The DOI matches its Crossref record; 37 of the 37 references it cites were found in a registry or an index.
MS2L: Multi-Task Self-Supervised Learning for Skeleton Based Action Recognition
Open accessAbstract
In this paper, we address self-supervised representation learning from human skeletons for action recognition. Previous methods, which usually learn feature presentations from a single reconstruction task, may come across the overfitting problem, and the features are not generalizable for action recognition. Instead, we propose to integrate multiple tasks to learn more general representations in a self-supervised manner. To realize this goal, we integrate motion prediction, jigsaw puzzle recognition, and contrastive learning to learn skeleton features from different aspects. Skeleton dynamics can be modeled through motion prediction by predicting the future sequence. And temporal patterns, which are critical for action recognition, are learned through solving jigsaw puzzles. We further regularize the feature space by contrastive learning. Besides, we explore different training strategies to utilize the knowledge from self-supervised tasks for action recognition. We evaluate our multi-task self-supervised learning approach with action classifiers trained under different configurations, including unsupervised, semi-supervised and fully-supervised settings. Our experiments on the NW-UCLA, NTU RGB+D, and PKUMMD datasets show remarkable performance for action recognition, demonstrating the superiority of our method in learning more discriminative and general features. Our project website is available at https://langlandslin.github.io/projects/MSL/.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Extensive results on NTU-RGB+D [40, 27] datasets demonstrate that our method indeed boosts the 3D action representation learning benefiting from cross-view consistency.”
From 3D Human Action Representation Learning via Cross-View Consistency Pursuit (2021) · cites this paper as [27]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“As shown in Table 4, for a single stream (i.e., joint stream), our AimCLR outperforms all other methods (Zheng et al. 2018; Lin et al. 2020; Rao et al. 2021; Su, Liu, and Shlizerman 2020; Nie, Liu, and Liu 2020; Li et al. 2021).”
From Contrastive Learning from Extremely Augmented Skeleton Sequences for Self-Supervised Action Recognition (2022) · cites this paper as Lin et al. 2020
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“The masking strategy in MAE only focuses on the spatial domain.”
From SkeletonMAE: Spatial-Temporal Masked Autoencoders for Self-Supervised Skeleton Action Recognition (2023) · cites this paper as [15]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“Notably, MAE experiment also shows that using a more random masking strategy is more beneficial to the final finetuning result.”
From SkeletonMAE: Spatial-Temporal Masked Autoencoders for Self-Supervised Skeleton Action Recognition (2023) · cites this paper as [15]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“The Vision Transformer (ViT) [10] architecture is used in MAE [15] to process the image input.”
From SkeletonMAE: Spatial-Temporal Masked Autoencoders for Self-Supervised Skeleton Action Recognition (2023) · cites this paper as [15]
Experimental — accuracy not yet measured
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
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 1×
“LSTM [13] 71.”
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition
Yan, S., Xiong, Y., & Lin, D. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.12328 · Cited by 5,410 · cited 5×
“30 ST-GCN [43] 81.”
BackgroundMethodInfluentialDimensionality Reduction by Learning an Invariant Mapping
Hadsell, R., Chopra, S., & LeCun, Y. (2006) · 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2 (CVPR'06) · doi:10.1109/cvpr.2006.100 · Cited by 5,336 · cited 1×
“search. Demands on action recognition are growing rapidly to facilitate the applications such as video surveillance, human-computer interaction, video understanding. In the past few years, many works [10, 38, 42] have been developed on recognition based on RGB videos and achieved many significant results. However, pro…”
BackgroundZhang, R., Isola, P., & Efros, A. A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46487-9_40 · Cited by 3,178 · cited 2×
“There are also other pretext tasks, such as colorizing grayscale images [47] or predicting image rotation angles [8, 44].”
BackgroundNTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis
Shahroudy, A., Liu, J., Ng, T.-T., & Wang, G. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.115 · Cited by 3,148 · cited 2×
“NTU RGB+D Dataset (NTU) [27] This is a large scale dataset including 56, 578 videos with 60 action labels and 25 joints for each body, including interactions with pairs and individual activities.”
MethodSelf-citationUnsupervised Learning of Visual Representations by Solving Jigsaw Puzzles
Noroozi, M., & Favaro, P. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46466-4_5 · Cited by 2,775 · cited 2×
“Following the idea, the works in [23, 24, 41] predict a permutation of multiple shuffled image patches to better model spatial relationships, which are called jigsaw puzzles.”
BackgroundMethodWhy Does Unsupervised Pre-training Help Deep Learning?
Erhan, D., Bengio, Y., Courville, A. C., Manzagol, P.-A., Vincent, P., & Samy Bengio, A. (2010) · Journal of Machine Learning Research · Cited by 2,631
Unsupervised Visual Representation Learning by Context Prediction
Doersch, C., Gupta, A., & Efros, A. A. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.167 · Cited by 2,629 · cited 2×
“[3] proposed to train a convolutional neural network to reorder perturbed image patches.”
BackgroundTwo-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition
Shi, L., Zhang, Y., Cheng, J., & Lu, H. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.01230 · Cited by 2,071 · cited 1×
“ion of the skeleton data, Yan et al. [43] employed graph convolution networks to represent skeleton data on graphs. To make the graph representation more flexible, attention mechanisms are applied in [28, 29] to adaptively capture discriminative features based on spatial configurations and temporal dynamics. Though th…”
BackgroundHuman Action Recognition by Representing 3D Skeletons as Points in a Lie Group
Vemulapalli, R., Arrate, F., & Chellappa, R. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.82 · Cited by 1,517 · cited 1×
“ve pairs. 2.2 Skeleton-Based Action Recognition Early skeleton-based action recognition methods are generally based on hand-crafted features by utilizing the geometry relationships of skeletal joints [9, 20, 34, 35, 38]. Recent methods for skeleton-based action recognition pay more attention to utilizing deep networks…”
MethodInfluential
Research works printed but not linked: 14
- Ting Chen Simon Kornblith Mohammad Norouzi and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. arXiv preprint arXiv:2002.05709 (2020). Ting Chen Simon Kornblith Mohammad Norouzi and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. arXiv preprint arXiv:2002.05709 (2020). · arXiv
- Du Yong, 2015, Proc. IEEE Conference on Computer Vision and Pattern Recognition. 1110--1118
- Gidaris Spyros, 2018, Proc. Int'l Conference on Learning Representations.
- Hadsell R., Proc. IEEE Conference on Computer Vision and Pattern Recognition. 1735--1742
- He Kaiming, 2019, Proc. IEEE Int?l Conference on Computer Vision. 4918--4927
- Eric Jang Coline Devin Vincent Vanhoucke and Sergey Levine. 2018. Grasp2vec: Learning object representations from self-supervised grasping. arXiv preprint arXiv:1811.06964 (2018). Eric Jang Coline Devin Vincent Vanhoucke and Sergey Levine. 2018. Grasp2vec: Learning object representations from self-supervised grasping. arXiv preprint arXiv:1811.06964 (2018). · arXiv
- Sebastian Ruder. 2017. An overview of multi-task learning in deep neural networks. arXiv preprint arXiv:1706.05098 (2017). Sebastian Ruder. 2017. An overview of multi-task learning in deep neural networks. arXiv preprint arXiv:1706.05098 (2017). · arXiv
- Song Sijie, 2017, Proc. AAAI Conference on Artificial Intelligence. 4263--4270
- Su Kun, 2019, PREDICT CLUSTER: Unsupervised Skeleton Based Action Recognition. ArXiv abs/1911.12409 · arXiv
- Vondrick Carl, 2016, Proc. Advances in Neural Information Processing Systems. 613--621
- Yan Sijie, 2018, Proc. AAAI Conference on Artificial Intelligence. 7444--7452
- Zhai Xiaohua, 2019, Proc. IEEE Int'l Conference on Computer Vision. 1476--1485
- Zheng Nenggan, 2018, Proc. AAAI Conference on Artificial Intelligence. 2644--2651
- Zhu Wentao, 2016, Proc. AAAI Conference on Artificial Intelligence. 3697--3703
Cited by
MotionBERT: A Unified Perspective on Learning Human Motion Representations
Zhu, W., Ma, X., Liu, Z., Liu, L., Wu, W., & Wang, Y. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01385 · Cited by 348
“From the action recognition perspective, a variety of pretext tasks are designed to learn motion representations in a self-supervised manner, including future prediction [100], jigsaw puzzle [60], skeletoncontrastive [107], speed change [101], cross-view consistency [62], and contrast-reconstruction [117].”
Background3D Human Action Representation Learning via Cross-View Consistency Pursuit
Li, L., Wang, M., Ni, B., Wang, H., Yang, J., & Zhang, W. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00471 · Cited by 236
“MS(2)L [26] proposed a multi-task learning scheme for action representation learning.”
BackgroundMethodInfluentialRao, H., Xu, S., Hu, X., Cheng, J., & Hu, B. (2021) · Information Sciences · doi:10.1016/j.ins.2021.04.023 · Cited by 235
GMSS: Graph-Based Multi-Task Self-Supervised Learning for EEG Emotion Recognition
Li, Y., Chen, J., Li, F., Fu, B., Wu, H., Ji, Y., Zhou, Y., Niu, Y., Shi, G., & Zheng, W. (2023) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2022.3170428 · Cited by 207
“In addition, the contrastive learning method was applied in the field of video processing, and achieved excellent performance at the time it was proposed [52], [53].”
MethodKumar, P., Chauhan, S., & Awasthi, L. K. (2024) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-023-09986-x · Cited by 206
“MS2L [321] integrates multiple pretext tasks in order to learn a better representation, in contrast to the previously mentioned methods which only adopt a single reconstruction task.”
Metadata
- Authors
- Lin, L., Song, S., Yang, W., & Liu, J.
- Affiliations
- Peking University (CN)
- Venue
- Proceedings of the 28th ACM International Conference on Multimedia
- Publisher
- ACM
- Published
- 2020-10-12
- Pages
- 2490-2498
- OpenAlex
- W3092424783
- Cited by
- 223
- References
- 33
- Citations by year
- 2026: 18 · 2025: 53 · 2024: 44 · 2023: 54 · 2022: 37 · 2021: 16 · 2020: 1
- FWCI
- 8.32
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- ACM International Conference on Multimedia (ACM MM) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Human Pose and Action Recognition
- Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Human Motion and Animation
- Physical Sciences › Computer Science › Human-Computer Interaction › Hand Gesture Recognition Systems
- Keywords
- self-supervised representation learning, human action recognition, multi-task learning, motion prediction, contrastive learning, overfitting, temporal patterns, semi-supervised learning, unsupervised learning
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: MS2L: Multi-Task Self-Supervised Learning for Skeleton Based Action Recognition. https://app.cochat.ai/papers/10.1145/3394171.3413548
- The paper
- Lin, L., Song, S., Yang, W., & Liu, J. (2020). MS2L: Multi-Task Self-Supervised Learning for Skeleton Based Action Recognition. Proceedings of the 28th ACM International Conference on Multimedia. https://doi.org/10.1145/3394171.3413548
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.