-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtensorflow_spiral_cnn.py
317 lines (286 loc) · 12.7 KB
/
tensorflow_spiral_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
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
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
import math
import numpy as np
from tensorflow.python.framework import tensor_shape
from tensorflow.python.keras.engine.base_layer import InputSpec
from tensorflow.python.ops import nn_ops
from tensorflow.python.keras.utils import conv_utils
from tensorflow.python.keras.utils import tf_utils
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import nn
from tensorflow.python.keras import activations
from tensorflow.python.keras import constraints
from tensorflow.python.keras import initializers
from tensorflow.python.keras import regularizers
from tensorflow.python.layers import base
from tensorflow.python.ops import init_ops
from tensorflow.python.keras.layers.convolutional import Conv
import tensorflow as tf
import random
class BasicSpiralConv2D(Conv):
def __init__(self,
filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format=None,
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer='glorot_uniform',
bias_initializer='zeros',
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
training=True,
spiral_prob=None,
plus=False,
**kwargs):
self.training = training
self.spiral_prob = spiral_prob
self.plus = plus
super(BasicSpiralConv2D, self).__init__(
rank=2,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activations.get(activation),
use_bias=use_bias,
kernel_initializer=initializers.get(kernel_initializer),
bias_initializer=initializers.get(bias_initializer),
kernel_regularizer=regularizers.get(kernel_regularizer),
bias_regularizer=regularizers.get(bias_regularizer),
activity_regularizer=regularizers.get(activity_regularizer),
kernel_constraint=constraints.get(kernel_constraint),
bias_constraint=constraints.get(bias_constraint),
**kwargs)
def create_spiral_constants(self, shape, spiral_numbers, training=True):
'''
:param spiral
shape: #output*input*height*width
:return:
train:
shape: spiral_number*input*height*width
test:
shape: output*spiral_number*input*height*width
'''
height = shape[2]
width = shape[3]
spiral_kernels = []
for s in range(spiral_numbers):
if self.plus == True and training == True:
const = np.ones(shape[2:], dtype=np.float32)
for h in range(height):
for w in range(width):
if h < s or h > (height - s - 1):
const[h, :] = 0
elif w < s or w > (width - s - 1):
const[:, w] = 0
spiral_kernels.append(const)
elif self.plus == False or training == False:
const = np.ones(shape[1:], dtype=np.float32)
for h in range(height):
for w in range(width):
if h < s or h > (height - s - 1):
const[:, h, :] = 0
elif w < s or w > (width - s - 1):
const[:, :, w] = 0
spiral_kernels.append(const)
if training:
return np.array(spiral_kernels)
elif not training:
return np.repeat(spiral_kernels[:-1], repeats=shape[0], axis=1).reshape(((spiral_numbers-1), *shape))
# def create_spiral_coeff(self, spiral_numbers):
# if spiral_numbers == 2:
# return 2.
# elif spiral_numbers == 3:
# return 1.89
def create_categorical_dist(self, spiral_numbers):
maxprob = tf.constant(1.)
remain = (maxprob - self.spiral_prob)*10
if spiral_numbers == 2:
return tf.stack([self.spiral_prob*10, remain])
elif spiral_numbers == 3:
return tf.stack([self.spiral_prob*10, remain/2, remain/2])
def build(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape)
if self.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
if input_shape[channel_axis].value is None:
raise ValueError('The channel dimension of the inputs '
'should be defined. Found `None`.')
input_dim = int(input_shape[channel_axis])
kernel_shape = self.kernel_size + (input_dim, self.filters)
self.original_kernel = self.add_weight(
name='original_kernel',
shape=kernel_shape,
initializer=self.kernel_initializer,
regularizer=self.kernel_regularizer,
constraint=self.kernel_constraint,
trainable=True,
dtype=self.dtype)
if self.use_bias:
self.bias = self.add_weight(
name='bias',
shape=(self.filters,),
initializer=self.bias_initializer,
regularizer=self.bias_regularizer,
constraint=self.bias_constraint,
trainable=True,
dtype=self.dtype)
else:
self.bias = None
self.input_spec = InputSpec(ndim=self.rank + 2,
axes={channel_axis: input_dim})
if self.padding == 'causal':
op_padding = 'valid'
else:
op_padding = self.padding
self._convolution_op = nn_ops.Convolution(input_shape,
filter_shape=self.original_kernel.get_shape(),
dilation_rate=self.dilation_rate,
strides=self.strides,
padding=op_padding.upper(),
data_format=conv_utils.convert_data_format(self.data_format,
self.rank + 2))
height = kernel_shape[1]
self.spiral_numbers = int(math.ceil(height / 2.))+1
spiral_constants = self.create_spiral_constants(kernel_shape[::-1], self.spiral_numbers, training=self.training)
self.tf_spiral_constants = tf.constant(spiral_constants, name='spiral_constants')
self.prob = self.create_categorical_dist(self.spiral_numbers)
self.normalized_prob = (self.prob/tf.reduce_sum(self.prob))[:-1]
tf.identity(self.normalized_prob, name='normalized_prob')
self.normalized_prob = tf.reshape(tensor=self.normalized_prob, shape=[self.normalized_prob.shape[0], 1, 1, 1, 1])
if self.plus == False:
self.random_indices = tf.transpose(tf.multinomial(tf.log([self.prob]), self.filters), [1,0])
print('random indices shape', self.random_indices.shape)
elif self.plus == True:
print("PPPPPPLUUUUUUSSSSSSS")
self.random_indices = tf.multinomial(tf.log([self.prob]*self.filters), input_dim)
self.random_indices = tf.expand_dims(self.random_indices, 2, 'random_index')
print('random indices shape', self.random_indices.shape)
self.build = True
def softmax(self, x):
"""Compute softmax values for each sets of scores in x."""
e_x = np.exp(x - np.max(x))
return e_x / e_x.sum(axis=0)
def call(self, inputs):
self.my_kernel = tf.identity(self.original_kernel, name='my_kernel')
if self.training == True:
self.dropout_kernel = tf.gather_nd(self.tf_spiral_constants, self.random_indices, name='dropout_kernel')
self.dropout_kernel = tf.transpose(self.dropout_kernel, [3, 2, 1, 0])
self.kernel = tf.multiply(self.dropout_kernel, self.my_kernel, name='conv_kernel')
return super(BasicSpiralConv2D, self).call(inputs)
elif self.training == False:
print("EVALUATING")
self.dropout_kernel = tf.transpose(self.tf_spiral_constants, [0, 4, 3, 2, 1])*self.normalized_prob
self.my_spiral_kernels = tf.multiply(self.my_kernel, self.dropout_kernel)
#print('spiral shape', self.my_spiral_kernels.shape)
self.my_spiral_kernels=tf.transpose(self.my_spiral_kernels, [0, 4, 3, 2, 1])
msk_shape = self.my_spiral_kernels.shape
final_output_channels = msk_shape[0]*msk_shape[1]
self.kernel = tf.reshape(tensor=self.my_spiral_kernels, shape=(final_output_channels, *msk_shape[2:]))
#print('kernel shape', self.kernel.shape)
#print('input shape', inputs.shape)
#print('stides', self.strides)
self.kernel = tf.transpose(self.kernel, [3, 2, 1, 0])
output = super(BasicSpiralConv2D, self).call(inputs)
#print('output shape', output.shape)
new_shape = (-1, (self.spiral_numbers-1), output.shape[1]//(self.spiral_numbers-1), *output.shape[-2:])
#print('new shape', new_shape)
output = tf.reshape(tensor=output, shape=new_shape)
#print('new shaped', output.shape)
layer_output = tf.reduce_sum(output, axis=1)
#print(layer_output.shape)
return layer_output
class SpiralConv2D(BasicSpiralConv2D, base.Layer):
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
spiral_prob=None,
**kwargs):
super(SpiralConv2D, self).__init__(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
spiral_prob=spiral_prob,
**kwargs)
def spiral_conv2d(inputs,
filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
training=True,
spiral_prob=None,
reuse=None):
layer = SpiralConv2D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
training=training,
spiral_prob=spiral_prob,
_reuse=reuse,
_scope=name,
)
return layer.apply(inputs)