Table 4.

Classification accuracy on CK+.

Method	Accuracy Rate
MSR [60]	91.4%
3DCNN-DAP [61]	92.4%
LBP+ORB features [55]	93.2%
Inception [9]	93.2%
Deep Features + HOG [57]	94.17%
IB-CNN [37]	95.1%
IACNN [38]	95.37%
DTAGN [62]	97.2%
ST-RNN [63]	97.2%
PPDN [64]	97.3%
Dynamic cascaded classifier [65]	97.8%
The proposed algorithm	98.0%