-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
296 lines (260 loc) · 12.1 KB
/
model.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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
from mxnet import gluon as g
import mxnet as mx
from utils import gauss_kernel
import os
def _conv3x3(channels, stride, in_channels):
return g.nn.Conv2D(channels, kernel_size=3, strides=stride, padding=1, use_bias=False, in_channels=in_channels)
class BasicBlockV2(g.nn.HybridBlock):
r"""BasicBlock V2 from
`"Identity Mappings in Deep Residual Networks"
<https://arxiv.org/abs/1603.05027>`_ paper.
This is used for ResNet V2 for 18, 34 layers.
Parameters
----------
channels : int
Number of output channels.
stride : int
Stride size.
downsample : bool, default False
Whether to downsample the input.
in_channels : int, default 0
Number of input channels. Default is 0, to infer from the graph.
"""
def __init__(self, channels, stride, downsample=False, in_channels=0, **kwargs):
super(BasicBlockV2, self).__init__(**kwargs)
self.bn1 = g.nn.InstanceNorm()
self.conv1 = _conv3x3(channels, stride, in_channels)
self.bn2 = g.nn.InstanceNorm()
self.conv2 = _conv3x3(channels, 1, channels)
if downsample:
self.downsample = g.nn.Conv2D(channels, 1, stride, use_bias=False,
in_channels=in_channels)
else:
self.downsample = None
def hybrid_forward(self, F, x):
residual = x
x = self.bn1(x)
x = F.Activation(x, act_type='relu')
if self.downsample:
residual = self.downsample(x)
x = self.conv1(x)
x = self.bn2(x)
x = F.Activation(x, act_type='relu')
x = self.conv2(x)
return x + residual
class ResNetV2(g.nn.HybridBlock):
r"""ResNet V2 model from
`"Identity Mappings in Deep Residual Networks"
<https://arxiv.org/abs/1603.05027>`_ paper.
Parameters
----------
block : HybridBlock
Class for the residual block. Options are BasicBlockV1, BottleneckV1.
layers : list of int
Numbers of layers in each block
channels : list of int
Numbers of channels in each block. Length should be one larger than layers list.
classes : int, default 1000
Number of classification classes.
thumbnail : bool, default False
Enable thumbnail.
"""
def __init__(self, block, layers, channels, **kwargs):
super(ResNetV2, self).__init__(**kwargs)
assert len(layers) == len(channels) - 1
with self.name_scope():
self.features = g.nn.HybridSequential(prefix='generator')
self.features.add(g.nn.Conv2D(channels=64, kernel_size=9, strides=1, padding=1, use_bias=True))
self.features.add(g.nn.Activation('relu'))
in_channels = channels[0]
for i, num_layer in enumerate(layers):
stride = 1 if i == 0 else 2
self.features.add(self._make_layer(block, num_layer, channels[i+1],
stride, i+1, in_channels=in_channels))
in_channels = channels[i+1]
self.features.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=3, use_bias=True))
self.features.add(g.nn.Activation('relu'))
self.features.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=3, use_bias=True))
self.features.add(g.nn.Activation('relu'))
self.features.add(g.nn.Conv2D(channels=3, kernel_size=9, strides=1, padding=3, use_bias=True))
self.features.add(g.nn.Activation('tanh'))
def _make_layer(self, block, layers, channels, stride, stage_index, in_channels=0):
layer = g.nn.HybridSequential(prefix='stage%d_'%stage_index)
with layer.name_scope():
layer.add(block(channels, stride, channels != in_channels, in_channels=in_channels,
prefix=''))
for _ in range(layers-1):
layer.add(block(channels, 1, False, in_channels=channels, prefix=''))
return layer
def hybrid_forward(self, F, x):
x = self.features(x)
return x
# Specification
resnet_spec = {11: ('basic_block', [1, 1, 1, 1], [64, 64, 64, 64])}
resnet_block_versions = [{'basic_block': BasicBlockV2}]
# Constructor
def get_resnet(num_layers, **kwargs):
r"""ResNet V1 model from `"Deep Residual Learning for Image Recognition"
<http://arxiv.org/abs/1512.03385>`_ paper.
ResNet V2 model from `"Identity Mappings in Deep Residual Networks"
<https://arxiv.org/abs/1603.05027>`_ paper.
Parameters
----------
version : int
Version of ResNet. Options are 1, 2.
num_layers : int
Numbers of layers. Options are 18, 34, 50, 101, 152.
pretrained : bool, default False
Whether to load the pretrained weights for model.
ctx : Context, default CPU
The context in which to load the pretrained weights.
root : str, default '~/.mxnet/models'
Location for keeping the model parameters.
"""
assert num_layers in resnet_spec, \
"Invalid number of layers: %d. Options are %s"%(
num_layers, str(resnet_spec.keys()))
block_type, layers, channels = resnet_spec[num_layers]
resnet_class = ResNetV2
block_class = BasicBlockV2
net = resnet_class(block_class, layers, channels, **kwargs)
return net
def resnet11_v2(**kwargs):
r"""ResNet-50 V1 model from `"Deep Residual Learning for Image Recognition"
<http://arxiv.org/abs/1512.03385>`_ paper.
Parameters
----------
pretrained : bool, default False
Whether to load the pretrained weights for model.
ctx : Context, default CPU
The context in which to load the pretrained weights.
root : str, default '~/.mxnet/models'
Location for keeping the model parameters.
"""
return get_resnet(11, **kwargs)
class blur(g.nn.HybridBlock):
def __init__(self):
super(blur, self).__init__()
def hybrid_forward(self, F, x, kernel_var):
# self.kernel_var = mx.nd.transpose(mx.nd.array(self.kernel_var), (3, 2, 0, 1))
x = F.Convolution(data=x, weight=kernel_var, num_filter=3, kernel=(21, 21), stride=(1, 1), pad=(1, 1), num_group=3,no_bias=True)
return x
class resnet(g.nn.HybridBlock):
def __init__(self):
super(resnet, self).__init__()
with self.name_scope():
self.body1 = g.nn.HybridSequential('generator_1')
self.body1.add(g.nn.Conv2D(channels=64, kernel_size=9, strides=1, padding=1, use_bias=True))
self.body1.add(g.nn.Activation('relu'))
self.body_res1 = g.nn.HybridSequential('generator_res1')
self.body_res1.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res1.add(g.nn.InstanceNorm())
self.body_res1.add(g.nn.Activation('relu'))
self.body_res1.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res1.add(g.nn.InstanceNorm())
self.body_res1.add(g.nn.Activation('relu'))
self.body_res2 = g.nn.HybridSequential('generator_res2')
self.body_res2.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res2.add(g.nn.InstanceNorm())
self.body_res2.add(g.nn.Activation('relu'))
self.body_res2.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res2.add(g.nn.InstanceNorm())
self.body_res2.add(g.nn.Activation('relu'))
self.body_res3 = g.nn.HybridSequential('generator_res3')
self.body_res3.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res3.add(g.nn.InstanceNorm())
self.body_res3.add(g.nn.Activation('relu'))
self.body_res3.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res3.add(g.nn.InstanceNorm())
self.body_res3.add(g.nn.Activation('relu'))
self.body_res4 = g.nn.HybridSequential('generator_res4')
self.body_res4.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res4.add(g.nn.InstanceNorm())
self.body_res4.add(g.nn.Activation('relu'))
self.body_res4.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=1, use_bias=True))
self.body_res4.add(g.nn.InstanceNorm())
self.body_res4.add(g.nn.Activation('relu'))
self.body2 = g.nn.HybridSequential('generator_2')
self.body2.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=3, use_bias=True))
self.body2.add(g.nn.Activation('relu'))
self.body2.add(g.nn.Conv2D(channels=64, kernel_size=3, strides=1, padding=3, use_bias=True))
self.body2.add(g.nn.Activation('relu'))
self.body3 = g.nn.HybridSequential('generator_final')
self.body3.add(g.nn.Conv2D(channels=3, kernel_size=9, strides=1, padding=3, use_bias=True))
self.body3.add(g.nn.Activation('tanh'))
def forward(self, x):
x = self.body1(x)
residual1 = self.body_res1(x)
x = residual1 + x
residual2 = self.body_res2(x)
x = residual2 + x
residual3 = self.body_res3(x)
x = residual3 + x
residual4 = self.body_res4(x)
x = residual4 + x
x = self.body2(x)
enhanced = self.body3(x) * 0.58 + 0.5
return enhanced
class adversarial(g.nn.HybridBlock):
def __init__(self):
super(adversarial, self).__init__()
with self.name_scope():
self.body = g.nn.HybridSequential('discriminator')
self.body.add(g.nn.Conv2D(channels=48, kernel_size=11, strides=4, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.Conv2D(channels=128, kernel_size=5, strides=2, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.InstanceNorm())
self.body.add(g.nn.Conv2D(channels=192, kernel_size=3, strides=1, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.InstanceNorm())
self.body.add(g.nn.Conv2D(channels=192, kernel_size=3, strides=1, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.InstanceNorm())
self.body.add(g.nn.Conv2D(channels=128, kernel_size=3, strides=2, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.InstanceNorm())
self.body.add(g.nn.Flatten())
self.body.add(g.nn.Dense(1024, use_bias=True))
self.body.add(g.nn.LeakyReLU(0.2))
self.body.add(g.nn.Dense(2, use_bias=True))
def hybrid_forward(self, F, x):
adv_out_logits = self.body(x)
adv_out = F.softmax(adv_out_logits)
return adv_out
# def adversarial(image_):
# with tf.variable_scope("discriminator"):
# conv1 = _conv_layer(image_, 48, 11, 4, batch_nn=False)
# conv2 = _conv_layer(conv1, 128, 5, 2)
# conv3 = _conv_layer(conv2, 192, 3, 1)
# conv4 = _conv_layer(conv3, 192, 3, 1)
# conv5 = _conv_layer(conv4, 128, 3, 2)
#
# flat_size = 128 * 7 * 7
# conv5_flat = tf.reshape(conv5, [-1, flat_size])
#
# W_fc = tf.Variable(tf.truncated_normal([flat_size, 1024], stddev=0.01))
# bias_fc = tf.Variable(tf.constant(0.01, shape=[1024]))
#
# fc = leaky_relu(tf.matmul(conv5_flat, W_fc) + bias_fc)
#
# W_out = tf.Variable(tf.truncated_normal([1024, 2], stddev=0.01))
# bias_out = tf.Variable(tf.constant(0.01, shape=[2]))
#
# adv_out = tf.nn.softmax(tf.matmul(fc, W_out) + bias_out)
#
# return adv_out
#
#
# def _conv_init_vars(net, out_channels, filter_size, transpose=False):
# _, rows, cols, in_channels = [i.value for i in net.get_shape()]
#
# if not transpose:
# weights_shape = [filter_size, filter_size, in_channels, out_channels]
# else:
# weights_shape = [filter_size, filter_size, out_channels, in_channels]
#
# weights_init = tf.Variable(tf.truncated_normal(weights_shape, stddev=0.01, seed=1), dtype=tf.float32)
#
#
# return weights_init