Распакуйте слой tf.keras.Model

Question

import tensorflow as tf

input_shape = (224, 224, 3)
inputs = tf.keras.layers.Input(shape=input_shape)
base_model = tf.keras.applications.ResNet50(input_shape=input_shape, include_top=False, weights='imagenet')

# Get the output of conv4
hidden_layer = base_model.get_layer('conv4_block6_out').output
stack = tf.keras.Model(inputs=base_model.input, outputs=[hidden_layer], name='conv4_block6_out')
conv4_block6_out = stack(inputs)
x = tf.keras.layers.Softmax()(conv4_block6_out)
model = tf.keras.Model(inputs, x)
model.summary()

Вывод:

Model: "model_3"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 224, 224, 3)]     0         
_________________________________________________________________
conv4_block6_out (Model)     (None, 14, 14, 1024)      8589184   
_________________________________________________________________
softmax_4 (Softmax)          (None, 14, 14, 1024)      0         
=================================================================
Total params: 8,589,184
Trainable params: 8,558,592
Non-trainable params: 30,592

Конечная модель имеет только 3 слоя.

Мой вопрос: как сделать все слои conv4_block6_out (Model) первыми Слои конечной модели?

Заранее спасибо.

Answer 1

Нашел ответ здесь .

import tensorflow as tf

input_shape = (224, 224, 3)
base_model = tf.keras.applications.ResNet50(input_shape=input_shape, include_top=False, weights='imagenet')

# Get the output of conv4
conv4_block6_out = base_model.get_layer('conv4_block6_out').output
x = tf.keras.layers.Softmax()(conv4_block6_out)
model = tf.keras.Model(base_model.input, x)
model.summary()

Результаты:

Model: "model_3"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_2 (InputLayer)            [(None, 224, 224, 3) 0                                            
__________________________________________________________________________________________________
conv1_pad (ZeroPadding2D)       (None, 230, 230, 3)  0           input_2[0][0]                    
__________________________________________________________________________________________________
conv1_conv (Conv2D)             (None, 112, 112, 64) 9472        conv1_pad[0][0]                  
__________________________________________________________________________________________________
conv1_bn (BatchNormalization)   (None, 112, 112, 64) 256         conv1_conv[0][0]                 
__________________________________________________________________________________________________
conv1_relu (Activation)         (None, 112, 112, 64) 0           conv1_bn[0][0]                   
...
__________________________________________________________________________________________________
conv4_block6_out (Activation)   (None, 14, 14, 1024) 0           conv4_block6_add[0][0]           
__________________________________________________________________________________________________
softmax_4 (Softmax)             (None, 14, 14, 1024) 0           conv4_block6_out[0][0]           
==================================================================================================
Total params: 8,589,184
Trainable params: 8,558,592
Non-trainable params: 30,592