arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.00324cs.CV
收藏

利用鲁棒的时空特征改进基于小波变换的动作识别

Improving Skeleton-based Action Recognitionwith Robust Spatial and Temporal Features

Simon Fraser University(西蒙弗雷泽大学)

Zeshi Yang, Kangkang Yin

摘要:

近年来,基于小波变换的动作识别在计算机视觉领域取得了重要进展。现有的算法大多是基于图卷积网络(GCN)的,目标是改进GCN骨干层的网络结构。在本文中,我们提出了一种新的机制来学习更鲁棒的鉴别功能在空间和时间。更具体地说,我们在网络的最后一层增加了一个判别特征学习(DFL)分支,以提取判别空间和时间特征来帮助学习。我们还正式提倡使用方向不变特征(DIF)作为神经网络的输入。我们表明,动作识别的准确性,可以提高这些强大的功能时,学习和使用。我们在NTU-RGBD 60、NTU-RGBD 120、SYSU 3DHOI和Eketon-Kinetics四个数据集上将我们的结果与ST-GCN和相关方法的结果进行了比较。

英文摘要:

Recently skeleton-based action recognition has made signif-icant progresses in the computer vision community. Most state-of-the-art algorithms are based on Graph Convolutional Networks (GCN), andtarget at improving the network structure of the backbone GCN lay-ers. In this paper, we propose a novel mechanism to learn more robustdiscriminative features in space and time. More specifically, we add aDiscriminative Feature Learning (DFL) branch to the last layers of thenetwork to extract discriminative spatial and temporal features to helpregularize the learning. We also formally advocate the use of Direction-Invariant Features (DIF) as input to the neural networks. We show thataction recognition accuracy can be improved when these robust featuresare learned and used. We compare our results with those of ST-GCNand related methods on four datasets: NTU-RGBD60, NTU-RGBD120,SYSU 3DHOI 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。

↑