Table 2.

Performance comparison on the AffectNet-7 dataset.

Methods	Accuracy (%)
Separate-Loss [54]	58.89
FMPN [55]	61.25
LDL-ALSG [56]	59.35
VGG-FACE [57]	60.00
OADN [58]	61.89
DDA-Loss [40]	62.34
EfficientFace [51]	63.70
MViT [53]	64.57
ResNet-18	56.97
DAN (ours)	65.69