Improved Human Action Recognition Algorithm Based on Two-Stream Faster Region Convolutional Neural Network

Ruyi Guo; Jie Jin; Gaohua Liu; Kaiyan Liu; Shiqi Jiang

doi:10.3788/LOP57.241506

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 24 >Page 241506 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 24, 241506 (2020)

Improved Human Action Recognition Algorithm Based on Two-Stream Faster Region Convolutional Neural Network

Ruyi Guo, Jie Jin^*, Gaohua Liu, Kaiyan Liu, and Shiqi Jiang

Author Affiliations

School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China

show less

DOI: 10.3788/LOP57.241506 Cite this Article Set citation alerts

Ruyi Guo, Jie Jin, Gaohua Liu, Kaiyan Liu, Shiqi Jiang. Improved Human Action Recognition Algorithm Based on Two-Stream Faster Region Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241506 Copy Citation Text

show less

Fig. 1. Overall structure of our network

Download full size

Fig. 2. Structure of SE-VGG16

Download full size

Fig. 3. Structure of Improved SE-VGG16

Download full size

Fig. 4. Detection results of our algorithm. (a) Walking with dog; (b) diving; (c) biking;(d) floor gymnastics

Download full size

Method	mAP
IDT	86.4
C3D	85.2
TDD	90.3
Two-stream	88.0
ST-ResNet	93.4
Ours	92.8

Table 1. mAP of different algorithms on UCF101 data set unit: %

Method	VGG16	SE-VGG16	Improved SE-VGG16	IOU loss	CIOU loss	mAP
Two-stream Faster RCNN	√			√		90.5
Two-stream Faster RCNN		√		√		92.1
Two-stream Faster RCNN			√	√		92.5
Two-stream Faster RCNN			√		√	92.8

Table 2. Influence of different modules on our algorithm unit: %

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information