resnet网络特征提取过程可视化

我们在训练图片时，是不是要看看具体提取时的每个特征图提取的样子，找了很多，终于功夫不负有心人，找到了，通过修改的代码： 

resnet代码：

import torch
import torch.nn as nn
from torchvision.models.utils import load_state_dict_from_url
import math
model_urls = {'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth','resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth','resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth','resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth','resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
}class GhostModule(nn.Module):def __init__(self, inp, oup, kernel_size=1, ratio=2, dw_size=3, stride=1, relu=True):super(GhostModule, self).__init__()self.oup = oupinit_channels = math.ceil(oup / ratio)new_channels = init_channels*(ratio-1)self.primary_conv = nn.Sequential(nn.Conv2d(inp, init_channels, kernel_size, stride, kernel_size//2, bias=False),nn.BatchNorm2d(init_channels),nn.ReLU(inplace=True) if relu else nn.Sequential(),)self.cheap_operation = nn.Sequential(nn.Conv2d(init_channels, new_channels, dw_size, 1, dw_size//2, groups=init_channels, bias=False),nn.BatchNorm2d(new_channels),nn.ReLU(inplace=True) if relu else nn.Sequential(),)def forward(self, x):x1 = self.primary_conv(x)x2 = self.cheap_operation(x1)out = torch.cat([x1,x2], dim=1)return out[:,:self.oup,:,:]def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,padding=dilation, groups=groups, bias=False, dilation=dilation)def conv1x1(in_planes, out_planes, stride=1):return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)class BasicBlock(nn.Module):expansion = 1def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,base_width=64, dilation=1, norm_layer=None):super(BasicBlock, self).__init__()if norm_layer is None:norm_layer = nn.BatchNorm2dif groups != 1 or base_width != 64:raise ValueError('BasicBlock only supports groups=1 and base_width=64')if dilation > 1:raise NotImplementedError("Dilation > 1 not supported in BasicBlock")# Both self.conv1 and self.downsample layers downsample the input when stride != 1self.conv1 = conv3x3(inplanes, planes, stride)self.bn1 = norm_layer(planes)self.relu = nn.ReLU(inplace=True)self.conv2 = conv3x3(planes, planes)self.bn2 = norm_layer(planes)self.downsample = downsampleself.stride = stridedef forward(self, x):identity = xout = self.conv1(x)out = self.bn1(out)out = self.relu(out)out = self.conv2(out)out = self.bn2(out)if self.downsample is not None:identity = self.downsample(x)out += identityout = self.relu(out)return outclass Bottleneck(nn.Module):expansion = 4def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,base_width=64, dilation=1, norm_layer=None):super(Bottleneck, self).__init__()if norm_layer is None:norm_layer = nn.BatchNorm2dwidth = int(planes * (base_width / 64.)) * groups# Both self.conv2 and self.downsample layers downsample the input when stride != 1self.conv1 = conv1x1(inplanes, width)self.bn1 = norm_layer(width)self.conv2 = conv3x3(width, width, stride, groups, dilation)self.bn2 = norm_layer(width)self.conv3 = conv1x1(width, planes * self.expansion)self.bn3 = norm_layer(planes * self.expansion)self.relu = nn.ReLU(inplace=True)self.downsample = downsampleself.stride = stridedef forward(self, x):identity = xout = self.conv1(x)out = self.bn1(out)out = self.relu(out)out = self.conv2(out)out = self.bn2(out)out = self.relu(out)out = self.conv3(out)out = self.bn3(out)if self.downsample is not None:identity = self.downsample(x)out += identityout = self.relu(out)return outclass ResNet(nn.Module):def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,groups=1, width_per_group=64, replace_stride_with_dilation=None,norm_layer=None):super(ResNet, self).__init__()if norm_layer is None:norm_layer = nn.BatchNorm2dself._norm_layer = norm_layerself.inplanes = 64self.dilation = 1if replace_stride_with_dilation is None:replace_stride_with_dilation = [False, False, False]if len(replace_stride_with_dilation) != 3:raise ValueError("replace_stride_with_dilation should be None ""or a 3-element tuple, got {}".format(replace_stride_with_dilation))self.block = blockself.groups = groupsself.base_width = width_per_group# 224,224,3 -> 112,112,64self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,bias=False)self.bn1 = norm_layer(self.inplanes)self.relu = nn.ReLU(inplace=True)# 112,112,64 -> 56,56,64self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)# 56,56,64 -> 56,56,256self.layer1 = self._make_layer(block, 64, layers[0])# 56,56,256 -> 28,28,512self.layer2 = self._make_layer(block, 128, layers[1], stride=2,dilate=replace_stride_with_dilation[0])# 28,28,512 -> 14,14,1024self.layer3 = self._make_layer(block, 256, layers[2], stride=2,dilate=replace_stride_with_dilation[1])# 14,14,1024 -> 7,7,2048self.layer4 = self._make_layer(block, 512, layers[3], stride=2,dilate=replace_stride_with_dilation[2])# 7,7,2048 -> 2048self.avgpool = nn.AdaptiveAvgPool2d((1, 1))# 2048 -> num_classesself.fc = nn.Linear(512 * block.expansion, num_classes)for m in self.modules():if isinstance(m, nn.Conv2d):nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):nn.init.constant_(m.weight, 1)nn.init.constant_(m.bias, 0)if zero_init_residual:for m in self.modules():if isinstance(m, Bottleneck):nn.init.constant_(m.bn3.weight, 0)def _make_layer(self, block, planes, blocks, stride=1, dilate=False):norm_layer = self._norm_layerdownsample = Noneprevious_dilation = self.dilationif dilate:self.dilation *= stridestride = 1if stride != 1 or self.inplanes != planes * block.expansion:downsample = nn.Sequential(conv1x1(self.inplanes, planes * block.expansion, stride),norm_layer(planes * block.expansion),)layers = []# Conv_blocklayers.append(block(self.inplanes, planes, stride, downsample, self.groups,self.base_width, previous_dilation, norm_layer))self.inplanes = planes * block.expansionfor _ in range(1, blocks):# identity_blocklayers.append(block(self.inplanes, planes, groups=self.groups,base_width=self.base_width, dilation=self.dilation,norm_layer=norm_layer))return nn.Sequential(*layers)def forward(self, x):x = self.conv1(x)x = self.bn1(x)x = self.relu(x)x = self.maxpool(x)x = self.layer1(x)x = self.layer2(x)x = self.layer3(x)x = self.layer4(x)x = self.avgpool(x)x = torch.flatten(x, 1)x = self.fc(x)return xdef freeze_backbone(self):backbone = [self.conv1, self.bn1, self.layer1, self.layer2, self.layer3, self.layer4]for module in backbone:for param in module.parameters():param.requires_grad = Falsedef Unfreeze_backbone(self):backbone = [self.conv1, self.bn1, self.layer1, self.layer2, self.layer3, self.layer4]for module in backbone:for param in module.parameters():param.requires_grad = Truedef resnet18(pretrained=False, progress=True, num_classes=1000):model = ResNet(BasicBlock, [2, 2, 2, 2])if pretrained:state_dict = load_state_dict_from_url(model_urls['resnet18'], model_dir='./model_data',progress=progress)model.load_state_dict(state_dict)if num_classes!=1000:model.fc = nn.Linear(512 * model.block.expansion, num_classes)return modeldef resnet34(pretrained=False, progress=True, num_classes=1000):model = ResNet(BasicBlock, [3, 4, 6, 3])if pretrained:state_dict = load_state_dict_from_url(model_urls['resnet34'], model_dir='./model_data',progress=progress)model.load_state_dict(state_dict)if num_classes!=1000:model.fc = nn.Linear(512 * model.block.expansion, num_classes)return modeldef resnet50(pretrained=False, progress=True, num_classes=1000):model = ResNet(Bottleneck, [3, 4, 6, 3])if pretrained:state_dict = load_state_dict_from_url(model_urls['resnet50'], model_dir='./model_data',progress=progress)model.load_state_dict(state_dict)if num_classes!=1000:model.fc = nn.Linear(512 * model.block.expansion, num_classes)return modeldef resnet101(pretrained=False, progress=True, num_classes=1000):model = ResNet(Bottleneck, [3, 4, 23, 3])if pretrained:state_dict = load_state_dict_from_url(model_urls['resnet101'], model_dir='./model_data',progress=progress)model.load_state_dict(state_dict)if num_classes!=1000:model.fc = nn.Linear(512 * model.block.expansion, num_classes)return modeldef resnet152(pretrained=False, progress=True, num_classes=1000):model = ResNet(Bottleneck, [3, 8, 36, 3])if pretrained:state_dict = load_state_dict_from_url(model_urls['resnet152'], model_dir='./model_data',progress=progress)model.load_state_dict(state_dict)if num_classes!=1000:model.fc = nn.Linear(512 * model.block.expansion, num_classes)return model

可以查看网络的结构：

#--------------------------------------------#
#   该部分代码只用于看网络结构，并非测试代码
#--------------------------------------------#
import torch
from thop import clever_format, profile
from torchsummary import summaryfrom nets import get_model_from_name
# from nets import resnet_cbam # 使用哪个引入哪个即可
if __name__ == "__main__":input_shape = [224, 224]num_classes = 3  #写自己的分类个数，如果是训练图像分割，要多分一个背景，比如猫狗两种，num_classes=2+1# backbone    = "mobilenetv2"backbone = "resnet50"device  = torch.device("cuda" if torch.cuda.is_available() else "cpu")model   = get_model_from_name[backbone](num_classes=num_classes, pretrained=False).to(device)summary(model, (3, input_shape[0], input_shape[1]))dummy_input     = torch.randn(1, 3, input_shape[0], input_shape[1]).to(device)flops, params   = profile(model.to(device), (dummy_input, ), verbose=False)#--------------------------------------------------------##   flops * 2是因为profile没有将卷积作为两个operations#   有些论文将卷积算乘法、加法两个operations。此时乘2#   有些论文只考虑乘法的运算次数，忽略加法。此时不乘2#   本代码选择乘2，参考YOLOX。#--------------------------------------------------------#flops           = flops * 2flops, params   = clever_format([flops, params], "%.3f")print('Total GFLOPS: %s' % (flops))print('Total params: %s' % (params))

结构如下：（resnet50的） 

其中两个残差模块Conv Block用于改变通道数，Identity Block用于加深网络

第一个残差块4个卷积，第二个3个

（4+3+3 ） +  （ 4+3+3+3）   +  （ 4+3+3+3+3+3+3）   +   （ 4+3+3）==52 

再加上一上来有个7*7卷积==53个

 ----------------------------------------------------------------
list         Layer (type)               Output Shape         Param #
================================================================
0          Conv2d-1         [-1, 64, 112, 112]           9,408
       BatchNorm2d-2         [-1, 64, 112, 112]             128
              ReLU-3         [-1, 64, 112, 112]               0
         MaxPool2d-4           [-1, 64, 56, 56]               0
1          Conv2d-5           [-1, 64, 56, 56]           4,096
       BatchNorm2d-6           [-1, 64, 56, 56]             128
              ReLU-7           [-1, 64, 56, 56]               0
 2           Conv2d-8           [-1, 64, 56, 56]          36,864
       BatchNorm2d-9           [-1, 64, 56, 56]             128
             ReLU-10           [-1, 64, 56, 56]               0
3           Conv2d-11          [-1, 256, 56, 56]          16,384
      BatchNorm2d-12          [-1, 256, 56, 56]             512
 4          Conv2d-13          [-1, 256, 56, 56]          16,384
      BatchNorm2d-14          [-1, 256, 56, 56]             512
             ReLU-15          [-1, 256, 56, 56]               0
       Bottleneck-16          [-1, 256, 56, 56]               0
5           Conv2d-17           [-1, 64, 56, 56]          16,384
      BatchNorm2d-18           [-1, 64, 56, 56]             128
             ReLU-19           [-1, 64, 56, 56]               0
6          Conv2d-20           [-1, 64, 56, 56]          36,864
      BatchNorm2d-21           [-1, 64, 56, 56]             128
             ReLU-22           [-1, 64, 56, 56]               0
7         Conv2d-23          [-1, 256, 56, 56]          16,384
      BatchNorm2d-24          [-1, 256, 56, 56]             512
             ReLU-25          [-1, 256, 56, 56]               0
       Bottleneck-26          [-1, 256, 56, 56]               0
8          Conv2d-27           [-1, 64, 56, 56]          16,384
      BatchNorm2d-28           [-1, 64, 56, 56]             128
             ReLU-29           [-1, 64, 56, 56]               0
9         Conv2d-30           [-1, 64, 56, 56]          36,864
      BatchNorm2d-31           [-1, 64, 56, 56]             128
             ReLU-32           [-1, 64, 56, 56]               0
10          Conv2d-33          [-1, 256, 56, 56]          16,384
      BatchNorm2d-34          [-1, 256, 56, 56]             512
             ReLU-35          [-1, 256, 56, 56]               0
       Bottleneck-36          [-1, 256, 56, 56]               0
11           Conv2d-37          [-1, 128, 56, 56]          32,768
      BatchNorm2d-38          [-1, 128, 56, 56]             256
             ReLU-39          [-1, 128, 56, 56]               0
 12          Conv2d-40          [-1, 128, 28, 28]         147,456
      BatchNorm2d-41          [-1, 128, 28, 28]             256
             ReLU-42          [-1, 128, 28, 28]               0
13          Conv2d-43          [-1, 512, 28, 28]          65,536
      BatchNorm2d-44          [-1, 512, 28, 28]           1,024
 14          Conv2d-45          [-1, 512, 28, 28]         131,072
      BatchNorm2d-46          [-1, 512, 28, 28]           1,024
             ReLU-47          [-1, 512, 28, 28]               0
       Bottleneck-48          [-1, 512, 28, 28]               0
 15          Conv2d-49          [-1, 128, 28, 28]          65,536
      BatchNorm2d-50          [-1, 128, 28, 28]             256
             ReLU-51          [-1, 128, 28, 28]               0
16           Conv2d-52          [-1, 128, 28, 28]         147,456
      BatchNorm2d-53          [-1, 128, 28, 28]             256
             ReLU-54          [-1, 128, 28, 28]               0
17           Conv2d-55          [-1, 512, 28, 28]          65,536
      BatchNorm2d-56          [-1, 512, 28, 28]           1,024
             ReLU-57          [-1, 512, 28, 28]               0
       Bottleneck-58          [-1, 512, 28, 28]               0
18           Conv2d-59          [-1, 128, 28, 28]          65,536
      BatchNorm2d-60          [-1, 128, 28, 28]             256
             ReLU-61          [-1, 128, 28, 28]               0
 19          Conv2d-62          [-1, 128, 28, 28]         147,456
      BatchNorm2d-63          [-1, 128, 28, 28]             256
             ReLU-64          [-1, 128, 28, 28]               0
20           Conv2d-65          [-1, 512, 28, 28]          65,536
      BatchNorm2d-66          [-1, 512, 28, 28]           1,024
             ReLU-67          [-1, 512, 28, 28]               0
       Bottleneck-68          [-1, 512, 28, 28]               0
 21          Conv2d-69          [-1, 128, 28, 28]          65,536
      BatchNorm2d-70          [-1, 128, 28, 28]             256
             ReLU-71          [-1, 128, 28, 28]               0
22           Conv2d-72          [-1, 128, 28, 28]         147,456
      BatchNorm2d-73          [-1, 128, 28, 28]             256
             ReLU-74          [-1, 128, 28, 28]               0
23           Conv2d-75          [-1, 512, 28, 28]          65,536
      BatchNorm2d-76          [-1, 512, 28, 28]           1,024
             ReLU-77          [-1, 512, 28, 28]               0
       Bottleneck-78          [-1, 512, 28, 28]               0
24           Conv2d-79          [-1, 256, 28, 28]         131,072
      BatchNorm2d-80          [-1, 256, 28, 28]             512
             ReLU-81          [-1, 256, 28, 28]               0
25           Conv2d-82          [-1, 256, 14, 14]         589,824
      BatchNorm2d-83          [-1, 256, 14, 14]             512
             ReLU-84          [-1, 256, 14, 14]               0
26           Conv2d-85         [-1, 1024, 14, 14]         262,144
      BatchNorm2d-86         [-1, 1024, 14, 14]           2,048
 27          Conv2d-87         [-1, 1024, 14, 14]         524,288
      BatchNorm2d-88         [-1, 1024, 14, 14]           2,048
             ReLU-89         [-1, 1024, 14, 14]               0
       Bottleneck-90         [-1, 1024, 14, 14]               0
  28         Conv2d-91          [-1, 256, 14, 14]         262,144
      BatchNorm2d-92          [-1, 256, 14, 14]             512
             ReLU-93          [-1, 256, 14, 14]               0
 29          Conv2d-94          [-1, 256, 14, 14]         589,824
      BatchNorm2d-95          [-1, 256, 14, 14]             512
             ReLU-96          [-1, 256, 14, 14]               0
30           Conv2d-97         [-1, 1024, 14, 14]         262,144
      BatchNorm2d-98         [-1, 1024, 14, 14]           2,048
             ReLU-99         [-1, 1024, 14, 14]               0
      Bottleneck-100         [-1, 1024, 14, 14]               0
 31         Conv2d-101          [-1, 256, 14, 14]         262,144
     BatchNorm2d-102          [-1, 256, 14, 14]             512
            ReLU-103          [-1, 256, 14, 14]               0
 32         Conv2d-104          [-1, 256, 14, 14]         589,824
     BatchNorm2d-105          [-1, 256, 14, 14]             512
            ReLU-106          [-1, 256, 14, 14]               0
 33         Conv2d-107         [-1, 1024, 14, 14]         262,144
     BatchNorm2d-108         [-1, 1024, 14, 14]           2,048
            ReLU-109         [-1, 1024, 14, 14]               0
      Bottleneck-110         [-1, 1024, 14, 14]               0
 34         Conv2d-111          [-1, 256, 14, 14]         262,144
     BatchNorm2d-112          [-1, 256, 14, 14]             512
            ReLU-113          [-1, 256, 14, 14]               0
 35         Conv2d-114          [-1, 256, 14, 14]         589,824
     BatchNorm2d-115          [-1, 256, 14, 14]             512
            ReLU-116          [-1, 256, 14, 14]               0
 36         Conv2d-117         [-1, 1024, 14, 14]         262,144
     BatchNorm2d-118         [-1, 1024, 14, 14]           2,048
            ReLU-119         [-1, 1024, 14, 14]               0
      Bottleneck-120         [-1, 1024, 14, 14]               0
37          Conv2d-121          [-1, 256, 14, 14]         262,144
     BatchNorm2d-122          [-1, 256, 14, 14]             512
            ReLU-123          [-1, 256, 14, 14]               0
38          Conv2d-124          [-1, 256, 14, 14]         589,824
     BatchNorm2d-125          [-1, 256, 14, 14]             512
            ReLU-126          [-1, 256, 14, 14]               0
 39         Conv2d-127         [-1, 1024, 14, 14]         262,144
     BatchNorm2d-128         [-1, 1024, 14, 14]           2,048
            ReLU-129         [-1, 1024, 14, 14]               0
      Bottleneck-130         [-1, 1024, 14, 14]               0
40          Conv2d-131          [-1, 256, 14, 14]         262,144
     BatchNorm2d-132          [-1, 256, 14, 14]             512
            ReLU-133          [-1, 256, 14, 14]               0
 41         Conv2d-134          [-1, 256, 14, 14]         589,824
     BatchNorm2d-135          [-1, 256, 14, 14]             512
            ReLU-136          [-1, 256, 14, 14]               0
 42         Conv2d-137         [-1, 1024, 14, 14]         262,144
     BatchNorm2d-138         [-1, 1024, 14, 14]           2,048
            ReLU-139         [-1, 1024, 14, 14]               0
      Bottleneck-140         [-1, 1024, 14, 14]               0
 43         Conv2d-141          [-1, 512, 14, 14]         524,288
     BatchNorm2d-142          [-1, 512, 14, 14]           1,024
            ReLU-143          [-1, 512, 14, 14]               0
 44         Conv2d-144            [-1, 512, 7, 7]       2,359,296
     BatchNorm2d-145            [-1, 512, 7, 7]           1,024
            ReLU-146            [-1, 512, 7, 7]               0
 45         Conv2d-147           [-1, 2048, 7, 7]       1,048,576
     BatchNorm2d-148           [-1, 2048, 7, 7]           4,096
 46         Conv2d-149           [-1, 2048, 7, 7]       2,097,152
     BatchNorm2d-150           [-1, 2048, 7, 7]           4,096
            ReLU-151           [-1, 2048, 7, 7]               0
      Bottleneck-152           [-1, 2048, 7, 7]               0
 47         Conv2d-153            [-1, 512, 7, 7]       1,048,576
     BatchNorm2d-154            [-1, 512, 7, 7]           1,024
            ReLU-155            [-1, 512, 7, 7]               0
 48         Conv2d-156            [-1, 512, 7, 7]       2,359,296
     BatchNorm2d-157            [-1, 512, 7, 7]           1,024
            ReLU-158            [-1, 512, 7, 7]               0
 49         Conv2d-159           [-1, 2048, 7, 7]       1,048,576
     BatchNorm2d-160           [-1, 2048, 7, 7]           4,096
            ReLU-161           [-1, 2048, 7, 7]               0
      Bottleneck-162           [-1, 2048, 7, 7]               0
50          Conv2d-163            [-1, 512, 7, 7]       1,048,576
     BatchNorm2d-164            [-1, 512, 7, 7]           1,024
            ReLU-165            [-1, 512, 7, 7]               0
 51         Conv2d-166            [-1, 512, 7, 7]       2,359,296
     BatchNorm2d-167            [-1, 512, 7, 7]           1,024
            ReLU-168            [-1, 512, 7, 7]               0
 52         Conv2d-169           [-1, 2048, 7, 7]       1,048,576
     BatchNorm2d-170           [-1, 2048, 7, 7]           4,096
            ReLU-171           [-1, 2048, 7, 7]               0
      Bottleneck-172           [-1, 2048, 7, 7]               0
AdaptiveAvgPool2d-173           [-1, 2048, 1, 1]               0
          Linear-174                    [-1, 3]           6,147
================================================================
Total params: 23,514,179
Trainable params: 23,514,179
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 286.55
Params size (MB): 89.70
Estimated Total Size (MB): 376.82
----------------------------------------------------------------
Total GFLOPS: 8.263G
Total params: 23.514M

可视化代码：

此段代码参考：https://blog.csdn.net/qq_34769162/article/details/115567093

# https://blog.csdn.net/qq_34769162/article/details/115567093
import numpy as npimport torch
import torchvision
from PIL import Image
from torchvision import transforms as Timport matplotlib.pyplot as plt
import pylabimport torch
import torchvisionfeature_extractor = torchvision.models.resnet50(pretrained=True)
if torch.cuda.is_available():feature_extractor.cuda()device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')class SaveOutput:def __init__(self):self.outputs = []def __call__(self, module, module_in, module_out):self.outputs.append(module_out)def clear(self):self.outputs = []save_output = SaveOutput()hook_handles = []for layer in feature_extractor.modules():if isinstance(layer, torch.nn.Conv2d):handle = layer.register_forward_hook(save_output)hook_handles.append(handle)from PIL import Image
from torchvision import transforms as Timage = Image.open('img/rot.jpg')
transform = T.Compose([T.Resize((224, 224)), T.ToTensor()])
X = transform(image).unsqueeze(dim=0).to(device)out = feature_extractor(X)print(len(save_output.outputs))
# 选择看的层数
# a_list = [0, 1, 6, 15, 28, 35]
a_list = [0, 1, 5, 11, 15, 24,28,43,47]# 43:512,14,14
#47:512, 7, 7
for i in a_list:print(save_output.outputs[i].cpu().detach().squeeze(0).shape)def grid_gray_image(imgs, each_row: int):'''imgs shape: batch * size (e.g., 64x32x32, 64 is the number of the gray images, and (32, 32) is the size of each gray image)'''row_num = imgs.shape[0]//each_rowfor i in range(row_num):img = imgs[i*each_row]img = (img - img.min()) / (img.max() - img.min())for j in range(1, each_row):tmp_img = imgs[i*each_row+j]tmp_img = (tmp_img - tmp_img.min()) / (tmp_img.max() - tmp_img.min())img = np.hstack((img, tmp_img))if i == 0:ans = imgelse:ans = np.vstack((ans, img))return ans# a_list = [0, 1, 5, 11, 15, 24,28,43,47]
img0 = save_output.outputs[0].cpu().detach().squeeze(0)
img0 = grid_gray_image(img0.numpy(), 8)
img1 = save_output.outputs[1].cpu().detach().squeeze(0)
img1 = grid_gray_image(img1.numpy(), 8)
img5 = save_output.outputs[5].cpu().detach().squeeze(0)
img5 = grid_gray_image(img5.numpy(), 8)
img11 = save_output.outputs[11].cpu().detach().squeeze(0)
img11 = grid_gray_image(img11.numpy(), 16)
img15 = save_output.outputs[15].cpu().detach().squeeze(0)
img15 = grid_gray_image(img15.numpy(), 16)
img24 = save_output.outputs[24].cpu().detach().squeeze(0)
img24 = grid_gray_image(img24.numpy(), 16)
img28 = save_output.outputs[28].cpu().detach().squeeze(0)
img28 = grid_gray_image(img28.numpy(), 16)
img43 = save_output.outputs[43].cpu().detach().squeeze(0)
img43 = grid_gray_image(img43.numpy(), 16)
img47 = save_output.outputs[47].cpu().detach().squeeze(0)
img47 = grid_gray_image(img47.numpy(), 16)# 64,112,112
plt.figure(figsize=(15, 15))
plt.imshow(img0, cmap='gray')#64,56,56
plt.figure(figsize=(15, 15))
plt.imshow(img1, cmap='gray')#64,56,56
plt.figure(figsize=(15, 15))
plt.imshow(img5, cmap='gray')#128，56，56
plt.figure(figsize=(30, 15))
plt.imshow(img11, cmap='gray')#128，28，28
plt.figure(figsize=(30, 15))
plt.imshow(img15, cmap='gray')#256，28，28
plt.figure(figsize=(30, 30))
plt.imshow(img24, cmap='gray')#256，14，14
plt.figure(figsize=(30, 30))
plt.imshow(img28, cmap='gray')#512，14，14
plt.figure(figsize=(45, 45))
plt.imshow(img43, cmap='gray')#512，7，7
plt.figure(figsize=(45, 45))
plt.imshow(img47, cmap='gray')pylab.show()

原图：

resnet50来说，首个卷积层的卷积核为7*7，将输入的三通道彩色图像通道增加至64，尺寸从224*224对折为112*112.

从输入图片3，224，224——>64, 112, 112

 我们对首个卷积层的提取结果进行可视化：

最后一层太抽象看不清了，512，7，7