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

Table 5

Slowfast fusion.

	CS (top 1)	CV (top 1)
Dilated-slow-only	78.05%	88.19%
Dilated-fast-only	75.32%	86.53%
Dilated-slowfastnet	82.12%	92.85%