-
Notifications
You must be signed in to change notification settings - Fork 3
/
CNN.py
122 lines (100 loc) · 3.65 KB
/
CNN.py
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
# -*- coding: utf-8 -*-
"""CNN.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1bsDqsK60vP02Uk8Baei6mIIlzIGaKhE9
#CNN
"""
import torch.nn as nn
def conv_block(in_channels, out_channels, pooling=False):
'''
params: in_channels: (int) number of input channels
params: out_channels: (int) number of output channels
params: pooling: (bool) use pooling or not
return: convolutional layers
'''
conv_layers = nn.Sequential(
nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1, stride=1),
nn.BatchNorm2d(out_channels),
nn.ReLU()
)
if pooling:
conv_layers.add_module('max_pooling',nn.MaxPool2d(2))
return conv_layers
class ResNet9(nn.Module):
def __init__(self, in_channels, num_classes):
super().__init__()
#1st Block
self.conv1 = conv_block(in_channels, 64)#input size 1*128*128
self.conv2 = conv_block(64, 128, True) #After pooling 64*64*64
#Residual layer
self.res1 = nn.Sequential(conv_block(128,128), conv_block(128,128))
#2nd Block
self.conv3 = conv_block(128, 256, True) #After pooling 256*32*32
self.conv4 = conv_block(256, 512, True) #After pooling 512*16*16
#Residual layer
self.res2 = nn.Sequential(conv_block(512,512), conv_block(512,512))
#Linear Network
self.linear = nn.Sequential(
nn.MaxPool2d(16), #After pooling 512*1*1
nn.Flatten(), # 512
nn.Linear(512, num_classes),
nn.LogSoftmax()
)
def forward(self,x):
#Block-1
out = self.conv1(x)
out = self.conv2(out)
res1 = self.res1(out) + out
#Block-2
out = self.conv3(res1)
out = self.conv4(out)
res2 = self.res2(out) + out
#Linear network
out = self.linear(res2)
return out
class ResNet15(nn.Module):
def __init__(self, in_channels, num_classes):
super().__init__()
#1st Block
self.conv1 = conv_block(in_channels, 64) #inputs size 1*128*128
self.conv2 = conv_block(64, 128, True) #After pooling 64*64*64
#Residual layer
self.res1 = nn.Sequential(conv_block(128,128), conv_block(128,128))
#2nd Block
self.conv3 = conv_block(128, 256, True) #After pooling 256*32*32
self.conv4 = conv_block(256, 512, True) #After pooling 512*16*16
#Residual layer
self.res2 = nn.Sequential(conv_block(512,512), conv_block(512,512))
#3rd Block
self.conv5 = conv_block(512, 512, True) #After pooling 512*8*8
self.conv6 = conv_block(512, 1024, True) #After pooling 1024*4*4
#Residual layer
self.res3 = nn.Sequential(conv_block(1024,1024), conv_block(1024,1024))
#Linear Network
self.linear = nn.Sequential(
nn.MaxPool2d(4), #After pooling 1024*1*1
nn.Flatten(), # 1024
nn.Linear(1024, 512),
nn.ReLU(),
nn.Linear(512,128),
nn.ReLU(),
nn.Linear(128,num_classes),
nn.LogSoftmax()
)
def forward(self,x):
#Block-1
out = self.conv1(x)
out = self.conv2(out)
res1 = self.res1(out) + out
#Block-2
out = self.conv3(res1)
out = self.conv4(out)
res2 = self.res2(out) + out
#Block-3
out = self.conv5(res2)
out = self.conv6(out)
res3 = self.res3(out) + out
#Linear network
out = self.linear(res3)
return out