arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2007.14690cs.CV
收藏

Dynamic GCN:用于基于骨架的动作识别的上下文丰富拓扑学习

Dynamic GCN: Context-enriched Topology Learning for Skeleton-based Action Recognition

Hikvision Research Institute(海康威视研究院);College of Information Science & Electronic Engineering, Zhejiang University(浙江大学信息科学与电子工程学院)

Fanfan Ye, Shiliang Pu, Qiaoyong Zhong, Chao Li, Di Xie, Huiming Tang

摘要:

图卷积网络(GCN)在基于图的动作识别中引起了越来越多的关注。其关键在于图结构的设计,它对骨架拓扑信息进行编码。在本文中,我们提出了动态GCN,其中一个新的卷积神经网络命名为上下文编码网络(CeN)的自动学习骨架拓扑结构。特别地,当学习两个关节之间的依赖性时,来自其余关节的上下文特征以全局方式并入。CeN非常轻量级但有效,可以嵌入到图卷积层中。通过堆叠多个支持CEN的图卷积层,我们构建了动态GCN。值得注意的是,作为Ceon的优点,动态图拓扑结构是针对不同的输入样本以及各种深度的图卷积层构建的。此外,三个可选的上下文建模架构进行了很好的探索,这可能是未来的研究图拓扑学习的指导方针。CeN仅为基线模型带来约7%的额外FLOP,而动态GCN实现了更好的性能,与现有方法相比,$2\times$的FLOP少了$4\times$。通过进一步结合静态物理身体连接和运动模态,我们在三个大规模基准测试中实现了最先进的性能,即NTU-RGB+D,NTU-RGB+D 120和Atrieton-Kinetics。

英文摘要:

Graph Convolutional Networks (GCNs) have attracted increasing interests for the task of skeleton-based action recognition. The key lies in the design of the graph structure, which encodes skeleton topology information. In this paper, we propose Dynamic GCN, in which a novel convolutional neural network named Contextencoding Network (CeN) is introduced to learn skeleton topology automatically. In particular, when learning the dependency between two joints, contextual features from the rest joints are incorporated in a global manner. CeN is extremely lightweight yet effective, and can be embedded into a graph convolutional layer. By stacking multiple CeN-enabled graph convolutional layers, we build Dynamic GCN. Notably, as a merit of CeN, dynamic graph topologies are constructed for different input samples as well as graph convolutional layers of various depths. Besides, three alternative context modeling architectures are well explored, which may serve as a guideline for future research on graph topology learning. CeN brings only ~7% extra FLOPs for the baseline model, and Dynamic GCN achieves better performance with $2\times$~$4\times$ fewer FLOPs than existing methods. By further combining static physical body connections and motion modalities, we achieve state-of-the-art performance on three large-scale benchmarks, namely NTU-RGB+D, NTU-RGB+D 120 and Skeleton-Kinetics.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑