Laser & Optoelectronics Progress, Volume. 57, Issue 24, 241003(2020)
Human-Body Action Recognition Based on Dense Trajectories and Video Saliency
Figures & Tables(9)
Fig. 6. Accuracy comparison of each class by DT and our method. (a) UCF Sports; (b) YouTube
Table 1. Experimental environment
View tableTable 1. Experimental environment
Experimental environment Detail information OS Ubuntu14.04 CPU Intel(R) i7-8700 @3.20 GHz GPU Nvidia GeForce GTX 1060 3 GB RAM 16 GB Compiler Matlab2016
Table 2. Comparison of mean accuracy by DT and our method unit: %
View tableTable 2. Comparison of mean accuracy by DT and our method unit: %
Datasets DT S-Traj UCF Sports 88.2 90.3 YouTube 84.1 89.6
Table 3. Results comparison of our method and the state-of-the-art method unit: %
View tableTable 3. Results comparison of our method and the state-of-the-art method unit: %
UCF Sports YouTube Method Mean accuracy Method Mean accuracy Wang et al[ 6 ]89.10 Wang et al[ 6 ]85.40 Yi et al[ 13 ]90.08 Yang et al[ 22 ]88.00 Somasundaram et al[ 14 ]87.30 Peng et al[ 23 ]87.60 Li et al[ 15 ]93.40 Guo et al[ 24 ]89.50 Cho et al[ 21 ]89.70 Duan et al[ 25 ]90.00 Our method 90.30 Our method 89.60
Tools
Get Citation
Copy Citation Text
Deyong Gao, Zibing Kang, Song Wang, Yangping Wang. Human-Body Action Recognition Based on Dense Trajectories and Video Saliency[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241003
Paper Information
Category: Image Processing
Received: Mar. 30, 2020
Accepted: May. 29, 2020
Published Online: Dec. 2, 2020
The Author Email: Kang Zibing (914764692@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20