????????resnet在深度學(xué)習(xí)領(lǐng)域的重要性不言而喻，自從15年resnet提出后，被各種深度學(xué)習(xí)模型大量引用。得益于其殘差結(jié)構(gòu)的設(shè)計(jì)，使得深度學(xué)習(xí)模型可以訓(xùn)練更深層的網(wǎng)絡(luò)。常見的resnet有resnet18、resnet34、resnet50、resnet101、resnet152幾種結(jié)構(gòu)，resnet殘差網(wǎng)絡(luò)由一個卷積塊和四個殘差塊組成，每個殘差塊包含多個殘差結(jié)構(gòu)。從殘差結(jié)構(gòu)上看，resnet18和resnet34結(jié)構(gòu)近似，使用相同的殘差結(jié)構(gòu)，僅在殘差結(jié)構(gòu)數(shù)量上有區(qū)別，都是使用兩個卷積層加一個殘差連接層作為殘差結(jié)構(gòu)。

resnet18與resnet34的殘差結(jié)構(gòu)：

? ? ? ? 輸入特征圖x經(jīng)過兩個卷積核大小為3*3的卷積層，使用padding為1的填充保持特征圖大小不變，每個卷積層后接BN層和Relu層，完成兩次卷積核將結(jié)果與輸入特征圖x直接相加。resnet50到resnet152使用另一種殘差結(jié)構(gòu)，這種殘差結(jié)構(gòu)使用兩個1*1的卷積核加一個3*3的卷積核。

	resnet18	resnet34	resnet50	resnet101	resnet152
block1	1*conv	1*conv	1*conv	1*conv	1*conv
block2	2*res(2*conv)	3*res(2*conv)	3*res(3*conv)	3*res(3*conv)	3*res(3*conv)
block3	2*residual(2*conv)	4*res(2*conv)	4*res(3*conv)	4*res(3*conv)	8*res(3*conv)
block4	2*residual(2*conv)	6*res(2*conv)	6*res(3*conv)	23*res(3*conv)	36*res(3*conv)
block5	2*residual(2*conv)	3*res(2*conv)	3*res(3*conv)	3*res(3*conv)	3*res(3*conv)
FN	linear	linear	linear	linear	linear

resnet18的模型實(shí)現(xiàn)：

import torch.nn as nn from torch.nn import functional as F# 殘差單元 class Residual(nn.Module):def __init__(self, input_channel, out_channel, use_conv1x1=False, strides=1):super().__init__()self.conv1 = nn.Conv2d(input_channel, out_channel, kernel_size=3, stride=strides, padding=1)self.conv2 = nn.Conv2d(out_channel, out_channel, kernel_size=3, stride=1, padding=1)if use_conv1x1:self.conv3 = nn.Conv2d(input_channel, out_channel, kernel_size=1, stride=strides)else:self.conv3 = Noneself.bn1 = nn.BatchNorm2d(out_channel)self.bn2 = nn.BatchNorm2d(out_channel)self.relu = nn.ReLU(inplace=True)def forward(self, X):Y = self.relu(self.bn1(self.conv1(X)))Y = self.bn2(self.conv2(Y))if self.conv3:X = self.conv3(X)Y += Xreturn F.relu(Y)# 多個殘差單元組成的殘差塊 def res_block(input_channel, out_channel, num_residuals, first_block=False):blk = []for i in range(num_residuals):if i == 0 and not first_block:blk.append(Residual(input_channel, out_channel, use_conv1x1=True, strides=2))else:blk.append(Residual(out_channel, out_channel))return blkdef resnet18(num_channel, classes):block_1 = nn.Sequential(nn.Conv2d(num_channel, 64, kernel_size=7, stride=2, padding=3), nn.BatchNorm2d(64),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1) )block_2 = nn.Sequential(*res_block(64, 64, 2, first_block=True))block_3 = nn.Sequential(*res_block(64, 128, 2))block_4 = nn.Sequential(*res_block(128, 256, 2))block_5 = nn.Sequential(*res_block(256, 512, 2))model = nn.Sequential(block_1, block_2, block_3, block_4, block_5,nn.AdaptiveAvgPool2d((1, 1)),nn.Flatten(),nn.Linear(512, classes))return model

?resnet34的實(shí)現(xiàn)：

def resnet34(classes):b1 = nn.Sequential(nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3), # [3,224,224]-->[1,112,112]nn.BatchNorm2d(64),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1) # [1,110,110]-->[1,56,56])b2 = nn.Sequential(*res_block(64, 64, 3, first_block=True))b3 = nn.Sequential(*res_block(64, 128, 4))b4 = nn.Sequential(*res_block(128, 256, 6))b5 = nn.Sequential(*res_block(256, 512, 3))model = nn.Sequential(b1, b2, b3, b4, b5,nn.AdaptiveAvgPool2d((1, 1)),nn.Flatten(),nn.Linear(512, classes))return model

???????其中殘差連接1*1的卷積核，在每個殘差塊第一個殘差結(jié)構(gòu)中對輸入特征圖使用用于改變輸入特征圖通道數(shù)量，對除第一個殘差塊之外的其他殘差塊輸入特征圖進(jìn)行特征縮放。當(dāng)輸入圖像為[3,224,224]，經(jīng)過第一個卷積層，特征圖變?yōu)閇64,56,56]。由于之前的卷積層和池化層已經(jīng)進(jìn)行了兩次特征縮放，在設(shè)計(jì)的時候，在第一個殘差塊沒有對特征圖做縮放，也沒有改變通道數(shù)。之后的每個殘差塊，都會對特征圖進(jìn)行大小減半，通道加倍，因此在后面三個殘差塊中，第一個殘差單元需要對輸入通道進(jìn)行1*1卷積核加步長為2的操作，以改變特征圖大小和通道數(shù)，同時對其中第一個卷積層使用步長為2，以保證最后相加時特征圖大小和通道數(shù)一致。

resnet50模型實(shí)現(xiàn)：

import torch import torch.nn as nn from torch.nn import functional as F# 殘差單元 class Residual(nn.Module):def __init__(self, input_channel, out_channel, use_conv1x1=False, strides=1):super().__init__()self.conv1 = nn.Conv2d(input_channel, int(out_channel / 4), kernel_size=1, stride=strides)self.conv2 = nn.Conv2d(int(out_channel / 4), int(out_channel / 4), kernel_size=3, stride=1, padding=1)self.conv3 = nn.Conv2d(int(out_channel / 4), out_channel, kernel_size=1)if use_conv1x1:self.conv4 = nn.Conv2d(input_channel, out_channel, kernel_size=1, stride=strides)else:self.conv4 = Noneself.bn1 = nn.BatchNorm2d(int(out_channel / 4))self.bn2 = nn.BatchNorm2d(out_channel)self.relu = nn.ReLU(inplace=True)def forward(self, X):Y = self.relu(self.bn1(self.conv1(X)))Y = self.relu(self.bn1(self.conv2(Y)))Y = self.bn2(self.conv3(Y))if self.conv4:X = self.conv4(X)Y += Xreturn F.relu(Y)# 多個殘差單元組成的殘差塊 def res_block(input_channel, out_channel, num_residuals, first_block=False):blk = []for i in range(num_residuals):if i == 0 and not first_block:blk.append(Residual(input_channel, out_channel, use_conv1x1=True, strides=2))input_channel = out_channelif i == 0 and first_block:blk.append(Residual(input_channel, out_channel, use_conv1x1=True, strides=1))input_channel = out_channelelse:blk.append(Residual(input_channel, out_channel))return blkdef resnet50(num_channel, classes):b1 = nn.Sequential(nn.Conv2d(num_channel, 64, kernel_size=7, stride=2, padding=3),nn.BatchNorm2d(64),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1))b2 = nn.Sequential(*res_block(64, 256, 3, first_block=True))b3 = nn.Sequential(*res_block(256, 512, 4))b4 = nn.Sequential(*res_block(512, 1024, 6))b5 = nn.Sequential(*res_block(1024, 2048, 3))model = nn.Sequential(b1, b2, b3, b4, b5,nn.AdaptiveAvgPool2d((1, 1)),nn.Flatten(),nn.Linear(2048, classes))return modelnet = resnet50(num_channel=3, classes=2) x = torch.rand(size=(1, 3, 224, 224), dtype=torch.float32) for layer in net:print(layer)x = layer(x)print(x.shape)

?resnet101實(shí)現(xiàn)：

def resnet101(num_channel, classes):b1 = nn.Sequential(nn.Conv2d(num_channel, 64, kernel_size=7, stride=2, padding=3), # [3,224,224]-->[64,112,112]nn.BatchNorm2d(64),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1) # [1,110,110]-->[1,56,56])b2 = nn.Sequential(*res_block(64, 256, 3, first_block=True))b3 = nn.Sequential(*res_block(256, 512, 4))b4 = nn.Sequential(*res_block(512, 1024, 23))b5 = nn.Sequential(*res_block(1024, 2048, 3))model = nn.Sequential(b1, b2, b3, b4, b5,nn.AdaptiveAvgPool2d((1, 1)),nn.Flatten(),nn.Linear(2048, classes))return model

resnet152實(shí)現(xiàn)：

def resnet152(num_channel, classes):b1 = nn.Sequential(nn.Conv2d(num_channel, 64, kernel_size=7, stride=2, padding=3), # [3,224,224]-->[64,112,112]nn.BatchNorm2d(64),nn.ReLU(),nn.MaxPool2d(kernel_size=3, stride=2, padding=1) # [64,112,112]-->[64,56,56])b2 = nn.Sequential(*res_block(64, 256, 3, first_block=True))b3 = nn.Sequential(*res_block(256, 512, 8))b4 = nn.Sequential(*res_block(512, 1024, 36))b5 = nn.Sequential(*res_block(1024, 2048, 3))model = nn.Sequential(b1, b2, b3, b4, b5,nn.AdaptiveAvgPool2d((1, 1)),nn.Flatten(),nn.Linear(2048, classes))return model

? ? ? resnet50之后的網(wǎng)絡(luò)出了采用3個卷結(jié)合，在殘差結(jié)構(gòu)內(nèi)部也有一些區(qū)別，在殘差結(jié)構(gòu)內(nèi)第一個卷積層和第二個卷積層輸出通道為殘差塊輸出通道的四分之一，并且第一個殘差塊需要對特征圖通道進(jìn)行變換，那么在代碼實(shí)現(xiàn)上需要對第一個殘差塊的第一個殘差結(jié)構(gòu)使用1*1卷積核進(jìn)行輸入特征圖的通道變換。

總結(jié)

以上是生活随笔為你收集整理的resnet18到resnet152模型pytorch实现的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。