檢視內神經網路架構

發表於 更新於 分類於 ， ， ， 文章字數： 179 所需閱讀時間 ≈ 1 分鐘

torchsummary

1
2
3
4
5
6
7
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = SuperPointNet_gauss2()
model = model.to(device)

# check keras-like model summary using torchsummary
from torchsummary import summary
summary(model, input_size=(1, 240, 320))

模型參數

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
   Layer (type)             Output Shape           Param #
       Conv2d-1         [-1, 64, 240, 320]             640
  BatchNorm2d-2         [-1, 64, 240, 320]             128
         ReLU-3         [-1, 64, 240, 320]               0
       Conv2d-4         [-1, 64, 240, 320]          36,928
  BatchNorm2d-5         [-1, 64, 240, 320]             128
         ReLU-6         [-1, 64, 240, 320]               0
  double_conv-7         [-1, 64, 240, 320]               0
       inconv-8         [-1, 64, 240, 320]               0
    MaxPool2d-9         [-1, 64, 120, 160]               0
      Conv2d-10         [-1, 64, 120, 160]          36,928
 BatchNorm2d-11         [-1, 64, 120, 160]             128
        ReLU-12         [-1, 64, 120, 160]               0
      Conv2d-13         [-1, 64, 120, 160]          36,928
 BatchNorm2d-14         [-1, 64, 120, 160]             128
        ReLU-15         [-1, 64, 120, 160]               0
 double_conv-16         [-1, 64, 120, 160]               0
        down-17         [-1, 64, 120, 160]               0
   MaxPool2d-18           [-1, 64, 60, 80]               0
      Conv2d-19          [-1, 128, 60, 80]          73,856
 BatchNorm2d-20          [-1, 128, 60, 80]             256
        ReLU-21          [-1, 128, 60, 80]               0
      Conv2d-22          [-1, 128, 60, 80]         147,584
 BatchNorm2d-23          [-1, 128, 60, 80]             256
        ReLU-24          [-1, 128, 60, 80]               0
 double_conv-25          [-1, 128, 60, 80]               0
        down-26          [-1, 128, 60, 80]               0
   MaxPool2d-27          [-1, 128, 30, 40]               0
      Conv2d-28          [-1, 128, 30, 40]         147,584
 BatchNorm2d-29          [-1, 128, 30, 40]             256
        ReLU-30          [-1, 128, 30, 40]               0
      Conv2d-31          [-1, 128, 30, 40]         147,584
 BatchNorm2d-32          [-1, 128, 30, 40]             256
        ReLU-33          [-1, 128, 30, 40]               0
 double_conv-34          [-1, 128, 30, 40]               0
        down-35          [-1, 128, 30, 40]               0
      Conv2d-36          [-1, 256, 30, 40]         295,168
 BatchNorm2d-37          [-1, 256, 30, 40]             512
        ReLU-38          [-1, 256, 30, 40]               0
      Conv2d-39           [-1, 65, 30, 40]          16,705
 BatchNorm2d-40           [-1, 65, 30, 40]             130
      Conv2d-41          [-1, 256, 30, 40]         295,168
 BatchNorm2d-42          [-1, 256, 30, 40]             512
        ReLU-43          [-1, 256, 30, 40]               0
      Conv2d-44          [-1, 256, 30, 40]          65,792
 BatchNorm2d-45          [-1, 256, 30, 40]             512
      Conv2d-46          [-1, 256, 30, 40]          65,792
      Conv2d-47          [-1, 256, 30, 40]          65,792
      Conv2d-48          [-1, 256, 30, 40]          65,792
LinearAttentionX-49      [-1, 256, 30, 40]               0

參考資料