Multiscaled Multi-Head Attention-based Video Transformer Network for Hand Gesture Recognition

Debashis Ghosh; Mallika Garg; Pyari Mohan Pradhan

REVIEW 4 major objections 6 minor 30 references

Multiscaled Multi-Head Attention-based Video Transformer Network for Hand Gesture Recognition

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A video transformer with halved head dimensions reports 88.22% on NVGesture and 99.10% on Briareo.

desk verdict Incremental transformer tweak with a plausible but under-specified multiscale head design; the gains versus the baseline are real but confounded by roughly twice the parameters, and the architecture has an unresolved dimension mismatch. read the letter →

arxiv 2501.00935 v1 pith:ZLZ3WJOG submitted 2025-01-01 cs.CV cs.HC

Mallika Garg , Debashis Ghosh , Pyari Mohan Pradhan This is my paper · ORCID

classification cs.CVcs.HC

keywords DynamichandgesturerecognitionVideotransformerMulti-headattentionMultiscalePyramidMultimodalfusionNVGestureBriareo

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 letter tries to show that a simple change to the transformer's multi-head attention, letting each head attend at a different resolution by halving the query and key dimension head by head, improves dynamic hand gesture recognition. The proposed MsMHA-VTN builds a pyramid of attention scales, uses ResNet-18 frame features as in the baseline video transformer, and reports 88.22% overall accuracy on NVGesture and 99.10% on Briareo. The authors also show that fusing RGB, depth, infrared, surface normals, and optical flow at the decision level helps, and that surface normals alone are a strong single modality. If the claim is right, it means scale diversity inside the attention heads, rather than a larger or more complex network, is enough to capture the pose and size variations that make gestures hard to recognize.

What carries the argument

The central object is the Multiscaled Multi-Head Attention (MsMHA) module. In each of the six transformer stages, eight heads compute scaled dot-product attention as in Eq. (1), but instead of a common $d_k$, the query and key tensors are built with dimensions $L \times D$, $L \times D/2$, ..., $L \times D/128$, forming a pyramid of decreasing resolutions. The head outputs are concatenated and projected through the output weight $W^O$ of Eq. (3), and the paper claims this lets each stage see gesture structure at multiple scales while keeping the later heads much smaller.

What would settle it

One concrete check: reimplement the model with the head sizes of Table I and verify that Concat(head1, ..., head8) $W^O$ is a defined matrix multiplication; if it is not, the mechanism as described cannot be what produced the reported accuracies. A second check: train the same pipeline on NVGesture with eight equal-dimension heads and the same total parameter budget; if the equal-head version matches or beats 88.22%, the improvement is not attributable to the multiscale pyramid.

Watch

Extended reading notes

Core claim

The central discovery claimed by the paper is that a video transformer can learn gesture features at multiple scales by making the eight attention heads progressively smaller: head 1 works at $L \times D$, head 2 at $L \times D/2$, and so on down to $L \times D/128$, instead of giving every head the same dimension. Six stacked stages of this Multiscaled Multi-Head Attention (MsMHA) extract frame-level features from a ResNet-18 backbone, and the authors report that this design beats the original transformer on every modality they test and yields overall accuracies of 88.22% on NVGesture and 99.10% on Briareo. The paper interprets these results as evidence that pyramidal, multiscale attention handles the variation in hand pose, size, and shape that makes dynamic gesture recognition hard.

Load-bearing premise

The load-bearing premise is that the eight differently sized attention heads can be concatenated and multiplied by a single output projection $W^O$, exactly as written in Eq. (3); the paper does not specify the size of that projection, and if the shapes do not match, the proposed transformer layer is not well defined.

Editorial extensions

If this is right

The standard transformer baseline [12] is improved on every single modality tested: color, depth, optical flow, normals, and infrared all score higher with the pyramid attention.
Surface normals computed from depth are a particularly informative input: 86.21% on NVGesture and 98.67% on Briareo with normals alone, already close to the best multimodal scores.
Late fusion helps on NVGesture, where all five input streams give the best result (88.22%), but on Briareo three streams (color, infrared, normals) are enough, and adding more modalities does not improve accuracy.
The reported accuracies compare favorably with published methods on both datasets, including recurrent 3D CNNs, multi-stream CNNs, and other transformer or neural-architecture-search based systems.

Reading between the lines

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

Beyond the paper's datasets, the same pyramid-head design could be tested on other video classification tasks with scale variation, such as sign language recognition or action recognition; the paper only evaluates hand gestures.
Because the later heads are much smaller, MsMHA likely uses fewer parameters and less compute per stage than a standard eight-head transformer, though the paper does not report parameter counts or FLOPs; a direct efficiency comparison would be needed to confirm.
The strong normals-only results suggest a practical single-sensor deployment path: if depth-based normals alone nearly match multimodal fusion on Briareo, one can drop RGB and IR streams in settings where a depth camera is available.
An ablation that keeps total parameters fixed while varying head dimensions would separate the multiscale effect from capacity effects; the paper does not include such an ablation.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The short version: this paper has a small, concrete idea—give each attention head a different Q/K/V dimension, halving head by head, so the layer sees a pyramid of scales—and it reports a few points of improvement over the vanilla Transformer baseline on two public gesture datasets. The idea is not identical to MViT or CrossViT; those use pooling and cross-attention, whereas this is a per-head dimension pyramid. That is a legitimate variant, even if it is more of a parameter choice than a new principle.

What it does well: the evaluation is on public benchmarks (NVGesture, Briareo), the comparison list is broad, and the late-fusion results are internally consistent. On Briareo, the proposed 99.10% is genuinely above the Transformer baseline (95.80% on normals, 97.20% on ir+normals). For a letter, the experimental setup is acceptable apart from what is missing.

Now the soft spots, in decreasing order.

First, the method as written does not close dimensionally. Eq. 3 still defines W^O as h*d_v x d_model, but Table I shows head dimensions D, D/2, ..., D/128. The concatenated width is about 1.99D, not 8*(D/8)=D, and the paper never says what W^O shape is or how the differently sized head tensors are concatenated. This is fixable by declaring W^O in R^{(sum_i d_i) x d_model}, but as printed the layer is underspecified.

Second, the headline comparison is confounded. If you implement the pyramid as described, the QKV projections and the output projection contain roughly twice the parameters of the baseline Transformer [12]. The paper's own tables show color 81.42 vs 76.50 and optical flow 85.30 vs 72.00, but no single-scale or equal-capacity control is run, so those gains could just be capacity, not multiscale attention per se. This is the strongest reason not to take the causal claim at face value.

Third, the SOTA claim is overstated. On NVGesture, NAS1+NAS2 reaches 88.38%, above the paper's 88.22%, and the color-only result (81.42%) is below NAS1 (83.61%). The conclusion's 'significantly improve... compared to some existing methods' is technically true only against the vanilla transformer and a few others.

There are also no seeds or error bars, so the point accuracies are single-run numbers. That is minor for a letter but worth noting.

Who should read this: someone tracking transformer variants for video, or a reviewer wanting a worked example of the capacity-confound critique. It deserves a serious referee—the dimensional issue and the confound are checkable, and the idea is coherent enough to warrant a revision. I would not cite it yet, but I would expect a competent referee to send it back for more work before accepting.

Recommended call: engage, but require a corrected equation, a capacity-matched baseline, and at least a couple of seeds before the multiscale claim can stand.

Referee Report

4 major / 6 minor

Summary. The paper proposes MsMHA-VTN, a video transformer for dynamic hand gesture recognition in which the per-head query/key/value dimensions are progressively halved across heads, forming a pyramid of multiscale attention. Frame-level features from a ResNet-18 backbone are processed through six transformer stages, and late fusion combines up to five input modalities (RGB, depth, infrared, surface normals, optical flow). The authors report an overall accuracy of 88.22% on NVGesture and 99.10% on Briareo, and compare with prior methods including the baseline Transformer of Eusanio et al.

Significance. The proposed multiscale-head idea is a reasonable and potentially useful modification of the standard transformer for gesture recognition, and the evaluation against two public benchmarks is a strength. If the architectural details are clarified and the gains are shown to be due to the multiscale mechanism rather than increased parameter count, the work would be a solid contribution. However, the current manuscript does not fully define the proposed layer, and the reported results are not statistically substantiated. The paper's own comparisons also do not support the claimed state-of-the-art status on NVGesture. With these fixes, the contribution could be of interest to the gesture recognition community.

major comments (4)

[Section II-B, Eqs. (3)-(5), Table I] The proposed MsMHA layer is not fully defined. Eq. (3) reuses the standard W^O ∈ R^{h·d_v × d_model}, but the concatenated output of the eight heads has dimension L × (D + D/2 + ... + D/128) ≈ L × 1.99D, not L × h·d_v. The manuscript never specifies the shape of W^O for MsMHA, nor how the query, key, and value tensors of differing widths are aligned in the attention computation. Without this specification, the architecture described in Section II-B cannot be reproduced, and the reported results cannot be clearly attributed to the described mechanism.
[Section III-B, Tables II-IV] The accuracy improvements over the baseline Transformer [12] are confounded with a large increase in attention-layer parameters. Because the multiscale heads have widths D, D/2, ..., D/128, the total QKV projection width and the output projection width are approximately twice those of the standard MHA baseline. No equal-capacity or single-scale control is reported, so the claim that multiscale attention per se improves accuracy is not supported. The authors should include a baseline with the same total attention width but uniform head sizes, or otherwise control for capacity.
[Abstract, Section III-B, Tables III-IV] The claim of state-of-the-art performance is contradicted by the paper's own results. On NVGesture color-only input, NAS1 achieves 83.61% (Table III) versus 81.42% for MsMHA-VTN; on multi-modal NVGesture, NAS1+NAS2 achieves 88.38% (Table IV) versus 88.22% for MsMHA-VTN. The state-of-the-art claim should be removed or restricted to specific modalities and datasets (e.g., surface normals on NVGesture and the overall accuracy on Briareo).
[Conclusion, Section III-B] The statement that MsMHA-VTN 'significantly improves' performance is not supported by any statistical evidence. All reported accuracies are single-run point estimates without standard deviations, confidence intervals, or significance tests. The authors should either provide repeated-run statistics or replace 'significantly' with a more modest description of the observed differences.

minor comments (6)

[Table IV] The row for the five-modality MsMHA-VTN result is typeset incorrectly: it reads 'color + depth + normals + 88.22%ir + flow' instead of listing the modalities and the accuracy separately.
[Figure 2 caption] The caption contains a typo: 'Mutiscaled' should be 'Multiscaled'.
[Section II-B, Eq. (5)] The notation Q_{jM/2}, K_{jM/2}, V_{jM/2} is unclear; it should be written with explicit superscripts or subscripts (e.g., Q_j^{M/2}) and the meaning of M as 'dimension of the previous head tensor' should be defined more precisely.
[Section II-B] The text states that the dimensions of the key and query vectors vary, but Table I lists the 'Attention Tensor size' for each head without specifying whether the value vectors also follow the same pyramid. Please clarify whether all three of Q, K, and V are scaled, or only Q and K.
[Section III-A] The description of how the ResNet-18 backbone features are extracted (e.g., output dimension, temporal pooling, positional encoding added to the transformer input) is omitted; please provide these details or refer to the exact configuration in [12].
[Section II-B] The paper does not report the number of parameters or FLOPs of the proposed model compared with the baseline, which would help assess the capacity confound raised in Major Comment 2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported accuracies are empirical results on external benchmarks, and the multiscale-head design is specified a priori rather than fitted to the test labels.

full rationale

The paper does not exhibit a circular derivation chain. The central claim is that the proposed MsMHA-VTN architecture, with per-head attention dimensions decreasing as D, D/2, ..., D/128 as specified in Table I, achieves 88.22% on NVGesture and 99.10% on Briareo. These numbers are produced by training on standard public benchmarks and comparing against published methods, including the Transformer baseline [12]; they are not derived from the model's assumptions or from fitted parameters renamed as predictions. No parameter is fitted to the test data, and no equation in the paper defines the target accuracy in terms of the model's own inputs. There is no load-bearing self-citation: the prior transformer works are cited as external baselines or architectural predecessors, not as an authority that forces the proposed design. The dimensional inconsistency in Eq. (3) regarding the concatenation of heads with different widths is a correctness risk, not a circularity, because the empirical evaluation would still be meaningful if the implementation resolves that bookkeeping. Similarly, the fact that NAS1+NAS2 [14] reports 88.38% versus the proposed 88.22% on NVGesture in Table IV concerns whether the state-of-the-art claim is accurate, not whether the derivation is circular. The absence of an equal-capacity baseline is a confounding-control issue in attributing gains to multiscale attention, but it does not make the reported benchmark numbers equivalent to the model's inputs by construction. Given the external, data-driven evaluation, the appropriate finding is no significant circularity.

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

The proposed method introduces no new physical entities. It relies on standard attention mechanisms, ImageNet pretrained features, late fusion by score summation, and prior dataset protocols. The main uncounted degrees of freedom are the hand-picked architecture constants: head scaling factor 1/2, eight heads, six stages, 40 frames, and a learning rate schedule copied from [12]. The ad hoc premise that differently sized heads can be concatenated into one output projection is the most fragile assumption.

free parameters (5)

per-head dimension scaling factor = 1/2 per head
Introduced in Section II-B and Table I to define head sizes D, D/2, ..., D/128; no ablation or search is reported, so it is a hand-chosen constant.
number of attention heads = 8
Set in Section II-A following the standard transformer; no sensitivity analysis on gesture data is provided.
number of transformer stages = 6
The architecture repeats the MsMHA block for six stages (T1-T6 in Fig. 1); no ablation justifies this depth.
number of input frames = 40
Section III-A states 40 frames per sequence as input; frame-count sensitivity is not analyzed.
learning rate schedule = 1e-4, decay at epochs 50 and 75
Copied from baseline [12]; no tuning or comparison is reported.

assumptions (5)

standard math Softmax scaled dot-product attention (Eq. 1) is a valid and differentiable module.
Background assumption from [5], used without proof.
domain assumption ImageNet-pretrained ResNet-18 features are a good frame-level representation for hand gestures.
Section II-C adopts the backbone and initialization of [12] and does not test alternative backbones.
domain assumption Summing classification scores across modalities (Eq. 6) is an effective fusion rule.
Late fusion is the only fusion strategy tried; no comparison to feature-level fusion is provided.
domain assumption Public dataset splits and evaluation protocols from [4] and [29] are unchanged and directly comparable.
The paper does not describe preprocessing or split details; comparisons rely on numbers reported by prior works.
ad hoc to paper Heads with decreasing key/query dimensions can be concatenated through one output projection W^O.
This is the defining property of MsMHA, but Eq. 3 and Table I do not specify how a variable-dimension concat is mapped back to d_model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Multiscaled Multi-Head Attention-based Video Transformer Network for Hand Gesture Recognition." pith.science (2026). https://pith.science/paper/ZLZ3WJOG

@misc{pith2026250100935,
  author       = {Pith},
  title        = {Pith review of: Multiscaled Multi-Head Attention-based Video Transformer Network for Hand Gesture Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZLZ3WJOG}},
  note         = {Machine review of arXiv:2501.00935}
}

read the original abstract

Dynamic gesture recognition is one of the challenging research areas due to variations in pose, size, and shape of the signer's hand. In this letter, Multiscaled Multi-Head Attention Video Transformer Network (MsMHA-VTN) for dynamic hand gesture recognition is proposed. A pyramidal hierarchy of multiscale features is extracted using the transformer multiscaled head attention model. The proposed model employs different attention dimensions for each head of the transformer which enables it to provide attention at the multiscale level. Further, in addition to single modality, recognition performance using multiple modalities is examined. Extensive experiments demonstrate the superior performance of the proposed MsMHA-VTN with an overall accuracy of 88.22\% and 99.10\% on NVGesture and Briareo datasets, respectively.

Figures

Figures reproduced from arXiv: 2501.00935 by the authors.

**Figure 2.** The proposed Mutiscaled Multi-Head pyramid attention. Pyramid [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 25 canonical work pages

[12]

A Transformer-based network for dynamic hand gesture recognition,

A. D. Eusanio, A. Simoni, S. Pini, G. Borghi, R. Vezzani, and R. Cucchiara, “A Transformer-based network for dynamic hand gesture recognition,” in Proc. International Conference on 3D Vision (3DV) , IEEE, 2020, pp. 623-632

work page 2020
[1]

Exploiting recurrent neural networks and leap motion controller for the recognition of sign language and semaphoric hand gestures,

D. Avola, M. Bernardi, L. Cinque, G. L. Foresti, and C. Massaroni, “Exploiting recurrent neural networks and leap motion controller for the recognition of sign language and semaphoric hand gestures,” IEEE Transactions on Multimedia , 21, no. 1, 2018, pp. 234-245

work page 2018
[2]

Attention in convolutional LSTM for gesture recognition,

L. Zhang, G. Zhu, L. Mei, P. Shen, S. A. A. Shah, and M. Bennamoun, “Attention in convolutional LSTM for gesture recognition,” in Proc. 32nd International Conference on Neural Information Processing Sys- tems, 2018, pp. 1957-1966

work page 2018
[3]

Attention-based gated recurrent unit for gesture recognition,

Khodabandelou, G., P. G. Jung , Y . Amirat, and S. Mohammed , “Attention-based gated recurrent unit for gesture recognition,” IEEE Transactions on Automation Science and Engineering, 18, no. 2, 2020, pp. 495-507

work page 2020
[4]

Online detection and classification of dynamic hand gestures with recurrent 3D convolutional neural network,

P. Molchanov, X. Yang, S. Gupta, K. Kim, S. Tyree, and J. Kautz, “Online detection and classification of dynamic hand gestures with recurrent 3D convolutional neural network,” in Proc. IEEE conference on computer vision and pattern recognition , 2016, pp. 4207-4215

work page 2016
[5]

Attention is all you need,

A. Vaswani, N Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. Advances in neural information processing systems , 2017, pp. 5998- 6008

work page 2017
[6]

An image is worth 16 ×16 words: Transformers for image recognition at scale,

A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszko- reit, and N. Houlsby, “An image is worth 16 ×16 words: Transformers for image recognition at scale,” in Proc. International Conference on Learning Representations, 2021

work page 2021
[7]

Training data-efficient image transformers & distillation through attention,

H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablayrolles, and H. J’egou, “Training data-efficient image transformers & distillation through attention,” arXiv preprint, 2020, arXiv:2012.12877

arXiv 2020

Show all 30 references

[8]

Video transformer network,

D. Neimark, O. Bar, M. Zohar, and D. Asselmann, “Video transformer network,” arXiv preprint, 2021, arXiv:2102.00719

2021 arXiv
[9]

Crossvit: Cross attention multi- scale vision transformer for image classification,

Chen, Chun-Fu, Q. Fan, and R. Panda, “Crossvit: Cross attention multi- scale vision transformer for image classification,” arXiv preprint, 2021, arXiv:2103.14899

2021 arXiv
[10]

Multiscale vision transformers,

H. Fan, B. Xiong, K. Mangalam, Y . Li, Z. Yan, J. Malik, and C. Feichtenhofer, “Multiscale vision transformers,” arXiv preprint arXiv:2104.11227, Apr 2021

2021 arXiv
[11]

Improved multiscale vision transformers for classification and detection

Y . Li , C. Y . Wu, H. Fan, K. Mangalam , B. Xiong, J. Malik, C. Fe- ichtenhofer, “Improved multiscale vision transformers for classification and detection”, arXiv preprint arXiv:2112.01526, Dec 2021

2021 arXiv
[13]

Incorporating relative position information in transformer-based sign language recognition and translation,

N. Aloysius, M. Geetha, and P. Nedungadi, “Incorporating relative position information in transformer-based sign language recognition and translation,” IEEE Access , 9, 2021, pp. 145929-145942

2021
[14]

Searching multi-rate and multi-modal temporal enhanced networks for gesture recognition,

Z. Yu, B. Zhou, J. Wan, P. Wang, H. Chen, X. Liu, S. Z. Li, and G. Zhao, “Searching multi-rate and multi-modal temporal enhanced networks for gesture recognition,” IEEE Transactions on Image Processing , 2021

2021
[15]

TMMF: Temporal Multi-Modal Fusion for single-stage continuous gesture recog- nition,

H. Gammulle, S. Denman, S. Sridharan, and C. Fookes, “TMMF: Temporal Multi-Modal Fusion for single-stage continuous gesture recog- nition,” IEEE Transactions on Image Processing, 30, 2021, pp. 7689- 7701

2021
[16]

Imagenet: A large-scale hierarchical image database

Deng, J., Dong, W., Socher, R., Li, L.J., Li, K. and Fei-Fei, L. “Imagenet: A large-scale hierarchical image database.” in IEEE conference on computer vision and pattern recognition , pp. 248-255, 2009

2009
[17]

Two-stream convolutional networks for action recognition in videos,

K. Simonyan, and A. Zisserman, “Two-stream convolutional networks for action recognition in videos,” arXiv preprint arXiv:1406.2199 , 2014

2014 arXiv
[18]

A robust and efficient video representation for action recognition,

H. Wang, D. Oneata, J. Verbeek, and C. Schmid, “A robust and efficient video representation for action recognition,” International journal of computer vision , 119(3), pp. 219-238, 2016

2016
[19]

Res3atn-deep 3D residual attention network for hand gesture recognition in videos,

N. Dhingra, and A. Kunz, “Res3atn-deep 3D residual attention network for hand gesture recognition in videos,” in Proc. IEEE International Conference on 3D Vision (3DV) , 2019 pp. 491-501

2019
[20]

Learning spatiotemporal features with 3d convolutional networks,

D. Tran, L. Bourdev, R. Fergus, L. Torresani, and M. Paluri, “Learning spatiotemporal features with 3d convolutional networks,” in Proc. IEEE international conference on computer vision , 2015, pp. 4489-4497

2015
[21]

Multi-task and multi-modal learning for RGB dynamic gesture recognition,

D. Fan, H. Lu, S. Xu, and S. Cao, “Multi-task and multi-modal learning for RGB dynamic gesture recognition,” IEEE Sensors Journal , 21, no. 23, 2021, pp. 27026-27036

2021
[22]

Making convolutional networks recurrent for visual sequence learning,

X. Yang, P. Molchanov, and J. Kautz, “Making convolutional networks recurrent for visual sequence learning,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition , Jun. 2018, pp. 6469–6478

2018
[23]

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

J. Carreira, and A. Zisserman, “Quo vadis, action recognition? A new model and the kinetics dataset,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 6299-6308

2017
[24]

Real-time hand ges- ture detection and classification using convolutional neural networks,

O. K ¨op¨ukl¨u, A. Gunduz, N. Kose, and G. Rigoll, “Real-time hand ges- ture detection and classification using convolutional neural networks,” in Proc. IEEE International Conference on Automatic Face & Gesture Recognition, 2019, IEEE, pp. 1-8

2019
[25]

Improving the perfor- mance of unimodal dynamic hand-gesture recognition with multimodal training,

M. Abavisani, H. R. V . Joze, and V . M. Patel, “Improving the perfor- mance of unimodal dynamic hand-gesture recognition with multimodal training,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2019, pp. 1165–1174

2019
[26]

Super normal vector for activity recognition using depth sequences,

X. Yang and Y . Tian., “Super normal vector for activity recognition using depth sequences,” in P . IEEE conference on computer vision and pattern recognition (CVPR) , 2014, pp. 804–811

2014
[27]

Motion fused frames: Data level fusion strategy for hand gesture recognition,

O. Kopuklu, N. Kose, and G. Rigoll., “Motion fused frames: Data level fusion strategy for hand gesture recognition,” in Proc. IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , 2018, pp. 2103–2111

2018
[28]

Dynamic hand gesture recognition based on short-term sampling neural networks,

W. Zhang, J. Wang, and F. Lan, “Dynamic hand gesture recognition based on short-term sampling neural networks,” EEE/CAA Journal of Automatica Sinica , 8, no. 1 (2020): 110-120

2020
[29]

Hand gestures for the human-car interaction: The Briareo dataset,

F. Manganaro, S. Pini, G. Borghi, R. Vezzani, and R. Cucchiara, “Hand gestures for the human-car interaction: The Briareo dataset,” in Proc. International Conference on Image Analysis and Processing (ICIAP) , Springer, 2019, pp. 560–571

2019
[30]

Multimodal hand gesture classification for the human–car interaction,

A. D. Eusanio, A. Simoni, S. Pini, G. Borghi, R. Vezzani, R. Cucchiara, “Multimodal hand gesture classification for the human–car interaction,” Informatics, 2020, 7, 31. https://doi.org/10.3390/informatics7030031

2020 doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [12]

A Transformer-based network for dynamic hand gesture recognition,

A. D. Eusanio, A. Simoni, S. Pini, G. Borghi, R. Vezzani, and R. Cucchiara, “A Transformer-based network for dynamic hand gesture recognition,” in Proc. International Conference on 3D Vision (3DV) , IEEE, 2020, pp. 623-632

work page 2020

[2] [1]

Exploiting recurrent neural networks and leap motion controller for the recognition of sign language and semaphoric hand gestures,

D. Avola, M. Bernardi, L. Cinque, G. L. Foresti, and C. Massaroni, “Exploiting recurrent neural networks and leap motion controller for the recognition of sign language and semaphoric hand gestures,” IEEE Transactions on Multimedia , 21, no. 1, 2018, pp. 234-245

work page 2018

[3] [2]

Attention in convolutional LSTM for gesture recognition,

L. Zhang, G. Zhu, L. Mei, P. Shen, S. A. A. Shah, and M. Bennamoun, “Attention in convolutional LSTM for gesture recognition,” in Proc. 32nd International Conference on Neural Information Processing Sys- tems, 2018, pp. 1957-1966

work page 2018

[4] [3]

Attention-based gated recurrent unit for gesture recognition,

Khodabandelou, G., P. G. Jung , Y . Amirat, and S. Mohammed , “Attention-based gated recurrent unit for gesture recognition,” IEEE Transactions on Automation Science and Engineering, 18, no. 2, 2020, pp. 495-507

work page 2020

[5] [4]

Online detection and classification of dynamic hand gestures with recurrent 3D convolutional neural network,

P. Molchanov, X. Yang, S. Gupta, K. Kim, S. Tyree, and J. Kautz, “Online detection and classification of dynamic hand gestures with recurrent 3D convolutional neural network,” in Proc. IEEE conference on computer vision and pattern recognition , 2016, pp. 4207-4215

work page 2016

[6] [5]

Attention is all you need,

A. Vaswani, N Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. Advances in neural information processing systems , 2017, pp. 5998- 6008

work page 2017

[7] [6]

An image is worth 16 ×16 words: Transformers for image recognition at scale,

A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszko- reit, and N. Houlsby, “An image is worth 16 ×16 words: Transformers for image recognition at scale,” in Proc. International Conference on Learning Representations, 2021

work page 2021

[8] [7]

Training data-efficient image transformers & distillation through attention,

H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablayrolles, and H. J’egou, “Training data-efficient image transformers & distillation through attention,” arXiv preprint, 2020, arXiv:2012.12877

arXiv 2020