Table 2.

Output size, numbers and sizes of filters, number of strides, and padding in our deep residual CNN structure (3* represents that 3 pixels are included as padding in left, right, up, and down positions of input image of 224 × 224 × 3 whereas 1* shows that 1 pixel is included as padding in left, right, up, and down positions of feature map) (2/1** means 2 at the 1st iteration and 1 from the 2nd iteration).

Layer Name		Size of Feature Map	Number of Filters	Size of Filters	Number of Strides	Amount of Padding	Number of Iterations
Image input layer		224 (height) × 224 (width) × 3 (channel)
Conv1		112 × 112 × 64	64	7 × 7 × 3	2	3*	1
Max pool		56 × 56 × 64	1	3 × 3	2	0	1
Conv2	Conv2-1	56 × 56 × 64	64	1 × 1 × 64	1	0	3
	Conv2-2	56 × 56 × 64	64	3 × 3 × 64	1	1*
	Conv2-3	56 × 56 × 256	256	1 × 1 × 64	1	0
	Conv2-4 (Shortcut)	56 × 56 × 256	256	1 × 1 × 64	1	0
Conv3	Conv3-1	28 × 28 × 128	128	1 × 1 × 256	2/1**	0	4
	Conv3-2 (Bottleneck)	28 × 28 × 128	128	3 × 3 × 128	1	1*
	Conv3-3	28 × 28 × 512	512	1 × 1 × 128	1	0
	Conv3-4 (Shortcut)	28 × 28 × 512	512	1 × 1 × 256	2	0
Conv4	Conv4-1	14 × 14 × 256	256	1 × 1 × 512	2/1**	0	6
	Conv4-2 (Bottleneck)	14 × 14 × 256	256	3 × 3 × 256	1	1*
	Conv4-3	14 × 14 × 1024	1024	1 × 1 × 256	1	0
	Conv4-4 (Shortcut)	14 × 14 × 1024	1024	1 × 1 × 512	2	0
Conv5	Conv5-1	7 × 7 × 512	512	1 × 1 × 1024	2/1**	0	3
	Conv5-2 (Bottleneck)	7 × 7 × 512	512	3 × 3 × 512	1	1*
	Conv5-3	7 × 7 × 2048	2048	1 × 1 × 512	1	0
	Conv5-4 (Shortcut)	7 × 7 × 2048	2048	1 × 1 × 1024	2	0
AVG pool		1 × 1 × 2048	1	7 × 7	1	0	1
FC layer		2					1
Softmax		2					1