The DOI matches its Crossref record; 31 of the 32 references it cites were found in a registry or an index.
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition
Open accessAbstract
Dynamics of human body skeletons convey significant information for human action recognition. Conventional approaches for modeling skeletons usually rely on hand-crafted parts or traversal rules, thus resulting in limited expressive power and difficulties of generalization. In this work, we propose a novel model of dynamic skeletons called Spatial-Temporal Graph Convolutional Networks (ST-GCN), which moves beyond the limitations of previous methods by automatically learning both the spatial and temporal patterns from data. This formulation not only leads to greater expressive power but also stronger generalization capability. On two large datasets, Kinetics and NTU-RGBD, it achieves substantial improvements over mainstream methods.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“In tasks such as image object recognition, the hierarchical representation and locality are usually achieved by the intrinsic properties of convolutional neural networks (Krizhevsky, Sutskever, and Hinton 2012), rather than manually assigning object parts.”
MethodSemi-supervised classification with graph convolutional networks
Kipf, T. N., & Welling, M. (2016) · arXiv · doi:10.48550/arxiv.1609.02907 · Cited by 24,853 · cited 8×
“In the single frame case, ST-GCN with the first partitioning strategy can be implemented with the following formula (Kipf and Welling 2017) where Here the weight vectors of multiple output channels are stacked to form the weight matrix W .”
BackgroundMethodInfluentialLearning Spatiotemporal Features with 3D Convolutional Networks
Tran, D., Bourdev, L., Fergus, R., Torresani, L., & Paluri, M. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.510 · Cited by 9,593 · cited 2×
“In general, human action can be recognized from multiple modalities(Simonyan and Zisserman 2014; Tran et al. 2015; Wang, Qiao, and Tang 2015; Wang et al. 2016; Zhao et al. 2017), such as appearance, depth, optical flows, and body skeletons (Du, Wang, and Wang 2015; Liu et al.”
BackgroundRealtime Multi-person 2D Pose Estimation Using Part Affinity Fields
Cao, Z., Simon, T., Wei, S.-E., & Sheikh, Y. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.143 · Cited by 7,539 · cited 3×
“Skeleton and joint trajectories of human bodies are robust to illumination change and scene variation, and they are easy to obtain owing to the highly accurate depth sensors or pose estimation algorithms (Shotton et al. 2011; Cao et al. 2017a).”
BackgroundMethodDeformable Convolutional Networks
Dai, J., Qi, H., Xiong, Y., Li, Y., Zhang, G., Hu, H., & Wei, Y. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.89 · Cited by 7,513 · cited 1×
“More detailed explanation and other applications of this formulation can be found in (Dai et al. 2017).”
BackgroundSelf-citationAdvances In Deep Learning On Graphs (Gsp'18 Workshop)Preprint
Defferrard, M., Bresson, X., & Vandergheynst, P. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1606.09375 · Cited by 7,021 · cited 2×
“…graphs of arbitrary structures, have received increasing attention and successfully been adopted in a number of applications, such as image classification (Bruna et al. 2014), document classifi-cation (Defferrard, Bresson, and Vandergheynst 2016), and semi-supervised learning (Kipf and Welling 2017).”
BackgroundTwo-Stream Convolutional Networks for Action Recognition in Videos
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1406.2199 · Cited by 6,756 · cited 1×
“In general, human action can be recognized from multiple modalities(Simonyan and Zisserman 2014; Tran et al. 2015; Wang, Qiao, and Tang 2015; Wang et al. 2016; Zhao et al. 2017), such as appearance, depth, optical flows, and body skeletons (Du, Wang, and Wang 2015; Liu et al. 2016).”
BackgroundThe Kinetics Human Action Video DatasetPreprint
Kay, W., Carreira, J., Simonyan, K., Zhang, B., Hillier, C., Vijayanarasimhan, S., Viola, F., Green, T., Back, T., Natsev, P., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1705.06950 · Cited by 5,004 · cited 13×
“Weevaluate the recognition performance by top-1 and top-5 classification accuracy as recommended by the dataset authors (Kay et al. 2017).”
BackgroundMethodInfluentialTemporal Segment Networks: Towards Good Practices for Deep Action Recognition
Wang, L., Xiong, Y., Wang, Z., Qiao, Y., Lin, D., Tang, X., & Gool, L. V. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-46484-8_2 · Cited by 4,232 · cited 4×
“We train the standard TSN (Wang et al. 2016) models from scratches on Kinetics with RGB and optical flow models.”
BackgroundMethodInfluentialSelf-citationSpectral Networks and Locally Connected Networks on Graphs
Bruna, J., Zaremba, W., Szlam, A., & LeCun, Y. (2013) · arXiv · doi:10.48550/arxiv.1312.6203 · Cited by 3,983
Research works printed but not linked: 4
- [Cao et al. 2017b] Cao, Z.; Simon, T.; Wei, S.-E.; and Sheikh, Y. 2017b. Realtime multi-person 2d pose estimation using part affinity fields. In CVPR .
- [Defferrard, Bresson, and Vandergheynst 2016] Defferrard, M.; Bresson, X.; and Vandergheynst, P. 2016. Convolutional neural networks on graphs with fast localized spectral filtering. In NIPS .
- [Hussein et al. 2013] Hussein, M. E.; Torki, M.; Gowayyed, M. A.; and El-Saban, M. 2013. Human action recognition using a temporal hierarchy of covariance descriptors on 3d joint locations. In IJCAI .
- [Wang et al. 2012] Wang, J.; Liu, Z.; Wu, Y.; and Yuan, J. 2012. Mining actionlet ensemble for action recognition with depth cameras. In CVPR . IEEE.
Cited by
Sorted by citations
Items 701–800 of 1,007
Yasin, H., Hussain, M., & Weber⋆, A. (2020) · Sensors · doi:10.3390/s20082226 · Cited by 43
Motion Matters: Motion-guided Modulation Network for Skeleton-based Micro-Action Recognition
Gu, J., Li, K., Wang, F., Wei, Y.-Y., Wu, Z.-L., Fan, H.-H., & Wang, M. (2025) · ACM Multimedia · doi:10.1145/3746027.3754722 · Cited by 38
Hadamard Product in Deep Learning: Introduction, Advances and Challenges
Chrysos, G. G., Wu, Y., Pascanu, R., Torr, P. H. S., & Cevher, V. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3560423 · Cited by 43
“In [260], [261], [262], a Hadamard product between a weighting matrix and the edges of a graph is performed for importance weighting.”
Yang, B., Zhu, J., Hu, C.-P., Yu, Z., Hu, H.-Y., & Ni, R.-R. (2024) · IEEE Transactions on Transportation Electrification · doi:10.1109/tte.2024.3360966 · Cited by 38
“Yan et al. [20] were the first to introduce the concept of spatio-temporal graph convolution.”
A human activity recognition method based on Vision Transformer
Han, H.-Y., Zeng, H., Kuang, L.-Q., Han, X., & Xue, H. (2024) · Scientific Reports · doi:10.1038/s41598-024-65850-3 · Cited by 38
Arbitrary Motion Style Transfer with Multi-Condition Motion Latent Diffusion Model
Song, W., Jin, X., Li, S., Chen, C., Hao, A., Hou, X., Li, N., & Qin, H. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.00084 · Cited by 41
“We train a content classifier as a feature extractor using [49] on a subset of the HumanML3D test set with annotated content labels.”
Zhao, H., Cui, H., Qu, K., Zhu, J., Li, H., Cui, Z., & Wu, Y. (2024) · Biosystems Engineering · doi:10.1016/j.biosystemseng.2024.02.011 · Cited by 40
GCNext: Towards the Unity of Graph Convolutions for Human Motion Prediction
Wang, X., Cui, Q., Chen, C., & Liu, M. (2023) · AAAI Conference on Artificial Intelligence · doi:10.48550/arxiv.2312.11850 · Cited by 38
“Spatial-temporal graph convolu-tions capture both spatial and temporal relationships (Yan, Xiong, and Lin 2018; Liu et al. 2020; Zhong et al. 2022; Li et al. 2021).”
BackgroundZhu, W., Liu, Y.-H., Wang, P., Zhang, M., Wang, T., & Yi, Y. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.109772 · Cited by 38
Dynamic attention augmented graph network for video accident anticipation
Song, W.-F., Li, S., Chang, T., Xie, K.-P., Hao, A.-M., & Qin, H. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110071 · Cited by 38
Ibh, M., Graßhof, S., Witzner, D., & Madeleine, P. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00548 · Cited by 40
“Graph convolutional networks (GCN) are a popular method for skeleton-based action recognition [35, 40].”
BackgroundMethodLearning Heterogeneous Spatial–Temporal Context for Skeleton-Based Action Recognition
Gao, X., Yang, Y., Wu, Y., & Du, S. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3252172 · Cited by 42
“After that, due to the importance of topology in graph convolution, most of these methods [26], [31], [34], [42] focus on learning action-related adaptive correlations between joints.”
BackgroundMethodInfluentialRDGCN: Reasonably Dense Graph Convolution Network for Pedestrian Trajectory Prediction
Sang, H.-F., Chen, W.-X., Wang, J.-Y., & Zhao, Z.-S. (2023) · Measurement · doi:10.1016/j.measurement.2023.112675 · Cited by 38
Li, N., Zhong, J.-X., Shu, X.-J., & Guo, H. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.01.026 · Cited by 38
Zhou, R. H., Zhou, H.-Y., Gao, H., Tomizuka, M., Li, J., & Xu, Z. (2022) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra46639.2022.9811585 · Cited by 33
“The group-level trajectory encoder is an STGCN proposed in [28] and used in [29].”
BackgroundMethodXu, J.-Y., Pradhan, A., & Duraisamy, K. (2021) · Neural Information Processing Systems · doi:10.48550/arxiv.2109.09510 · Cited by 38
“Applications include cloud classification [25, 26], action recognition [27] and control [28], traffic forecasting [29, 30], quantum chemistry [24].”
BackgroundFiltjens, B., Ginis, P., Nieuwboer, A., Slaets, P., & Vanrumste, B. (2022) · Journal of NeuroEngineering and Rehabilitation · doi:10.1186/s12984-022-01025-3 · Cited by 43
“The adjusted input is passed through several blocks of ST-GCN [50].”
MethodResultInfluentialFacial Expression Recognition Using Spatial-Temporal Semantic Graph Network
Zhou, J., Zhang, X., Liu, Y. M., & Lan, X. (2020) · Proceedings - International Conference on Image Processing · doi:10.1109/icip40778.2020.9191181 · Cited by 41
“Inspired by [13], the edges of the semantic facial graph are transformed into a N × N size adjacency matrix A , where the value of each element e ij is generated by Eq.”
MethodGraph Interaction Networks for Relation Transfer in Human Activity Videos
Tang, Y.-S., Wei, Y., Yu, X.-M., Lu, J.-W., & Zhou, J. (2020) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2020.2973301 · Cited by 38
“Discussion on the Insight of GINs: Recent works have shown the great importance of relation information for various video analysis tasks, especially for skeleton-based action recogniton [16], [35]–[39] and group activity recognition [57], [58].”
BackgroundMethodInfluentialLearning Representations From Skeletal Self-Similarities for Cross-View Action Recognition
Shao, Z., Li, Y., & Zhang, H. (2020) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.2965574 · Cited by 41
A comprehensive survey of action quality assessment: Method and benchmark
Zhou, K.-L., Cai, R., Wang, L.-Y., Shum, H. P. H., & Liang, X.-H. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2026.113933 · Cited by 37
Video-Based Fall Detection Using Human Pose and Constrained Generative Adversarial Network
Wu, L., Huang, C., Fei, L., Zhao, S.-P., Zhao, J., Cui, Z.-W., & Xu, Y. (2024) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2023.3303258 · Cited by 37
Motion Complement and Temporal Multifocusing for Skeleton-Based Action Recognition
Wu, C., Wu, X.-J., Xu, T.-Y., Shen, Z.-W., & Kittler, J. (2024) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2023.3236430 · Cited by 37
“Yan et al. [13] extended the conventional spatial graph convolution and proposed a spatiotemporal graph convolution structure to better reflect the intuitive understanding of spatiotemporal information characterizing actions.”
BackgroundMethodInfluentialUnified Pose Sequence Modeling
Foo, L. G., Li, T., Rahmani, H., Ke, Q., & Liu, J. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01251 · Cited by 44
“Following previous works [88], we extract 2D pose sequences using the OpenPose [6] toolbox.”
BackgroundMethodInfluentialCascading Pose Features with CNN-LSTM for Multiview Human Action Recognition
Malik, N. U. R., Abu–Bakar, S. A. R., Sheikh, U. U., Channa, A., & Popescu, N. (2023) · Signals · doi:10.3390/signals4010002 · Cited by 47
Node-Oriented Spectral Filtering for Graph Neural Networks
Zheng, S., Zhu, Z., Liu, Z., Li, Y., & Zhao, Y. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3324937 · Cited by 33
“Meanwhile, GNNs have also demonstrated remarkable capabilities in a number of graph-related applications, including but not limited to recommendation system [? 7], disease prediction [5, 3], drug discovery [9, 8], and action recognition [11, 10].”
BackgroundMOTFR: Multiple Object Tracking Based on Feature Recoding
Kong, J., Mo, E., Jiang, M., & Liu, T. (2022) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2022.3182709 · Cited by 37
“A good way to model the interaction among countable embeddings is graph neural network [33]–[35].”
MethodWu, T., Scaglione, A., & Arnold, D. B. (2022) · IEEE Transactions on Smart Grid · doi:10.1109/tsg.2023.3332591 · Cited by 37
One Model is Not Enough: Ensembles for Isolated Sign Language Recognition
Hrúz, M., Gruber, I., Kanis, J., Boháček, M., Hlaváč, M., & Krňoul, Z. (2022) · Sensors · doi:10.3390/s22135043 · Cited by 39
“Inspired by Yan et al. [26], who used a spatio-temporal graph convolutional network (GCN) for action recognition, Vázquez-Enríquez et al. [27] used GCNs also for SLR.”
MethodLearn to Predict How Humans Manipulate Large-sized Objects from Interactive Motions
Wan, W., Yang, L., Liu, L.-J., Zhang, Z., Jia, R., Choi, Y.-K., Pan, J., Theobalt, C., Komura, T., & Wang, W.-P. (2022) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2022.3151614 · Cited by 37
“We adopt the spatial-temporal graph convolution proposed in [37] to learn motion features and predict the future interactive motions based on the input sequence.”
InfluentialLee, G., & Kim, J. (2022) · Applied Sciences · doi:10.3390/app12042168 · Cited by 37
Inception Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition
Jiang, M., Dong, J., Ma, D., Sun, J., He, J.-M., & Lang, L. (2022) · 2022 International Symposium on Control Engineering and Robotics (ISCER) · doi:10.1109/iscer55570.2022.00042 · Cited by 37
Learning Dynamical Human-Joint Affinity for 3D Pose Estimation in Videos
Zhang, J., Wang, Y., Zhou, Z., Luan, T., Wang, Z., & Qiao, Y. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3109517 · Cited by 43
“Recently, it has been adopted for human-centric computer vision tasks [4], [28],by treating human skeleton as a specific graph.”
Learning Sparse Interaction Graphs of Partially Detected Pedestrians for Trajectory Prediction
Huang, Z., Li, R., Shin, K., & Driggs-Campbell, K. (2021) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2021.3138547 · Cited by 37
“We analyze the effect of recursive trajectory prediction by comparison between Masked LSTM and temporal convolution network (TCN), which has been applied to encode temporal relationship and make sequential prediction [16], [40].”
MethodSTAR: Sparse Transformer-based Action RecognitionPreprint
Shi, F., Lee, C., Qiu, L., Zhao, Y., Shen, T., Muralidhar, S., Han, T., Zhu, S. C., & Narayanan, V. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.07089 · Cited by 32
“The model consists of spatial and temporal convolution modules similar to conventional convolutional filters used for images (Yan, Xiong, and Lin 2018).”
BackgroundMethodInfluentialEAGLE-Eye: Extreme-pose Action Grader using detaiL bird’s-Eye view
Nekoui, M., Cruz, F. O. T., & Cheng, L. (2021) · IEEE Workshop/Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00044 · Cited by 37
Gesnouin, J., Pechberti, S., Bresson, G., Stanciulescu, B., & Moutarde, F. (2020) · Algorithms · doi:10.3390/a13120331 · Cited by 41
“Yan et al. [47] extend the spatial convolutions of graphs into spatio-temporal convolutions.”
BackgroundZhao, L., Wang, Y.-X., Zhao, J., Yuan, L.-Z., Sun, J. J., Schroff, F., Adam, H., Peng, X., Metaxas, D. N., & Liu, T. (2020) · Computer Vision and Pattern Recognition · doi:10.1109/cvpr46437.2021.01260 · Cited by 37
“For fully-supervised baselines taking 2D poses, we compare our approach with three state-of-the-art methods based on temporal convolutions: Res-TCN [20], ST-GCN [53], and HCN [23].”
BackgroundMethodInfluentialLan, G., Heit, B., Scargill, T., & Gorlatova, M. A. (2020) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/3384419.3430774 · Cited by 35
Miao, S., Chen, L., Hu, R., & Luo, Y. (2022) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3550331 · Cited by 36
“In recent years, skeleton-based HAR models have been developed from CNN-RNN-based models [24, 25, 56] to GCN-based models [8, 29, 41, 42, 51, 55, 57].”
MethodLi, N., Chang, F., & Liu, C. (2022) · Neurocomputing · doi:10.1016/j.neucom.2021.12.023 · Cited by 36
Liu, R.-Y., Liu, Y., Wu, M., Xin, W., Miao, Q.-G., Liu, X.-Z., & Li, L. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.111377 · Cited by 36
Zhou, Y., Xu, T.-Y., Wu, C., Wu, X., & Kittler, J. (2024) · IEEE International Conference on Computer Vision · doi:10.1109/iccv51701.2025.01175 · Cited by 36
“Regarding modelling the spatio-temporal relevance, ST-GCN [18] proposes to represent the topology with an adjacency matrix of 3 subsets.”
InfluentialFedGTP: Exploiting Inter-Client Spatial Dependency in Federated Graph-based Traffic Prediction
Yang, L., Chen, W.-K., He, X., Wei, S., Xu, Y., Zhou, Z., & Tong, Y. (2024) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3637528.3671613 · Cited by 23
“For temporal modeling , both recurrent [5, 14, 24] and convolutional neural networks [48, 50] are frequently used.”
InfluentialScaling up Multimodal Pre-Training for Sign Language Understanding
Zhou, W.-G., Zhao, W.-C., Hu, H.-Z., Li, Z.-C., & Li, H.-Q. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3599313 · Cited by 36
“For unpaired samples from Phoenix14 [18], we only feed them into the sign pose branch to participate in masked pose modeling.”
Two-path target-aware contrastive regression for action quality assessment
Ke, X., Xu, H., Lin, X., & Guo, W. (2024) · Information Sciences · doi:10.1016/j.ins.2024.120347 · Cited by 30
Zhou, W., Liu, Y., Zhao, L., Xu, S., & Wang, C. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3314051 · Cited by 40
“Unlike GCN-LSTM models [25], [26], which are commonly used GCN-based spatial-temporal models, ST-GCNN uses GCN to extract features across both spatial and temporal domains, thereby avoiding the parameter inefficiency and expensive cost of recurrent architectures (e.g., LSTM) [27].”
MethodGraph-Based Progressive Fusion Network for Multi-Modality Vehicle Re-Identification
He, Q., Lu, Z., Wang, Z., & Hu, H. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3285758 · Cited by 40
“Yan et al. [21] first applied GCN to human action recognition and proposed a dynamic skeleton model ST-GCN.”
BackgroundDeepfake Video Detection via Facial Action Dependencies Estimation
Tan, L., Wang, Y., Wang, J., Yang, L., Chen, X., & Guo, Y. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i4.25658 · Cited by 23
“By considering the structure of human skeletons, GCNs are adopted to model the movement of linked joints to recognize human actions (Jain et al. 2016; Yan, Xiong, and Lin 2018).”
MethodLocalised Adaptive Spatial-Temporal Graph Neural Network
Duan, W., He, X., Zhou, Z., Thiele, L., & Rao, H. (2023) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3580305.3599418 · Cited by 33
“STGCN [26] composes a spatial-temporal model for skeleton-based action recognition using a 1D convolutional layer and a Partition Graph Convolution (PGC) layer.”
BackgroundMethodDeep Learning for Human Activity Recognition on 3D Human Skeleton: Survey and Comparative Study
Hung-Cuong, N., Nguyen, T.-H., Scherer, R., & Le, V.-H. (2023) · Sensors · doi:10.3390/s23115121 · Cited by 59
“In the study, the authors only present and analyze the ST-GCN (Spatial Temporal Graph Convolutional Networks) [16] and 3D CNNs (3D Convolutional Neural Networks) hybrid with some architecture [17,18].”
MethodMotion Guided Attention Learning for Self-Supervised 3D Human Action Recognition
Yang, Y., Liu, G., & Gao, X. (2022) · IEEE transactions on circuits and systems for video technology (Print) · doi:10.1109/tcsvt.2022.3194350 · Cited by 36
“1) Supervised and Unsupervised Methods: To further validate the advances of our proposed MG-AL, we compare our model with the state-of-the-arts supervised methods [31], [41], [51] and unsupervised methods [9], [10], [15], [46].”
BackgroundMethodModeling the Uncertainty for Self-supervised 3D Skeleton Action Representation Learning
Su, Y., Lin, G., Sun, R., Hao, Y., & Wu, Q. (2021) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3474085.3475248 · Cited by 31
“ST-GCN [51] propose the spatial-temporal graph convolution aim at modeling dynamic skeletons sequences.”
BackgroundMethodInfluentialAnomaly Detection in Multi-Agent Trajectories for Automated Driving
Wiederer, J., Bouazizi, A., Troina, M., Kressel, U., & Belagiannis, V. (2021) · Conference on Robot Learning · doi:10.48550/arxiv.2110.07922 · Cited by 36
“As introduced in [33], graph convolutions are applied on 115 each frame individually first.”
MethodPoseGTAC: Graph Transformer Encoder-Decoder with Atrous Convolution for 3D Human Pose Estimation
Zhu, Y., Xu, X., Shen, F.-M., Ji, Y., Gao, L.-L., & Shen, H. (2021) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2021/188 · Cited by 36
Mixed graph convolution and residual transformation network for skeleton-based action recognition
Liu, S., Bai, X., Fang, M., Li, L., & Hung, C. (2021) · Applied Intelligence · doi:10.1007/s10489-021-02517-w · Cited by 36
“[1] proposed an ST-GCNnetwork and applied graph convolution to human action recognition for the first time, and it was considered a groundbreaking approach for skeleton based action recognition.”
BackgroundMethodRecognizing Actions in Videos from Unseen Viewpoints
Piergiovanni, A., & Ryoo, M. S. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00411 · Cited by 27
“Previous works have studied representations of 3D pose for skeleton action recognition [9, 45, 22, 23], such as joint angles [29] showing the difficulty of this task.”
BackgroundLiu, C., Li, X., Li, Q., Xue, Y.-X., Liu, H.-J., & Gao, Y. (2021) · Neurocomputing · doi:10.1016/j.neucom.2020.10.016 · Cited by 36
Cheng, Y.-B., Chen, X., Zhang, D., & Lin, L. (2021) · ACM International Conference on Multimedia in Asia (MMAsia) · doi:10.1145/3444685.3446289 · Cited by 31
“We follow [31] to perform data pre-processing by repeating the samples up to 300 frames and normalizing their views.”
MethodGround‐Based Cloud Classification Using Task‐Based Graph Convolutional Network
Liu, S., Li, M., Zhang, Z., Cao, X., & Durrani, T. (2020) · Geophysical Research Letters · doi:10.1029/2020gl087338 · Cited by 36
“Recently, graph convolutional networks (GCNs), which are designed to learn graph relations with convolution, have achieved huge success in various tasks, for instance, action recognition (Shi et al., 2019; Yan et al., 2018), text classification (Defferrard et al.”
BackgroundTempel, F., Ihlen, E. A. F., Adde, L., & Strümke, I. (2025) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2025.109838 · Cited by 33
MaskCLR: Attention-Guided Contrastive Learning for Robust Action Representation Learning
Abdelfattah, M., Hassan, M., & Alahi, A. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01767 · Cited by 23
“Significant performance gains were obtained by employing Graph Neural Networks (GCNs) as a feature extractor on heuristically designed fixed graphs, which was first introduced in ST-GCN [48].”
MethodSafety-aware human-centric collaborative assembly
Yi, S., Liu, S.-C., Yang, Y., Yan, S.-J., Guo, D.-Q., Wang, X., & Wang, L.-H. (2024) · Advanced Engineering Informatics · doi:10.1016/j.aei.2024.102371 · Cited by 35
Self-citationDeep Reinforcement Learning in Human Activity Recognition: A Survey and Outlook
Nikpour, B., Sinodinos, D., & Armanfard, N. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3360990 · Cited by 40
“The original Kinetics data have only RGB information from videos, but Yan et al. [79] created a skeleton version of these data using the OpenPose toolbox.”
SBI-DHGR: Skeleton-based intelligent dynamic hand gestures recognition
Narayan, S., Mazumdar, A. P., & Vipparthi, S. (2023) · Expert systems with applications · doi:10.1016/j.eswa.2023.120735 · Cited by 35
Chi, S., Chi, H., Huang, Q., & Ramani, K. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3466212 · Cited by 33
Unraveling a Decade: A Comprehensive Survey on Isolated Sign Language Recognition
Sarhan, N., & Frintrop, S. (2023) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw60793.2023.00345 · Cited by 32
“In recent research, there is a noticeable trend towards skeleton-based methods, following the progress in human action recognition with spatial-temporal Graph Convolutional Networks (GCN) [55, 42].”
BackgroundMulti-cue temporal modeling for skeleton-based sign language recognition
Özdemir, O., Baytaş, İ. M., & Akarun, L. (2023) · Frontiers in Neuroscience · doi:10.3389/fnins.2023.1148191 · Cited by 39
“Following Yan et al. (2018), we construct a spatio-temporal skeleton graph G = ( V , E ) using upper body and hand skeleton joints with 35 unique joints for the skeleton sequence of length T .”
BackgroundMethodInfluentialAdaptive dependency learning graph neural networks
Sriramulu, A., Fourrier, N., & Bergmeir, C. (2023) · Information Sciences · doi:10.1016/j.ins.2022.12.086 · Cited by 34
“In spatio-temporal GNNs, the graph convolution module models the spatial dependencies while 1D convolution layers model the temporal patterns in the data [44].”
BackgroundLeveraging Spatio-Temporal Dependency for Skeleton-Based Action Recognition
Lee, J., Lee, M., Cho, S., Woo, S.-M., Jang, S.-J., & Lee, S. (2022) · IEEE International Conference on Computer Vision · doi:10.1109/iccv51070.2023.00941 · Cited by 35
“1 compares the temporal flows of the spatio-temporal GCN for existing methods [2, 24, 34] and our proposed method.”
BackgroundMethodInfluentialLe, V.-H. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-13921-w · Cited by 40
“…baseline of deep learning methods for human segmentation (Mask R-CNN [40], PointRend [57], TridentNet [68], TensorMask [8], CenterMask [66]) and tracking, 3D-HPE (RepNet [131], MediaPipe Pose [3], Lifting from the Deep [128], V2V-PoseNet [3]), and HAR (ST-GCN [151], DD-net [152], PA-GesGCN [120]).”
BackgroundMethodZhu, A., Ke, Q.-H., Gong, M.-M., & Bailey, J. (2022) · IEEE Workshop/Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00598 · Cited by 35
Two-Person Graph Convolutional Network for Skeleton-Based Human Interaction Recognition
Li, Z., Li, Y., Tang, L., Zhang, T., & Su, J. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3232373 · Cited by 38
Graph-Time Convolutional Neural Networks: Architecture and Theoretical Analysis
Sabbaqi, M., & Isufi, E. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3311912 · Cited by 31
Skeleton-based Action Recognition via Adaptive Cross-Form Learning
Wang, X., Dai, Y., Gao, L.-L., & Song, J.-K. (2022) · ACM Multimedia · doi:10.1145/3503161.3547811 · Cited by 35
“To date, recent skeleton-based action recognition methods have achieved great progress and can be roughly divided into three categories: 1) Recurrent Neural Network (RNN) based methods [12, 21], 2) Convolutional Neural Network (CNN) based methods [4] and 3) Graph Convolutional Network (GCN) based methods [1, 3, 7, 14,…”
BackgroundMethodInfluentialSpatial Attention-Based 3D Graph Convolutional Neural Network for Sign Language Recognition
Al-Hammadi, M. H., Bencherif, M., Alsulaiman, M., Muhammad, G., Mekhtiche, M., Abdul, W., Alohali, Y. A., Alrayes, T. S., Mathkour, H., Faisal, M., et al. (2022) · Italian National Conference on Sensors · doi:10.3390/s22124558 · Cited by 35
“Furthermore, we investigated three partitioning strategies suggested by [29] for more enhancement of spatial representation.”
InfluentialA Survey on the Use of Graph Convolutional Networks for Combating Fake News
Varlamis, I., Michail, D., Glykou, F., & Tsantilas, P. (2022) · Future Internet · doi:10.3390/fi14030070 · Cited by 35
“Graph convolutions can be applied in any type of data that can be represented as a graph, including text corpora [25], video or images [26], in an attempt to capture the intrinsic semantic information hidden in the documents or the spatio-temporal information hidden in the consecutive images of a video.”
BackgroundGE-STDGN: a novel spatio-temporal weather prediction model based on graph evolution
Ni, Q., Wang, Y., & Fang, Y. (2021) · Applied Intelligence · doi:10.1007/s10489-021-02824-2 · Cited by 32
“GNN is a neural network model for analyze the graph structure data in real life and has emerged in many fields [32, 33].”
BackgroundSkeleton Graph Scattering Networks for 3D Skeleton-based Human Motion Prediction
Li, M., Chen, S.-H., Liu, Z., Zhang, Z., Xie, L.-X., Tian, Q., & Zhang, Y. (2021) · 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · doi:10.1109/iccvw54120.2021.00101 · Cited by 35
“Graphs represent numerous data with non-grid structures by depicting the vertices relations [56, 46], which could be leveraged in various scenarios such as social networks [49], human behaviors [45, 55, 60, 56, 33, 47, 19, 59, 5, 29, 10, 36], and dynamic system analysis [26, 58, 21, 28, 32, 20].”
BackgroundhSMAL: Detailed Horse Shape and Pose Reconstruction for Motion Pattern RecognitionPreprint
Li, C., Ghorbani, N., Broomé, S., Rashid, M., Black, M. J., Hernlund, E., Kjellström, H., & Zuffi, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.10102 · Cited by 21
“This is done by training separate ST-GCN classifiers, on the sequence of raw 3D mocap point posi- tions, and the sequence of pose parameters and joint positions extracted from the fitted hSMAL model from side view reconstruction in each frame.”
MethodInfluentialMa, C.-Y., Fan, J.-M., Yao, J.-Y., & Zhang, T. (2021) · Applied Sciences · doi:10.3390/app11104426 · Cited by 35
“Another contribution was inspired by the application of graph convolution neural networks in the field of skeleton-based human action recognition [6].”
BackgroundMethodWang, H., Yang, L., Rong, X., Feng, J., & Tian, Y. (2021) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00381 · Cited by 28
“[38] further proposed a 3D graph convolution model to directly learn from the 3D point cloud sequence, which allows direct temporal inferring and decreases the convolution operation size.”
BackgroundTopology-learnable graph convolution for skeleton-based action recognition
Zhu, G.-M., Zhang, L., Li, H.-S., Shen, P.-Y., Shah, S. A. A., & Bennamoun (2020) · Pattern Recognition Letters · doi:10.1016/j.patrec.2020.05.005 · Cited by 35
Yang, H., Gu, Y., Zhu, J., Hu, K., & Zhang, X. (2020) · IEEE Access · doi:10.1109/access.2020.2964115 · Cited by 46
“Current models for skeletal action recognition [4], [26] use 3D coordinates as features at each vertex.”
BackgroundMethodInfluentialSkeleton based Zero Shot Action Recognition in Joint Pose-Language Semantic Space
Jasani, B. A., & Mazagonwalla, A. (2019) · arXiv.org · doi:10.48550/arxiv.1911.11344 · Cited by 35
“ST-GCN takes as input the time series of body pose of the action performer in the given video, which can be generated using a human body pose detector.”
BackgroundMethodInfluentialAction Machine: Toward Person-Centric Action Recognition in Videos
Zhu, J., Zou, W., Zhu, Z. D., Xu, L., & Huang, G. (2019) · IEEE Signal Processing Letters · doi:10.1109/lsp.2019.2942739 · Cited by 42
Emotion Recognition Based on Multi-View Body Gestures
Shen, Z., Cheng, J., Hu, X., & Dong, Q. (2019) · International Conference on Information Photonics · doi:10.1109/icip.2019.8803460 · Cited by 35
“According to Figure 3, although the overall accuracies are same, TSN and ST-GCN are perform different on different categories.”
MethodInfluentialMulti-View Time-Series Hypergraph Neural Network for Action Recognition
Ma, N., Wu, Z., Feng, Y., Wang, C., & Gao, Y. (2024) · IEEE Transactions on Image Processing · doi:10.1109/tip.2024.3391913 · Cited by 40
“To obtain high-order semantics of human body joints, some existing work [10], [11], [12], [13] have constructed graphs by direct and indirect kinematic dependencies in spatial and temporal, and utilize skeleton-based graph convolutional network (GCN) to overcome illumination and view problems in…”
BackgroundMethodInfluentialMerge-and-Split Graph Convolutional Network for Skeleton-Based Interaction Recognition
Wang, H., Wang, Y., Yan, S., Du, X., Gao, Y., & Liu, H. (2024) · Cyborg and Bionic Systems · doi:10.34133/cbsystems.0102 · Cited by 38
“Yan et al. [23] proposed ST-GCN, which first uses topological maps to extract and aggregate the spatiotemporal information of bones.”
Self-citationHulk: A Universal Knowledge Translator for Human-Centric Tasks
Wang, Y., Wu, Y., He, W., Guo, X., Zhu, F., Bai, L., Zhao, R., Wu, J., He, T., Ouyang, W., et al. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3552604 · Cited by 32
“For example, GCN-based methods [48], [49], [50], [51], [52], [53], [54] and transformer-based methods [55], [56], [57] have achieved satisfactory results in skeleton-based action recognition tasks, as they can effectively model the positional relationships among human joints.”
Joint Skeletal and Semantic Embedding Loss for Micro-gesture Classification
Li, K., Guo, D., Chen, G., Peng, X.-L., & Wang, M. (2023) · MiGA@IJCAI · doi:10.48550/arxiv.2307.10624 · Cited by 34
“Method Modality iMiGUE dataset SMG dataset Top-1 (%) Top-5 (%) Top-1 (%) Top-5 (%) ST-GCN [12] Joint 46.”
Motion Question Answering via Modular Motion Programs
Endo, M., Hsu, J., Li, J., & Wu, J. (2023) · International Conference on Machine Learning · doi:10.48550/arxiv.2305.08953 · Cited by 34
“In recent years, action recognition for human motion has been extensively studied (Yan et al., 2018; Asghari-Esfeden et al., 2020; Caetano et al., 2019; Cai et al., 2021; Chen et al., 2021a; Cheng et al., 2020; Choutas et al., 2018; Du et al., 2015; Ke et al., 2017; Liu et al., 2020; Shi et al.,…”
GaitRef: Gait Recognition with Refined Sequential Skeletons
Zhu, H., Zheng, W., Zheng, Z., & Nevatia, R. (2023) · IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb57857.2023.10448634 · Cited by 38
“For the silhouette feature encoder in GaitMix , we follow ST-GCN [33] for encoding the skeletons into the same embedding dimension N out as the silhouette feature encoder.”
MethodTemporal Pattern Attention for Multivariate Time Series of Tennis Strokes Classification
Skublewska-Paszkowska, M., & Powroźnik, P. (2023) · Italian National Conference on Sensors · doi:10.3390/s23052422 · Cited by 34
“HAR is a challenging task used in numerous applications.”
MethodInfluentialChen, X.-Y., Kan, S.-C., Zhang, F., Cen, Y., Zhang, L.-N., & Zhang, D. (2022) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2022.103707 · Cited by 34
Du, C., Graham, S., Depp, C., & Nguyen, T. Q. (2021) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society · doi:10.1109/embc46164.2021.9629569 · Cited by 34
“Since GCN [21] has been validated to learn features from body landmarks for various applications, such as action recognition [21]–[24], we propose to use a GCN [21] as a feature extractor to efficiently exploit the spatial information of body movements.”
BackgroundMethodInfluentialDecoupling Long- and Short-Term Patterns in Spatiotemporal Inference
Hu, J., Liang, Y., Fan, Z., Liu, L., Yin, Y., & Zimmermann, R. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3293814 · Cited by 23
“They either couple GNNs with RNNs [18, 19] or TCNs [16, 23, 25, 28].”
BackgroundInfluentialRepresenting Videos as Discriminative Sub-graphs for Action Recognition
Li, D., Qiu, Z., Pan, Y., Yao, T., Li, H., & Mei, T. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00332 · Cited by 33
“For example, ST-GCN [51] proposes a spatio-temporal graph to model the structured information among the human joints.”
Learning Three Dimensional Tennis Shots Using Graph Convolutional Networks
Skublewska‐Paszkowska, M., Powroźnik, P., & Łukasik, E. (2020) · Sensors · doi:10.3390/s20216094 · Cited by 34
“The above relationship, for a single frame, can be described in the following equation [6]: where ) and W corresponds to the weight vector.”
BackgroundMethodInfluentialAction Recognition via Pose-Based Graph Convolutional Networks with Intermediate Dense Supervision
Shi, L., Zhang, Y., Cheng, J., & Lu, H.-Q. (2019) · Pattern Recognition · doi:10.1016/j.patcog.2021.108170 · Cited by 34
“There are two streams for constructing GCNs: 1) the spectral-stream approaches formulate the GCNs in the spectral domain, which are based on the analogy between the classical Fourier transforms and projections onto the eigenvectors of the graph Laplacian operator [25, 8]; 2) the spatial-stream approaches provide filte…”
BackgroundMethodInfluential
Metadata
- Authors
- Yan, S., Xiong, Y., & Lin, D.
- Affiliations
- Chinese University of Hong Kong (HK)
- Venue
- Proceedings of the AAAI Conference on Artificial Intelligence
- Published
- 2018-04-27
- Volume
- 32
- Issue
- 1
- ISSN
- 2159-5399 (ISSN-L)
- 2374-3468
- OpenAlex
- W2963076818
- Cited by
- 5,410
- References
- 32
- Citations by year
- 2026: 552 · 2025: 895 · 2024: 915 · 2023: 915 · 2022: 755 · 2021: 742 · 2020: 414 · 2019: 193 · 2018: 29
- FWCI
- 103.84
- Type
- conference paper
- Language
- en
- Open access
- diamond
- doi.org
- Locations
- Proceedings of the AAAI Conference on Artificial Intelligence (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- AAAI Publications (The Association for the Advancement of Artificial Intelligence (AAAI)) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Chinese University of Hong Kong
- SenseTime Group
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Human Pose and Action Recognition
- Physical Sciences › Computer Science › Human-Computer Interaction › Hand Gesture Recognition Systems
- Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Human Motion and Animation
- Keywords
- skeleton-based action recognition, spatio-temporal graph convolutional network, graph convolutional networks, temporal modeling, spatial modeling, generalization
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
