Dynamic Channel-Aware Subgraph Interactive Networks for Skeleton-Based Action Recognition

Wu, Kunlun; Gong, Xun

doi:10.1109/lsp.2022.3227816

Cited by 4 publications

(5 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the CNN-based method, the recognition accuracy of our model is 1% higher than the MSSTNet [39] . Among the GCN-based methods, our model improves the recognition accuracy by 6.7%, 0.5%, 2.4% and 0.2%, respectively, compared with 2s-AGCN [7], DCA-SGIN [43], IA-ASGCN [42] and ACC-GCN [44]. The comparison results of three datasets demonstrate that our model has superior recognition performance and strong generalization on datasets of various scales.…”

Section: E Comparison With Other Methodsmentioning

confidence: 86%

“…X-view(%) ST-GCN [6] 81.5 88.3 3.10 2s-AGCN [7] 88.5 95.1 6.94 DCA-SGIN [43] 87.2 88.7 -AS-GCN [22] 86.8 94.2 9.50 MS-AAGCN [23] 90.0 96.2 15.04 DC-GCN [41] 91.1 96.7 -FLAGCN [37] 89.4 94.8 -CTR-GCN [24] 92.4 96.8 1.44 DGNN [8] 89.9 96.1 26.24 Dynamic GCN [10] 91.5 96.0 14.4 MSSTNet [39] 89.6 95.3 39.6 2s-ICE-GCN [40] 92.0 96.2 -Ours 92.5 96.9 1.00 Methods NTU RGB+D 120 Year X-sub(%) X-set(%) ST-GCN [6] 70.7 73.2 2018 2s-AGCN [7] 82.5 84.2 2019 AS-GCN [22] 77.9 78.5 2019 Dynamic GCN [10] 87.3 88.6 2020 CTR-GCN [24] 88.9 90.6 2021 IA-ASGCN [42] 85.4 87.4 2022 DCA-SGIN [43] 87.2 88.7 2022 DC-GCN [41] 87.1 88.6 2023 MSSTNet [39] 85.3 86.0 2023 2s-ICE-GCN [40] 89.1 90.2 2023 DC-GCN [41] 87.1 88.6 2023 Ours 88.9 90.6…”

Section: Methods Ntu Rgb+d Params(m) X-sub(%)mentioning

confidence: 99%

“…Northwestern-UCLA Year Accuracy(%) Ensemble TS-LSTM [46] 89.2 2017 2s-AGC-LSTM [45] 93.3 2019 2s-AGCN [24] 89.6 2019 DCA-SGIN [43] 95.8 2022 IA-ASGCN [42] 93.9 2022 MSSTNet [39] 95.3 2023 ACC-GCN [44] 96.…”

Section: Methodsmentioning

confidence: 99%

“…And we do comparisons of model parameters on NTU RGB+D dataset, which proves the lightweight of our model. The compared models are GCN-based methods including [6][7][8] [10] [22][23][24] [37] [40][41][42][43][44], CNN-based method including [39], and RNN-based methods including [45][46]. For a comprehensive comparison, the metrics reported by EfficientGCN [47] are used for the evaluation, and the results are shown in Tables 5, 6, and 7.…”

Section: E Comparison With Other Methodsmentioning

confidence: 99%

See 3 more Smart Citations

Multi-Scale Adaptive Graph Convolution Network for Skeleton-Based Action Recognition

Hu,

Fang,

Han

et al. 2024

IEEE Access

View full text Add to dashboard Cite

The skeleton-based action recognition technology can effectively avoid the background interference and occlusion problems in the image. However, the recognition of similar actions is still a challenge. In this paper, a multi-scale dynamic topological modeling method (MDTM) is proposed to solve this problem. The topological modeling through the convolution kernel generated from the original data, increases the connection of the convolution process to the original data compared to the previous randomly generated ones, effectively distinguishing similar actions. In addition, MDTM uses a multi-scale temporal convolutional network to obtain a wider receptive field, which can effectively extract the temporal information in the action. At the same time, a dynamic topology learning method is utilized to design a spatiotemporal information extractor that can effectively extract the spatiotemporal information in the action to dynamically adjust the topology structure. Extensive experiments have performed on three large-scale datasets, NTU RGB + D 60, NTU RGB + D 120, and NW-UCLA to validate the effectiveness of MDTM. The results show that MART-GCN performs better than the others in terms of accuracy and number of parameters.

show abstract

Section: E Comparison With Other Methodsmentioning

confidence: 86%

Section: Methods Ntu Rgb+d Params(m) X-sub(%)mentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: E Comparison With Other Methodsmentioning

confidence: 99%

See 2 more Smart Citations

Multi-Scale Adaptive Graph Convolution Network for Skeleton-Based Action Recognition

Hu,

Fang,

Han

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For example, obscured hands or legs, this type of motion information is easily lost. For the skeleton-based approaches [10][11][12][13][14][15][16], lightweight skeleton data can make the model computationally less costly, but it is easy to misunderstand similar motion trajectory of the action without image visual information.…”

Section: Introductionmentioning

confidence: 99%

3sG: Three‐stage guidance for indoor human action recognition

Nan,

Ye,

et al. 2024

IET Image Processing

View full text Add to dashboard Cite

Inference using skeleton to steer RGB videos is applicable to fine‐grained activities in indoor human action recognition (IHAR). However, existing methods that explore only spatial alignment are prone to bias, resulting in limited performance. The authors propose a Three‐stage Guidance (3sG) framework, leveraging skeleton knowledge to promote RGB in three stages. First, a soft shading image is proposed for alleviating background noise in videos, allowing the network to directly focus more on the motion region. Second, the authors propose to extract RGB frames of interest to reduce the computational effort. Furthermore, to explore more fully the complementary information between skeletons and RGB, the skeleton is coupled to the frame representation in a different spatial–temporal sharing pattern. Third, the global skeleton and skeleton‐guided RGB features are fed into the shared classifiers, which approximate the logit distributions of the two to enhance the performance in RGB unimodal. Finally, a fusion strategy that utilizes two learnable parameters to adaptively integrate the skeleton with the RGB is proposed. 3sG outperforms the state‐of‐the‐art results on the Toyota Smarthome dataset while it is more efficient than similar methods on the NTU RGB+D dataset.

show abstract