Pith. sign in

REVIEW 4 major objections 5 minor 53 references

DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read DSTSA-GCN claims that per-channel and per-frame topology graphs, plus multiscale temporal convolution, push skeleton gesture recognition to 97.74% on SHREC'17 and 95.04% on DHG.

desk verdict A competent incremental GCN paper whose parallel channel-wise/temporal-wise topology design is genuinely new, but whose SHREC'17 SOTA claim rests on test-set tuning and a margin of a fraction of one test sequence. read the letter →

arxiv 2501.12086 v1 pith:ZET2A7CK submitted 2025-01-21 cs.CV

Hu Cui , Renjing Huang , Ruoyu Zhang , Tessai Hayama This is my paper · ORCID
classification cs.CV
keywords skeleton-basedgesturerecognitiongraphconvolutionalnetworksdynamictopologymodelingchannel-wiseconvolutiontemporal-wisemulti-scaletemporalhumanactionspatio-temporal
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

The paper is trying to establish that skeleton-based gesture recognition improves when graph convolution does not force one adjacency pattern across all channels and all frames. The proposed DSTSA-GCN builds separate dynamic topologies for channel groups and for temporal positions, and it pairs these with a multi-scale temporal convolution. The authors report that this design reaches top accuracies on SHREC'17 Track (97.74% for 14 gestures, 95.37% for 28 gestures) and DHG-14/28 (95.04% and 93.57%), and competitive results on NTU RGB+D and NTU RGB+D 120. A careful reader would care because the grouped construction keeps the parameter count roughly independent of the number of groups, so the added expressiveness does not come from a large increase in model size.

What carries the argument

The load-bearing object is the pair of grouped topology-modeling operations, GC-GC and GT-GC, together with MS-TCN. For a channel group, the adjacency matrix is $\hat{A}_c = \alpha A_c^d(X) + A_g$, where $A_c^d(X)$ is a sample-dependent distance matrix computed from transformed joint features, $\alpha$ is a trainable scalar, and $A_g$ is a static graph shared within the group. GT-GC applies the same formula after pooling over channels, producing frame-specific graphs. The grouping trick keeps the number of static graphs equal to the number of groups $K$ rather than the number of channels, so parameter count stays nearly flat while each channel still gets a distinct dynamic topology. MS-TCN contributes multiscale temporal receptive fields through parallel dilated temporal convolutions.

What would settle it

Retrain DSTSA-GCN on SHREC'17 with all design choices (K, theta, STCA variant, MS-TCN branch set, static initializer) selected by cross-validation on the training split alone, then evaluate once on the fixed test split; if accuracy falls materially below 97.74% and 95.37%, the reported gain is at least partly an artifact of test-set selection.

Watch

Extended reading notes

Core claim

The central claim is that spatial and temporal topology should both be dynamic and non-shared, and that grouping lets the model capture multiscale joint interactions without inflating parameters. In the Group Channel-wise Graph Convolution, each channel receives its own adjacency matrix built from a learnable static graph plus a sample-dependent graph computed from distances between transformed joint features; the Group Temporal-wise Graph Convolution applies the same construction along the temporal axis, so different frames can use different interaction graphs. A Spatio-Temporal Coordinate-Aware module injects position information during feature transformation, and the Multi-Scale Temporal Convolution replaces the single temporal kernel with parallel branches at dilations {1,2,3,4} plus a max-pooling branch and a shortcut. Ablations on SHREC'17 attribute the largest single gain to the channel-wise module, and fused four-modality results on gesture and action benchmarks exceed the results of CTR-GCN and TD-GCN in most reported settings.

Load-bearing premise

The load-bearing premise is that the SHREC'17 test split was not used to select the number of groups, the activation function, the STCA design, the MS-TCN branches, and the static-topology initialization; if those choices were tuned on the test set, the reported 97.74% and 95.37% accuracies are optimistic.

Editorial extensions

If this is right

  • On SHREC'17 Track, fusing joint, bone, joint-motion, and bone-motion modalities gives 97.74% for 14 gestures and 95.37% for 28 gestures, above the closest prior GCN results reported in Table 7.
  • On DHG-14/28, the same fusion reaches 95.04% and 93.57%, improvements of 1.94 and 3.07 points over CTR-GCN.
  • On NTU RGB+D, the model reaches 92.78% (Cross-Subject) and 97.03% (Cross-View); on NTU RGB+D 120 it reaches 89.12% (Cross-Subject) and 90.97% (Cross-Setup), matching or exceeding several recent baselines.
  • The grouped static graphs and MS-TCN lower parameter and FLOP counts relative to a standard TCN while improving accuracy, so the gain is not a parameter-count effect.
  • Ablations show each component contributes: removing GC-GC costs the most accuracy, and the combined STCA module outperforms separate spatial-only or temporal-only coordinate-aware modules.

Reading between the lines

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

  • The paper does not test whether its hyperparameters were chosen on a validation split rather than the SHREC'17 test split; if a validation-only retraining reproduces the gains, the same grouped topology idea likely transfers to other graph time-series where connectivity changes over time, such as human-object interaction or multi-agent motion.
  • The frame-specific topology learned by GT-GC could serve as an interpretability tool: class-activation maps and temporal topology visualizations suggest that particular frames carry distinct joint interactions, and one could quantify how much each frame's graph contributes to the prediction.
  • Because the static graph is randomly initialized and learns dataset-level topology, the grouping strategy may be a general remedy for the local-bias problem of deep GCNs beyond skeleton data, applicable whenever a fixed adjacency matrix overly weights near neighbors.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript presents DSTSA-GCN, a graph convolutional network for skeleton-based gesture and action recognition. The architecture combines a grouped channel-wise graph convolution (GC-GC) and a grouped temporal-wise graph convolution (GT-GC) with a multi-scale temporal convolution (MS-TCN), and introduces a Spatio-Temporal Coordinate-Aware (STCA) feature transformation. The authors report fused accuracies of 97.74% and 95.37% on SHREC'17 Track, 95.04% and 93.57% on DHG-14/28, 92.78% and 97.03% on NTU-60, and 89.12% and 90.97% on NTU-120, and claim state-of-the-art performance on these benchmarks.

Significance. The proposed modules are clearly motivated, and the ablation study (Tables 1-5) is thorough in exploring each design choice. The per-subject DHG results in Table 6 and the parameter/FLOP reporting are useful. If the reported accuracies are unbiased, the method is competitive with recent GCN baselines at similar parameter counts. However, the headline SHREC'17 result is compromised by test-set-based model selection, and the state-of-the-art claim is not supported by the authors' own tables for several benchmarks. The method's technical contribution is plausible, but the empirical evidence needs strengthening.

major comments (4)
  1. [4.3, Tables 2-5] The SHREC'17 test split is used to select the number of groups K, the activation θ, the STCA construction, the MS-TCN branch configuration, and the static-topology initialization (Tables 2-5), and the same split is then used to report the final accuracies in Table 7 (97.74% and 95.37%). Since no validation split is described in Section 4.2, these hyperparameters are effectively fitted to the test set, making the SHREC'17 comparison with baselines that did not tune on this split optimistic. The paper should either introduce a validation-based selection procedure or explicitly quantify the expected optimism, for example by reporting results for a fixed default configuration.
  2. [Abstract; Section 4.4, Tables 7-8] The abstract states that DSTSA-GCN achieves state-of-the-art performance on SHREC'17, DHG-14/28, NTU-RGB+D, and NTU-RGB+D-120, but the paper's own tables contradict this: Table 8 shows that on NTU-60 X-Sub DSTSA-GCN (92.78%) is below BlockGCN and LG-SGNet (93.1%), and on NTU-120 X-Sub and X-Setup the method is below BlockGCN (89.12 vs 90.3 and 90.97 vs 91.5), while Table 7 shows the method is second-best on SHREC'17 28-gesture (95.37 vs 95.84 for STDA-GCN). The performance claims should be revised to match the reported numbers.
  3. [3.3, Eq. 22] Section 3.3 states that 'the number of parameters in our approach is independent of the number of groups,' but the static topology tensor A ∈ R^{K×V×V} in Eq. 22 introduces K·V^2 parameters, which scales linearly with K. This is also inconsistent with Table 2, where the parameter counts change with K (1.96M, 1.97M, 1.99M, 1.97M). The statement should be corrected or clarified to specify which parameter subset is group-independent.
  4. [4.2; Tables 7-8] No error bars or multiple-seed results are reported for any dataset. The SHREC'17 and NTU results are single-split evaluations, and the DHG result is a mean over 20 leave-one-subject-out folds without a standard deviation. Since many of the reported improvements over baselines are between 0.1 and 0.5 percentage points, the statistical significance of the claimed gains is not established. At minimum, the authors should report standard deviations or multiple runs.
minor comments (5)
  1. [Eq. 20] Equation 20 is missing the multiplication sign between the feature and adjacency matrices; as typeset it is ambiguous.
  2. [Eq. 14] Equation 14 (TGP) does not specify over which dimension the softmax is taken; the summation over t and the softmax over a scalar expression need clarification.
  3. [2.2] Section 2.2 repeats the citations [10, 11, 12, 13, 14] twice in one sentence; the duplicate citation list should be removed.
  4. [Abstract] The code link in the abstract is given as a URL without a release date; please confirm the repository is public at the time of publication.
  5. [Table 7] In Table 7, single-modality DSTSA-GCN results are reported for SHREC'17 but not for DHG-14/28; including them would help isolate the contribution of the fusion.

Circularity Check

1 steps flagged · score 5.0 of 10

SHREC'17 state-of-the-art numbers are partly fitted: the same SHREC'17 test split is used to select K, theta, STCA, MS-TCN branches, and static-topology initialization, then reported as the headline result.

  1. fitted input called prediction [Section 4.1 (dataset split), Section 4.3 Tables 2-5 (config selection), Section 4.4 Table 7 (headline SHREC'17 result)]
    "We first performed an exhaustive ablation study on the SHREC17 dataset to validate the effectiveness of the proposed model components. ... The dataset is divided into 1960 sequences for training and 840 sequences for testing, following the evaluation protocol in[41],[44] and [45]."

    Tables 2-5 select the architecture by maximizing SHREC'17 test accuracy: K and the activation in Table 2, the STCA construction in Table 3, the MS-TCN branch set H in Table 4, and the static-topology initialization strategy in Table 5. The same 840-sequence test split is then reused in Table 7 to report DSTSA-GCN's 97.74%/95.37% SHREC'17 accuracies. The reported quantity is therefore the selection criterion itself, not an independent prediction; comparisons to baselines that did not tune on this split are biased in favor of the selected configuration. This is the fitted-input-called-prediction pattern: the headline SHREC'17 claim partially reduces to the configuration search.

full rationale

This is an empirical architecture paper, so there is no theoretical derivation chain in which Eq. X reduces to Eq. Y. The central circular element is empirical: the SHREC'17 test split is used both as the model-selection criterion (K, activation, STCA layout, MS-TCN branches, static-topology initialization) and as the reported evaluation, making the headline SHREC'17 numbers optimistic by construction. The paper describes a fixed 1960/840 split and no held-out validation set, so the Tables 2-5 choices are explicitly test-set choices. This warrants a partial-circularity score near the middle of the scale. The score is not higher because the DHG-14/28 and NTU-60/120 results use separate splits/protocols and provide partially independent evidence for the architecture, and because no load-bearing self-citation chain or uniqueness argument is used; the self-citations [39,40] are ordinary related-work references. The strongest SOTA claim, however, rests on the SHREC'17 numbers that were selected on the same test split.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The model's accuracy rests almost entirely on learned parameters and hand-selected hyperparameters; there is no closed-form derivation. The only structural priors are standard graph-convolution assumptions and the paper's own modeling choices. No new physical entities are introduced.

free parameters (7)
  • Static adjacency matrices A (K groups per layer) = K=8, random initialization
    Learned through backprop and fixed at inference (Eqs. 16-18); Table 5 compares four initializations on the SHREC'17 test set, then random is chosen for all experiments.
  • Dynamic topology transforms Phi1, Phi2, Phi3 and activation theta = theta=tanh, K=8
    Eqs. 13-15 and 25-27 define sample-dependent adjacency as a learned function of input; the activation and group count are selected by SHREC'17 test accuracy in Table 2.
  • STCA attention parameters and reduction ratio r = learned, r not specified numerically
    Eqs. 8-12 define the STCA module; the construction, whether parallel, series, or coupled, is selected in Table 3 on SHREC'17.
  • Fusion scalars a and b in Eq. 5 = learned
    Trained online to weight GC-GC and GT-GC outputs; no analysis of their converged values is given.
  • MS-TCN branch configuration = H: M,S,g1,g2,g3,g4
    Table 4 ablates branch sets on SHREC'17 and selects configuration H for all final models.
  • Training hyperparameters = C=64, frames 150 or 64, batch 64 or 32, SGD lr=0.1, wd=0.0004, momentum 0.9, 170 epochs, warmup 20
    Section 4.2 lists these choices; they are customary for the domain but not derived from theory.
  • Number of groups K = 8
    Table 2: K=8 gives 96.67% and 94.17%, the best 28gesture result and tied best 14gesture result.
assumptions (6)
  • standard math Graph convolution as defined by Kipf and Welling is a valid feature aggregation operator for skeleton graphs.
    Section 3.1, Eq. 1-2, uses normalized adjacency convolution without further justification.
  • domain assumption The skeleton's natural joint connections are an appropriate prior graph for gesture and action recognition.
    Section 2.2 and Fig. 1 inherit this from ST-GCN; no evidence beyond benchmark performance.
  • ad hoc to paper The difference between learned feature vectors, theta(Phi1(x_i) - Phi2(x_j)), is a valid measure of topological relationship between joints or frames.
    Eqs. 15 and 27 define topology as nonlinearly transformed pairwise distance; this is a modeling choice without derivation.
  • ad hoc to paper Random-initialized static topologies avoid local bias and are preferable to distance or partition initializations.
    Table 5 shows distance initialization gives 96.81% and 93.82% while random gives 96.67% and 94.17%; the paper selects random for all remaining experiments.
  • ad hoc to paper Learning curves on the SHREC'17 test split are a valid way to choose model hyperparameters and then report final accuracy on the same split.
    Tables 2-5 use SHREC'17 test accuracy to select K, theta, STCA form, MS-TCN branches, and initialization; Table 7 then reports SHREC'17 as an evaluation result.
  • domain assumption Learned topologies trained on one subject set generalize to unseen subjects and camera setups.
    All benchmark evaluations assume the train and test distribution shift is captured by the provided protocol.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling." pith.science (2026). https://pith.science/paper/ZET2A7CK

@misc{pith2026250112086,
  author       = {Pith},
  title        = {Pith review of: DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZET2A7CK}},
  note         = {Machine review of arXiv:2501.12086}
}
read the original abstract

Graph convolutional networks (GCNs) have emerged as a powerful tool for skeleton-based action and gesture recognition, thanks to their ability to model spatial and temporal dependencies in skeleton data. However, existing GCN-based methods face critical limitations: (1) they lack effective spatio-temporal topology modeling that captures dynamic variations in skeletal motion, and (2) they struggle to model multiscale structural relationships beyond local joint connectivity. To address these issues, we propose a novel framework called Dynamic Spatial-Temporal Semantic Awareness Graph Convolutional Network (DSTSA-GCN). DSTSA-GCN introduces three key modules: Group Channel-wise Graph Convolution (GC-GC), Group Temporal-wise Graph Convolution (GT-GC), and Multi-Scale Temporal Convolution (MS-TCN). GC-GC and GT-GC operate in parallel to independently model channel-specific and frame-specific correlations, enabling robust topology learning that accounts for temporal variations. Additionally, both modules employ a grouping strategy to adaptively capture multiscale structural relationships. Complementing this, MS-TCN enhances temporal modeling through group-wise temporal convolutions with diverse receptive fields. Extensive experiments demonstrate that DSTSA-GCN significantly improves the topology modeling capabilities of GCNs, achieving state-of-the-art performance on benchmark datasets for gesture and action recognition, including SHREC17 Track, DHG-14\/28, NTU-RGB+D, and NTU-RGB+D-120.

Figures

Figures reproduced from arXiv: 2501.12086 by the authors.

Figure 1
Figure 1. Illustration of the construction of different skeletal topologies. Solid arrows indicate shared topology in the channel or temporal dimension, while dashed lines indicate non-shared. Yellow arrows indicate the direction of the channel and green arrows indicate the direction of the temporal dimension. c and d is based on the dimension-specific of b. a is parameterized in training stage and does not vary in the infere… view at source ↗
Figure 2
Figure 2. The architecture of Dynamic Spatial-Temporal Semantic Awareness Graph Convolutional Network. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualization of the static topology graphs learned from di [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (4 more)
Figure 1
Figure 1. Figure 1: In [PITH_FULL_IMAGE:figures/full_fig_p011_1.png]
Figure 4
Figure 4. Figure 4: Visualization comparison of temporal-wise topology graphs (last layer) with CTR-GCN and TD-GCN. Gesture class : Grap. Blue boxes [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Class activation mapping results for action sample: Grap, [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Visualization of the CAM results projected onto the skeletal gesture sequence (Tap). The yellow joints indicate the 10 most weighted [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 45 canonical work pages

  1. [1]

    X. Shen, S. Yuan, H. Sheng, H. Du, X. Yu, Auslan-daily: Aus- tralian sign language translation for daily communication and news, Advances in Neural Information Processing Systems 36 (2024)

    work page 2024

  2. [2]

    A. S. M. Miah, M. A. M. Hasan, Y . Tomioka, J. Shin, Hand gesture recognition for multi-culture sign language using graph and general deep learning network, IEEE Open Journal of the Computer Society (2024)

    work page 2024

  3. [3]

    Desai, L

    A. Desai, L. Berger, F. Minakov, N. Milano, C. Singh, K. Pumphrey, R. Ladner, H. Daum ´e III, A. X. Lu, N. Caselli, et al., Asl citizen: a community-sourced dataset for advancing isolated sign language recognition, Advances in Neural Infor- mation Processing Systems 36 (2024)

    work page 2024

  4. [4]

    J. Qi, L. Ma, Z. Cui, Y . Yu, Computer vision-based hand gesture recognition for human-robot interaction: a review, Complex & Intelligent Systems 10 (1) (2024) 1581–1606

    work page 2024

  5. [5]

    Z. Yu, C. Lu, Y . Zhang, L. Jing, Gesture-controlled robotic arm for agricultural harvesting using a data glove with bending sen- sor and optitrack systems, Micromachines 15 (7) (2024) 918

    work page 2024

  6. [6]

    Di Qi, F

    T. Di Qi, F. L. Cibrian, M. Raswan, T. Kay, H. M. Camarillo- Abad, Y . Wen, Towards intuitive 3d interactions in virtual re- ality: A deep learning-based dual-hand gesture recognition ap- proach, IEEE Access (2024)

    work page 2024

  7. [7]

    M. Modaberi, The role of gesture-based interaction in improving user satisfaction for touchless interfaces, International Journal of Advanced Human Computer Interaction 2 (2) (2024) 20–32

    work page 2024

  8. [8]

    S. Yan, Y . Xiong, D. Lin, Spatial temporal graph convolutional networks for skeleton-based action recognition, in: Proceedings of the AAAI conference on artificial intelligence, V ol. 32, 2018

    work page 2018

Show all 53 references
  1. [9]

    T. N. Kipf, M. Welling, Semi-supervised classification with graph convolutional networks, arXiv preprint arXiv:1609.02907 (2016)

    2016 arXiv

  2. [10]

    L. Shi, Y . Zhang, J. Cheng, H. Lu, Two-stream adaptive graph convolutional networks for skeleton-based action recognition, in: Proceedings of the IEEE /CVF conference on computer vi- sion and pattern recognition, 2019, pp. 12026–12035

    2019

  3. [11]

    M. Li, S. Chen, X. Chen, Y . Zhang, Y . Wang, Q. Tian, Actional- structural graph convolutional networks for skeleton-based ac- tion recognition, in: Proceedings of the IEEE /CVF conference on computer vision and pattern recognition, 2019, pp. 3595– 3603

    2019

  4. [12]

    F. Ye, S. Pu, Q. Zhong, C. Li, D. Xie, H. Tang, Dynamic gcn: Context-enriched topology learning for skeleton-based action recognition, in: Proceedings of the 28th ACM international con- ference on multimedia, 2020, pp. 55–63

    2020

  5. [13]

    Zhang, C

    P. Zhang, C. Lan, W. Zeng, J. Xing, J. Xue, N. Zheng, Semantics-guided neural networks for e fficient skeleton-based human action recognition, in: proceedings of the IEEE /CVF conference on computer vision and pattern recognition, 2020, pp. 1112–1121

    2020

  6. [14]

    Cheng, Y

    K. Cheng, Y . Zhang, X. He, W. Chen, J. Cheng, H. Lu, Skeleton- based action recognition with shift graph convolutional network, in: Proceedings of the IEEE /CVF conference on computer vi- sion and pattern recognition, 2020, pp. 183–192

    2020

  7. [15]

    Cheng, Y

    K. Cheng, Y . Zhang, C. Cao, L. Shi, J. Cheng, H. Lu, De- coupling gcn with dropgraph module for skeleton-based action recognition, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIV 16, Springer, 2020, pp. 536–553

    2020

  8. [16]

    Y . Chen, Z. Zhang, C. Yuan, B. Li, Y . Deng, W. Hu, Channel- wise topology refinement graph convolution for skeleton-based action recognition, in: Proceedings of the IEEE /CVF interna- tional conference on computer vision, 2021, pp. 13359–13368

    2021

  9. [17]

    H. Duan, J. Wang, K. Chen, D. Lin, Dg-stgcn: Dynamic spatial- temporal modeling for skeleton-based action recognition, arXiv preprint arXiv:2210.05895 (2022)

    2022 arXiv

  10. [18]

    Veli ˇckovi´c, Everything is connected: Graph neural networks, Current Opinion in Structural Biology 79 (2023) 102538

    P. Veli ˇckovi´c, Everything is connected: Graph neural networks, Current Opinion in Structural Biology 79 (2023) 102538

    2023

  11. [19]

    K. Xu, W. Hu, J. Leskovec, S. Jegelka, How powerful are graph neural networks?, arXiv preprint arXiv:1810.00826 (2018)

    2018 arXiv

  12. [20]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, S. Y . Philip, A com- prehensive survey on graph neural networks, IEEE transactions on neural networks and learning systems 32 (1) (2020) 4–24

    2020

  13. [21]

    S. Zhu, S. Pan, C. Zhou, J. Wu, Y . Cao, B. Wang, Graph geom- etry interaction learning, Advances in Neural Information Pro- cessing Systems 33 (2020) 7548–7558

    2020

  14. [22]

    Y . Liu, K. Ding, Q. Lu, F. Li, L. Y . Zhang, S. Pan, Towards self- interpretable graph-level anomaly detection, Advances in Neu- ral Information Processing Systems 36 (2024)

    2024

  15. [23]

    Zheng, M

    X. Zheng, M. Zhang, C. Chen, Q. V . H. Nguyen, X. Zhu, S. Pan, Structure-free graph condensation: From large-scale graphs to condensed graph-free data, Advances in Neural Information Processing Systems 36 (2024)

    2024

  16. [24]

    Zheng, M

    X. Zheng, M. Zhang, C. Chen, S. Molaei, C. Zhou, S. Pan, Gn- nevaluator: Evaluating gnn performance on unseen graphs with- out labels, Advances in Neural Information Processing Systems 36 (2024)

    2024

  17. [25]

    X. Wang, M. Zhang, How powerful are spectral graph neu- ral networks, in: International conference on machine learning, PMLR, 2022, pp. 23341–23362

    2022

  18. [26]

    D. Bo, X. Wang, Y . Liu, Y . Fang, Y . Li, C. Shi, A survey on spec- tral graph neural networks, arXiv preprint arXiv:2302.05631 (2023)

    2023 arXiv

  19. [27]

    Danel, P

    T. Danel, P. Spurek, J. Tabor, M. ´Smieja, Ł. Struski, A. Słowik, Ł. Maziarka, Spatial graph convolutional networks, in: Interna- tional Conference on Neural Information Processing, Springer, 2020, pp. 668–675

    2020

  20. [28]

    H. Pei, B. Wei, K. C.-C. Chang, Y . Lei, B. Yang, Geom- gcn: Geometric graph convolutional networks, arXiv preprint arXiv:2002.05287 (2020)

    2020 arXiv

  21. [29]

    De Smedt, H

    Q. De Smedt, H. Wannous, J.-P. Vandeborre, Skeleton-based dy- namic hand gesture recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition work- shops, 2016, pp. 1–9

    2016

  22. [30]

    K. Lai, S. N. Yanushkevich, Cnn + rnn depth and skeleton based dynamic hand gesture recognition, in: 2018 24th inter- 14 national conference on pattern recognition (ICPR), IEEE, 2018, pp. 3451–3456

    2018

  23. [31]

    De Smedt, H

    Q. De Smedt, H. Wannous, J.-P. Vandeborre, Heterogeneous hand gesture recognition using 3d dynamic skeletal data, Com- puter Vision and Image Understanding 181 (2019) 60–72

    2019

  24. [32]

    Z. Chen, S. Li, B. Yang, Q. Li, H. Liu, Multi-scale spatial temporal graph convolutional network for skeleton-based action recognition, in: Proceedings of the AAAI conference on artifi- cial intelligence, V ol. 35, 2021, pp. 1113–1122

    2021

  25. [33]

    D. Feng, Z. Wu, J. Zhang, T. Ren, Multi-scale spatial tempo- ral graph neural network for skeleton-based action recognition, IEEE Access 9 (2021) 58256–58265

    2021

  26. [34]

    J. Liu, X. Wang, C. Wang, Y . Gao, M. Liu, Temporal decoupling graph convolutional network for skeleton-based gesture recog- nition, IEEE Transactions on Multimedia 26 (2023) 811–823

    2023

  27. [35]

    R. D. Hjelm, A. Fedorov, S. Lavoie-Marchildon, K. Grewal, P. Bachman, A. Trischler, Y . Bengio, Learning deep representa- tions by mutual information estimation and maximization, arXiv preprint arXiv:1808.06670 (2018)

    2018 arXiv

  28. [36]

    A. A. Alemi, I. Fischer, J. V . Dillon, K. Murphy, Deep varia- tional information bottleneck, arXiv preprint arXiv:1612.00410 (2016)

    2016 arXiv

  29. [37]

    H.-g. Chi, M. H. Ha, S. Chi, S. W. Lee, Q. Huang, K. Ramani, Infogcn: Representation learning for human skeleton-based ac- tion recognition, in: Proceedings of the IEEE /CVF conference on computer vision and pattern recognition, 2022, pp. 20186– 20196

    2022

  30. [38]

    Z. Liu, H. Zhang, Z. Chen, Z. Wang, W. Ouyang, Disentan- gling and unifying graph convolutions for skeleton-based action recognition, in: Proceedings of the IEEE /CVF conference on computer vision and pattern recognition, 2020, pp. 143–152

    2020

  31. [39]

    H. Cui, T. Hayama, Joint-partition group attention for skeleton- based action recognition, Signal Processing 224 (2024) 109592

    2024

  32. [40]

    H. Cui, T. Hayama, Stsd: spatial–temporal semantic decompo- sition transformer for skeleton-based action recognition, Multi- media Systems 30 (1) (2024) 43

    2024

  33. [41]

    De Smedt, H

    Q. De Smedt, H. Wannous, J.-P. Vandeborre, J. Guerry, B. L. Saux, D. Filliat, 3d hand gesture recognition using a depth and skeletal dataset: Shrec’17 track, in: Proceedings of the Work- shop on 3D Object Retrieval, 2017, pp. 33–38

    2017

  34. [42]

    Shahroudy, J

    A. Shahroudy, J. Liu, T.-T. Ng, G. Wang, Ntu rgb + d: A large scale dataset for 3d human activity analysis, in: Proceedings of the IEEE conference on computer vision and pattern recogni- tion, 2016, pp. 1010–1019

    2016

  35. [43]

    J. Liu, A. Shahroudy, M. Perez, G. Wang, L.-Y . Duan, A. C. Kot, Ntu rgb+ d 120: A large-scale benchmark for 3d human activity understanding, IEEE transactions on pattern analysis and ma- chine intelligence 42 (10) (2019) 2684–2701

    2019

  36. [44]

    J.-H. Song, K. Kong, S.-J. Kang, Dynamic hand gesture recog- nition using improved spatio-temporal graph convolutional net- work, IEEE Transactions on Circuits and Systems for Video Technology 32 (9) (2022) 6227–6239

    2022

  37. [45]

    L. Shi, Y . Zhang, J. Cheng, H. Lu, Decoupled spatial-temporal attention network for skeleton-based action-gesture recognition, in: Asian Conference on Computer Vision, 2020, pp. 38–53

    2020

  38. [46]

    Y . Li, Z. He, X. Ye, Z. He, K. Han, Spatial temporal graph con- volutional networks for skeleton-based dynamic hand gesture recognition, EURASIP Journal on Image and Video Processing 2019 (2019) 1–7

    2019

  39. [47]

    J. Liu, Y . Liu, Y . Wang, V . Prinet, S. Xiang, C. Pan, Decoupled representation learning for skeleton-based gesture recognition, in: Proceedings of the IEEE /CVF conference on computer vi- sion and pattern recognition, 2020, pp. 5751–5760

    2020

  40. [48]

    X. Han, Y . Cui, X. Chen, Y . Lu, W. Hu, Spatio-temporal dy- namic attention graph convolutional network based on skeleton gesture recognition, Electronics 13 (18) (2024) 3733

    2024

  41. [49]

    Peng, P.-H

    S.-H. Peng, P.-H. Tsai, An e fficient graph convolution network for skeleton-based dynamic hand gesture recognition, IEEE Transactions on Cognitive and Developmental Systems 15 (4) (2023) 2179–2189

    2023

  42. [50]

    H. Tian, X. Ma, X. Li, Y . Li, Skeleton-based action recognition with select-assemble-normalize graph convolutional networks, IEEE Transactions on Multimedia (2023)

    2023

  43. [51]

    Y . Zhou, X. Yan, Z.-Q. Cheng, Y . Yan, Q. Dai, X.-S. Hua, Blockgcn: Redefine topology awareness for skeleton-based ac- tion recognition, in: 2024 IEEE /CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 2049–2058. doi:10.1109/CVPR52733.2024.00200

    2024

  44. [52]

    Z. Wu, Y . Ding, L. Wan, T. Li, F. Nian, Local and global self- attention enhanced graph convolutional network for skeleton- based action recognition, Pattern Recognition 159 (2025) 111106

    2025

  45. [53]

    B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, A. Torralba, Learn- ing deep features for discriminative localization, in: Proceed- ings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2921–2929. 15

    2016

Pith tools

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