Table 1 The dimension of each layer in ResNet18-attention.

Input

6 × 1024

6 × 1024

Fully connected layer

\(\left[6\times 1024\right]\times 1\)

Conv1

6 × 1024

6 × 1024 × 64

\(\left[1\times 3, 64\right]\times 1\)

Attention model (SE)

6 × 1024 × 64

6 × 1024 × 64

SE model Encoder

Conv2

6 × 1024 × 64

6 × 512 × 64

\(\left[\begin{array}{c}1\times 3, 64\\ 1\times 3, 64\end{array}\right]\times 2\)

Maximum pool

6 × 512 × 64

6 × 256 × 64

Kernel size: \(\left[1\times \text{2,64}\right]\times 1\)

Strides:\(\left[1\times \text{2,64}\right]\)

Conv3

6 × 256 × 64

6 × 256 × 128

\(\left[\begin{array}{c}1\times 3, 128\\ 1\times 3, 128\end{array}\right]\times 2\)

Maximum pool

6 × 256 × 128

6 × 128 × 128

Kernel size: \(\left[1\times \text{2,64}\right]\times 1\)

Strides:\(\left[1\times \text{2,64}\right]\)

Conv4

6 × 128 × 128

6 × 128 × 256

\(\left[\begin{array}{c}1\times 3, 256\\ 1\times 3, 256\end{array}\right]\times 2\)

Maximum pool

6 × 128 × 256

6 × 64 × 256

Kernel size: \(\left[1\times \text{2,64}\right]\times 1\)

Strides:\(\left[1\times \text{2,64}\right]\)

Conv5

6 × 64 × 256

6 × 64 × 512

\(\left[\begin{array}{c}1\times 3, 512\\ 1\times 3, 512\end{array}\right]\times 2\)

Upsample layer

6 × 64 × 512

6 × 64 × 1024

2x

Attention model (SE)

6 × 64 × 1024

6 × 1024 × 64

SE model Decoder

SoftMax

6 × 1024 × 64

7 × 1

Average pool, 7-class, SoftMax