-
Notifications
You must be signed in to change notification settings - Fork 6
/
mydnn.py
165 lines (138 loc) · 5.48 KB
/
mydnn.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
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
'''
deep learning networks
'''
import torch
import torch.nn as nn
import torch.nn.functional as F
class NetTranslation5(nn.Module):
'''The chosen one! In the Jupyter Notebook, it's name is NetTranslation5_norm
Image translation. Comparing with version 4, it adds a layer so it is symetric, it also has group normalization
the first CNN is the same as NetTranslation,
the second one uses the output of the first CNN and output the # of Tx
Assuming the input image is 1 x 100 x 100
'''
def __init__(self):
super(NetTranslation5, self).__init__()
self.conv1 = nn.Conv2d(1, 8, 5, padding=2) # TUNE: a larger filter decrease miss, decrease localization error
self.conv2 = nn.Conv2d(8, 32, 5, padding=2)
self.conv3 = nn.Conv2d(32, 8, 5, padding=2)
self.conv4 = nn.Conv2d(8, 1, 5, padding=2)
self.norm1 = nn.GroupNorm(4, 8)
self.norm2 = nn.GroupNorm(16, 32)
self.norm3 = nn.GroupNorm(4, 8)
def forward(self, x):
x = F.relu(self.norm1(self.conv1(x)))
x = F.relu(self.norm2(self.conv2(x)))
x = F.relu(self.norm3(self.conv3(x)))
y = self.conv4(x)
return y
class PowerPredictor5(nn.Module):
'''The input is 1 x 21 x 21, the output is a scaler between 0.5 and 5.5
No fully connected layer in the end
The CHOSEN one.
'''
def __init__(self):
super(PowerPredictor5, self).__init__()
self.conv1 = nn.Conv2d(1, 32, 5)
self.conv2 = nn.Conv2d(32, 128, 5)
self.conv3 = nn.Conv2d(128, 32, 5)
self.conv4 = nn.Conv2d(32, 8, 5)
self.conv5 = nn.Conv2d(8, 1, 5)
self.norm1 = nn.BatchNorm2d(32)
self.norm2 = nn.BatchNorm2d(128)
self.norm3 = nn.BatchNorm2d(32)
self.norm4 = nn.BatchNorm2d(8)
def forward(self, x):
x = F.relu(self.norm1(self.conv1(x)))
x = F.relu(self.norm2(self.conv2(x)))
x = F.relu(self.norm3(self.conv3(x)))
x = F.relu(self.norm4(self.conv4(x)))
x = self.conv5(x)
x = x.view(-1, self.num_flat_features(x))
return x
def num_flat_features(self, x):
size = x.size()[1:]
num_features = 1
for s in size:
num_features *= s
return num_features
class SubtractNet3(nn.Module):
'''The chosen one! Simply adding layers helps
'''
def __init__(self):
super(SubtractNet3, self).__init__()
self.conv1 = nn.Conv2d(2, 8, 5, padding=2) # TUNE: a larger filter decrease miss, decrease localization error
self.conv2 = nn.Conv2d(8, 16, 5, padding=2)
self.conv3 = nn.Conv2d(16, 32, 5, padding=2)
self.conv4 = nn.Conv2d(32, 32, 5, padding=2)
self.conv5 = nn.Conv2d(32, 16, 5, padding=2)
self.conv6 = nn.Conv2d(16, 8, 5, padding=2)
self.conv7 = nn.Conv2d(8, 2, 5, padding=2)
self.conv8 = nn.Conv2d(2, 1, 5, padding=2)
self.norm1 = nn.GroupNorm(4, 8)
self.norm2 = nn.GroupNorm(8, 16)
self.norm3 = nn.GroupNorm(16, 32)
self.norm4 = nn.GroupNorm(16, 32)
self.norm5 = nn.GroupNorm(8, 16)
self.norm6 = nn.GroupNorm(4, 8)
self.norm7 = nn.GroupNorm(1, 2)
def forward(self, x):
x = F.relu(self.norm1(self.conv1(x)))
x = F.relu(self.norm2(self.conv2(x)))
x = F.relu(self.norm3(self.conv3(x)))
x = F.relu(self.norm4(self.conv4(x)))
x = F.relu(self.norm5(self.conv5(x)))
x = F.relu(self.norm6(self.conv6(x)))
x = F.relu(self.norm7(self.conv7(x)))
y = self.conv8(x)
return y
########## Below are for DeepTxFinder ###############
class CNN_NoTx(nn.Module):
"""this CNN predicts # of TX """
def __init__(self, max_ntx):
super(CNN_NoTx, self).__init__()
self.conv1 = nn.Conv2d(1, 32, kernel_size=(3, 3))
self.conv2 = nn.Conv2d(32, 64, kernel_size=(3, 3))
self.conv3 = nn.Conv2d(64, 128, kernel_size=(3, 3))
self.conv4 = nn.Conv2d(128, 128, kernel_size=(3, 3))
self.flat = nn.Flatten()
self.drop = nn.Dropout(p=0.5)
self.dense = nn.Linear(8192, 512)
self.dense1 = nn.Linear(512, max_ntx)
def forward(self, x):
x = F.max_pool2d(F.relu(self.conv1(x)), 2)
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = F.max_pool2d(F.relu(self.conv3(x)), 2)
x = F.relu(self.conv4(x))
x = self.flat(x)
x = self.drop(x)
x = F.relu(self.dense(x))
x = self.dense1(x)
return x
def num_flat_features(self, x):
size = x.size()[1:]
num_features = 1
for s in size:
num_features *= s
return num_features
class CNN_i(nn.Module):
def __init__(self, ntx):
super(CNN_i, self).__init__()
self.conv1 = nn.Conv2d(1, 32, kernel_size=(3, 3))
self.conv2 = nn.Conv2d(32, 64, kernel_size=(3, 3))
self.conv3 = nn.Conv2d(64, 128, kernel_size=(3, 3))
self.conv4 = nn.Conv2d(128, 128, kernel_size=(3, 3))
self.flat = nn.Flatten()
self.drop = nn.Dropout(p=0.5)
self.dense = nn.Linear(8192, 512)
self.dense1 = nn.Linear(512, 2*ntx)
def forward(self, x):
x = F.max_pool2d(F.relu(self.conv1(x)), 2)
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = F.max_pool2d(F.relu(self.conv3(x)), 2)
x = F.relu(self.conv4(x))
x = self.flat(x)
x = self.drop(x)
x = F.relu(self.dense(x))
x = self.dense1(x)
return x