forked from Shuangfei/s3pool
-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_net.py
357 lines (289 loc) · 14.4 KB
/
build_net.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
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
import sys
import os
import time
import string
import random
import pickle
import numpy as np
import theano
import theano.tensor as T
import lasagne
from lasagne.layers import *
from stochastic_pool import stochastic_max_pool_bc01, weighted_max_pool_bc01
from lib.ops import deconv
class DeconvLayer(lasagne.layers.conv.BaseConvLayer):
def __init__(self, incoming, num_filters, filter_size, stride=(1, 1),
crop=0, untie_biases=False,
W=lasagne.init.GlorotUniform(), b=lasagne.init.Constant(0.),
nonlinearity=lasagne.nonlinearities.rectify, flip_filters=False, **kwargs):
super(DeconvLayer, self).__init__(incoming, num_filters, filter_size, stride, crop,
untie_biases, W, b, nonlinearity, flip_filters,
n=2, **kwargs)
def get_W_shape(self):
num_input_channels = self.input_shape[1]
return (num_input_channels, self.num_filters) + self.filter_size
def get_output_shape_for(self, input_shape):
return (input_shape[0], self.num_filters, 2*input_shape[2], 2*input_shape[3])
def convolve(self, input, **kwargs):
return deconv(input, self.W, subsample=(2, 2), border_mode='half')
class StochasticPool2DLayer(lasagne.layers.Layer):
def __init__(self, incoming, pool_size=2, maxpool=True, grid_size=None, **kwargs):
super(StochasticPool2DLayer, self).__init__(incoming, **kwargs)
self.rng = T.shared_randomstreams.RandomStreams(123)
self.pool_size = pool_size
self.maxpool = maxpool
if grid_size:
self.grid_size = grid_size
else:
self.grid_size = pool_size
def get_output_shape_for(self, input_shape):
return (input_shape[0], input_shape[1],
input_shape[2]/self.pool_size, input_shape[3]/self.pool_size)
def get_output_for(self, input, deterministic=False, **kwargs):
if self.maxpool:
input = T.signal.pool.pool_2d(input,
ds=(self.pool_size,)*2,
ignore_border=True,
st=(1,1),
mode='max')
if deterministic:
input = T.signal.pool.pool_2d(input,
ds=(self.pool_size,)*2,
ignore_border=True,
st=(self.pool_size,)*2,
padding=(self.pool_size/2,)*2,
mode='average_exc_pad')
return input
# return input[:, :, ::self.pool_size, ::self.pool_size]
else:
w, h = self.input_shape[2:]
n_w, n_h = w / self.grid_size, h / self.grid_size
n_sample_per_grid = self.grid_size / self.pool_size
idx_w = []
idx_h = []
for i in range(n_w):
offset = self.grid_size * i
if i < n_w - 1:
this_n = self.grid_size
else:
this_n = input.shape[2] - offset
this_idx = T.sort(self.rng.permutation(size=(1,), n=this_n)[0, :n_sample_per_grid])
idx_w.append(offset + this_idx)
for i in range(n_h):
offset = self.grid_size * i
if i < n_h - 1:
this_n = self.grid_size
else:
this_n = input.shape[3] - offset
this_idx = T.sort(self.rng.permutation(size=(1,), n=this_n)[0, :n_sample_per_grid])
idx_h.append(offset + this_idx)
idx_w = T.concatenate(idx_w, axis=0)
idx_h = T.concatenate(idx_h, axis=0)
output = input[:, :, idx_w][:, :, :, idx_h]
return output
class ZeilerPool2DLayer(Layer):
def __init__(self, incoming, pool_size=2, **kwargs):
super(ZeilerPool2DLayer, self).__init__(incoming, **kwargs)
self.pool_size = pool_size
def get_output_shape_for(self, input_shape):
return (input_shape[0], input_shape[1],
input_shape[2]/self.pool_size, input_shape[3]/self.pool_size)
def get_output_for(self, input, deterministic=False, **kwargs):
if deterministic:
pool_fn = weighted_max_pool_bc01
else:
pool_fn = stochastic_max_pool_bc01
return pool_fn(input, (self.pool_size,)*2, (self.pool_size,)*2, self.input_shape[2:])
def build_nin(input_var, option, ny=10, visualize=False, **kwargs):
if option.startswith('stochastic'):
grid_sizes = [int(s) for s in option.split('-')[1:]]
option = option.split('-')[0]
net = InputLayer((None, 3, 32, 32), input_var=input_var)
net = batch_norm(Conv2DLayer(net, num_filters=192, filter_size=5, pad='same', flip_filters=False))
net = batch_norm(NINLayer(net, num_units=160))
net = batch_norm(NINLayer(net, num_units=96))
if option == 'standard':
net = MaxPool2DLayer(net, pool_size=2)
net = DropoutLayer(net, p=0.5)
elif option == 'stochastic':
net = StochasticPool2DLayer(net, pool_size=2, maxpool=True, grid_size=grid_sizes[0])
inv_net1 = batch_norm(DeconvLayer(net, 128, (5,5)))
inv_net1 = batch_norm(Conv2DLayer(inv_net1, 3, (5,5), pad='same', nonlinearity=None))
elif option == 'zeiler':
net = ZeilerPool2DLayer(net, pool_size=2)
else:
raise NotImplementedError
net = batch_norm(Conv2DLayer(net, num_filters=192, filter_size=5, pad='same', flip_filters=False))
net = batch_norm(NINLayer(net, num_units=192))
net = batch_norm(NINLayer(net, num_units=192))
if option == 'standard':
net = MaxPool2DLayer(net, pool_size=2)
net = DropoutLayer(net, p=0.5)
elif option == 'stochastic':
net = StochasticPool2DLayer(net, pool_size=2, maxpool=True, grid_size=grid_sizes[1])
inv_net2 = batch_norm(DeconvLayer(net, 128, (5,5)))
inv_net2 = batch_norm(Conv2DLayer(inv_net2, 128, (5, 5), pad='same'))
inv_net2 = batch_norm(DeconvLayer(inv_net2, 128, (5, 5)))
inv_net2 = batch_norm(Conv2DLayer(inv_net2, 3, (5, 5), pad='same', nonlinearity=None))
elif option == 'zeiler':
net = ZeilerPool2DLayer(net, pool_size=2)
else:
raise NotImplementedError
net = batch_norm(Conv2DLayer(net, num_filters=192, filter_size=3, pad='same', flip_filters=False))
net = batch_norm(NINLayer(net, num_units=192))
net = batch_norm(DenseLayer(GlobalPoolLayer(net), num_units=ny, nonlinearity=T.nnet.softmax))
if not visualize:
return net
else:
return net, inv_net1, inv_net2
def build_resnet(input_var, option, n=3, ny=10, visualize=False, **kwargs):
sys.setrecursionlimit(10000)
if option.startswith('stochastic'):
grid_sizes = [int(s) for s in option.split('-')[1:]]
option = option.split('-')[0]
def residual_block(l, increase_dim=False, down_sample=False):
input_num_filters = l.output_shape[1]
if increase_dim:
out_num_filters = input_num_filters*2
else:
out_num_filters = input_num_filters
if down_sample:
first_stride = (2,2)
else:
first_stride = (1,1)
stack_1 = batch_norm(Conv2DLayer(l, num_filters=out_num_filters,
filter_size=(3,3), stride=first_stride,
nonlinearity=T.nnet.relu, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
stack_2 = batch_norm(Conv2DLayer(stack_1, num_filters=out_num_filters,
filter_size=(3,3), stride=(1,1),
nonlinearity=None, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
# add shortcut connections
if down_sample:
l = ExpressionLayer(l, lambda X: X[:, :, ::2, ::2],
lambda s: (s[0], s[1], s[2]//2, s[3]//2))
if increase_dim:
l = PadLayer(l, [out_num_filters//4,0,0], batch_ndim=1)
block = NonlinearityLayer(ElemwiseSumLayer([stack_2, l]),nonlinearity=T.nnet.relu)
return block
l_in = InputLayer(shape=(None, 3, 32, 32), input_var=input_var)
l = batch_norm(Conv2DLayer(l_in, num_filters=32,
filter_size=(3,3), stride=(1,1),
nonlinearity=T.nnet.relu, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
for _ in range(n):
l = residual_block(l)
########################################################################################
if option == 'standard':
down_sample = True
elif option == 'stochastic':
l = StochasticPool2DLayer(l, pool_size=2, maxpool=True, grid_size=grid_sizes[0])
inv_net1 = batch_norm(DeconvLayer(l, 128, (5,5)))
inv_net1 = batch_norm(Conv2DLayer(inv_net1, 3, (5,5), pad='same', nonlinearity=None))
down_sample = False
elif option == 'zeiler':
l = ZeilerPool2DLayer(l, pool_size=2)
down_sample = False
else:
raise NotImplementedError
l = residual_block(l, increase_dim=True, down_sample=down_sample)
for _ in range(1,n):
l = residual_block(l)
########################################################################################
if option == 'standard':
down_sample = True
elif option == 'stochastic':
l = StochasticPool2DLayer(l, pool_size=2, maxpool=True, grid_size=grid_sizes[1])
inv_net2 = batch_norm(DeconvLayer(l, 128, (5,5)))
inv_net2 = batch_norm(Conv2DLayer(inv_net2, 128, (5, 5), pad='same'))
inv_net2 = batch_norm(DeconvLayer(inv_net2, 128, (5, 5)))
inv_net2 = batch_norm(Conv2DLayer(inv_net2, 3, (5, 5), pad='same', nonlinearity=None))
down_sample = False
elif option == 'zeiler':
l = ZeilerPool2DLayer(l, pool_size=2)
down_sample = False
else:
raise NotImplementedError
l = residual_block(l, increase_dim=True, down_sample=down_sample)
for _ in range(1,n):
l = residual_block(l)
########################################################################################
l = GlobalPoolLayer(l)
network = DenseLayer(l, num_units=ny, W=lasagne.init.HeNormal(), nonlinearity=T.nnet.softmax)
if not visualize:
return network
else:
return network, inv_net1, inv_net2
def build_stl10resnet(input_var, option, ny=10, **kwargs):
sys.setrecursionlimit(10000)
if option.startswith('stochastic'):
grid_sizes = [int(s) for s in option.split('-')[1:]]
option = option.split('-')[0]
def residual_block(l, increase_dim=False, down_sample=False):
input_num_filters = l.output_shape[1]
if increase_dim:
out_num_filters = input_num_filters*2
else:
out_num_filters = input_num_filters
if down_sample:
first_stride = (2,2)
else:
first_stride = (1,1)
stack_1 = batch_norm(Conv2DLayer(l, num_filters=out_num_filters,
filter_size=(3,3), stride=first_stride,
nonlinearity=T.nnet.relu, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
stack_2 = batch_norm(Conv2DLayer(stack_1, num_filters=out_num_filters,
filter_size=(3,3), stride=(1,1),
nonlinearity=None, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
# add shortcut connections
if down_sample:
l = ExpressionLayer(l, lambda X: X[:, :, ::2, ::2],
lambda s: (s[0], s[1], s[2]//2, s[3]//2))
if increase_dim:
l = PadLayer(l, [out_num_filters//4,0,0], batch_ndim=1)
block = NonlinearityLayer(ElemwiseSumLayer([stack_2, l]),nonlinearity=T.nnet.relu)
return block
########################################################################################
l_in = InputLayer(shape=(None, 3, 96, 96), input_var=input_var)
l = batch_norm(Conv2DLayer(l_in, num_filters=32,
filter_size=(5,5), stride=(1,1),
nonlinearity=T.nnet.relu, pad='same',
W=lasagne.init.HeNormal(gain='relu'), flip_filters=False))
########################################################################################
for i in range(4):
if i == 0:
increase_dim = False
else:
increase_dim = True
if option == 'standard':
down_sample = True
elif option == 'stochastic':
l = StochasticPool2DLayer(l, pool_size=2, maxpool=True, grid_size=grid_sizes[i])
down_sample = False
elif option == 'zeiler':
l = ZeilerPool2DLayer(l, pool_size=2)
down_sample = False
else:
raise NotImplementedError
l = residual_block(l, increase_dim=increase_dim, down_sample=down_sample)
l = residual_block(l)
########################################################################################
l = GlobalPoolLayer(l)
network = DenseLayer(l, num_units=ny, W=lasagne.init.HeNormal(), nonlinearity=T.nnet.softmax)
return network
def build_net(option, arch, ny, visualize=False):
# Prepare Theano variables for inputs and targets
input_var = T.tensor4('inputs')
target_var = T.ivector('targets')
# Create neural network model
print "Building model and compiling functions..."
if arch.startswith('resnet'):
network = build_resnet(input_var, option, n=int(arch.split('-')[1]), ny=ny, visualize=visualize)
elif arch == 'nin':
network = build_nin(input_var, option, ny=ny, visualize=visualize)
elif arch == 'stl10resnet':
network = build_stl10resnet(input_var, option, ny=ny)
return input_var, target_var, network