Pith. sign in

REVIEW 4 major objections 8 minor 1 cited by

Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition

T0 review · 4 major / 8 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Neuron learns evolving skeleton prototypes to recognize unseen actions from text descriptions.

desk verdict The evaluation metric makes the reported SOTA gains non-comparable to baselines, but the mechanisms are worth a serious look. read the letter →

arxiv 2411.11288 v2 pith:245KA335 submitted 2024-11-18 cs.CV

Yang Chen , Jingcai Guo , Song Guo , Dacheng Tao This is my paper · ORCID
classification cs.CV
keywords zero-shotskeletonactionrecognitiongeneralizedlearningcontext-awaresideinformationmicro-prototypesspatialcompressiontemporalmemoryLLM-generateddescriptionscross-modalalignment
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 aims to improve zero-shot skeleton action recognition by replacing static, one-line semantic descriptions with context-rich side information generated through multi-turn LLM prompting, and by aligning skeleton features to this semantics through learnable spatial and temporal micro-prototypes that grow iteratively. It claims this dynamic, controllable alignment from micro to macro perspectives captures fine-grained cross-modal correspondences more effectively than prior global or static alignment schemes. If correct, the method should transfer knowledge from seen to unseen action categories more reliably, reducing the confusion between visually similar actions. The paper reports state-of-the-art accuracy on NTU RGB+D, NTU RGB+D 120, and PKU-MMD under both zero-shot and generalized zero-shot protocols.

What carries the argument

The central object is the spatial-temporal micro-prototype pair: learnable attribute vectors initialized randomly and updated over three phases, intended to capture structure-related spatial patterns and regularity-dependent temporal patterns. Spatial prototypes are refined by a spatial compression mechanism that thresholds joint-prototype similarity scores with a gradually increasing retention ratio alpha, while temporal prototypes are updated by a temporal memory mechanism with sigmoid-gated recall and remember branches to prevent forgetting across phases. These prototypes are guided at each phase by context-aware side information, produced via multi-turn LLM prompting and embedded with a frozen CLIP text encoder, and aligned to skeleton features through per-phase contrastive losses, providing the micro-to-macro progressive alignment that carries the generalization claim.

What would settle it

Take the context-aware descriptions generated for a confusable pair such as 'walking apart' versus 'walking towards' or 'sneeze/cough' versus 'nausea/vomiting', embed them with the same CLIP text encoder, and compute their cosine similarity; if the similarity is near 1.0 at all three phases, the semantic space cannot support the claimed fine-grained separation, and the method's reported gains would have to come from elsewhere. A complementary test is to rerun the full pipeline with the multi-turn contextual descriptions replaced by one-turn single descriptions while keeping everything else fixed; a small accuracy drop on the hardest split would indicate that the context-aware side information is not actually load-bearing.

Watch

Extended reading notes

Core claim

The paper's central claim is that zero-shot skeleton action recognition improves when semantics are both context-aware and evolving, and when skeleton features are aligned to them through growing prototypes rather than through a single static embedding. Concretely, the proposed Neuron framework first queries an LLM in multiple turns to generate spatial descriptions at coarse, mid, and fine granularities and temporal descriptions at start, mid, and end phases, producing heterogeneous, context-rich side information for each action class. It then constructs spatial and temporal micro-prototypes: the spatial prototype grows through a spatial compression mechanism that keeps only the top-alpha joint-prototype similarity scores per phase (alpha increasing), while the temporal prototype grows through a temporal memory mechanism that combines recall of previous phases with absorption of new temporal patterns via sigmoid-gated updates. At each phase, the prototypes are pooled into a skeleton representation and aligned to the corresponding CLIP-embedded semantics through contrastive losses, progressively refining the cross-modal alignment. The paper argues that this stepwise, controllable alignment avoids shortcut feature learning and knowledge oblivion, and the reported experiments show accuracy gains over previous methods on three benchmarks.

Load-bearing premise

The load-bearing premise is that the frozen CLIP text encoder, when fed multi-turn LLM-generated spatial and temporal action descriptions, produces semantic embeddings that are discriminative enough to separate fine-grained skeleton actions, so that the per-phase contrastive alignment can succeed.

Editorial extensions

If this is right

  • For zero-shot skeleton action recognition, richer context-aware side information improves accuracy on both seen and unseen classes, reducing the domain bias problem in generalized zero-shot settings.
  • The spatial compression mechanism, by discarding least relevant prototype attributes per phase, reduces shortcut feature learning and improves generalization, as shown in the component ablation.
  • The temporal memory mechanism mitigates knowledge oblivion across phases, which the paper links to better recognition of temporally similar actions such as 'walking apart' versus 'walking towards'.
  • Performance remains stable across a range of prototype counts, suggesting the framework is not highly sensitive to the number of spatial or temporal micro-prototypes chosen.

Reading between the lines

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

  • Since the method relies on a frozen CLIP text encoder, a direct test of its load-bearing premise would be to compute pairwise cosine similarities of the CLIP embeddings of generated descriptions for confusable action pairs; if those similarities remain very high at all phases, the semantic space alone cannot support the reported fine-grained alignment.
  • The paper does not independently validate the semantic quality of the multi-turn LLM descriptions; replacing the LLM with a weaker model, or shuffling the descriptions across classes, would provide a clean diagnosis of how much of the gain comes from the context-aware side information versus the evolving prototype alignment.
  • The micro-to-macro prototype growth idea may transfer to other skeleton-language tasks such as text-based motion retrieval or instruction-conditioned action generation, where stepwise semantic guidance could similarly help control alignment.
  • The reported failure on 'sneeze/cough' versus 'nausea/vomiting' suggests that the current context-aware descriptions still carry homogeneous semantics for extremely similar abstract actions; further refining the prompting strategy or using complementary visual descriptions might be a testable extension.
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 / 8 minor

Summary. The paper proposes Neuron, a zero-shot skeleton action recognition framework that combines multi-turn LLM-generated "context-aware side information" with spatial and temporal micro-prototypes that evolve over several phases. The spatial stream uses a compression mechanism to retain only top-α joint-prototype similarities, while the temporal stream splits the sequence into three phases and uses a memory mechanism to combine recall and new information. Training aligns pooled prototype features with CLIP text embeddings of the generated semantics at each phase. The paper reports state-of-the-art ZSL and GZSL results on NTU RGB+D 60, NTU RGB+D 120, and PKU-MMD, together with ablations and qualitative visualizations.

Significance. The core idea—using multi-turn LLM descriptions to guide progressively refined micro-prototypes—is a reasonable and potentially useful extension of recent fine-grained zero-shot skeleton recognition work, and the paper provides a substantial set of experiments, ablations, and code. If the results are correctly measured, the observed margins over STAR and other baselines would be notable. However, the central state-of-the-art claim is compromised as written by a nonstandard evaluation rule and by unresolved inconsistencies in the objective and hyperparameters. The significance of the empirical contribution therefore cannot be confirmed from the current manuscript.

major comments (4)
  1. [Sections 3.5 and 4.2, Eqs. (9)-(10)] The final prediction is defined as the two-element set \hat{y}^* = {\hat{y}_s, \hat{y}_t}, and the reported Top-1 accuracy is Acc = (1/N) \sum_i I[y_i \in \hat{y}^*]. This is not the standard single-label Top-1 accuracy used by RelationNet, STAR, and the other baselines in Tables 2-4: a test sample is counted correct if either stream independently produces the right class, giving the method two chances per sample. Consequently the reported SOTA gains, such as +5.5 ZSL on NTU60 Xsub 55/5 over STAR, are not measured on the same basis as the baselines. The authors need to supply a deterministic fusion rule that collapses the two stream predictions into one label (for example, summing the two similarity scores, or a learned confidence gate) and recompute all tables under that single-label rule.
  2. [Section 4.3 and Section 3.4, Eq. (8)] The implementation details state that "the hyperparameter \lambda_s and \lambda_t are set to 0.0003 and 0.0002", but the overall objective in Eq. (8) contains no such coefficients: it is a plain sum of L_s^e and L_t^e. Either the equations are missing weighting terms, or the text is describing unused parameters. In addition, the spatial compression threshold α is described only as "incrementally increased" without a schedule or actual values. Both issues block exact reproduction of the reported results and should be corrected.
  3. [Tables 2-4, Section 4.4, and Table 5] All reported comparisons are single-run results without error bars or standard deviations. Table 5 says three different seen-unseen partitions were averaged "to minimize variance", but no variance is reported. Given that the central claim is state-of-the-art performance, the authors should report mean and standard deviation over multiple runs (at least with different seeds) for the main tables, and for the averaged settings in Table 5.
  4. [Section 3.2] The context-aware side information is a central contribution, but the generation process is not reproducible from the paper: no multi-turn prompt templates are provided, no example generated descriptions are shown, and the number of descriptions N_a per phase is never specified. Since the method's separation and claimed benefit depend on these LLM-generated semantics, the authors should include the exact prompting protocol and sample outputs (or a link to them) so that reviewers and readers can judge the semantic quality and reproduce the experiments.
minor comments (8)
  1. [Table 1] The row for the proposed method contains the typo "Neruon"; it should be "Neuron".
  2. [Table 2] The header contains "A cc" in several places; this should be "Acc".
  3. [Table 3] The first column header reads "Xusb"; it should be "Xsub".
  4. [Section 3.5] The sentence "In the inference stage, we can contain the spatial-temporal features" should be "we can obtain".
  5. [Section 3.3, Eq. (3)] The temporal feature F_t^0 is not defined: Section 3.3 defines temporal segments as F_t = {F_t^1, ..., F_t^{N_e}}, so Eq. (3) should use F_t^1 (and similarly in Eq. (4)).
  6. [Eqs. (6)-(7) vs. Eqs. (9)-(10)] The inference equations introduce functions \rho_s, \rho_t, and \phi that are not defined in the training objective, where the corresponding projection functions are \phi_s, \phi_t, and \psi. The notation should be unified or explicitly distinguished.
  7. [Sections 2.1 and 2.2] The paragraphs end with the literal text "[Summary]:"; these appear to be leftover drafting placeholders and should be removed or converted into normal prose.
  8. [Section 4.3 and Figure 3] The text states that the best performance is obtained with N_s around 100 and N_t around 50, yet the implementation sets both N_s and N_t to 80. Please clarify which values were used for the reported results.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Neuron's results rest on a standard ZSL training and evaluation pipeline; the set-valued prediction metric is a comparability concern, not a circular reduction.

full rationale

The paper's derivation chain is an empirical deep-learning pipeline, not a closed-form derivation or a uniqueness argument. The spatial and temporal micro-prototypes are trained with the cross-modal alignment losses in Eqs. 6-8 on seen classes, and inference in Eqs. 9-10 is a standard nearest-semantic-embedding rule; no parameter is fitted to the reported test accuracy, and the unseen-class predictions are not defined in terms of the accuracy metric. Self-citations, chiefly to STAR [6], are used for comparison, data preprocessing, and the illustrative notion of skeleton attributes; they do not carry the load of the SOTA claim. The paper itself acknowledges residual semantic homogeneity in Sec. 4.6, which is a limitation rather than a circular step. The one notable issue is that Sec. 3.5 defines the final prediction as the set y* = {ys, yt} and Sec. 4.2 defines 'Top-1' accuracy as Acc = (1/N) sum_i I[yi in y*], which gives the model two chances to match the label and is not the single-label Top-1 accuracy used by the baselines in Tables 2-4; this is a metric-comparability and correctness concern, not a circularity, because the accuracy is not equivalent by construction to any training input or fitted constant. Under the circularity taxonomy, there is no self-definitional, fitted-input, self-citation-load-bearing, uniqueness-imported, or ansatz-smuggled step, so no circularity is found.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The method's central claim rests on hand-tuned hyperparameters (phase count, prototype counts, alpha schedule, calibration factors) and on the reliability of pre-trained CLIP and LLM outputs. No new physical entities are introduced; micro-prototypes are learnable parameters, not external postulates.

free parameters (7)
  • Number of phases N_e = 3
    Hand-chosen; no ablation on phase count reported.
  • Number of spatial micro-prototypes N_s = 80 (Fig. 3 optimum near 100)
    Hyperparameter; sensitivity shown but optimal value not used in final experiments.
  • Number of temporal micro-prototypes N_t = 80 (Fig. 3 optimum near 50)
    Hyperparameter; sensitivity shown but optimal value not used in final experiments.
  • Spatial compression threshold alpha schedule = Not specified
    Top-alpha fraction in each phase; paper says it increases but gives no values.
  • GZSL calibration factors gamma_s, gamma_t = Not specified
    Used in Eq. 9-10; values not reported.
  • Number of LLM state descriptions N_a = Not specified
    Section 3.2 introduces N_a descriptions per category; the value is never given.
  • lambda_s and lambda_t = 0.0003 and 0.0002 (not used in losses)
    Listed in Section 4.3 but no lambda appears in Eqs. 6-8; likely a leftover, indicating inconsistency.
assumptions (4)
  • domain assumption Pre-trained CLIP text encoder embeds skeleton action semantics discriminatively
    Section 3.2 uses CLIP ViT-L/14@336px to embed LLM descriptions; if CLIP lacks fine-grained action distinctions, the alignment losses cannot separate similar classes.
  • domain assumption Shift-GCN skeleton encoder pretrained on seen categories provides transferable features
    Section 4.3 uses Shift-GCN pretrained on seen categories; the method assumes these features support zero-shot transfer to unseen categories.
  • domain assumption Multi-turn LLM (GPT-4o) outputs faithfully describe skeleton motion
    Section 3.2 generates semantics via multi-turn prompts without verification; hallucinated or mistaken descriptions would corrupt prototypes.
  • ad hoc to paper Equal three-way temporal split aligns with the three semantic phases
    Section 3.3 splits F_t into three equal segments; no evidence that these correspond to the LLM's start/mid/end temporal descriptions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition." pith.science (2026). https://pith.science/paper/245KA335

@misc{pith2026241111288,
  author       = {Pith},
  title        = {Pith review of: Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/245KA335}},
  note         = {Machine review of arXiv:2411.11288}
}
read the original abstract

Zero-shot skeleton action recognition is a non-trivial task that requires robust unseen generalization with prior knowledge from only seen classes and shared semantics. Existing methods typically build the skeleton-semantics interactions by uncontrollable mappings and conspicuous representations, thereby can hardly capture the intricate and fine-grained relationship for effective cross-modal transferability. To address these issues, we propose a novel dyNamically Evolving dUal skeleton-semantic syneRgistic framework with the guidance of cOntext-aware side informatioN (dubbed Neuron), to explore more fine-grained cross-modal correspondence from micro to macro perspectives at both spatial and temporal levels, respectively. Concretely, 1) we first construct the spatial-temporal evolving micro-prototypes and integrate dynamic context-aware side information to capture the intricate and synergistic skeleton-semantic correlations step-by-step, progressively refining cross-model alignment; and 2) we introduce the spatial compression and temporal memory mechanisms to guide the growth of spatial-temporal micro-prototypes, enabling them to absorb structure-related spatial representations and regularity-dependent temporal patterns. Notably, such processes are analogous to the learning and growth of neurons, equipping the framework with the capacity to generalize to novel unseen action categories. Extensive experiments on various benchmark datasets demonstrated the superiority of the proposed method.

Figures

Figures reproduced from arXiv: 2411.11288 by the authors.

Figure 1
Figure 1. Method sketches: (a) Existing methods employ homoge [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The pipeline of the proposed method. (a) represents the evolving spatial-temporal representation learning process. (b) shows the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The influence of hyper-parameters on the NTU 60. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: (a) - (c) represent the t-SNE visualization of the skeleton spatial spaces for unseen categories on different phases. The color [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: (a) - (c) represent the visualization of updated temporal micro-prototype in each phase for a randomly selected skeleton sequence. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Unseen action classification accuracy compared with [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. DynaPURLS: Dynamic Refinement of Part-Aware Representations for Skeleton-Based Zero-Shot Action Recognition

    cs.CV 2025-12 conditional novelty 6.0 of 10

    DynaPURLS adapts textual action descriptions at inference time using the model's own confident predictions, improving zero-shot skeleton action recognition accuracy on NTU60/120 and PKU-MMD over static-matching baselines.

Reference graph

Works this paper leans on

42 extracted references · 32 canonical work pages · cited by 1 Pith paper

  1. [1]

    Multi-modal learning algorithms and net- work architectures for information extraction and retrieval

    Maurits Bleeker. Multi-modal learning algorithms and net- work architectures for information extraction and retrieval. In Proceedings of the 30th ACM International Conference on Multimedia, pages 6925–6929, 2022. 3

    work page 2022

  2. [2]

    Mmnet: A model-based multimodal net- work for human action recognition in rgb-d videos

    XB Bruce, Yan Liu, Xiang Zhang, Sheng-hua Zhong, and Keith CC Chan. Mmnet: A model-based multimodal net- work for human action recognition in rgb-d videos. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(3):3522–3538, 2022. 3

    work page 2022

  3. [3]

    A dendrite method for cluster analysis

    Tadeusz Cali ´nski and Jerzy Harabasz. A dendrite method for cluster analysis. Communications in Statistics-theory and Methods, 3(1):1–27, 1974. 7, 8

    work page 1974

  4. [4]

    An empirical study and analysis of generalized zero- shot learning for object recognition in the wild

    Wei-Lun Chao, Soravit Changpinyo, Boqing Gong, and Fei Sha. An empirical study and analysis of generalized zero- shot learning for object recognition in the wild. In Com- puter Vision–ECCV 2016: 14th European Conference, Am- sterdam, The Netherlands, October 11-14, 2016, Proceed- ings, Part II 14, pages 52–68. Springer, 2016. 5

    work page 2016

  5. [5]

    Channel-wise topology refinement graph convolution for skeleton-based action recognition

    Yuxin Chen, Ziqi Zhang, Chunfeng Yuan, Bing Li, Ying Deng, and Weiming Hu. Channel-wise topology refinement graph convolution for skeleton-based action recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 13359–13368, 2021. 1

    2021

  6. [6]

    Fine-grained side information guided dual-prompts for zero-shot skeleton action recognition

    Yang Chen, Jingcai Guo, Tian He, Xiaocheng Lu, and Ling Wang. Fine-grained side information guided dual-prompts for zero-shot skeleton action recognition. In Proceedings of the 32nd ACM International Conference on Multimedia , pages 778–786, 2024. 1, 2, 3, 4, 5, 6, 7, 8

    work page 2024

  7. [7]

    Vision-language meets the skele- ton: Progressively distillation with cross-modal knowledge for 3d action representation learning

    Yang Chen, Tian He, Junfeng Fu, Ling Wang, Jingcai Guo, Ting Hu, and Hong Cheng. Vision-language meets the skele- ton: Progressively distillation with cross-modal knowledge for 3d action representation learning. IEEE Transactions on Multimedia, 2024. 1, 3

    work page 2024

  8. [8]

    Skeleton-based action recognition with shift graph convolutional network

    Ke Cheng, Yifan Zhang, Xiangyu He, Weihan Chen, Jian Cheng, and Hanqing Lu. Skeleton-based action recognition with shift graph convolutional network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 183–192, 2020. 1, 6

    work page 2020

Show all 42 references
  1. [9]

    Pku-mmd: A large scale benchmark for continu- ous multi-modal human action understanding.arXiv preprint arXiv:1703.07475, 2017

    Liu Chunhui, Hu Yueyu, Li Yanghao, Song Sijie, and Liu Jiaying. Pku-mmd: A large scale benchmark for continu- ous multi-modal human action understanding.arXiv preprint arXiv:1703.07475, 2017. 5

    2017 arXiv

  2. [10]

    Multi-modal gait recognition via effective spatial-temporal feature fusion

    Yufeng Cui and Yimei Kang. Multi-modal gait recognition via effective spatial-temporal feature fusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 17949–17957, 2023. 3

    2023

  3. [11]

    Beyond question-based biases: Assessing multimodal shortcut learning in visual question answering

    Corentin Dancette, Remi Cadene, Damien Teney, and Matthieu Cord. Beyond question-based biases: Assessing multimodal shortcut learning in visual question answering. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 1574–1583, 2021. 3

    2021

  4. [12]

    A method for com- paring two hierarchical clusterings

    Edward B Fowlkes and Colin L Mallows. A method for com- paring two hierarchical clusterings. Journal of the American statistical association, 78(383):553–569, 1983. 7, 8

    1983

  5. [13]

    Syntactically guided generative embeddings for zero-shot skeleton action recognition

    Pranay Gupta, Divyanshu Sharma, and Ravi Kiran Sarvadev- abhatla. Syntactically guided generative embeddings for zero-shot skeleton action recognition. In2021 IEEE Interna- tional Conference on Image Processing (ICIP) , pages 439–

  6. [14]

    Enhancing skeleton-based action recogni- tion with language descriptions from pre-trained large mul- timodal models

    Tian He, Yang Chen, Xu Gao, Ling Wang, Ting Hu, and Hong Cheng. Enhancing skeleton-based action recogni- tion with language descriptions from pre-trained large mul- timodal models. IEEE Transactions on Circuits and Systems for Video Technology, 2024. 3

    2024

  7. [15]

    An expert-knowledge-based graph convolutional network for skeleton-based physical rehabilitation exercises assessment

    Tian He, Yang Chen, Ling Wang, and Hong Cheng. An expert-knowledge-based graph convolutional network for skeleton-based physical rehabilitation exercises assessment. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2024. 1

    2024

  8. [16]

    Comparing partitions

    Lawrence Hubert and Phipps Arabie. Comparing partitions. Journal of classification, 2:193–218, 1985. 7, 8

    1985

  9. [17]

    Learning robust visual-semantic embed- dings

    Yao-Hung Hubert Tsai, Liang-Kang Huang, and Ruslan Salakhutdinov. Learning robust visual-semantic embed- dings. In Proceedings of the IEEE International conference on Computer Vision, pages 3571–3580, 2017. 6, 7

    2017

  10. [18]

    Skeleton based zero shot action recognition in joint pose-language semantic space

    Bhavan Jasani and Afshaan Mazagonwalla. Skeleton based zero shot action recognition in joint pose-language semantic space. arXiv preprint arXiv:1911.11344, 2019. 2, 3, 4

    1911 arXiv

  11. [19]

    Cross- modal learning with 3d deformable attention for action recognition

    Sangwon Kim, Dasom Ahn, and Byoung Chul Ko. Cross- modal learning with 3d deformable attention for action recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10265–10275, 2023. 3

    2023

  12. [20]

    Multi-semantic fusion model for generalized zero-shot skeleton-based action recognition

    Ming-Zhe Li, Zhen Jia, Zhang Zhang, Zhanyu Ma, and Liang Wang. Multi-semantic fusion model for generalized zero-shot skeleton-based action recognition. InInternational Conference on Image and Graphics, pages 68–80. Springer,

  13. [21]

    Sa-dvae: Improv- ing zero-shot skeleton-based action recognition by disentan- gled variational autoencoders

    Sheng-Wei Li, Zi-Xiang Wei, Wei-Jie Chen, Yi-Hsin Yu, Chih-Yuan Yang, and Jane Yung-jen Hsu. Sa-dvae: Improv- ing zero-shot skeleton-based action recognition by disentan- gled variational autoencoders. In European Conference on Computer Vision, pages 447–462. Springer, 2025. 1...

    2025

  14. [22]

    Ntu rgb+ d 120: A large- scale benchmark for 3d human activity understanding

    Jun Liu, Amir Shahroudy, Mauricio Perez, Gang Wang, Ling-Yu Duan, and Alex C Kot. Ntu rgb+ d 120: A large- scale benchmark for 3d human activity understanding. IEEE transactions on pattern analysis and machine intelligence , 42(10):2684–2701, 2019. 5

    2019

  15. [23]

    Multi-modality co-learning for efficient skeleton-based action recognition

    Jinfu Liu, Chen Chen, and Mengyuan Liu. Multi-modality co-learning for efficient skeleton-based action recognition. arXiv preprint arXiv:2407.15706, 2024. 3

    2024 arXiv

  16. [24]

    Enhanc- ing multi-modal fusion in visual dialog via sample debias- ing and feature interaction.Information Fusion, 107:102302,

    Chenyu Lu, Jun Yin, Hao Yang, and Shiliang Sun. Enhanc- ing multi-modal fusion in visual dialog via sample debias- ing and feature interaction.Information Fusion, 107:102302,

  17. [25]

    Cross- modal contrastive pre-training for few-shot skeleton action recognition

    Mingqi Lu, Siyuan Yang, Xiaobo Lu, and Jun Liu. Cross- modal contrastive pre-training for few-shot skeleton action recognition. IEEE Transactions on Circuits and Systems for Video Technology, 2024. 3

    2024

  18. [26]

    Cmd: Self-supervised 3d action representation learning with cross-modal mutual distillation

    Yunyao Mao, Wengang Zhou, Zhenbo Lu, Jiajun Deng, and Houqiang Li. Cmd: Self-supervised 3d action representation learning with cross-modal mutual distillation. In European Conference on Computer Vision , pages 734–752. Springer,

  19. [27]

    Progressive cross- modal knowledge distillation for human action recognition

    Jianyuan Ni, Anne HH Ngu, and Yan Yan. Progressive cross- modal knowledge distillation for human action recognition. In Proceedings of the 30th ACM International Conference on Multimedia, pages 5903–5912, 2022. 3

    2022

  20. [28]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  21. [29]

    Silhouettes: a graphical aid to the inter- pretation and validation of cluster analysis

    Peter J Rousseeuw. Silhouettes: a graphical aid to the inter- pretation and validation of cluster analysis. Journal of com- putational and applied mathematics, 20:53–65, 1987. 7, 8

    1987

  22. [30]

    Prompt- guided zero-shot anomaly action recognition using pre- trained deep skeleton features

    Fumiaki Sato, Ryo Hachiuma, and Taiki Sekii. Prompt- guided zero-shot anomaly action recognition using pre- trained deep skeleton features. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6471–6480, 2023. 1

    2023

  23. [31]

    Generalized zero-shot learn- ing via aligned variational autoencoders

    Edgar Schonfeld, Sayna Ebrahimi, Samarth Sinha, Trevor Darrell, and Zeynep Akata. Generalized zero-shot learn- ing via aligned variational autoencoders. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 54–57, 2019. 6, 7

    2019

  24. [32]

    Ntu rgb+ d: A large scale dataset for 3d human activity anal- ysis

    Amir Shahroudy, Jun Liu, Tian-Tsong Ng, and Gang Wang. Ntu rgb+ d: A large scale dataset for 3d human activity anal- ysis. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1010–1019, 2016. 5

    2016

  25. [33]

    Modality compensation network: Cross-modal adaptation for action recognition

    Sijie Song, Jiaying Liu, Yanghao Li, and Zongming Guo. Modality compensation network: Cross-modal adaptation for action recognition. IEEE Transactions on Image Pro- cessing, 29:3957–3969, 2020. 3

    2020

  26. [34]

    Uni- fied multi-modal unsupervised representation learning for skeleton-based action understanding

    Shengkai Sun, Daizong Liu, Jianfeng Dong, Xiaoye Qu, Junyu Gao, Xun Yang, Xun Wang, and Meng Wang. Uni- fied multi-modal unsupervised representation learning for skeleton-based action understanding. In Proceedings of the 31st ACM International Conference on Multimedia , pages ...

    2023

  27. [35]

    Fine-grained action retrieval through multiple parts- of-speech embeddings

    Michael Wray, Diane Larlus, Gabriela Csurka, and Dima Damen. Fine-grained action retrieval through multiple parts- of-speech embeddings. In Proceedings of the IEEE/CVF in- ternational conference on computer vision , pages 450–459,

  28. [36]

    Generative action description prompts for skeleton-based action recognition

    Wangmeng Xiang, Chao Li, Yuxuan Zhou, Biao Wang, and Lei Zhang. Generative action description prompts for skeleton-based action recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 10276–10285, 2023. 3

    2023

  29. [37]

    Finesports: A multi-person hierarchical sports video dataset for fine-grained action understanding

    Jinglin Xu, Guohao Zhao, Sibo Yin, Wenhao Zhou, and Yuxin Peng. Finesports: A multi-person hierarchical sports video dataset for fine-grained action understanding. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 21773–21782, 2024. 1

    2024

  30. [38]

    Learn- ing discriminative representations for skeleton based action recognition

    Huanyu Zhou, Qingjie Liu, and Yunhong Wang. Learn- ing discriminative representations for skeleton based action recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10608– 10617, 2023. 1, 4

    2023

  31. [39]

    Zero-shot skeleton-based action recogni- tion via mutual information estimation and maximization

    Yujie Zhou, Wenwen Qiang, Anyi Rao, Ning Lin, Bing Su, and Jiaqi Wang. Zero-shot skeleton-based action recogni- tion via mutual information estimation and maximization. In Proceedings of the 31st ACM International Conference on Multimedia, pages 5302–5310, 2023. 2, 3, 4, 5, 6, 7

    2023

  32. [40]

    Blockgcn: Redefine topology aware- ness for skeleton-based action recognition

    Yuxuan Zhou, Xudong Yan, Zhi-Qi Cheng, Yan Yan, Qi Dai, and Xian-Sheng Hua. Blockgcn: Redefine topology aware- ness for skeleton-based action recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2049–2058, 2024. 1

    2024

  33. [41]

    Part-aware unified representation of language and skele- ton for zero-shot action recognition

    Anqi Zhu, Qiuhong Ke, Mingming Gong, and James Bai- ley. Part-aware unified representation of language and skele- ton for zero-shot action recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18761–18770, 2024. 1, 2, 3, 4, 6

    2024

  34. [443]

    2, 3, 4, 6, 7

    IEEE, 2021. 2, 3, 4, 6, 7

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.