pytorch: 计算ConvTranspose1d输出特征大小

问题：如何经过convTransposed1d输出指定大小的特征？

import torch from torch import nn import torch.nn.functional as F   conv1 = nn.Conv1d(1, 2, 3, padding=1) conv2 = nn.Conv1d(in_channels=2, out_channels=4, kernel_size=3, padding=1) #转置卷积 dconv1 = nn.ConvTranspose1d(4, 1, kernel_size=3, stride=2, padding=1, output_padding=1)  x = torch.randn(16, 1, 8) print(x.size())  x1 = conv1(x) x2 = conv2(x1) print(x2.size())  x3 = dconv1(x2) print(x3.size())  ''' torch.Size([16, 1, 8]) torch.Size([16, 4, 8]) #conv2输出特征图大小 torch.Size([16, 1, 16]) #转置卷积输出特征图大小 '''   

计算转置卷积输出特征大小公式
输入：

(

N

,

C

i

n

,

L

i

n

)

(N, C_{in}, L_{in})

(N,Cin,Lin)
输出：

(

N

,

C

o

u

t

,

L

o

u

t

)

(N, C_{out}, L_{out})

(N,Cout,Lout)

计算

L

o

u

t

L_{out}

Lout大小：

L

o

u

t

=

(

L

i

n

−

1

)

×

s

t

r

i

d

e

−

2

×

p

a

d

d

i

n

g

+

d

i

l

a

t

i

o

n

×

(

k

e

r

n

e

l

s

i

z

e

−

1

)

+

o

u

t

p

u

t

p

a

d

d

i

n

g

+

1

L_{out}=(L_{in}-1)×stride-2×padding+dilation×(kernelsize-1)+outputpadding+1

Lout=(Lin−1)×stride−2×padding+dilation×(kernelsize−1)+outputpadding+1

dilation默认为1， 上式简写为：

L

o

u

t

=

(

L

i

n

−

1

)

×

s

t

r

i

d

e

−

2

×

p

a

d

d

i

n

g

+

k

e

r

n

e

l

s

i

z

e

+

o

u

t

p

u

t

p

a

d

d

i

n

g

L_{out}=(L_{in}-1)×stride-2×padding+kernelsize+outputpadding

Lout=(Lin−1)×stride−2×padding+kernelsize+outputpadding
kernel_size固定，由stride, padding, outputpadding共同决定输出特征大小。

问题：输出特定的

L

o

u

t

L_{out}

Lout大小
假如

L

i

n

=

8

,

L

o

u

t

=

23

,

k

e

r

n

e

l

s

i

z

e

=

3

,

o

u

t

p

u

t

p

a

d

d

i

n

g

=

1

L_{in}=8,L_{out}=23,kernelsize=3,outputpadding=1

Lin=8,Lout=23,kernelsize=3,outputpadding=1，根据上式可以求得：

s

t

r

i

d

e

=

3

,

p

a

d

d

i

n

g

=

1

stride=3,padding=1

stride=3,padding=1

#转置卷积 dconv1 = nn.ConvTranspose1d(1, 1, kernel_size=3, stride=3, padding=1, output_padding=1)  x = torch.randn(16, 1, 8) print(x.size()) #torch.Size([16, 1, 23])   x3 = dconv1(x) print(x3.size()) #torch.Size([16, 1, 23])  

下面两图为演示conv1d，在padding和不padding下的输出特征图大小

不带padding

带padding