Two-Stream RNN/CNN for Action Recognition in 3D Videos

被引:0
作者
Zhao, Rui [1 ,2 ,5 ]
Ali, Haider [3 ,5 ]
van der Smagt, Patrick [4 ]
机构
[1] Siemens AG, Munich, Germany
[2] Ludwig Maximilian Univ Munich, Munich, Germany
[3] Johns Hopkins Univ, Ctr Imaging Sci, Baltimore, MD USA
[4] Volkswagen Grp, AI Res, Data Lab, Wolfsburg, Germany
[5] German Aerosp Ctr DLR, Inst Robot & Mechatron, D-82234 Wessling, Germany
来源
2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS) | 2017年
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The recognition of actions from video sequences has many applications in health monitoring, assisted living, surveillance, and smart homes. Despite advances in sensing, in particular related to 3D video, the methodologies to process the data are still subject to research. We demonstrate superior results by a system which combines recurrent neural networks with convolutional neural networks in a voting approach. The gated-recurrent-unit-based neural networks are particularly well-suited to distinguish actions based on long-term information from optical tracking data; the 3D-CNNs focus more on detailed, recent information from video data. The resulting features are merged in an SVM which then classifies the movement. In this architecture, our method improves recognition rates of state-of-the-art methods by 14% on standard data sets.
引用
收藏
页码:4260 / 4267
页数:8
相关论文
共 71 条
[1]  
ABADI M, 2015, TENSORFLOW LARGE SCA, DOI DOI 10.48550/ARXIV.1605.08695
[2]  
Amiri S. M., 2013, 2013 IEEE 15 INT C E, P606
[3]  
[Anonymous], 2015, ACTION RECOGNITION D
[4]  
[Anonymous], 2015, PROC CVPR IEEE
[5]  
[Anonymous], 2012, IEEE COMP SOC C COMP, DOI DOI 10.1109/CVPRW.2012.6239234
[6]  
[Anonymous], 2014, Asian Conference on Computer Vision
[7]  
Bloom V., 2014, WORKSH EUR C COMP VI, V1, P698
[8]  
Bloom V, 2013, LECT NOTES COMPUT SC, V8212, P64, DOI 10.1007/978-3-319-02714-2_6
[9]  
Carletti V, 2013, LECT NOTES COMPUT SC, V8158, P436, DOI 10.1007/978-3-642-41190-8_47
[10]  
Chen C, 2015, IEEE IMAGE PROC, P168, DOI 10.1109/ICIP.2015.7350781