-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy patharchitectures.py
150 lines (125 loc) · 4.97 KB
/
architectures.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
from densenet import DenseNet, DenseNetIN
""" Architectures of DenseNet as described in the paper:
https://arxiv.org/pdf/1608.06993.pdf
Name legend:
DenseNet40k12 - 40 layers (12,12,12), k=12
DenseNet100k12 - 100 layers (32,32,32), k=12
DenseNet100k24 - 100 layers (32,32,32), k=24
DenseNetBC100k12 - 100 layers (32,32,32), k=12, bottleneck True and compression 0.5
DenseNetBC250k24 - 250 layers (82,82,82), k=12, bottleneck True and compression 0.5
DenseNetBC190k40 - 190 layers (62,62,62), k=12, bottleneck True and compression 0.5
DenseNetIN121 - 121 layers (6,12,24,16), k=32, bottleneck True, compression 0.5, additional max pool
DenseNetIN169 - 169 layers (6,12,32,32), k=32, bottleneck True, compression 0.5, additional max pool
DenseNetIN201 - 201 layers (6,12,48,32), k=32, bottleneck True, compression 0.5, additional max pool
DenseNetIN264 - 264 layers (6,12,64,48), k=32, bottleneck True, compression 0.5, additional max pool
Each network has default set the defining parameters. You only need to define img_dim and nb_classes.
"""
def DenseNet40k12(img_dim, nb_classes, growth_rate=12, nb_filters=16,
nb_layers=[12,12,12],
init_kernel_size=(3,3),
bottleneck=False,
compression_factor=1,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNet100k12(img_dim, nb_classes, growth_rate=12, nb_filters=16,
nb_layers=[32,32,32],
init_kernel_size=(3,3),
bottleneck=False,
compression_factor=1,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNet100k24(img_dim, nb_classes, growth_rate=24, nb_filters=16,
nb_layers=[32,32,32],
init_kernel_size=(3,3),
bottleneck=False,
compression_factor=1,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetBC100k12(img_dim, nb_classes, growth_rate=12, nb_filters=16,
nb_layers=[32,32,32],
init_kernel_size=(3,3),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetBC250k24(img_dim, nb_classes, growth_rate=24, nb_filters=16,
nb_layers=[82,82,82],
init_kernel_size=(3,3),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetBC100k40(img_dim, nb_classes, growth_rate=40, nb_filters=16,
nb_layers=[62,62,62],
init_kernel_size=(3,3),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNet(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
# ImageNet
def DenseNetIN121(img_dim,nb_classes,
growth_rate=32,
nb_filters=64,
nb_layers=[6,12,24,16],
init_kernel_size=(7,7),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNetIN(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetIN169(img_dim,nb_classes,
growth_rate=32,
nb_filters=64,
nb_layers=[6,12,32,32],
init_kernel_size=(7,7),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNetIN(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetIN201(img_dim,nb_classes,
growth_rate=32,
nb_filters=64,
nb_layers=[6,12,48,32],
init_kernel_size=(7,7),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNetIN(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model
def DenseNetIN264(img_dim,nb_classes,
growth_rate=32,
nb_filters=64,
nb_layers=[6,12,64,48],
init_kernel_size=(7,7),
bottleneck=True,
compression_factor=0.5,
weight_decay=1E-4):
model = DenseNetIN(img_dim,growth_rate,nb_classes,nb_filters,
nb_layers, init_kernel_size, bottleneck, compression_factor,
weight_decay)
return model