.png){kind=icon}
返回
Dynamic spatial-temporal topology graph network for skeleton-based action recognition
.png){kind=icon}
PRE
AI
DOI：10.1007/s00530-024-01531-5.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Over the past few years, skeleton-based action recognition has gained significant attention for its simple yet robust representation of the human body structure. Many researchers have employed Graph Convolutional Network (GCN) to explore discriminative features of skeletons, achieving notable success. Nevertheless, conventional GCNs may be unable to efficiently capture joint dependencies in spatial-temporal dimensions due to their shared topology for different frames and channels, which are crucial for action understanding. To address this issue, we propose a Dynamic Spatial-Temporal Topology Graph Network (DST-GNet), which generates distinct topologies for different frames and channels, exploring abundant joint correlations of action sequences. Specifically, a data-driven Temporal Topology Matrix Learning (TTML) module is designed to capture intrinsic body correlations in each frame, boosting the model's generalization and flexibility. A Temporal-Specific Aggregation (TSA) module is also proposed for temporal-specific feature aggregation. Furthermore, we introduce an optimized Multi-Level Temporal Aggregation (MLTA) module to aggregate motion information along the temporal dimension, reducing the model complexity significantly. Experimental results demonstrate that our DST-GNet achieves comparable performance on three large-scale 3D human action datasets: NTU-RGB+D, NTU-RGB+D 120 and Northwestern-UCLA.
Keyword:
3D skeleton-based action recognition
Graph convolutional network
Spatial-temporal feature
Graph inference
期刊
IF:
3.1
论文数:
2.8K
被引数:
2.7K
机构
引用论文
Evaluation of the Effect of Systolic Blood Pressure and Pulse Pressure on Cognitive Function: The Women's Health and Aging Study II.png)
PLoS ONE
OAAI
PLoS ONEIF0
OAAI分享收藏Crystalline‐State Reaction with Allosteric Effect in Spin‐Crossover, Interpenetrated Networks with Magnetic and Optical Bistability具有磁和光学双稳态的自旋交叉，互穿网络中具有变构效应的晶态反应
PREAI
分享收藏分享收藏
查看更多内容