期刊
2022 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, ICIP
卷 -, 期 -, 页码 3301-3305出版社
IEEE
DOI: 10.1109/ICIP46576.2022.9897522
关键词
Hyperbolic geometry; dynamic graphs; graph convolutional networks; human action recognition
资金
- Academy of Finland [336116, 345122, 328115]
This work introduces compact hyperbolic space ST-GCNs, which outperform their corresponding Euclidean counterparts, improve the performance of large Euclidean models, reduce the total number of model parameters and model size. Experimental results demonstrate the promising performance of these hyperbolic networks in human action recognition tasks.
Spatial-temporal graph convolutional networks (ST-GCNs) have been successfully applied for dynamic graphs representation learning, such as modeling skeleton-based human actions. However, ST-GCNs embed these non-Euclidean graph structures into Euclidean space, which is not the natural space to represent such structures as embedding them in this space incurs a large distortion. In this work, we make use of hyperbolic non-Euclidean geometry and construct compact ST-GCNs in the hyperbolic space. It can be shown that hyperbolic ST-GCNs (HST-GCNs) outperform the corresponding Euclidean counterparts. Additionally, these compact hyperbolic models can be used to increase the performance of large complex Euclidean models. Moreover, we show that the same or even better performance of large Euclidean models can be achieved by fusing the scores of smaller Euclidean models and a compact hyperbolic model. This in turn leads to reducing the total number of model parameters and hence model size. To validate the performance of these hyperbolic networks, we conducted extensive experiments on NTU RGB+D, NTU RGB+D 120 and Kinectics-Skeleton datasets for human action recognition.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据