Table 2.

The configurations of trainable layers for the proposed FER model.

VGG14		VGG17		ResNet18		ResNet34		ResNet50		ResNet101		ResNeXt34		SE-ResNet34		SE-ResNeXt34
$3\times 3,64$	$\times 1$	$3\times 3,64$	$\times 2$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$	$7\times 7,64$	$\times 1$
$3\times 3,238$	$\times 2$	$3\times 3,128$	$\times 2$	$\left.\begin{array}{c}\hfill 3\times 3,64\\ \hfill 3\times 3,64\end{array}\right\}$	$\times 2$	$\left.\begin{array}{c}\hfill 3\times 3,64\\ \hfill 3\times 3,64\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 1\times 1,64\\ \hfill 3\times 3,64\\ \hfill 1\times 1,256\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 1\times 1,64\\ \hfill 3\times 3,64\\ \hfill 1\times 1,256\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,64\\ \hfill 3\times 3,2\times 32\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,64\\ \hfill 3\times 3,64\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,64\\ \hfill 3\times 3,2\times 32\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 3$
$3\times 3,256$	$\times 3$	$3\times 3,256$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,128\\ \hfill 3\times 3,128\end{array}\right\}$	$\times 2$	$\left.\begin{array}{c}\hfill 3\times 3,128\\ \hfill 3\times 3,128\end{array}\right\}$	$\times 4$	$\left.\begin{array}{c}\hfill 1\times 1,128\\ \hfill 3\times 3,128\\ \hfill 1\times 1,512\end{array}\right\}$	$\times 4$	$\left.\begin{array}{c}\hfill 1\times 1,128\\ \hfill 3\times 3,128\\ \hfill 1\times 1,512\end{array}\right\}$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,128\\ \hfill 3\times 3,4\times 32\end{array}\right\}$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,128\\ \hfill 3\times 3,128\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,128\\ \hfill 3\times 3,4\times 32\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 4$
$3\times 3,512$	$\times 3$	$3\times 3,512$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,256\\ \hfill 3\times 3,256\end{array}\right\}$	$\times 2$	$\left.\begin{array}{c}\hfill 3\times 3,256\\ \hfill 3\times 3,256\end{array}\right\}$	$\times 6$	$\left.\begin{array}{c}\hfill 1\times 1,256\\ \hfill 3\times 3,256\\ \hfill 1\times 1,1024\end{array}\right\}$	$\times 6$	$\left.\begin{array}{c}\hfill 1\times 1,256\\ \hfill 3\times 3,256\\ \hfill 1\times 1,1024\end{array}\right\}$	$\times 23$	$\left.\begin{array}{c}\hfill 3\times 3,256\\ \hfill 3\times 3,8\times 32\end{array}\right\}$	$\times 6$	$\left.\begin{array}{c}\hfill 3\times 3,256\\ \hfill 3\times 3,256\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 6$	$\left.\begin{array}{c}\hfill 3\times 3,256\\ \hfill 3\times 3,8\times 32\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 6$
$3\times 3,512$	$\times 3$	$3\times 3,512$	$\times 4$	$\left.\begin{array}{c}\hfill 3\times 3,512\\ \hfill 3\times 3,512\end{array}\right\}$	$\times 2$	$\left.\begin{array}{c}\hfill 3\times 3,512\\ \hfill 3\times 3,512\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 1\times 1,512\\ \hfill 3\times 3,512\\ \hfill 1\times 1,2048\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 1\times 1,512\\ \hfill 3\times 3,512\\ \hfill 1\times 1,2048\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,512\\ \hfill 3\times 3,16\times 32\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,512\\ \hfill 3\times 3,512\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 3$	$\left.\begin{array}{c}\hfill 3\times 3,512\\ \hfill 3\times 3,16\times 32\\ \hfill \mathrm{SE}\mathrm{block}\end{array}\right\}$	$\times 3$
Fully connected with 2 units, tanh

Note: Two columns describe the information of each model: the left column shows the CNN filter and depth information and the right column shows the number of repetitions. The last row means the output layer that all have in common.