. 2022 Mar 23;22(7):2457. doi: 10.3390/s22072457

Table 1.

List of network architecture information and their trainable parameters, where N is the number of input images.

Fusion Block	Network Architecture	Layer Type	Filter Size	Input Channel	Output Channel	Number of Parameters
1	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	1	32	352
1	Pool	MaxPool	$1 \times 1$	32N	32	-
2	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	64	32	18,496
2	Pool	MaxPool	$1 \times 1$	32N	32	-
3	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	64	32	18,496
3	Pool	MaxPool	$1 \times 1$	32N	32	-
4	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	64	32	18,496
4	Pool	MaxPool	$1 \times 1$	32N	32	-
Bottleneck	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	64	32	18,496
	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	32	32	9280
	Conv.	ConvLayer	$3 \times 3$	32	32	9248
5	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	32	32	9280
5	Pool	MaxPool	$1 \times 1$	32N	32	-
6	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	96	32	27,712
6	Pool	MaxPool	$1 \times 1$	32N	32	-
7	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	96	32	27,712
7	Pool	MaxPool	$1 \times 1$	32N	32	-
8	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	96	32	27,712
8	Pool	MaxPool	$1 \times 1$	32N	32	-
Tail	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	96	32	27,712
	Pool	MaxPool	$1 \times 1$	32N	32	-
	Conv.	ConvLayer+ReLU+BN	$3 \times 3$	32	1	321