1.簡(jiǎn)介(torch.nn下的)

卷積層主要使用的有3類，用于處理不同維度的數(shù)據(jù)

參數(shù) Parameters：

in_channels(int) – 輸入信號(hào)的通道

out_channels(int) – 卷積產(chǎn)生的通道

kerner_size(int or tuple) - 卷積核的尺寸

stride(int or tuple, optional) - 卷積步長(zhǎng)

padding (int or tuple, optional)- 輸入的每一條邊補(bǔ)充0的層數(shù)

dilation(int or tuple, `optional``) – 卷積核元素之間的間距

groups(int, optional) – 從輸入通道到輸出通道的阻塞連接數(shù)

bias(bool, optional) - 如果bias=True，添加偏置

class torch.nn.Conv1d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True)

一維卷積層。用于計(jì)算ECG等一維數(shù)據(jù)。

input: (N,C_in,L_in) N為批次，C_in即為in_channels，即一批內(nèi)輸入一維數(shù)據(jù)個(gè)數(shù)，L_in是是一維數(shù)據(jù)基數(shù)

output: (N,C_out,L_out) N為批次，C_in即為out_channels，即一批內(nèi)輸出一維數(shù)據(jù)個(gè)數(shù)，L_out是一維數(shù)據(jù)基數(shù)

class torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True)

二維卷積層。用于計(jì)算CT斷層或MR斷層，或二維超聲圖像，自然圖像等二維數(shù)據(jù)。

self.conv1 = nn.Conv2d( # 1*28*28 -> 32*28*28
      in_channels=1,
      out_channels=32,
      kernel_size=5,
      stride=1,
      padding=2 #padding是需要計(jì)算的，padding=（stride-1）/2
    )

input: (N,C_in,H_in,W_in) N為批次，C_in即為in_channels，即一批內(nèi)輸入二維數(shù)據(jù)個(gè)數(shù)，H_in是二維數(shù)據(jù)行數(shù)，W_in是二維數(shù)據(jù)的列數(shù)

output: (N,C_out,H_out,W_out) N為批次，C_out即為out_channels，即一批內(nèi)輸出二維數(shù)據(jù)個(gè)數(shù)，H_out是二維數(shù)據(jù)行數(shù)，W_out是二維數(shù)據(jù)的列數(shù)

con2 = nn.Conv2d(1,16,5,1,2)
# con2(np.empty([1,1,28,28])) 只能接受tensor/variable
con2(torch.Tensor(1,1,28,28))
con2(Variable(torch.Tensor(1,1,28,28)))

class torch.nn.Conv3d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True)

三維卷積層。用于計(jì)算CT或MR等容積數(shù)據(jù)，視頻數(shù)據(jù)等三維數(shù)據(jù)。

input: (N,C_in,D_in,H_in,W_in)

output: (N,C_out,D_out,H_out,W_out)

2.簡(jiǎn)介(torch.nn.functional下的)

在torch.nn.functional下也有卷積層，但是和torch.nn下的卷積層的區(qū)別在于，functional下的是函數(shù)，不是實(shí)際的卷積層，而是有卷積層功能的卷積層函數(shù)，所以它并不會(huì)出現(xiàn)在網(wǎng)絡(luò)的圖結(jié)構(gòu)中。

torch.nn.functional.conv1d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1)

參數(shù)：

- input – 輸入張量的形狀 (minibatch x in_channels x iW)

- weight – 過(guò)濾器的形狀 (out_channels, in_channels, kW)

- bias – 可選偏置的形狀 (out_channels)

- stride – 卷積核的步長(zhǎng)，默認(rèn)為1

>>> filters = autograd.Variable(torch.randn(33, 16, 3))
>>> inputs = autograd.Variable(torch.randn(20, 16, 50))
>>> F.conv1d(inputs, filters)

torch.nn.functional.conv2d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1)

>>> # With square kernels and equal stride
>>> filters = autograd.Variable(torch.randn(8,4,3,3))
>>> inputs = autograd.Variable(torch.randn(1,4,5,5))
>>> F.conv2d(inputs, filters, padding=1)

torch.nn.functional.conv3d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1)

>>> filters = autograd.Variable(torch.randn(33, 16, 3, 3, 3))
>>> inputs = autograd.Variable(torch.randn(20, 16, 50, 10, 20))
>>> F.conv3d(inputs, filters)

以上這篇Pytorch之卷積層的使用詳解就是小編分享給大家的全部?jī)?nèi)容了，希望能給大家一個(gè)參考，也希望大家多多支持腳本之家。