Table 3.

Proposed network architecture for COVID-19 classification.

Layers	Output Size	Model
Convolution	112 × 112	7 × 7 conv, stride 2
Pooling	56 × 56	3 × 3 max pool, stride 2
$\begin{array}{c}\text{Dense Block}\\ (\text{1})\end{array}$	56 × 56	$\left[\begin{array}{c}1x1\text{conv}\\ 3x3\text{conv}\end{array}\right]x6$
Transition Layer	56 × 56	1 × 1 conv
(1)	28 × 28	2 × 2 average pool, stride 2
$\begin{array}{c}\text{Dense Block}\\ (\text{2})\end{array}$	28 × 28	$\left[\begin{array}{c}1x1\text{conv}\\ 3x3\text{conv}\end{array}\right]x12$
Transition Layer	28 × 28	1 × 1 conv
(2)	14 × 14	2 × 2 average pool, stride 2
$\begin{array}{c}\text{Dense Block}\\ (\text{3})\end{array}$	14 × 14	$\left[\begin{array}{c}1x1\text{conv}\\ 3x3\text{conv}\end{array}\right]x24$
Transition Layer	14 × 14	1 × 1 conv
(3)	7 × 7	2 × 2 average pool, stride 2
$\begin{array}{c}\text{Dense Block}\\ (\text{4})\end{array}$	7 × 7	$\left[\begin{array}{c}1x1\text{conv}\\ 3x3\text{conv}\end{array}\right]x16$
Average Pooling	1 × 1	7 × 7 global average pool
DNN	–	128 units, relu
Dropout	–	10%
Classification	–	3 category softmax