Pith. sign in

REVIEW 4 major objections 6 minor 93 references

Zero-Shot Skeleton-Based Action Recognition With Prototype-Guided Feature Alignment

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Zero-shot skeleton action recognition improves when unseen-class text features are replaced by prototypes built from confident test samples.

desk verdict A solid method paper with real contributions, but the headline gains are inflated by undisclosed test-time transduction and per-dataset α tuned on the test set. read the letter →

arxiv 2507.00566 v2 pith:NINNR2YV submitted 2025-07-01 cs.CV

Kai Zhou , Shuhai Zhang , Zeng You , Jinwu Hu , Mingkui Tan , Fei Liu This is my paper · ORCID
classification cs.CV
keywords zero-shotactionrecognitionskeleton-basedprototype-guidedfeaturealignmentcross-modalcontrastivelearningpseudo-labelentropyfilteringtext-skeletondistributiondiscrepancyprototypical
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 argues that zero-shot skeleton-based action recognition suffers from two correctable problems: skeleton features trained by a frozen cross-entropy pretrained encoder are not compact enough, and the text features of unseen classes are misaligned with skeleton features at test time. It proposes an end-to-end cross-modal contrastive training framework for the first problem, and a prototype-guided text feature alignment strategy for the second, in which each unseen class is represented by a prototype computed from the highest-confidence pseudo-labeled test skeletons. On NTU-60, NTU-120, and PKU-MMD, the full method reaches 93.17%, 71.38%, and 87.80%, which are absolute gains of 22.96, 12.53, and 18.54 points over the strongest prior baseline. The central point is that adjusting the class representations at test time, not just improving training, is what drives large zero-shot gains.

What carries the argument

The central mechanism is the prototype-guided text feature alignment strategy (Eqs. 8-11): for each unseen class k, build a normalized support set S_k from test skeleton features whose pseudo-label is k, filter by prediction entropy to a final support set Z_k, and define the class prototype c_{u,k} as the normalized centroid of Z_k, using it in place of the text feature w_{u,k} in the softmax similarity classifier. The supporting training-side mechanism is an end-to-end KL-divergence contrastive loss over a batch, where every sample in a class is a positive match for that class's text feature; this is what produces the intra-class compactness on which the entropy filter and prototype centroids rely.

What would settle it

Corrupt the pseudo-labels used to build the prototype support set, for example by randomly flipping an increasing fraction of labels to other unseen classes, and measure the final accuracy; if prototype reclassification does not degrade smoothly with this noise, or if a near-chance initial classifier still produces large gains, then the mechanism attributed to prototype alignment is not doing the claimed work.

Watch

Extended reading notes

Core claim

The central claim is that alignment bias between skeleton features and unseen-class text features can be reduced by substituting prototype features for text features at test time. Given the unlabeled test set, the method first pseudo-labels every test skeleton by cosine similarity to the original unseen-class text features, then for each class retains only the pseudo-labeled skeletons whose prediction entropy falls below a per-class threshold, and sets the class prototype to the normalized centroid of those skeletons; if the filtered set is empty, the original text feature is retained as a fallback. Final predictions are made by cosine similarity to these prototypes. Theorem 1 states that if normalized skeleton features follow von Mises–Fisher distributions with a common concentration parameter, then as the support set grows, assigning a sample to the prototype with highest cosine similarity is equivalent to assigning it to the class with highest class-conditional probability. The paper also claims that training the skeleton encoder and projection layer end-to-end with a bidirectional KL-divergence contrastive loss, instead of pretraining with cross-entropy and freezing the encoder, produces more compact skeleton features, and that these two changes together explain the reported improvements over previous methods.

Load-bearing premise

The load-bearing premise is that the unlabeled test set is available all at once and that the initial text-based pseudo-labels are accurate enough that entropy filtering leaves prototypes that are closer to the true skeleton class centers than the original text features are.

Editorial extensions

If this is right

  • If the end-to-end training claim is correct, zero-shot skeleton recognition can skip cross-entropy pretraining entirely, saving a training stage and reducing parameters (1.00M versus 2.38M for the compared baseline).
  • If the prototype replacement claim is correct, any static text feature can be improved by test-time adaptation whenever the unlabeled test set is available as a whole.
  • The entropy filter is robust across a wide range of the tolerance margin alpha, so the strategy does not require precise hyperparameter tuning to produce most of its gain.
  • The same framework extends directly to one-shot recognition, where a single labeled skeleton per unseen class replaces the text feature, giving competitive results on NTU-60, NTU-120, and PKU-MMD.
  • Complete and skeleton-focused LLM-generated action descriptions consistently beat bare class names, so text quality and test-time prototype alignment compound rather than substitute for each other.

Reading between the lines

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

  • Beyond the paper, the same prototype-substitution mechanism should transfer to zero-shot image or video classification, where unseen-class text features are also known to sit off-center relative to visual features.
  • The paper's Theorem 1 assumes clean pseudo-labels, so a natural extension is to bound final accuracy as a function of initial label noise; the entropy filter softens but does not remove that dependence.
  • Because the method needs all test samples before reclassifying, a practical follow-up is an online prototype bank that updates as samples arrive; comparing its accuracy with the batch version would quantify the transductive advantage.
  • A testable implication of the paper's reasoning is that iterative refinement, re-pseudo-labeling and re-centering prototypes several times, should further close the alignment gap whenever the initial text classifier is better than chance.
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

4 major / 6 minor

Summary. The paper proposes PGFA, a paradigm for zero-shot skeleton-based action recognition consisting of two components: (i) an end-to-end cross-modal contrastive training framework that aligns skeleton features with fixed Sentence-BERT text features using a KL-divergence-based bidirectional contrastive loss, and (ii) a prototype-guided text feature alignment strategy applied at test time, in which pseudo-labels are generated from text-feature similarities, entropy filtering selects high-confidence samples per unseen class, per-class prototypes are computed, and the same test samples are reclassified against these prototypes. The authors report large absolute gains over the SMIE baseline on NTU-60, NTU-120, and PKU-MMD, provide a von Mises-Fisher consistency theorem, and include ablations over description types, frameworks, the tolerance margin alpha, pseudo-labeling strategies, one-shot settings, and cross-dataset transfer.

Significance. If the reported results are reproducible under a clean evaluation protocol, PGFA would be a substantial advance: it simplifies zero-shot skeleton action recognition by removing the separate cross-entropy pretraining stage, and it demonstrates that adapting unseen text features with test-time prototypes can substantially improve accuracy. The manuscript has clear strengths: the code is released, the end-to-end contrastive training idea is well motivated, the ablations in Tables III-VI isolate several design choices, and the theoretical statement, while limited, is not circular and is correctly identified as a consistency result under idealized assumptions. The main significance risk is that the headline comparisons are confounded by a transductive test-time procedure and by per-dataset selection of alpha on the test set, so the claimed 22.96%, 12.53%, and 18.54% improvements over SMIE are not yet established as an advantage of the proposed alignment mechanism over generic transductive self-training.

major comments (4)
  1. [Section IV-C and Fig. 6] The tolerance margin alpha is selected per dataset by inspecting test accuracy curves and then reused to report the main results, with no held-out validation described. Specifically, alpha is set to 0.9, 0.4, and 1.0 for NTU-60, NTU-120, and PKU-MMD, respectively, and Fig. 6 marks the maxima of the test-set accuracy curves at exactly these values. This means the reported gains over SMIE partially reflect test-set tuning. Please provide a validation-based selection rule (e.g., a small labeled validation split or a fixed alpha across datasets and folds), or alternatively report accuracy as a function of alpha with confidence intervals and clarify which alpha values would be used in a deployment scenario.
  2. [Section III-C, Eqs. (8)-(11), and Table V] The proposed testing strategy is transductive: the entire unlabeled test set is used to build support sets and prototypes, and the same test samples are reclassified using Eq. (11). All baselines in Table I, including SMIE, are inductive and use only static text features. Table V shows that the prototype step alone adds roughly 13-17 points on NTU-60 even under the Pretraining&Fixed framework (73.65 to 87.23) and roughly 10-13 points under the End-to-End framework, so a substantial part of the headline improvement may come from generic transductive self-training rather than from the proposed alignment or end-to-end training. The paper should explicitly state that this is transductive zero-shot inference, compare against transductive baselines such as T3A or self-training with the same entropy filtering on the same learned features, and report the decomposition of the SMIE gap accordingly.
  3. [Supplementary Materials, Theorem 1 and Eq. (1)] Theorem 1 assumes that the n support samples v_i^(k) are drawn from the true von Mises-Fisher distribution of class k and that the prototype is the normalized mean of these true-class samples. In contrast, Eq. (8) constructs support sets from pseudo-labels that contain errors, and Eq. (9) filters them by entropy. The theorem therefore does not establish that the entropy-filtered pseudo-label prototypes improve classification over the original text features, nor does it analyze how pseudo-label noise propagates into the prototype. It is a standard law-of-large-numbers consistency statement under noiseless labels. Please qualify the theoretical claim accordingly, or extend the analysis to a noise model for pseudo-labels.
  4. [Section III-A and Algorithm 2] The problem formulation in Section III-A defines the testing set as Du = {(x_u_i, t_u_i)} but does not state that inference is performed in a transductive batch mode using all unlabeled test skeletons. Algorithm 2 explicitly loops over the entire test set before computing prototypes, and the future-work paragraph in Section V acknowledges that this is unsuitable for online scenarios. The paper should state the transductive assumption in the problem formulation, because the validity of the comparison with inductive baselines depends on this assumption being explicit and on the baselines being adapted or clearly labeled as not having access to the unlabeled test distribution.
minor comments (6)
  1. [Section III-B] In the paragraph on Complete Description, 'we follow the approach of SIME [32]' should read 'SMIE'.
  2. [Section III-C, Eq. (9)] The text above Eq. (9) says 'we create a final support set Zk for each seen class k', but this should refer to unseen classes; the notation and surrounding discussion are about unseen classes.
  3. [Table I and Table II captions] The table captions contain the typo 'COMPARSION' and should be 'COMPARISON'.
  4. [Section IV-D] In Setting I, the average accuracy over the three class splits is reported without standard deviations; please report the per-fold results or standard deviations, especially because the alpha values are selected on the aggregate test accuracy.
  5. [Section IV-E, weighted pseudo-labeling formula] The weighted pseudo-labeling formula in the ablation study sums over all test samples Du, whereas the default prototype in Eq. (10) is restricted to the per-class support set Zk; please clarify whether the weighted variant also uses entropy filtering or whether it is a deliberately different aggregation over the full test set.
  6. [Theorem 1 statement] The notation P(v|class k) in Eq. (12) denotes the von Mises-Fisher density value rather than a calibrated class posterior; please clarify this to avoid the impression that the theorem establishes a probabilistic guarantee about classification probabilities.

Circularity Check

2 steps flagged · score 4.0 of 10

Reported gains over SMIE are partially self-referential: α is tuned on the test set, and the prototype reclassification uses the same test set from which the prototypes are built.

  1. fitted input called prediction [Section III-C, Eqs. (8)-(11); results in Section IV-D, Table I]
    "we create a support set Sk for each unseen class k: Sk = { v u i /||v u i || | ˆyi = k, i ≤ Nu, i ∈ N+ } ... Built upon the prototype features, We reclassify each test skeleton sequence xu i by: ˜yi = arg max k exp(sim(vu i, cu,k)) ..."

    The final classifier is constructed from the very test samples it then labels: Eq. (8) puts every test feature into a support set according to its pseudo-label, Eq. (10) defines the class prototype as the centroid of those pseudo-labeled test features, and Eq. (11) reclassifies the same test samples against those prototypes. The reported accuracy is therefore a transductive self-consistency measure, not an inductive zero-shot prediction. Part of the gain over inductive baselines (e.g., the 13-16 point additions from feature alignment in Table V under the SMIE-like Pretraining&Fixed framework) is attributable to using the test distribution itself, not specifically to the paper's alignment mechanism.

  2. fitted input called prediction [Section IV-C Implementation Details; Section IV-E, Fig. 6]
    "The tolerance margin α, which controls the size of final support sets, is set to 0.9, 0.4, and 1.0 for the NTU-60, NTU-120, and PKU-MMD datasets, respectively. ... Fig. 6. Effect of tolerance margin α. ... (Max Acc=93.17%, α=0.9) (Max Acc=71.38%, α=0.4) (Max Acc=87.80%, α=1.0)"

    Alpha is chosen per dataset as the value that maximizes test accuracy in Fig. 6, and the accuracies at those alpha values (93.17%, 71.38%, 87.80%) are exactly the headline results reported for PGFA. Thus the claimed absolute improvements over SMIE are not independent predictions for unseen classes; they are test-set-optimized numbers. The ablation claim that the method is robust to alpha is also read from the same figure used to pick the best alpha, making the reported peak accuracy self-referential.

full rationale

The paper's methodological derivation is not circular: the end-to-end contrastive loss (Eq. 5), the prototype construction (Eqs. 8-10), and Theorem 1 are independent algorithmic and mathematical content, and no load-bearing self-citation chain is used. The prototype strategy is transparently transductive (Algorithm 2 takes the full test set as input), and Table V shows that the feature-alignment component adds a large gain even under the SMIE-like Pretraining&Fixed framework, so the central training contribution has independent support. However, the headline empirical claims are partially self-referential in two ways: alpha is selected per dataset from Fig. 6 to maximize test accuracy, and the prototypes in Eqs. (8)-(10) are built from and evaluated on the same test set. These make the reported absolute gains over inductive baselines an optimistic, test-set-tuned transductive number rather than a pure derivation from the proposed alignment mechanism. Score 4 reflects partial evaluation circularity with independent central content.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The method rests on four domain assumptions (vMF feature geometry, reliable text encoder, full test-set access, and accurate pseudo-labels) plus one fitted hyperparameter, alpha, which is tuned on the test set. The text encoder and LLM descriptions are external assets, not the paper's own contributions.

free parameters (1)
  • alpha (tolerance margin) = 0.9 (NTU-60), 0.4 (NTU-120), 1.0 (PKU-MMD)
    Controls the size of the final support set Zk in Eq. 9. It is selected per dataset to maximize test accuracy, as shown in Figure 6, with no held-out validation split described.
assumptions (5)
  • ad hoc to paper Unseen normalized skeleton features follow a von Mises-Fisher distribution with equal concentration parameter kappa.
    Assumed in Theorem 1 to justify prototype-based classification. No empirical tests confirm this for the actual features.
  • domain assumption The pre-trained text encoder (Sentence-BERT) provides semantically meaningful features for unseen action descriptions.
    The method's pseudo-labels and initial similarities depend on text features (Eqs. 6-7) not tuned to the skeleton domain.
  • domain assumption The entire unlabeled test set is available for computing prototypes.
    The testing strategy in Eqs. 8-11 uses all test samples, making the setting transductive, which the problem formulation does not state.
  • domain assumption Entropy-filtered pseudo-labels are accurate enough to yield useful prototypes.
    The effectiveness of Eq. 10 depends on this; Theorem 1 assumes clean class-conditional features and does not model pseudo-label noise.
  • domain assumption The ground-truth similarity vectors in Eq. 5 are normalized probability distributions.
    The KL loss requires normalized targets; the paper does not explain how multiple positives are normalized.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zero-Shot Skeleton-Based Action Recognition With Prototype-Guided Feature Alignment." pith.science (2026). https://pith.science/paper/NINNR2YV

@misc{pith2026250700566,
  author       = {Pith},
  title        = {Pith review of: Zero-Shot Skeleton-Based Action Recognition With Prototype-Guided Feature Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NINNR2YV}},
  note         = {Machine review of arXiv:2507.00566}
}
read the original abstract

Zero-shot skeleton-based action recognition aims to classify unseen skeleton-based human actions without prior exposure to such categories during training. This task is extremely challenging due to the difficulty in generalizing from known to unknown actions. Previous studies typically use two-stage training: pre-training skeleton encoders on seen action categories using cross-entropy loss and then aligning pre-extracted skeleton and text features, enabling knowledge transfer to unseen classes through skeleton-text alignment and language models' generalization. However, their efficacy is hindered by 1) insufficient discrimination for skeleton features, as the fixed skeleton encoder fails to capture necessary alignment information for effective skeleton-text alignment; 2) the neglect of alignment bias between skeleton and unseen text features during testing. To this end, we propose a prototype-guided feature alignment paradigm for zero-shot skeleton-based action recognition, termed PGFA. Specifically, we develop an end-to-end cross-modal contrastive training framework to improve skeleton-text alignment, ensuring sufficient discrimination for skeleton features. Additionally, we introduce a prototype-guided text feature alignment strategy to mitigate the adverse impact of the distribution discrepancy during testing. We provide a theoretical analysis to support our prototype-guided text feature alignment strategy and empirically evaluate our overall PGFA on three well-known datasets. Compared with the top competitor SMIE method, our PGFA achieves absolute accuracy improvements of 22.96%, 12.53%, and 18.54% on the NTU-60, NTU-120, and PKU-MMD datasets, respectively.

Figures

Figures reproduced from arXiv: 2507.00566 by the authors.

Figure 1
Figure 1. Illustration of traditional framework and our framework for zero-shot skeleton-based action recognition. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Training framework of proposed PGFA. During [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Different descriptions of action “drink water”. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Testing framework of proposed PGFA. During [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Illustration of prototype-guided text feature alignment strategy. Class [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Effect of tolerance margin α. The x-axis represents the value of α, while the y-axis represents the average accuracy for the three class splits under Setting I. is not conducive to subsequent skeleton-text alignment. By allowing finetuning of the skeleton encoder durin…
Figure 7
Figure 7. Figure 7: Visualization of skeleton features using t-SNE [ [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: The average Fisher Discrimination Ratio (FDR) of skeleton features [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 10
Figure 10. Figure 10: Visualization of prototype-guided text feature alignment in feature [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 1
Figure 1. Figure 1: The confusion matrix visualization for NTU-60 in Setting I compares three methods: SMIE (reimplemented) in the first column, our PGFA [PITH_FULL_IMAGE:figures/full_fig_p018_1.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

93 extracted references · 69 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

  2. [2]

    11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEcompsocconfonly \@IEEEauthordefaulttextstyle \@IEEEcompsocnotconfonly \@IEEEauthor...

  3. [3]

    Simonyan and A

    K. Simonyan and A. Zisserman, ``Two-stream convolutional networks for action recognition in videos,'' Advances in neural information processing systems, vol. 27, 2014

    2014

  4. [4]

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

    2015

  5. [5]

    L. Wang, Y. Xiong, Z. Wang, Y. Qiao, D. Lin, X. Tang, and L. Van Gool, ``Temporal segment networks: Towards good practices for deep action recognition,'' in European conference on computer vision. 1em plus 0.5em minus 0.4em Springer, 2016, pp. 20--36

    2016

  6. [6]

    Carreira and A

    J. Carreira and A. Zisserman, ``Quo vadis, action recognition? a new model and the kinetics dataset,'' in proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 6299--6308

    2017

  7. [7]

    Feichtenhofer, H

    C. Feichtenhofer, H. Fan, J. Malik, and K. He, ``Slowfast networks for video recognition,'' in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6202--6211

    2019

  8. [8]

    Z. Tong, Y. Song, J. Wang, and L. Wang, ``Videomae: Masked autoencoders are data-efficient learners for self-supervised video pre-training,'' Advances in neural information processing systems, vol. 35, pp. 10\,078--10\,093, 2022

    2022

Show all 93 references
  1. [9]

    C. Li, B. Zhang, C. Chen, Q. Ye, J. Han, G. Guo, and R. Ji, ``Deep manifold structure transfer for action recognition,'' IEEE transactions on image processing, vol. 28, no. 9, pp. 4646--4658, 2019

    2019

  2. [10]

    Y. Liu, K. Wang, G. Li, and L. Lin, ``Semantics-aware adaptive knowledge distillation for sensor-to-vision action recognition,'' IEEE Transactions on Image Processing, vol. 30, pp. 5573--5588, 2021

    2021

  3. [11]

    J. Qin, L. Liu, Z. Zhang, Y. Wang, and L. Shao, ``Compressive sequential learning for action similarity labeling,'' IEEE Transactions on Image Processing, vol. 25, no. 2, pp. 756--769, 2015

    2015

  4. [12]

    Y. Tang, J. Lu, Z. Wang, M. Yang, and J. Zhou, ``Learning semantics-preserving attention and contextual interaction for group activity recognition,'' IEEE Transactions on Image Processing, vol. 28, no. 10, pp. 4997--5012, 2019

    2019

  5. [13]

    W. Liu, X. Zhong, Z. Zhou, K. Jiang, Z. Wang, and C.-W. Lin, ``Dual-recommendation disentanglement network for view fuzz in action recognition,'' IEEE Transactions on Image Processing, vol. 32, pp. 2719--2733, 2023

    2023

  6. [14]

    Reily, F

    B. Reily, F. Han, L. E. Parker, and H. Zhang, ``Skeleton-based bio-inspired human activity prediction for real-time human--robot interaction,'' Autonomous Robots, vol. 42, pp. 1281--1298, 2018

    2018

  7. [15]

    Bandi and U

    C. Bandi and U. Thomas, ``Skeleton-based action recognition for human-robot interaction using self-attention mechanism,'' in 2021 16th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021). 1em plus 0.5em minus 0.4em IEEE, 2021, pp. 1--8

    2021

  8. [16]

    J. Yin, J. Han, C. Wang, B. Zhang, and X. Zeng, ``A skeleton-based action recognition system for medical condition detection,'' in 2019 IEEE Biomedical Circuits and Systems Conference (BioCAS). 1em plus 0.5em minus 0.4em IEEE, 2019, pp. 1--4

    2019

  9. [17]

    Noor and I

    N. Noor and I. K. Park, ``A lightweight skeleton-based 3d-cnn for real-time fall detection and action recognition,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2179--2188

    2023

  10. [18]

    Elaoud, W

    A. Elaoud, W. Barhoumi, E. Zagrouba, and B. Agrebi, ``Skeleton-based comparison of throwing motion for handball players,'' Journal of Ambient Intelligence and Humanized Computing, vol. 11, pp. 419--431, 2020

    2020

  11. [19]

    S. Yan, Y. Xiong, and D. Lin, ``Spatial temporal graph convolutional networks for skeleton-based action recognition,'' in Proceedings of the AAAI conference on artificial intelligence, vol. 32, no. 1, 2018

    2018

  12. [20]

    Wang and L

    H. Wang and L. Wang, ``Beyond joints: Learning representations from primitive geometries for skeleton-based action recognition and detection,'' IEEE Transactions on Image Processing, vol. 27, no. 9, pp. 4382--4394, 2018

    2018

  13. [21]

    L. Shi, Y. Zhang, J. Cheng, and H. Lu, ``Two-stream adaptive graph convolutional networks for skeleton-based action recognition,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 12\,026--12\,035

    2019

  14. [22]

    Zhang, C

    P. Zhang, C. Lan, J. Xing, W. Zeng, J. Xue, and N. Zheng, ``View adaptive neural networks for high performance skeleton-based human action recognition,'' IEEE transactions on pattern analysis and machine intelligence, vol. 41, no. 8, pp. 1963--1978, 2019

    1963

  15. [23]

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

    2020

  16. [24]

    X. Hao, J. Li, Y. Guo, T. Jiang, and M. Yu, ``Hypergraph neural network for skeleton-based action recognition,'' IEEE Transactions on Image Processing, vol. 30, pp. 2263--2275, 2021

    2021

  17. [25]

    Cheng, Y

    K. Cheng, Y. Zhang, X. He, J. Cheng, and H. Lu, ``Extremely lightweight skeleton-based action recognition with shiftgcn++,'' IEEE Transactions on Image Processing, vol. 30, pp. 7333--7348, 2021

    2021

  18. [26]

    Y. Chen, Z. Zhang, C. Yuan, B. Li, Y. Deng, and W. Hu, ``Channel-wise topology refinement graph convolution for skeleton-based action recognition,'' in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 13\,359--13\,368

    2021

  19. [27]

    H.-g. Chi, M. H. Ha, S. Chi, S. W. Lee, Q. Huang, and K. Ramani, ``Infogcn: Representation learning for human skeleton-based action recognition,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 20\,186--20\,196

    2022

  20. [28]

    H. Duan, Y. Zhao, K. Chen, D. Lin, and B. Dai, ``Revisiting skeleton-based action recognition,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 2969--2978

    2022

  21. [29]

    Y. Zhu, H. Shuai, G. Liu, and Q. Liu, ``Multilevel spatial--temporal excited graph network for skeleton-based action recognition,'' IEEE Transactions on Image Processing, vol. 32, pp. 496--508, 2022

    2022

  22. [30]

    Myung, N

    W. Myung, N. Su, J.-H. Xue, and G. Wang, ``Degcn: Deformable graph convolutional networks for skeleton-based action recognition,'' IEEE Transactions on Image Processing, vol. 33, pp. 2477--2490, 2024

    2024

  23. [31]

    X. Xu, T. Hospedales, and S. Gong, ``Transductive zero-shot action recognition by word-vector embedding,'' International Journal of Computer Vision, vol. 123, pp. 309--333, 2017

    2017

  24. [32]

    Jasani and A

    B. Jasani and A. Mazagonwalla, ``Skeleton based zero shot action recognition in joint pose-language semantic space,'' arXiv preprint arXiv:1911.11344, 2019

    1911 arXiv

  25. [33]

    Gupta, D

    P. Gupta, D. Sharma, and R. K. Sarvadevabhatla, ``Syntactically guided generative embeddings for zero-shot skeleton action recognition,'' in 2021 IEEE International Conference on Image Processing (ICIP). 1em plus 0.5em minus 0.4em IEEE, 2021, pp. 439--443

    2021

  26. [34]

    Y. Zhou, W. Qiang, A. Rao, N. Lin, B. Su, and J. Wang, ``Zero-shot skeleton-based action recognition via mutual information estimation and maximization,'' in Proceedings of the 31st ACM International Conference on Multimedia, 2023, pp. 5302--5310

    2023

  27. [35]

    A. Zhu, Q. Ke, M. Gong, and J. Bailey, ``Part-aware unified representation of language and skeleton for zero-shot action recognition,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 18\,761--18\,770

    2024

  28. [36]

    K. He, H. Fan, Y. Wu, S. Xie, and R. Girshick, ``Momentum contrast for unsupervised visual representation learning,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 9729--9738

    2020

  29. [37]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., ``Learning transferable visual models from natural language supervision,'' in International conference on machine learning. 1em plus 0.5em minus 0.4em PMLR, 202...

    2021

  30. [38]

    M. Wang, J. Xing, J. Mei, Y. Liu, and Y. Jiang, ``Actionclip: Adapting language-image pretrained models for video action recognition,'' IEEE Transactions on Neural Networks and Learning Systems, 2023

    2023

  31. [39]

    L. Xue, M. Gao, C. Xing, R. Mart \' n-Mart \' n, J. Wu, C. Xiong, R. Xu, J. C. Niebles, and S. Savarese, ``Ulip: Learning a unified representation of language, images, and point clouds for 3d understanding,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pat...

    2023

  32. [40]

    Hegde, J

    D. Hegde, J. M. J. Valanarasu, and V. Patel, ``Clip goes 3d: Leveraging prompt tuning for language grounded 3d recognition,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2028--2038

    2023

  33. [41]

    Snell, K

    J. Snell, K. Swersky, and R. Zemel, ``Prototypical networks for few-shot learning,'' Advances in neural information processing systems, vol. 30, 2017

    2017

  34. [42]

    Huang, B

    W. Huang, B. Xiao, J. Hu, and X. Bi, ``Location-aware transformer network for few-shot medical image segmentation,'' in 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 1150--1157

    2023

  35. [43]

    Rebuffi, A

    S.-A. Rebuffi, A. Kolesnikov, G. Sperl, and C. H. Lampert, ``icarl: Incremental classifier and representation learning,'' in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2017, pp. 2001--2010

    2017

  36. [44]

    H. Lin, Y. Zhang, Z. Qiu, S. Niu, C. Gan, Y. Liu, and M. Tan, ``Prototype-guided continual adaptation for class-incremental unsupervised domain adaptation,'' in European Conference on Computer Vision. 1em plus 0.5em minus 0.4em Springer, 2022, pp. 351--368

    2022

  37. [45]

    Iwasawa and Y

    Y. Iwasawa and Y. Matsuo, ``Test-time classifier adjustment module for model-agnostic domain generalization,'' Advances in Neural Information Processing Systems, vol. 34, pp. 2427--2440, 2021

    2021

  38. [46]

    Y. Du, W. Wang, and L. Wang, ``Hierarchical recurrent neural network for skeleton based action recognition,'' in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 1110--1118

    2015

  39. [47]

    S. Song, C. Lan, J. Xing, W. Zeng, and J. Liu, ``An end-to-end spatio-temporal attention model for human action recognition from skeleton data,'' in Proceedings of the AAAI conference on artificial intelligence, vol. 31, no. 1, 2017

    2017

  40. [48]

    Zhang, C

    P. Zhang, C. Lan, J. Xing, W. Zeng, J. Xue, and N. Zheng, ``View adaptive recurrent neural networks for high performance human action recognition from skeleton data,'' in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2117--2126

    2017

  41. [49]

    Choutas, P

    V. Choutas, P. Weinzaepfel, J. Revaud, and C. Schmid, ``Potion: Pose motion representation for action recognition,'' in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7024--7033

    2018

  42. [50]

    Caetano, J

    C. Caetano, J. Sena, F. Br \'e mond, J. A. Dos Santos, and W. R. Schwartz, ``Skelemotion: A new representation of skeleton joint sequences based on motion information for 3d action recognition,'' in 2019 16th IEEE international conference on advanced video and signal based sur...

    2019

  43. [51]

    K. Xu, F. Ye, Q. Zhong, and D. Xie, ``Topology-aware convolutional neural network for efficient skeleton-based action recognition,'' in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 3, 2022, pp. 2866--2874

    2022

  44. [52]

    Cheng, Y

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

    2020

  45. [53]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, . Kaiser, and I. Polosukhin, ``Attention is all you need,'' Advances in neural information processing systems, vol. 30, 2017

    2017

  46. [54]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., ``An image is worth 16x16 words: Transformers for image recognition at scale,'' arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  47. [55]

    Plizzari, M

    C. Plizzari, M. Cannici, and M. Matteucci, ``Spatial temporal transformer network for skeleton-based action recognition,'' in Pattern recognition. ICPR international workshops and challenges: virtual event, January 10--15, 2021, Proceedings, Part III. 1em plus 0.5em minus 0.4e...

    2021

  48. [56]

    L. Shi, Y. Zhang, J. Cheng, and H. Lu, ``Decoupled spatial-temporal attention network for skeleton-based action-gesture recognition,'' in Proceedings of the Asian conference on computer vision, 2020

    2020

  49. [57]

    Zhang, B

    Y. Zhang, B. Wu, W. Li, L. Duan, and C. Gan, ``Stst: Spatial-temporal specialized transformer for skeleton-based action recognition,'' in Proceedings of the 29th ACM International Conference on Multimedia, 2021, pp. 3229--3237

    2021

  50. [58]

    Frome, G

    A. Frome, G. S. Corrado, J. Shlens, S. Bengio, J. Dean, M. Ranzato, and T. Mikolov, ``Devise: A deep visual-semantic embedding model,'' Advances in neural information processing systems, vol. 26, 2013

    2013

  51. [59]

    F. Sung, Y. Yang, L. Zhang, T. Xiang, P. H. Torr, and T. M. Hospedales, ``Learning to compare: Relation network for few-shot learning,'' in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 1199--1208

    2018

  52. [60]

    Mikolov, I

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, ``Distributed representations of words and phrases and their compositionality,'' in Advances in Neural Information Processing Systems 26: 27th Annual Conference on Neural Information Processing Systems 2013. Procee...

    2013

  53. [61]

    Reimers and I

    N. Reimers and I. Gurevych, ``Sentence-bert: Sentence embeddings using siamese bert-networks,'' in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 2...

    2019

  54. [62]

    Hubert Tsai, L.-K

    Y.-H. Hubert Tsai, L.-K. Huang, and R. Salakhutdinov, ``Learning robust visual-semantic embeddings,'' in Proceedings of the IEEE International conference on Computer Vision, 2017, pp. 3571--3580

    2017

  55. [63]

    M. Wray, D. Larlus, G. Csurka, and D. Damen, ``Fine-grained action retrieval through multiple parts-of-speech embeddings,'' in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 450--459

    2019

  56. [64]

    Schonfeld, S

    E. Schonfeld, S. Ebrahimi, S. Sinha, T. Darrell, and Z. Akata, ``Generalized zero-and few-shot learning via aligned variational autoencoders,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 8247--8255

    2019

  57. [65]

    C. Jia, Y. Yang, Y. Xia, Y.-T. Chen, Z. Parekh, H. Pham, Q. Le, Y.-H. Sung, Z. Li, and T. Duerig, ``Scaling up visual and vision-language representation learning with noisy text supervision,'' in International conference on machine learning. 1em plus 0.5em minus 0.4em PMLR, 20...

    2021

  58. [66]

    X. Wang, Y. Yan, H.-M. Hu, B. Li, and H. Wang, ``Cross-modal contrastive learning network for few-shot action recognition,'' IEEE Transactions on Image Processing, 2024

    2024

  59. [67]

    Zhang, Z

    R. Zhang, Z. Guo, W. Zhang, K. Li, X. Miao, B. Cui, Y. Qiao, P. Gao, and H. Li, ``Pointclip: Point cloud understanding by clip,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 8552--8562

    2022

  60. [68]

    Tevet, B

    G. Tevet, B. Gordon, A. Hertz, A. H. Bermano, and D. Cohen-Or, ``Motionclip: Exposing human motion generation to clip space,'' in European Conference on Computer Vision. 1em plus 0.5em minus 0.4em Springer, 2022, pp. 358--374

    2022

  61. [69]

    Xiang, C

    W. Xiang, C. Li, Y. Zhou, B. Wang, and L. Zhang, ``Generative action description prompts for skeleton-based action recognition,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 10\,276--10\,285

    2023

  62. [70]

    H. Xu, L. Liu, Q. Bian, and Z. Yang, ``Semi-supervised semantic segmentation with prototype-based consistency regularization,'' Advances in neural information processing systems, vol. 35, pp. 26\,007--26\,020, 2022

    2022

  63. [71]

    W. Liu, Y. Wen, Z. Yu, and M. Yang, ``Large-margin softmax loss for convolutional neural networks,'' in International Conference on Machine Learning. 1em plus 0.5em minus 0.4em PMLR, 2016, pp. 507--516

    2016

  64. [72]

    Elsayed, D

    G. Elsayed, D. Krishnan, H. Mobahi, K. Regan, and S. Bengio, ``Large margin deep networks for classification,'' Advances in neural information processing systems, vol. 31, 2018

    2018

  65. [73]

    Khosla, P

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y. Tian, P. Isola, A. Maschinot, C. Liu, and D. Krishnan, ``Supervised contrastive learning,'' Advances in neural information processing systems, vol. 33, pp. 18\,661--18\,673, 2020

    2020

  66. [74]

    Z. Shi, H. Wang, and C.-S. Leung, ``Constrained center loss for convolutional neural networks,'' IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 2, pp. 1080--1088, 2021

    2021

  67. [75]

    A. v. d. Oord, Y. Li, and O. Vinyals, ``Representation learning with contrastive predictive coding,'' arXiv preprint arXiv:1807.03748, 2018

    2018 arXiv

  68. [76]

    Brown, B

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell et al., ``Language models are few-shot learners,'' Advances in neural information processing systems, vol. 33, pp. 1877--1901, 2020

    1901

  69. [77]

    Banerjee, I

    A. Banerjee, I. S. Dhillon, J. Ghosh, S. Sra, and G. Ridgeway, ``Clustering on the unit hypersphere using von mises-fisher distributions.'' Journal of Machine Learning Research, vol. 6, no. 9, 2005

    2005

  70. [78]

    A. T. Wood, ``Simulation of the von mises fisher distribution,'' Communications in statistics-simulation and computation, vol. 23, no. 1, pp. 157--164, 1994

    1994

  71. [79]

    Shahroudy, J

    A. Shahroudy, J. Liu, T.-T. Ng, and 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 recognition, 2016, pp. 1010--1019

    2016

  72. [80]

    J. Liu, A. Shahroudy, M. Perez, G. Wang, L.-Y. Duan, and A. C. Kot, ``Ntu rgb+ d 120: A large-scale benchmark for 3d human activity understanding,'' IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 10, pp. 2684--2701, 2019

    2019

  73. [81]

    J. Liu, S. Song, C. Liu, Y. Li, and Y. Hu, ``A benchmark dataset and comparison study for multi-modal human action analytics,'' ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), vol. 16, no. 2, pp. 1--24, 2020

    2020

  74. [82]

    L. Li, M. Wang, B. Ni, H. Wang, J. Yang, and W. Zhang, ``3d human action representation learning via cross-view consistency pursuit,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 4741--4750

    2021

  75. [83]

    S. Hu, Y. Tu, X. Han, C. He, G. Cui, X. Long, Z. Zheng, Y. Fang, Y. Huang, W. Zhao et al., ``Minicpm: Unveiling the potential of small language models with scalable training strategies,'' arXiv preprint arXiv:2404.06395, 2024

    2024 arXiv

  76. [84]

    A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan et al., ``Deepseek-v3 technical report,'' arXiv preprint arXiv:2412.19437, 2024

    2024 arXiv

  77. [85]

    H.-J. Ye, H. Hu, D.-C. Zhan, and F. Sha, ``Few-shot learning via embedding adaptation with set-to-set functions,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 8808--8817

    2020

  78. [86]

    Simon, P

    C. Simon, P. Koniusz, R. Nock, and M. Harandi, ``Adaptive subspaces for few-shot learning,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 4136--4145

    2020

  79. [87]

    C. Xu, Y. Fu, C. Liu, C. Wang, J. Li, F. Huang, L. Zhang, and X. Xue, ``Learning dynamic alignment via meta-filter for few-shot learning,'' in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 5182--5191

    2021

  80. [88]

    S. Yang, J. Liu, S. Lu, E. M. Hwa, and A. C. Kot, ``One-shot action recognition via multi-scale spatial-temporal skeleton matching,'' IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 7, pp. 5149--5156, 2024

    2024

  81. [89]

    J. Hu, Y. Hou, Z. Guo, and J. Gao, ``Global and local contrastive learning for self-supervised skeleton-based action recognition,'' IEEE Transactions on Circuits and Systems for Video Technology, 2024

    2024

  82. [90]

    Van der Maaten and G

    L. Van der Maaten and G. Hinton, ``Visualizing data using t-sne.'' Journal of machine learning research, vol. 9, no. 11, 2008

    2008

  83. [91]

    G. J. McLachlan, Discriminant analysis and statistical pattern recognition. 1em plus 0.5em minus 0.4em John Wiley & Sons, 2005

    2005

  84. [92]

    P. J. Rousseeuw, ``Silhouettes: a graphical aid to the interpretation and validation of cluster analysis,'' Journal of computational and applied mathematics, vol. 20, pp. 53--65, 1987

    1987

  85. [93]

    F. M. Dekking, C. Kraaikamp, H. P. Lopuha \"a , and L. E. Meester, A Modern Introduction to Probability and Statistics: Understanding why and how. 1em plus 0.5em minus 0.4em Springer Science & Business Media, 2006

    2006

Pith tools

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