Skip to main content

Table 6 The DenseNet-201 architectures

From: Rapid identification of COVID-19 severity in CT scans through classification of deep features

Layers Output size DenseNet-201
Convolution 112 × 112 × 64 7 × 7 conv, stride 2, padding 3
Pooling 55 × 55 × 64 3 × 3 maxpool, stride 2, padding 1
Dense block
(1)
55 × 55 × 32 \(\left[ {\begin{array}{*{20}c} { 1\times 1 {\text{conv, stride 1, padding 0}}} \\ { 3\times 3 {\text{conv, stride 1, padding 1}}} \\ \end{array} } \right] \times 6\)
Transition layer
(1)
55 × 55 × 128 1 × 1 conv, stride 1, padding 0
26 × 26 × 128 2 × 2 average pool, stride 2, padding 0
Dense block
(2)
26 × 26 × 32 \(\left[ {\begin{array}{*{20}c} { 1\times 1 {\text{conv, stride 1, padding 0}}} \\ { 3\times 3 {\text{conv, stride 1, padding 1}}} \\ \end{array} } \right] \times 1 2\)
Transition layer
(2)
26 × 26 × 256 1 × 1 conv, stride 1, padding 0
13 × 13 × 256 2 × 2 average pool, stride 2, padding 0
Dense block
(3)
11 × 11 × 32 \(\left[ {\begin{array}{*{20}c} { 1\times 1 {\text{conv, stride 1, padding 0}}} \\ { 3\times 3 {\text{conv, stride 1, padding 1}}} \\ \end{array} } \right] \times 4 8\)
Transition layer
(3)
11 × 11 × 896 1 × 1 conv, stride 1, padding 0
5 × 5 × 896 2 × 2 average pooling stride 2, padding 0
Dense block
(4)
5 × 5×32 \(\left[ {\begin{array}{*{20}c} { 1\times 1 {\text{conv, stride 1, padding 0}}} \\ { 3\times 3 {\text{conv, stride 1, padding 1}}} \\ \end{array} } \right] \times 3 2\)
Classification layer   7 × 7 global average pool
  1000D fully connected (FC-1000), softmax