-
Notifications
You must be signed in to change notification settings - Fork 184
/
convnets.py
341 lines (262 loc) · 12.5 KB
/
convnets.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
import numpy as np
from keras.models import Sequential, Model
from keras.layers import Flatten, Dense, Dropout, Reshape, Permute, Activation, \
Input, merge
from keras.layers.convolutional import Convolution2D, MaxPooling2D, ZeroPadding2D
from keras.optimizers import SGD
from customlayers import convolution2Dgroup, crosschannelnormalization, splittensor, \
Softmax4D
from imagenet_tool import synset_to_id, id_to_synset,synset_to_dfs_ids
from scipy.misc import imread, imresize, imsave
def convnet(network, weights_path=None, heatmap=False,
trainable=None):
"""
Returns a keras model for a CNN.
BEWARE !! : Since the different convnets have been trained in different settings, they don't take
data of the same shape. You should change the arguments of preprocess_img_batch for each CNN :
* For AlexNet, the data are of shape (227,227), and the colors in the RGB order (default)
* For VGG16 and VGG19, the data are of shape (224,224), and the colors in the BGR order
It can also be used to look at the hidden layers of the model.
It can be used that way :
>>> im = preprocess_img_batch(['cat.jpg'])
>>> # Test pretrained model
>>> model = convnet('vgg_16', 'weights/vgg16_weights.h5')
>>> sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
>>> model.compile(optimizer=sgd, loss='categorical_crossentropy')
>>> out = model.predict(im)
Parameters
--------------
network: str
The type of network chosen. For the moment, can be 'vgg_16' or 'vgg_19'
weights_path: str
Location of the pre-trained model. If not given, the model will be trained
heatmap: bool
Says wether the fully connected layers are transformed into Convolution2D layers,
to produce a heatmap instead of a
Returns
---------------
model:
The keras model for this convnet
output_dict:
Dict of feature layers, asked for in output_layers.
"""
# Select the network
if network == 'vgg_16':
convnet_init = VGG_16
elif network == 'vgg_19':
convnet_init = VGG_19
elif network == 'alexnet':
convnet_init = AlexNet
convnet = convnet_init(weights_path, heatmap=False)
if not heatmap:
return convnet
else:
convnet_heatmap = convnet_init(heatmap=True)
for layer in convnet_heatmap.layers:
if layer.name.startswith("conv"):
orig_layer = convnet.get_layer(layer.name)
layer.set_weights(orig_layer.get_weights())
elif layer.name.startswith("dense"):
orig_layer = convnet.get_layer(layer.name)
W,b = orig_layer.get_weights()
n_filter,previous_filter,ax1,ax2 = layer.get_weights()[0].shape
new_W = W.reshape((previous_filter,ax1,ax2,n_filter))
new_W = new_W.transpose((3,0,1,2))
new_W = new_W[:,:,::-1,::-1]
layer.set_weights([new_W,b])
return convnet_heatmap
return model
def VGG_16(weights_path=None, heatmap=False):
model = Sequential()
if heatmap:
model.add(ZeroPadding2D((1,1),input_shape=(3,None,None)))
else:
model.add(ZeroPadding2D((1,1),input_shape=(3,224,224)))
model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_2'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_2'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_3'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_3'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_3'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
if heatmap:
model.add(Convolution2D(4096,7,7,activation="relu",name="dense_1"))
model.add(Convolution2D(4096,1,1,activation="relu",name="dense_2"))
model.add(Convolution2D(1000,1,1,name="dense_3"))
model.add(Softmax4D(axis=1,name="softmax"))
else:
model.add(Flatten())
model.add(Dense(4096, activation='relu', name='dense_1'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu', name='dense_2'))
model.add(Dropout(0.5))
model.add(Dense(1000, name='dense_3'))
model.add(Activation("softmax",name="softmax"))
if weights_path:
model.load_weights(weights_path)
return model
def VGG_19(weights_path=None,heatmap=False):
model = Sequential()
if heatmap:
model.add(ZeroPadding2D((1,1),input_shape=(3,None,None)))
else:
model.add(ZeroPadding2D((1,1),input_shape=(3,224,224)))
model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_2'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_2'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_3'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_4'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_3'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_4'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_1'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_2'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_3'))
model.add(ZeroPadding2D((1,1)))
model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_4'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
if heatmap:
model.add(Convolution2D(4096,7,7,activation="relu",name="dense_1"))
model.add(Convolution2D(4096,1,1,activation="relu",name="dense_2"))
model.add(Convolution2D(1000,1,1,name="dense_3"))
model.add(Softmax4D(axis=1,name="softmax"))
else:
model.add(Flatten())
model.add(Dense(4096, activation='relu', name='dense_1'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu', name='dense_2'))
model.add(Dropout(0.5))
model.add(Dense(1000, name='dense_3'))
model.add(Activation("softmax"))
if weights_path:
model.load_weights(weights_path)
return model
def AlexNet(weights_path=None, heatmap=False):
if heatmap:
inputs = Input(shape=(3,None,None))
else:
inputs = Input(shape=(3,227,227))
conv_1 = Convolution2D(96, 11, 11,subsample=(4,4),activation='relu',
name='conv_1')(inputs)
conv_2 = MaxPooling2D((3, 3), strides=(2,2))(conv_1)
conv_2 = crosschannelnormalization(name="convpool_1")(conv_2)
conv_2 = ZeroPadding2D((2,2))(conv_2)
conv_2 = merge([
Convolution2D(128,5,5,activation="relu",name='conv_2_'+str(i+1))(
splittensor(ratio_split=2,id_split=i)(conv_2)
) for i in range(2)], mode='concat',concat_axis=1,name="conv_2")
conv_3 = MaxPooling2D((3, 3), strides=(2, 2))(conv_2)
conv_3 = crosschannelnormalization()(conv_3)
conv_3 = ZeroPadding2D((1,1))(conv_3)
conv_3 = Convolution2D(384,3,3,activation='relu',name='conv_3')(conv_3)
conv_4 = ZeroPadding2D((1,1))(conv_3)
conv_4 = merge([
Convolution2D(192,3,3,activation="relu",name='conv_4_'+str(i+1))(
splittensor(ratio_split=2,id_split=i)(conv_4)
) for i in range(2)], mode='concat',concat_axis=1,name="conv_4")
conv_5 = ZeroPadding2D((1,1))(conv_4)
conv_5 = merge([
Convolution2D(128,3,3,activation="relu",name='conv_5_'+str(i+1))(
splittensor(ratio_split=2,id_split=i)(conv_5)
) for i in range(2)], mode='concat',concat_axis=1,name="conv_5")
dense_1 = MaxPooling2D((3, 3), strides=(2,2),name="convpool_5")(conv_5)
if heatmap:
dense_1 = Convolution2D(4096,6,6,activation="relu",name="dense_1")(dense_1)
dense_2 = Convolution2D(4096,1,1,activation="relu",name="dense_2")(dense_1)
dense_3 = Convolution2D(1000, 1,1,name="dense_3")(dense_2)
prediction = Softmax4D(axis=1,name="softmax")(dense_3)
else:
dense_1 = Flatten(name="flatten")(dense_1)
dense_1 = Dense(4096, activation='relu',name='dense_1')(dense_1)
dense_2 = Dropout(0.5)(dense_1)
dense_2 = Dense(4096, activation='relu',name='dense_2')(dense_2)
dense_3 = Dropout(0.5)(dense_2)
dense_3 = Dense(1000,name='dense_3')(dense_3)
prediction = Activation("softmax",name="softmax")(dense_3)
model = Model(input=inputs, output=prediction)
if weights_path:
model.load_weights(weights_path)
return model
def preprocess_image_batch(image_paths, img_size=None, crop_size=None, color_mode="rgb", out=None):
img_list = []
for im_path in image_paths:
img = imread(im_path, mode='RGB')
if img_size:
img = imresize(img,img_size)
img = img.astype('float32')
# We permute the colors to get them in the BGR order
if color_mode=="bgr":
img[:,:,[0,1,2]] = img[:,:,[2,1,0]]
# We normalize the colors with the empirical means on the training set
img[:, :, 0] -= 123.68
img[:, :, 1] -= 116.779
img[:, :, 2] -= 103.939
img = img.transpose((2, 0, 1))
if crop_size:
img = img[:,(img_size[0]-crop_size[0])//2:(img_size[0]+crop_size[0])//2
,(img_size[1]-crop_size[1])//2:(img_size[1]+crop_size[1])//2]
img_list.append(img)
img_batch = np.stack(img_list, axis=0)
if not out is None:
out.append(img_batch)
else:
return img_batch
if __name__ == "__main__":
### Here is a script to compute the heatmap of the cars synsets.
## We find the synsets corresponding to cars on ImageNet website
s = "n02084071"
ids = synset_to_dfs_ids(s)
# Most of the synsets are not in the subset of the synsets used in ImageNet recognition task.
ids = np.array([id for id in ids if id != None])
im = preprocess_image_batch(['examples/dog.jpg'],color_mode="bgr")
# Test pretrained model
sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
model = convnet('alexnet',weights_path="weights/alexnet_weights.h5", heatmap=True)
model.compile(optimizer=sgd, loss='mse')
out = model.predict(im)
heatmap = out[0,ids,:,:].sum(axis=0)