-
Notifications
You must be signed in to change notification settings - Fork 1
/
net_archs.py
176 lines (137 loc) · 6.34 KB
/
net_archs.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
165
166
167
168
169
170
171
172
173
174
175
176
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import numpy
from models.crnn import *
import matplotlib.pyplot as plt
from models.instance_detector import SingleDetector, MultiDetector
from models.milpool import MILPooler
class CNN_MIL(BaseModel):
"""
conv_blocks + single detector + max-pooling
"""
def __init__(self, args):
super(CNN_MIL, self).__init__()
self.pooling = args.pooling
# num_features: should be input channel
self.input_bn = nn.BatchNorm2d(num_features=1)
self.conv_blocks = ConvBlock(filters=(32, 64, 128))
self.conv_full_frequency = nn.Sequential(
nn.Conv2d(in_channels=128, out_channels=256, kernel_size=(5, 1), padding=0),
nn.BatchNorm2d(num_features=256),
nn.ReLU()
)
self.instance_detector = SingleDetector(feature_dim=256, nb_class=args.nb_class)
self.pooler = MILPooler(pooling=self.pooling)
def forward(self, x):
"""
:param x: (batch, channel, frequency, time) = (batch, 1, 64, 400)
:return: global_prob: (Batch, Class), frame_prob: (Batch, Time, Class)
"""
x = self.input_bn(x) # ->(batch, 1, 64, 400)
x = self.conv_blocks(x) # ->(batch, 128, 8, 25)
x = self.conv_full_frequency(x) # ->(batch, 256, 1, 25) = (B, C, F, T)
instance_prob = self.instance_detector(x)
return self.pooler(instance_prob)
class CNN_MD_MIL(BaseModel):
"""
conv_blocks + multi-detector + max-pooling
"""
def __init__(self, args):
super(CNN_MD_MIL, self).__init__()
self.pooling = args.pooling
self.nb_detector = args.nb_detector
self.is_instance_softmax = args.is_instance_softmax
# num_features: should be input channel
self.input_bn = nn.BatchNorm2d(num_features=1)
self.conv_blocks = ConvBlock(filters=(32, 64, 128))
self.conv_full_frequency = nn.Sequential(
nn.Conv2d(in_channels=128, out_channels=256, kernel_size=(5, 1), padding=0),
nn.BatchNorm2d(num_features=256),
nn.ReLU()
)
assert self.nb_detector >= 2
self.instance_detector = MultiDetector(feature_dim=256,
nb_class=args.nb_class,
nb_detector=self.nb_detector,
is_instance_softmax=self.is_instance_softmax)
self.pooler = MILPooler(pooling=self.pooling)
def forward(self, x):
"""
:param x: (batch, channel, frequency, time) = (batch, 1, 64, 400)
:return: global_prob: (Batch, Class), frame_prob: (Batch, Time, Class)
"""
x = self.input_bn(x) # ->(batch, 1, 64, 400)
x = self.conv_blocks(x) # ->(batch, 128, 8, 25)
x = self.conv_full_frequency(x) # ->(batch, 256, 1, 25) = (B, C, F, T)
instance_prob = self.instance_detector(x)
return self.pooler(instance_prob)
class CNN_MTS_MIL(BaseModel):
"""
conv_blocks + multi-temporal-scale + single detector + max-pooling
"""
def __init__(self, args):
super(CNN_MTS_MIL, self).__init__()
self.pooling = args.pooling
# num_features: should be input channel
self.input_bn = nn.BatchNorm2d(num_features=1)
self.conv_blocks = ConvBlock(filters=(32, 64, 128))
self.conv_full_frequency = nn.Sequential(
nn.Conv2d(in_channels=128, out_channels=256, kernel_size=(5, 1), padding=0),
nn.BatchNorm2d(num_features=256),
nn.ReLU()
)
self.multi_resolution = MultiResolutionBlock(in_channels=256, out_channels=256,
combine_type=args.combine_type)
self.instance_detector = SingleDetector(feature_dim=256, nb_class=args.nb_class)
self.pooler = MILPooler(pooling=self.pooling)
def forward(self, x):
"""
:param x: (batch, channel, frequency, time) = (batch, 1, 64, 400)
:return: global_prob: (Batch, Class), frame_prob: (Batch, Time, Class)
"""
x = self.input_bn(x) # ->(batch, 1, 64, 400)
x = self.conv_blocks(x) # ->(batch, 128, 8, 25)
x = self.conv_full_frequency(x) # ->(batch, 256, 1, 25) = (B, C, F, T)
x = self.multi_resolution(x)
instance_prob = self.instance_detector(x)
return self.pooler(instance_prob)
class CNN_MTS_MD_MIL(BaseModel):
"""
conv_blocks + multi-temporal-scale + single detector + max-pooling
"""
def __init__(self, args):
super(CNN_MTS_MD_MIL, self).__init__()
self.pooling = args.pooling
self.nb_detector = args.nb_detector
self.is_instance_softmax = args.is_instance_softmax
# num_features: should be input channel
self.input_bn = nn.BatchNorm2d(num_features=1)
self.conv_blocks = ConvBlock(filters=(32, 64, 128))
self.conv_full_frequency = nn.Sequential(
nn.Conv2d(in_channels=128, out_channels=256, kernel_size=(5, 1), padding=0),
nn.BatchNorm2d(num_features=256),
nn.ReLU()
)
self.multi_resolution = MultiResolutionBlock(in_channels=256, out_channels=256,
combine_type=args.combine_type)
assert self.nb_detector >= 2
self.instance_detector = MultiDetector(feature_dim=256,
nb_class=args.nb_class,
nb_detector=self.nb_detector,
is_instance_softmax=self.is_instance_softmax)
self.pooler = MILPooler(pooling=self.pooling)
def forward(self, x):
"""
:param x: (batch, channel, frequency, time) = (batch, 1, 64, 400)
:return: global_prob: (Batch, Class), frame_prob: (Batch, Time, Class)
"""
x = self.input_bn(x) # ->(batch, 1, 64, 400)
x = self.conv_blocks(x) # ->(batch, 128, 8, 25)
x = self.conv_full_frequency(x) # ->(batch, 256, 1, 25) = (B, C, F, T)
x = self.multi_resolution(x)
instance_prob = self.instance_detector(x)
return self.pooler(instance_prob)
if __name__ == '__main__':
pass