Table 3.

Classification accuracy (%) of facial expressions on Affectnet validation set (the bold value indicates the best classification accuracy).

Methods	Accuracy
Support Vector Machine (Mollahosseini et al. [33])	30.00%
Microsoft Cognitive Services (Mollahosseini et al. [33])	37.00%
AlexNet (Mollahosseini et al. [33])	58.00%
Siqueira [34]	50.32%
Sharma et al. [30]	56.38%
Zeng et al. [31]	57.31%
Ours	58.50%