Объединены 1D-CNN и 2D-CNN - PullRequest
Новая
       7

Объединены 1D-CNN и 2D-CNN

1 голос
/ 17 февраля 2020

Я хочу построить объединенную модель CNN, используя 1D и 2D CNN, но я попробовал много способов ее построения, но этот работал со мной, но я не знаю, почему я получаю эту ошибку при использовании model_combined.summary(). Я приложил два изображения, которые содержат сводку 1D & 2D CNN сводку 1D CNN , сводку 2D CNN

Большое спасибо! 

ValueError                                Traceback (most recent call last)
<ipython-input-20-3c58e6d04c4d> in <module>()
     60 #opt = RMSprop(lr=0.001, rho=0.9)
     61 model_combined.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])
---> 62 print(model_combined.summary())

/usr/local/lib/python3.6/dist-packages/keras/engine/network.py in summary(self, line_length, positions, print_fn)
   1305         if not self.built:
   1306             raise ValueError(
-> 1307                 'This model has not yet been built. '
   1308                 'Build the model first by calling build() '
   1309                 'or calling fit() with some data. '

ValueError: This model has not yet been built. Build the model first by calling build() or calling fit() with some data. Or specify input_shape or batch_input_shape in the first layer for automatic build.

Вот код, 

from keras.models import Sequential, Model
from keras.layers.core import Dense, Activation
from keras.layers.convolutional import Conv2D , Conv1D
from keras.layers import Conv2D, Conv1D,MaxPooling2D, Reshape, Concatenate, Dropout , MaxPooling1D

# Optimizers
from keras.optimizers import Adagrad
from keras.optimizers import Adam
from keras.optimizers import SGD
from keras.optimizers import RMSprop
# ----------------------- 1D CNN ----------------------
model_1D = Sequential()
# 1
model_1D.add(Conv1D(32, kernel_size= 5 , strides=1, activation='relu')) # input shape after preprocessing
model_1D.add(MaxPooling1D(pool_size= 4, strides=4))
# 2 
model_1D.add(Conv1D(32, kernel_size= 5 , strides=1 , activation='relu'))
model_1D.add(MaxPooling1D(pool_size= 4, strides=4))
# 3
model_1D.add(Conv1D(64, kernel_size= 5 , strides=1 , activation='relu'))
model_1D.add(MaxPooling1D(pool_size= 4, strides=4))
# 4 
model_1D.add(Conv1D(64, kernel_size= 5 , strides=1 , activation='relu'))
model_1D.add(MaxPooling1D(pool_size= 2, strides=2))
# 5
model_1D.add(Conv1D(128, kernel_size= 5 , strides= 1 , activation='relu'))
model_1D.add(MaxPooling1D(pool_size= 2, strides= 2))
# 6
model_1D.add(Conv1D(128, kernel_size= 5 , strides= 1 , activation='relu'))
model_1D.add(MaxPooling1D(pool_size= 2, strides= 2))
model_1D.add(Dense(9 , activation='relu')) # K
# model_1D.add(Dense(7, activation='softmax'))

model_1D.build(input_shape = (None,7380, 128000))
#print(model_1D.summary())

# ----------------------- 2D CNN ----------------------

model_2D = Sequential()
model_2D.add(Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu'))
model_2D.add(Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu'))
model_2D.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
model_2D.add(Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu'))
model_2D.add(Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu'))
model_2D.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
model_2D.add(Dense(9 , activation='relu')) # K
# model_2D.add(Dense(7, activation='softmax'))
model_2D.build (input_shape = (7380, 128, 251, 1))
# print(model_2D.summary())

# ----------------------- Merged CNN ----------------------

merged = Concatenate([model_2D, model_1D])
model_combined = Sequential()
model_combined.add(merged)
model_combined.add(Dense(7, activation='softmax'))
opt = Adam(lr=0.0001)
#opt = Adagrad(lr=0.1, epsilon=1e-08, decay=0.0)
#opt = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
#opt = RMSprop(lr=0.001, rho=0.9)
model_combined.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])
print(model_combined.summary())

1 Ответ

0 голосов
/ 07 мая 2020

Вместо последовательного API вы можете попробовать функциональный API, предоставив форму ввода для 1D-CNN и 2D-CNN только на первом уровне. Затем, добавив выравнивание перед плотным слоем в 1D-CNN и 2D-CNN, вы решите свою проблему.
Вы можете следовать нижеприведенному измененному коду.
1D-CNN: 

# ----------------------- 1D CNN ----------------------
in_1D = Input((7380, 128000))
# 1
model_1D = Conv1D(32, kernel_size= 5 , strides=1, activation='relu')(in_1D)
model_1D = MaxPooling1D(pool_size= 4, strides=4)(model_1D)
# 2
model_1D = Conv1D(32, kernel_size= 5 , strides=1 , activation='relu')(model_1D)
model_1D = MaxPooling1D(pool_size= 4, strides=4)(model_1D)
# 3
model_1D = Conv1D(64, kernel_size= 5 , strides=1 , activation='relu')(model_1D)
model_1D = MaxPooling1D(pool_size= 4, strides=4)(model_1D)
# 4
model_1D = Conv1D(64, kernel_size= 5 , strides=1 , activation='relu')(model_1D)
model_1D = MaxPooling1D(pool_size= 2, strides=2)(model_1D)
# 5
model_1D = Conv1D(128, kernel_size= 5 , strides= 1 , activation='relu')(model_1D)
model_1D = MaxPooling1D(pool_size= 2, strides= 2)(model_1D)
# 6
model_1D = Conv1D(128, kernel_size= 5 , strides= 1 , activation='relu')(model_1D)
model_1D = MaxPooling1D(pool_size= 2, strides= 2)(model_1D)
model_1D = Flatten()(model_1D)
model_1D = Dense(9 , activation='softmax')(model_1D)

2D-CNN: 

# ----------------------- 2D CNN ----------------------
in_2D = Input((7380, 128, 251))
model_2D = Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu')(in_2D)
model_2D = Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu')(model_2D)
model_2D = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(model_2D)
model_2D = Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu')(model_2D)
model_2D = Conv2D(32, kernel_size=(3, 3) , strides=(1,1), activation='relu')(model_2D)
model_2D = MaxPooling2D(pool_size=(2, 2), strides=(2, 2))(model_2D)
model_2D = Flatten()(model_2D)
model_2D = Dense(9 , activation='relu')(model_2D)

Слияние: 

merged = Concatenate()([model_2D, model_1D])
output = Dense(7, activation='softmax')(merged)

model_combined = Model(inputs=[in_2D, in_1D], outputs=[output])

model_combined.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

model_combined.summary()

Выход: 

Model: "model_1"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            (None, 7380, 128000) 0                                            
__________________________________________________________________________________________________
conv1d_1 (Conv1D)               (None, 7376, 32)     20480032    input_1[0][0]                    
__________________________________________________________________________________________________
max_pooling1d_1 (MaxPooling1D)  (None, 1844, 32)     0           conv1d_1[0][0]                   
__________________________________________________________________________________________________
conv1d_2 (Conv1D)               (None, 1840, 32)     5152        max_pooling1d_1[0][0]            
__________________________________________________________________________________________________
max_pooling1d_2 (MaxPooling1D)  (None, 460, 32)      0           conv1d_2[0][0]                   
__________________________________________________________________________________________________
conv1d_3 (Conv1D)               (None, 456, 64)      10304       max_pooling1d_2[0][0]            
__________________________________________________________________________________________________
input_2 (InputLayer)            (None, 7380, 128, 25 0                                            
__________________________________________________________________________________________________
max_pooling1d_3 (MaxPooling1D)  (None, 114, 64)      0           conv1d_3[0][0]                   
__________________________________________________________________________________________________
conv2d_1 (Conv2D)               (None, 7378, 126, 32 72320       input_2[0][0]                    
__________________________________________________________________________________________________
conv1d_4 (Conv1D)               (None, 110, 64)      20544       max_pooling1d_3[0][0]            
__________________________________________________________________________________________________
conv2d_2 (Conv2D)               (None, 7376, 124, 32 9248        conv2d_1[0][0]                   
__________________________________________________________________________________________________
max_pooling1d_4 (MaxPooling1D)  (None, 55, 64)       0           conv1d_4[0][0]                   
__________________________________________________________________________________________________
max_pooling2d_1 (MaxPooling2D)  (None, 3688, 62, 32) 0           conv2d_2[0][0]                   
__________________________________________________________________________________________________
conv1d_5 (Conv1D)               (None, 51, 128)      41088       max_pooling1d_4[0][0]            
__________________________________________________________________________________________________
conv2d_3 (Conv2D)               (None, 3686, 60, 32) 9248        max_pooling2d_1[0][0]            
__________________________________________________________________________________________________
max_pooling1d_5 (MaxPooling1D)  (None, 25, 128)      0           conv1d_5[0][0]                   
__________________________________________________________________________________________________
conv2d_4 (Conv2D)               (None, 3684, 58, 32) 9248        conv2d_3[0][0]                   
__________________________________________________________________________________________________
conv1d_6 (Conv1D)               (None, 21, 128)      82048       max_pooling1d_5[0][0]            
__________________________________________________________________________________________________
max_pooling2d_2 (MaxPooling2D)  (None, 1842, 29, 32) 0           conv2d_4[0][0]                   
__________________________________________________________________________________________________
max_pooling1d_6 (MaxPooling1D)  (None, 10, 128)      0           conv1d_6[0][0]                   
__________________________________________________________________________________________________
flatten_2 (Flatten)             (None, 1709376)      0           max_pooling2d_2[0][0]            
__________________________________________________________________________________________________
flatten_1 (Flatten)             (None, 1280)         0           max_pooling1d_6[0][0]            
__________________________________________________________________________________________________
dense_2 (Dense)                 (None, 9)            15384393    flatten_2[0][0]                  
__________________________________________________________________________________________________
dense_1 (Dense)                 (None, 9)            11529       flatten_1[0][0]                  
__________________________________________________________________________________________________
concatenate_1 (Concatenate)     (None, 18)           0           dense_2[0][0]                    
                                                                 dense_1[0][0]                    
__________________________________________________________________________________________________
dense_3 (Dense)                 (None, 7)            133         concatenate_1[0][0]              
==================================================================================================
Total params: 36,135,287
Trainable params: 36,135,287
Non-trainable params: 0
...