appropriate surrogate for sign localization quality.
    The quantity is defined as intersection over window length and used to weight classification scores; its benefit is only shown empirically (+0.05%).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Training Strategies for Isolated Sign Language Recognition." pith.science (2026). https://pith.science/paper/JX4CNRPI

@misc{pith2026241211553,
  author       = {Pith},
  title        = {Pith review of: Training Strategies for Isolated Sign Language Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JX4CNRPI}},
  note         = {Machine review of arXiv:2412.11553}
}
read the original abstract

Accurate recognition and interpretation of sign language are crucial for enhancing communication accessibility for deaf and hard of hearing individuals. However, current approaches of Isolated Sign Language Recognition (ISLR) often face challenges such as low data quality and variability in gesturing speed. This paper introduces a comprehensive model training pipeline for ISLR designed to accommodate the distinctive characteristics and constraints of the Sign Language (SL) domain. The constructed pipeline incorporates carefully selected image and video augmentations to tackle the challenges of low data quality and varying sign speeds. Including an additional regression head combined with IoU-balanced classification loss enhances the model's awareness of the gesture and simplifies capturing temporal information. Extensive experiments demonstrate that the developed training pipeline easily adapts to different datasets and architectures. Additionally, the ablation study shows that each proposed component expands the potential to consider ISLR task specifics. The presented strategies enhance recognition performance across various ISLR benchmarks and achieve state-of-the-art results on the WLASL and Slovo datasets.

Figures

Figures reproduced from arXiv: 2412.11553 by the authors.

Figure 1
Figure 1. The process of receiving IoU scores. (a) Consider the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall training pipeline. Video-level and image-level augmentations are applied, and the neural network is further trained with [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Applying video augmentations to untrimmed videos. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 43 canonical work pages

  1. [1]

    Slow- fast Network for Continuous Sign Language Recognition

    Junseok Ahn, Youngjoon Jang, and Joon Son Chung. Slow- fast Network for Continuous Sign Language Recognition. In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 3920–3924. IEEE, 2024

    work page 2024

  2. [2]

    Visual Transformer Meets CutMix for Improved Accuracy, Communication Efficiency, and Data Privacy in Split Learning

    Sihun Baek, Jihong Park, Praneeth Vepakomma, Ramesh Raskar, Mehdi Bennis, and Seong-Lyun Kim. Visual trans- 8 former meets cutmix for improved accuracy, communica- tion efficiency, and data privacy in split learning. In arXiv preprint arXiv:2207.00234, 2022

    work page Pith review arXiv 2022

  3. [3]

    Multi-channel transformers for multi- articulatory sign language translation

    Necati Cihan Camgoz, Oscar Koller, Simon Hadfield, and Richard Bowden. Multi-channel transformers for multi- articulatory sign language translation. In Computer Vision– ECCV 2020 Workshops: Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16, pages 301–319. Springer, 2020

    work page 2020

  4. [4]

    Quo vadis, action recognition? a new model and the kinetics dataset

    Joao Carreira and Andrew Zisserman. Quo vadis, action recognition? a new model and the kinetics dataset. In pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6299–6308, 2017

    work page 2017

  5. [5]

    A simple multi-modality transfer learning baseline for sign language translation

    Yutong Chen, Fangyun Wei, Xiao Sun, Zhirong Wu, and Stephen Lin. A simple multi-modality transfer learning baseline for sign language translation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5120–5130, 2022

    2022

  6. [6]

    Two-stream network for sign language recognition and translation

    Yutong Chen, Ronglai Zuo, Fangyun Wei, Yu Wu, Shujie Liu, and Brian Mak. Two-stream network for sign language recognition and translation. In Advances in Neural Infor- mation Processing Systems, volume 35, pages 17043–17056, 2022

    work page 2022

  7. [7]

    Randaugment: Practical automated data augmen- tation with a reduced search space

    Ekin D Cubuk, Barret Zoph, Jonathon Shlens, and Quoc V Le. Randaugment: Practical automated data augmen- tation with a reduced search space. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, pages 702–703, 2020

    work page 2020

  8. [8]

    Re- altime sign language detection and recognition

    Aakash Deep, Aashutosh Litoriya, Akshay Ingole, Vaib- hav Asare, Shubham M Bhole, and Shantanu Pathak. Re- altime sign language detection and recognition. In 2022 2nd Asian Conference on Innovation in Technology (ASIAN- CON), pages 1–4. IEEE, 2022

    work page 2022

Show all 50 references
  1. [9]

    Systemic Biases in Sign Language AI Research: A Deaf-Led Call to Reevaluate Re- search Agendas

    Aashaka Desai, Maartje De Meulder, Julie A Hochgesang, Annemarie Kocab, and Alex X Lu. Systemic Biases in Sign Language AI Research: A Deaf-Led Call to Reevaluate Re- search Agendas. In arXiv preprint arXiv:2403.02563, 2024

    2024 arXiv

  2. [10]

    Extending Temporal Data Augmentation for Video Action Recognition

    Artjoms Gorpincenko and Michal Mackiewicz. Extending Temporal Data Augmentation for Video Action Recognition. In Computer Vision – ECCV 2022 Workshops , pages 116–

    2022

  3. [11]

    LSE-sign: A lexical database for spanish sign language

    Eva Gutierrez-Sigut, Brendan Costello, Cristina Baus, and Manuel Carreiras. LSE-sign: A lexical database for spanish sign language. In Behavior Research Methods, volume 48, pages 123–137. Springer, 2016

    2016

  4. [12]

    Self-mutual dis- tillation learning for continuous sign language recognition

    Aiming Hao, Yuecong Min, and Xilin Chen. Self-mutual dis- tillation learning for continuous sign language recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 11303–11312, 2021

    2021

  5. [13]

    Con- tinuous sign language recognition with correlation network

    Lianyu Hu, Liqing Gao, Zekang Liu, and Wei Feng. Con- tinuous sign language recognition with correlation network. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2529–2539, 2023

    2023

  6. [14]

    Robust estimation of a location parameter

    Peter J Huber. Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution , pages 492–518. Springer, 1992

    1992

  7. [15]

    A dataset for linguistic understand- ing, visual evaluation, and recognition of sign languages: The k-rsl

    Alfarabi Imashev, Medet Mukushev, Vadim Kimmelman, and Anara Sandygulova. A dataset for linguistic understand- ing, visual evaluation, and recognition of sign languages: The k-rsl. In Proceedings of the 24th conference on com- putational natural language learning, pages 631–640, 2020

    2020

  8. [16]

    A Survey of Sign Language Recog- nition Systems

    Vaishnavi Jadhav, Priyal Agarwal, Dhruvisha Mondhe, Ru- tuja Patil, and CS Lifna. A Survey of Sign Language Recog- nition Systems. In booktitle of Innovative Image Processing, volume 4, pages 237–246, 2022

    2022

  9. [17]

    ADDSL: hand gesture detection and sign lan- guage recognition on annotated danish sign language

    Sanyam Jain. ADDSL: hand gesture detection and sign lan- guage recognition on annotated danish sign language. In arXiv preprint arXiv:2305.09736, 2023

    2023 arXiv

  10. [18]

    Ms-asl: A large- scale data set and benchmark for understanding american sign language

    Hamid Reza Vaezi Joze and Oscar Koller. Ms-asl: A large- scale data set and benchmark for understanding american sign language. In arXiv preprint arXiv:1812.01053, 2018

    2018 arXiv

  11. [19]

    TheRuSLan: Database of Russian sign language

    Ildar Kagirov, Denis Ivanko, Dmitry Ryumin, Alexander Axyonov, and Alexey Karpov. TheRuSLan: Database of Russian sign language. In Proceedings of the Twelfth Lan- guage Resources and Evaluation Conference , pages 6079– 6085, 2020

    2020

  12. [20]

    Sign language translation with hierarchical spatio-temporal graph neural network

    Jichao Kan, Kun Hu, Markus Hagenbuchner, Ah Chung Tsoi, Mohammed Bennamoun, and Zhiyong Wang. Sign language translation with hierarchical spatio-temporal graph neural network. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 3367– 3376, 2022

    2022

  13. [21]

    Slovo: Russian Sign Language Dataset

    Alexander Kapitanov, Kvanchiani Karina, Alexander Na- gaev, and Petrova Elizaveta. Slovo: Russian Sign Language Dataset. In International Conference on Computer Vision Systems, pages 63–73. Springer, 2023

    2023

  14. [22]

    The ki- netics human action video dataset

    Will Kay, Joao Carreira, Karen Simonyan, Brian Zhang, Chloe Hillier, Sudheendra Vijayanarasimhan, Fabio Viola, Tim Green, Trevor Back, Paul Natsev, et al. The ki- netics human action video dataset. In arXiv preprint arXiv:1705.06950, 2017

    2017 arXiv

  15. [23]

    Word-level deep sign language recognition from video: A new large-scale dataset and methods comparison

    Dongxu Li, Cristian Rodriguez, Xin Yu, and Hongdong Li. Word-level deep sign language recognition from video: A new large-scale dataset and methods comparison. In Pro- ceedings of the IEEE/CVF winter conference on applications of computer vision, pages 1459–1469, 2020

    2020

  16. [24]

    Mvitv2: Improved multiscale vision transform- ers for classification and detection

    Yanghao Li, Chao-Yuan Wu, Haoqi Fan, Karttikeya Man- galam, Bo Xiong, Jitendra Malik, and Christoph Feicht- enhofer. Mvitv2: Improved multiscale vision transform- ers for classification and detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern rec...

    2022

  17. [25]

    Auxiliary tasks in multi- task learning

    Lukas Liebel and Marco K ¨orner. Auxiliary tasks in multi- task learning. In arXiv preprint arXiv:1805.06334, 2018

    2018 arXiv

  18. [26]

    Uniformaugment: A search-free proba- bilistic data augmentation approach

    Tom Ching LingChen, Ava Khonsari, Amirreza Lashkari, Mina Rafi Nazari, Jaspreet Singh Sambee, and Mario A Nascimento. Uniformaugment: A search-free proba- bilistic data augmentation approach. In arXiv preprint arXiv:2003.14348, 2020

    2003 arXiv

  19. [27]

    End-to-end temporal action detection with transformer

    Xiaolong Liu, Qimeng Wang, Yao Hu, Xu Tang, Shiwei Zhang, Song Bai, and Xiang Bai. End-to-end temporal action detection with transformer. In IEEE Transactions on Image Processing, volume 31, pages 5427–5441. IEEE, 2022

    2022

  20. [28]

    Decoupled Weight De- cay Regularization

    Ilya Loshchilov and Frank Hutter. Decoupled Weight De- cay Regularization. In International Conference on Learning Representations, 2019. 9

    2019

  21. [29]

    An attention-enhanced multi- scale and dual sign language recognition network based on a graph convolution network

    Lu Meng and Ronghui Li. An attention-enhanced multi- scale and dual sign language recognition network based on a graph convolution network. In Sensors, volume 21, page 1120, 2021

    2021

  22. [30]

    Visual alignment constraint for continuous sign language recognition

    Yuecong Min, Aiming Hao, Xiujuan Chai, and Xilin Chen. Visual alignment constraint for continuous sign language recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 11542–11551, 2021

    2021

  23. [31]

    FluentSigners-50: A signer independent benchmark dataset for sign language processing

    Medet Mukushev, Aidyn Ubingazhibov, Aigerim Kydyr- bekova, Alfarabi Imashev, Vadim Kimmelman, and Anara Sandygulova. FluentSigners-50: A signer independent benchmark dataset for sign language processing. In Plos one, volume 17, page e0273649. Public Library of Science San Fran...

    2022

  24. [32]

    Weakly-supervised fin- gerspelling recognition in british sign language videos

    KR Prajwal, Hannah Bull, Liliane Momeni, Samuel Albanie, G¨ul Varol, and Andrew Zisserman. Weakly-supervised fin- gerspelling recognition in british sign language videos. In arXiv preprint arXiv:2211.08954, 2022

    2022 arXiv

  25. [33]

    Sign language recognition based on notations and neural net- works

    Alexey Prikhodko, Mikhail Grif, and Maxim Bakaev. Sign language recognition based on notations and neural net- works. In Digital Transformation and Global Society: 5th International Conference, DTGS 2020, St. Petersburg, Rus- sia, June 17–19, 2020, Revised Selected Papers 5 , ...

    2020

  26. [34]

    Boosting continuous sign language recognition via cross modality augmentation

    Junfu Pu, Wengang Zhou, Hezhen Hu, and Houqiang Li. Boosting continuous sign language recognition via cross modality augmentation. In Proceedings of the 28th ACM international conference on multimedia , pages 1497–1505, 2020

    2020

  27. [35]

    LSA64: an Argentinian sign language dataset

    Franco Ronchetti, Facundo Manuel Quiroga, C ´esar Estre- bou, Laura Lanzarini, and Alejandro Rosete. LSA64: an Argentinian sign language dataset. In arXiv preprint arXiv:2310.17429, 2023

    2023 arXiv

  28. [36]

    Autsl: A large scale multi-modal turkish sign language dataset and baseline methods

    Ozge Mercanoglu Sincan and Hacer Yalim Keles. Autsl: A large scale multi-modal turkish sign language dataset and baseline methods. In IEEE access, volume 8, pages 181340– 181355. IEEE, 2020

    2020

  29. [37]

    Masked feature predic- tion for self-supervised visual pre-training

    Chen Wei, Haoqi Fan, Saining Xie, Chao-Yuan Wu, Alan Yuille, and Christoph Feichtenhofer. Masked feature predic- tion for self-supervised visual pre-training. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14668–14678, 2022

    2022

  30. [38]

    Iou-balanced loss functions for single-stage object de- tection

    Shengkai Wu, Jinrong Yang, Xinggang Wang, and Xiaoping Li. Iou-balanced loss functions for single-stage object de- tection. In Pattern Recognition Letters, volume 156, pages 96–103. Elsevier, 2022

    2022

  31. [39]

    Action recognition and localization with spatial and temporal con- texts

    Wanru Xu, Zhenjiang Miao, Jian Yu, and Qiang Ji. Action recognition and localization with spatial and temporal con- texts. In Neurocomputing, volume 333, pages 351–363. El- sevier, 2019

    2019

  32. [40]

    CutMix: Regu- larization Strategy to Train Strong Classifiers with Localiz- able Features

    Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo. CutMix: Regu- larization Strategy to Train Strong Classifiers with Localiz- able Features. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 6022–6031, 2019

    2019

  33. [41]

    Actionformer: Lo- calizing moments of actions with transformers

    Chen-Lin Zhang, Jianxin Wu, and Yin Li. Actionformer: Lo- calizing moments of actions with transformers. In European Conference on Computer Vision , pages 492–510. Springer, 2022

    2022

  34. [42]

    mixup: Beyond empirical risk minimiza- tion

    Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimiza- tion. In arXiv preprint arXiv:1710.09412, 2017

    2017 arXiv

  35. [43]

    Gloss-free sign language translation: Improving from visual- language pretraining

    Benjia Zhou, Zhigang Chen, Albert Clap ´es, Jun Wan, Yanyan Liang, Sergio Escalera, Zhen Lei, and Du Zhang. Gloss-free sign language translation: Improving from visual- language pretraining. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision , pages 2...

    2023

  36. [44]

    Spatial-temporal multi-cue network for continuous sign lan- guage recognition

    Hao Zhou, Wengang Zhou, Yun Zhou, and Houqiang Li. Spatial-temporal multi-cue network for continuous sign lan- guage recognition. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 13009–13016, 2020

    2020

  37. [45]

    Learning disentangled classification and lo- calization representations for temporal action localization

    Zixin Zhu, Le Wang, Wei Tang, Ziyi Liu, Nanning Zheng, and Gang Hua. Learning disentangled classification and lo- calization representations for temporal action localization. In Proceedings of the AAAI Conference on Artificial Intel- ligence, volume 36, pages 3644–3652, 2022

    2022

  38. [46]

    Improving continuous sign lan- guage recognition with consistency constraints and signer removal

    Ronglai Zuo and Brian Mak. Improving continuous sign lan- guage recognition with consistency constraints and signer removal. In ACM Transactions on Multimedia Computing, Communications and Applications, volume 20, pages 1–25. ACM New York, NY , 2024

    2024

  39. [47]

    Natural language-assisted sign language recognition

    Ronglai Zuo, Fangyun Wei, and Brian Mak. Natural language-assisted sign language recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 14890–14900, 2023. 10 Supplementary materials A. The Variable Parameters. Aug. type Augme...

    2023

  40. [48]

    The sign basket was the same as in the Slovo dataset [21]: 1,000 frequently used words with a video template from the SpreadTheSign website 9 were collected

    Mining. The sign basket was the same as in the Slovo dataset [21]: 1,000 frequently used words with a video template from the SpreadTheSign website 9 were collected. The signers were informed about further publication of submitted videos

  41. [49]

    The recorded videos were strictly validated, excluding videos performed incorrectly and low-resolution videos

    Validation. The recorded videos were strictly validated, excluding videos performed incorrectly and low-resolution videos. At least three crowdworkers confirmed the correctness of each video. 7https://platform.toloka.ai/ 8https://elementary.activebc.ru 9https://www.spreadthesi...

  42. [50]

    Crowdworkers indicated the start and end frames of the sign on each video to train the model to locate it

    Annotation. Crowdworkers indicated the start and end frames of the sign on each video to train the model to locate it. Three different crowdworkers were assigned to annotate each video, and their annotations were averaged to create a high-confidence, accurate annotation. Then,...

Pith tools

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