A Unified Multimodal De- and Re-Coupling Framework for RGB-D Motion Recognition
Open accessThe DOI matches its Crossref record; 86 of the 86 references it cites were found in a registry or an index.
Abstract
Motion recognition is a promising direction in computer vision, but the training of video classification models is much harder than images due to insufficient data and considerable parameters. To get around this, some works strive to explore multimodal cues from RGB-D data. Although improving motion recognition to some extent, these methods still face sub-optimal situations in the following aspects: (i) Data augmentation, i.e., the scale of the RGB-D datasets is still limited, and few efforts have been made to explore novel data augmentation strategies for videos; (ii) Optimization mechanism, i.e., the tightly space-time-entangled network structure brings more challenges to spatiotemporal information modeling; And (iii) cross-modal knowledge fusion, i.e., the high similarity between multimodal representations leads to insufficient late fusion. To alleviate these drawbacks, we propose to improve RGB-D-based motion recognition both from data and algorithm perspectives in this article. In more detail, firstly, we introduce a novel video data augmentation method dubbed ShuffleMix, which acts as a supplement to MixUp, to provide additional temporal regularization for motion recognition. Secondly, a Unified Multimodal De-coupling and multi-stage Re-coupling framework, termed UMDR, is proposed for video representation learning. Finally, a novel cross-modal Complement Feature Catcher (CFCer) is explored to mine potential commonalities features in multimodal information as the auxiliary fusion stream, to improve the late fusion results. The seamless combination of these novel designs forms a robust spatiotemporal representation and achieves better performance than state-of-the-art methods on four public motion datasets. Specifically, UMDR achieves unprecedented improvements of ↑ 4.5% on the Chalearn IsoGD dataset.
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
“However, existing works on missing modality robustness vary widely in terms of model architectures [32, 4], fusion methods [35, 63], training procedures [20, 32], and missing feature generation methods [23]. Due to this heterogeneity, a fair comparison based solely on model size/number of parameters is infeasible.”
From Robust Multimodal Learning with Missing Modalities via Parameter-Efficient Adaptation (2025) · cites this paper as [63]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“As shown in Table 5, our adaptation performs better compared to recent modality masking and generation based approach ActionMAE [23] and modality de- and re-coupling based approaches Motion-RGBD [91] and UMDR [63].”
From Robust Multimodal Learning with Missing Modalities via Parameter-Efficient Adaptation (2025) · cites this paper as [63]
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
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472 · cited 1×
“To achieve the goal in the former aspect, inspired by self-attention [46], we calculate dynamic attention maps based on the input to re-weight spatial features of each frame to highlight sensitive filters in it.”
MethodDistilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204
A survey on Image Data Augmentation for Deep Learning
Shorten, C., & Khoshgoftaar, T. M. (2019) · Journal of Big Data · doi:10.1186/s40537-019-0197-0 · Cited by 13,203 · cited 1×
“Data augmentation, a data-driven and informed regularization strategy that artificially increases the number of training samples to improve the generalization ability of models, has been widely explored for better training of deep neural networks [20], [21], [22], [23].”
BackgroundQuo Vadis, Action Recognition? A New Model and the Kinetics Dataset
Carreira, J., & Zisserman, A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.502 · Cited by 9,570
mixup: Beyond Empirical Risk Minimization
Zhang, H., Cisse, M., Dauphin, Y. N., & Lopez-Paz, D. (2017) · arXiv · doi:10.48550/arxiv.1710.09412 · Cited by 5,831 · cited 3×
“The latter refers to utilizing multiple samples to generate a new instance, including video MixUp [14] and CutMix [28].”
BackgroundMethodCutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features
Yun, S., Han, D., Chun, S., Oh, S. J., Yoo, Y., & Choe, J. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00612 · Cited by 5,400 · cited 1×
“The latter refers to utilizing multiple samples to generate a new instance, including video MixUp [14] and CutMix [28].”
BackgroundSGDR: Stochastic Gradient Descent with Warm Restarts
Loshchilov, I., & Hutter, F. (2016) · arXiv · doi:10.48550/arxiv.1608.03983 · Cited by 5,309 · cited 1×
“The learning rate is linearly ramped up to 0.01 during the first 5 epochs and then decayed with a cosine schedule [63].”
BackgroundInfluentialInception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11231 · Cited by 4,636 · cited 1×
“Similar to [45], the Steam module is the initial set of operations used to map low-level visual embedding before executing to the SMS module, written as: where Θ represents the learnable parameter matrices.”
MethodSlowFast Networks for Video Recognition
Feichtenhofer, C., Fan, H., Malik, J., & He, K. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00630 · Cited by 4,007 · cited 1×
“For each block, inspired by [11], [47], we construct multi-branch structures to tackle inputs with different temporal resolutions and feature dimensions accordingly: fast ( R ), moderate ( R T × C ), and slow ( R ), to perceive different motion variations, i.e., the high temporal resolution…”
MethodTemporal Segment Networks: Towards Good Practices for Deep Action Recognition
Wang, L., Xiong, Y., Wang, Z., Qiao, Y., Lin, D., Tang, X., & Van Gool, L. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46484-8_2 · Cited by 3,958
Research works printed but not linked: 1
- [82] O. K¨ op ¨ ukl¨ u, A. Gunduz, N. Kose, and G. Rigoll, 'Real-time hand gesture detection and classification using convolutional neural networks,' in 2019 14th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2019) . IEEE, 2019, pp. 1-8. -
Cited by
Gloss-free Sign Language Translation: Improving from Visual-Language Pretraining
Zhou, B., Chen, Z., Clapés, A., Wan, J., Liang, Y., Escalera, S., Lei, Z., & Zhang, D. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01908 · Cited by 236
“The goal of ISLR is to translate an isolated sign into a corresponding single sign language word [14, 15, 20, 22], which is somewhat similar to the isolated gesture recognition task [33, 18, 9, 38, 40, 39, 36].”
BackgroundSelf-citationRobust Multimodal Learning with Missing Modalities via Parameter-Efficient Adaptation
Reza, M. K., Prater-Bennette, A., & Asif, M. S. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3476487 · Cited by 78
Zhang, Y., & Wang, Y. (2025) · EURASIP Journal on Image and Video Processing · doi:10.1186/s13640-025-00677-0 · Cited by 25
Li, Y., Chen, H., Feng, G., & Miao, Q. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01917 · Cited by 24
“Although great progress has been made in this field, the attention has always been drawn to boost performance via better network structure [50, 48, 55, 56] or by introducing extra modalities of data [6, 18, 23, 19].”
BackgroundClassroom Behavior Detection Method Based on PLA-YOLO11n
Zhang, H., Zhou, G., He, W., & Deng, H. (2025) · Sensors · doi:10.3390/s25175386 · Cited by 14
“…multi-scale receptive fields; and (3) the integration of attention mechanisms to dynamically focus on key regions (e.g., hand-raising, head-down, interactive gestures) while suppressing background noise, thereby enhancing the accuracy and stability of action detection and classification [7–9].”
Metadata
- Authors
- Zhou, B., Wang, P., Wan, J., Liang, Y., & Wang, F.
- Affiliations
- Macau University of Science and Technology (MO)
- Amazon (United States) (US)
- Chinese Academy of Sciences (CN)
- Alibaba Group (United States) (US)
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2023-10
- Volume
- 45
- Issue
- 10
- Pages
- 11428-11442
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- 1939-3539
- OpenAlex
- W4376132550
- Cited by
- 38
- References
- 86
- Citations by year
- 2026: 7 · 2025: 19 · 2024: 9 · 2023: 3
- FWCI
- 3.45
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Key Research and Development Plan (2021YFE0205700)
- External cooperation key project of Chinese Academy Sciences (173211KYSB20200002)
- Science and Technology Development Fund of Macau (0123/2022/A3, 0070/2020/AMJ, 0004/2020/A1)
- Guangdong Provincial Key R&D Programme (2019B010148001)
- Open Research Projects of Zhejiang Lab (2021KH0AB07)
- CCF-Zhipu AI Large Model OF (202219)
- Alibaba 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
- multimodal fusion, mixup, spatio-temporal representation learning, decision fusion
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
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.