Table 1.

The developed DenseNet model's internal architecture, including relevant hyperparameters. ReLU, BN, fully connected, and softmax layers are not shown here.

Dense blocks	Layers name	Output size	Kernel size	# Filters	Stride	Padding
Primary convolution layer	Conv2d-1	128 × 128	7 × 7	64	2	3
Dense_Block-1	Conv2d-4	128 × 128	3 × 3	32	1	1
	Conv2d-6	128 × 128	3 × 3	32	1	1
	Conv2d-9	128 × 128	3 × 3	32	1	1
	Conv2d-12	128 × 128	3 × 3	32	1	1
	Conv2d-15	128 × 128	3 × 3	32	1	1
Transition layer -1	Conv2d-19	128 × 128	1 × 1	128	1	0
	AvgPool2d-22	64 × 64	2 × 2	128	2	0
Dense_Block-2	Conv2d-25	64 × 64	3 × 3	32	1	1
	Conv2d-27	64 × 64	3 × 3	32	1	1
	Conv2d-30	64 × 64	3 × 3	32	1	1
	Conv2d-33	64 × 64	3 × 3	32	1	1
	Conv2d-36	64 × 64	3 × 3	32	1	1
Transition layer -2	Conv2d-40	64 × 64	1 × 1	128	1	0
	AvgPool2d-43	32 × 32	2 × 2	128	2	0
Dense_Block-3	Conv2d-46	32 × 32	3 × 3	32	1	1
	Conv2d-48	32 × 32	3 × 3	32	1	1
	Conv2d-51	32 × 32	3 × 3	32	1	1
	Conv2d-54	32 × 32	3 × 3	32	1	1
	Conv2d-57	32 × 32	3 × 3	32	1	1
Transition layer -3	Conv2d-61	32 × 32	1 × 1	64	1	0
	AvgPool2d-64	16 × 16	2 × 2	64	2	0