Table 5.

Performance comparison (%) with previous methods on RAF-DB and FERPLUS.

Methods	RAF-DB	FERPLUS
gACNN (Li et al., 2019)	85.07	-
RAN (Wang K. et al., 2019)	86.90	88.55
SCN (Wang et al., 2020)	87.03	88.01
DACL (Farzaneh and Qi, 2021)	87.78	-
KTN (Li et al., 2021)	88.07	-
MA-Net (Zhao et al., 2021)	88.40	-
RUL (Zhang et al., 2021)	-	88.75
DMUE (She et al., 2021)	-	88.64
SeNet50 (Albanie et al., 2018)	-	88.80
FFRA-Net(Ours)	88.66	88.97

The bold values are outcomes from model runs described in this paper.