给keras层命名,并提取中间层输出值,保存到文档的实例

时间:2021-05-22

更新:

感谢评论区提供的方案。

采用model.summary(),model.get_config()和for循环均可获得Keras的层名。

示例如下图

对于keras特定层的命名,只需在层内添加 name 即可

model.add(Activation('softmax',name='dense_1') ) # 注意 name 要放于函数内#提取中间层from keras.models import Modelimport keraslayer_name = 'dense_1' #获取层的名称intermediate_layer_model = Model(inputs=model.input, outputs=model.get_layer(layer_name).output)#创建的新模型intermediate_output = intermediate_layer_model.predict(X_test)doc = open(r'C://Users//CCUT04//Desktop//1.txt','w')for i in intermediate_output: print(i) print(i , file = doc)doc.close()

补充知识:关于用keras提取NN中间layer输出

Build model...__________________________________________________________________________________________________Layer (type) Output Shape Param # Connected to ==================================================================================================main_input (InputLayer) (None, 89, 39) 0 __________________________________________________________________________________________________cropping1d_1 (Cropping1D) (None, 85, 39) 0 main_input[0][0] __________________________________________________________________________________________________cropping1d_2 (Cropping1D) (None, 85, 39) 0 main_input[0][0] __________________________________________________________________________________________________cropping1d_3 (Cropping1D) (None, 85, 39) 0 main_input[0][0] __________________________________________________________________________________________________cropping1d_4 (Cropping1D) (None, 85, 39) 0 main_input[0][0] __________________________________________________________________________________________________cropping1d_5 (Cropping1D) (None, 85, 39) 0 main_input[0][0] __________________________________________________________________________________________________concatenate_1 (Concatenate) (None, 85, 195) 0 cropping1d_1[0][0] cropping1d_2[0][0] cropping1d_3[0][0] cropping1d_4[0][0] cropping1d_5[0][0] __________________________________________________________________________________________________fc1 (BatchNormalization) (None, 85, 195) 780 concatenate_1[0][0] __________________________________________________________________________________________________fc2 (Bidirectional) (None, 85, 2048) 9994240 fc1[0][0] __________________________________________________________________________________________________fc3 (BatchNormalization) (None, 85, 2048) 8192 fc2[0][0] __________________________________________________________________________________________________global_average_pooling1d_1 (Glo (None, 2048) 0 fc3[0][0] __________________________________________________________________________________________________main_output (Dense) (None, 2) 4098 global_average_pooling1d_1[0][0] ==================================================================================================Total params: 10,007,310Trainable params: 10,002,824Non-trainable params: 4,486__________________________________________________________________________________________________

假设我网络层数是上面这个结构.

如果我想得到pooling的输出, keras上有两张方法。

intermediate_layer_model = Model(inputs=model.input,outputs=model.get_layer(str('global_average_pooling1d_1')).output)#model.summary()#model.get_layer(str('cropping1d_1'))intermediate_output = intermediate_layer_model.predict(data)

data是你的输入所用的数据....

from keras import backend as Kget_11rd_layer_output = K.function([model.layers[0].input], [model.layers[10].output])layer_output = get_11rd_layer_output([data])[0]

我这里第10层是Pooling层.

这两个代码的output是一样的..

一般我看人用的都是第二个...

以上这篇给keras层命名,并提取中间层输出值,保存到文档的实例就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持。

声明:本页内容来源网络,仅供用户参考;我单位不保证亦不表示资料全面及准确无误,也不保证亦不表示这些资料为最新信息,如因任何原因,本网内容或者用户因倚赖本网内容造成任何损失或损害,我单位将不会负任何法律责任。如涉及版权问题,请提交至online#300.cn邮箱联系删除。

相关文章