Skip to main content

Table 4 Structure and parameters of different branch networks

From: Automated fundus ultrasound image classification based on siamese convolutional neural networks with multi-attention

VGG16 branch(SVK_MA)

Alexnet branch(SAK_MA)

Resnet18 branch(SRK_MA)

Conv, 64, 3 × 3

Conv, 64, 11 × 11

Conv, 64, 7 × 7

Conv, 64, 3 × 3

Maxpool, 3 × 3

BatchNorm, 64

Maxpool, 2 × 2

Conv, 192, 5 × 5

Maxpool, 3 × 3

Conv, 128, 3 × 3

Maxpool, 3 × 3

Conv, 64, 3 × 3

Conv, 128, 3 × 3

Conv, 384, 3 × 3

BatchNorm, 64

Maxpool, 2 × 2

Conv, 256, 3 × 3

Conv, 64, 3 × 3

Conv, 256, 3 × 3

Conv, 256, 3 × 3

BatchNorm, 64

Conv, 256, 3 × 3

Maxpool, 3 × 3

Conv, 64, 3 × 3

Conv, 256, 3 × 3

Attention Model

BatchNorm, 64

Maxpool, 2 × 2

FC(4096)

Conv, 64, 3 × 3

Conv, 512, 3 × 3

FC(512)

BatchNorm, 64

Conv, 512, 3 × 3

FC(100)

Conv, 128, 3 × 3

Conv, 512, 3 × 3

 

BatchNorm, 128

Maxpool, 2 × 2

 

Conv, 128, 3 × 3

Conv, 512, 3 × 3

 

BatchNorm, 128

Conv, 512, 3 × 3

 

(downsample) Conv, 128, 1 × 1

Conv, 512, 3 × 3

 

BatchNorm, 128

Maxpool, 2 × 2

 

Conv, 128, 3 × 3

Attention Model

 

BatchNorm, 128

FC(4096)

 

Conv, 128, 3 × 3

FC(512)

 

BatchNorm, 128

FC(100)

 

Conv, 256, 3 × 3

  

BatchNorm, 256

  

Conv, 256, 3 × 3

  

BatchNorm, 256

  

(downsample) Conv, 256, 1 × 1

  

BatchNorm, 256

  

Conv, 256, 3 × 3

  

BatchNorm, 256

  

Conv, 256, 3 × 3

  

BatchNorm, 256

  

Conv, 512, 3 × 3

  

BatchNorm, 512

  

Conv, 512, 3 × 3

  

BatchNorm, 512

  

(downsample) Conv, 512, 1 × 1

  

BatchNorm, 512

  

Conv, 512, 3 × 3

  

BatchNorm, 512

  

Conv, 512, 3 × 3

  

BatchNorm, 512

  

Attention Model

  

FC(512)

  

FC(100)