快捷導(dǎo)航

在Keras中CNN聯(lián)合LSTM進(jìn)行分類實(shí)例

更新時(shí)間：2020年06月29日 10:31:25 作者：FQ_G

這篇文章主要介紹了在Keras中CNN聯(lián)合LSTM進(jìn)行分類實(shí)例，具有很好的參考價(jià)值，希望對(duì)大家有所幫助。一起跟隨小編過來看看吧

我就廢話不多說，大家還是直接看代碼吧~

def get_model():
  n_classes = 6
  inp=Input(shape=(40, 80))
  reshape=Reshape((1,40,80))(inp)
 #  pre=ZeroPadding2D(padding=(1, 1))(reshape)
  # 1
  conv1=Convolution2D(32, 3, 3, border_mode='same',init='glorot_uniform')(reshape)
  #model.add(Activation('relu'))
  l1=LeakyReLU(alpha=0.33)(conv1)
 
  conv2=ZeroPadding2D(padding=(1, 1))(l1)
  conv2=Convolution2D(32, 3, 3, border_mode='same',init='glorot_uniform')(conv2)
  #model.add(Activation('relu'))
  l2=LeakyReLU(alpha=0.33)(conv2)
 
  m2=MaxPooling2D((3, 3), strides=(3, 3))(l2)
  d2=Dropout(0.25)(m2)
  # 2
  conv3=ZeroPadding2D(padding=(1, 1))(d2)
  conv3=Convolution2D(64, 3, 3, border_mode='same',init='glorot_uniform')(conv3)
  #model.add(Activation('relu'))
  l3=LeakyReLU(alpha=0.33)(conv3)
 
  conv4=ZeroPadding2D(padding=(1, 1))(l3)
  conv4=Convolution2D(64, 3, 3, border_mode='same',init='glorot_uniform')(conv4)
  #model.add(Activation('relu'))
  l4=LeakyReLU(alpha=0.33)(conv4)
 
  m4=MaxPooling2D((3, 3), strides=(3, 3))(l4)
  d4=Dropout(0.25)(m4)
  # 3
  conv5=ZeroPadding2D(padding=(1, 1))(d4)
  conv5=Convolution2D(128, 3, 3, border_mode='same',init='glorot_uniform')(conv5)
  #model.add(Activation('relu'))
  l5=LeakyReLU(alpha=0.33)(conv5)
 
  conv6=ZeroPadding2D(padding=(1, 1))(l5)
  conv6=Convolution2D(128, 3, 3, border_mode='same',init='glorot_uniform')(conv6)
  #model.add(Activation('relu'))
  l6=LeakyReLU(alpha=0.33)(conv6)
 
  m6=MaxPooling2D((3, 3), strides=(3, 3))(l6)
  d6=Dropout(0.25)(m6)
  # 4
  conv7=ZeroPadding2D(padding=(1, 1))(d6)
  conv7=Convolution2D(256, 3, 3, border_mode='same',init='glorot_uniform')(conv7)
  #model.add(Activation('relu'))
  l7=LeakyReLU(alpha=0.33)(conv7)
 
  conv8=ZeroPadding2D(padding=(1, 1))(l7)
  conv8=Convolution2D(256, 3, 3, border_mode='same',init='glorot_uniform')(conv8)
  #model.add(Activation('relu'))
  l8=LeakyReLU(alpha=0.33)(conv8)
  g=GlobalMaxPooling2D()(l8)
  print("g=",g)
  #g1=Flatten()(g)
  lstm1=LSTM(
    input_shape=(40,80),
    output_dim=256,
    activation='tanh',
    return_sequences=False)(inp)
  dl1=Dropout(0.3)(lstm1)
  
  den1=Dense(200,activation="relu")(dl1)
  #model.add(Activation('relu'))
  #l11=LeakyReLU(alpha=0.33)(d11)
  dl2=Dropout(0.3)(den1)
 
#   lstm2=LSTM(
#     256,activation='tanh',
#     return_sequences=False)(lstm1)
#   dl2=Dropout(0.5)(lstm2)
  print("dl2=",dl1)
  g2=concatenate([g,dl2],axis=1)
  d10=Dense(1024)(g2)
  #model.add(Activation('relu'))
  l10=LeakyReLU(alpha=0.33)(d10)
  l10=Dropout(0.5)(l10)
  l11=Dense(n_classes, activation='softmax')(l10)
 
  model=Model(input=inp,outputs=l11)
  model.summary()
  #編譯model
  adam = keras.optimizers.Adam(lr = 0.0005, beta_1=0.95, beta_2=0.999,epsilon=1e-08)
  #adam = keras.optimizers.Adam(lr = 0.001, beta_1=0.95, beta_2=0.999,epsilon=1e-08)
  #sgd = keras.optimizers.SGD(lr = 0.001, decay = 1e-06, momentum = 0.9, nesterov = False)
 
  #reduce_lr = ReduceLROnPlateau(monitor = 'loss', factor = 0.1, patience = 2,verbose = 1, min_lr = 0.00000001, mode = 'min')
  model.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])
  
  return model

補(bǔ)充知識(shí)：keras中如何將不同的模型聯(lián)合起來（以cnn/lstm為例）

可能會(huì)遇到多種模型需要揉在一起，如cnn和lstm，而我一般在keras框架下開局就是一句

model = Sequential()

然后model.add ，model.add , ......到最后

model.compile(loss=["mae"], optimizer='adam',metrics=[mape])

這突然要把模型加起來，這可怎么辦？

以下示例代碼是將cnn和lstm聯(lián)合起來，先是由cnn模型卷積池化得到特征，再輸入到lstm模型中得到最終輸出

import os
import keras
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
from keras.models import Model
from keras.layers import *
from matplotlib import pyplot
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
from keras.layers import Dense,Dropout,Activation,Convolution2D,MaxPooling2D,Flatten
from keras.layers import LSTM
def design_model():
  # design network
  inp=Input(shape=(11,5))
  reshape=Reshape((11,5,1))(inp)
  conv1=Convolution2D(32,3,3,border_mode='same',init='glorot_uniform')(reshape)
  print(conv1)
  l1=Activation('relu')(conv1)
  conv2=Convolution2D(64,3,3, border_mode='same',)(l1)
  l2=Activation('relu')(conv2)
  print(l2)
  m2=MaxPooling2D(pool_size=(2, 2), border_mode='valid')(l2)
  print(m2)
  reshape1=Reshape((10,64))(m2)
  lstm1=LSTM(input_shape=(10,64),output_dim=30,activation='tanh',return_sequences=False)(reshape1)
  dl1=Dropout(0.3)(lstm1)
  # den1=Dense(100,activation="relu")(dl1)
  den2=Dense(1,activation="relu")(dl1)
  model=Model(input=inp,outputs=den2)
  model.summary() #打印出模型概況
  adam = keras.optimizers.Adam(lr = 0.001, beta_1=0.95, beta_2=0.999,epsilon=1e-08)
  model.compile(loss=["mae"], optimizer=adam,metrics=['mape'])
  return model
model=design_model()
history = model.fit(train_x, train_y, epochs=epochs, batch_size=batch_size, validation_data=[test_x, test_y],verbose=2, shuffle=True)
# #save LeNet_model_files after train
model.save('model_trained.h5')

以上示例代碼中cnn和lstm是串聯(lián)即cnn輸出作為lstm的輸入，一條路線到底

如果想實(shí)現(xiàn)并聯(lián)，即分開再匯總到一起

可用concatenate函數(shù)把cnn的輸出端和lstm的輸出端合并起來，后面再接上其他層，完成整個(gè)模型圖的構(gòu)建。