RGB-D Human Action Recognition of Deep Feature Enhancement and Fusion Using Two-Stream ConvNet

Figure 2

Schematic diagram of 25 joint points.