Skip to main content
EURASIP Journal on Image and Video Processing

Table 1 Specification of the generator network

From: Face image synthesis from facial parts

Num

Type

Kernel

Stride

Outputs

1

Input

–

–

3

1

Conv

5 × 5

1

64

1

Conv

3 × 3

2

128

1

Conv

3 × 3

1

128

1

Conv

3 × 3

2

256

6

Conv

3 × 3

1

256

1

Decon

3 × 3

2

256

1

Decon

3 × 3

1

128

1

Decon

3 × 3

2

128

1

Decon

3 × 3

1

64

1

Decon

3 × 3

1

3

  1. Each “conv.” denotes a convolutional-BatchNorm-LeakyReLu module. Each “deconv.” is followed by a BatchNorm layer. The last layer uses the tanh activation function
Back to article page