Table 7. ResNet configuration for ImageNet100 (He et al., 2016).

Layer	Output size	Network configuration
conv1	$224\times 224$	$3\times 3$, 64, stride 1
conv2 $\mathrm{\_}$x	$224\times 224$	$\left[\begin{array}{c}3\times 3,32\\ 3\times 3,32\end{array}\right]\times 2$
	$224\times 224$	None/FMAttn/CAM
conv3 $\mathrm{\_}$x	$112\times 112$	$\left[\begin{array}{c}3\times 3,64\\ 3\times 3,64\end{array}\right]\times 2$
	$112\times 112$	None/FMAttn/CAM
conv4 $\mathrm{\_}$x	$56\times 56$	$\left[\begin{array}{c}3\times 3,64\\ 3\times 3,64\end{array}\right]\times 2$
	$56\times 56$	None/FMAttn/CAM
conv5 $\mathrm{\_}$x	$28\times 28$	$\left[\begin{array}{c}3\times 3,128\\ 3\times 3,128\end{array}\right]\times 2$
	$28\times 28$	None/FMAttn/CAM
conv6 $\mathrm{\_}$x	$14\times 14$	$\left[\begin{array}{c}3\times 3,128\\ 3\times 3,128\end{array}\right]\times 2$
	$14\times 14$	None/FMAttn/CAM
	$4\times 4$	Adaptive average pooling
	$1\times 1$	FC-100

Notes:

The first block of convN $\mathrm{\_}$x is followed by a downsample layer, except for conv2 $\mathrm{\_}$x.

The first convolutional layer of conv3_x and conv4_x has a stride of 2. The rest are all 1.

Each convolutional layer is followed by a BN layer and ReLU layer.

[ $m\times m,n$]: Convolution kernel size $m\times m$, n channels.

Output size: $width\times height$.