Skip to main content
EURASIP Journal on Image and Video Processing

Table 2 Specification of the discriminator network

From: Face image synthesis from facial parts

Num

Type

Kernel

Stride

Outputs

1

Input

–

–

3

1

Conv

3 × 3

2

32

1

Conv

3 × 3

2

64

1

Conv

3 × 3

2

128

1

Conv

3 × 3

2

256

1

Conv

3 × 3

2

512

1

FC

–

–

1024

1

FC

–

–

2

  1. Except for the last type, each “conv.” denotes a convolutional-LeakyReLu module. The term “FC” denotes the fully connected layer. The last layer uses the sigmoid activation function
Back to article page