Table 8.

Relevant papers performance using the valence–arousal model.

Ref.	Architecture	Valence CCC per Database	Arousal CCC per Database
[143]	CNN	0.791—AVEC2016	0.805—AVEC2016
[15]	RNN	0.676—RECOLA	0.446—RECOLA
[148]	CNN, RNN	0.535—Aff-Wild, Aff-Wild2	0.365—Aff-Wild, Aff-Wild2
[46]	CNN	0.71—AffectNet, 0.75—SEWA, 0.57—AFEW-VA	0.63—AffectNet, 0.52—SEWA, 0.56—AFEW-VA
[165]	LSTM	0.068—LIRIS-ACCEDE	0.128—LIRIS-ACCEDE
[47]	CNN	0.408—AffectNet	0.373—AffectNet
[166]	ANN	0.75—SEWA, 0.438—Aff-Wild2	0.64—SEWA, 0.498—Aff-Wild2
[42]	2D CNN–LSTM	0.625—RAF-DB	0.557—RAF-DB
[50]	CNN–RNN	0.505—Aff-Wild2	0.475—Aff-Wild2