Question

我是机器学习的新手。我在微调VGG16模型上遵循了tutorial。

此代码可以很好地加载模型：

vgg_model = tensorflow.keras.applications.vgg16.VGG16()

但出现此错误：

TypeError: The added layer must be an instance of class Layer. Found: <tensorflow.python.keras.engine.input_layer.InputLayer object at 0x000001FA104CBB70>

运行此代码时：

model = Sequential()
for layer in vgg_model.layers[:-1]:
    model.add(layer)

依赖项：

Keras 2.2.3
Tensorflow 1.12.0
tensorflow-gpu1.12.0
Python 3.6.0

我正在关注此blog，但我想使用VGG16。

任何解决此问题的帮助将不胜感激。非常感谢。

Answer 1

这将不起作用，因为已将tensorflow.keras图层添加到了keras模型中。

vgg_model = tensorflow.keras.applications.vgg16.VGG16()
model = keras.Sequential()
model.add(vgg_model.layers[0])

实例化tensorflow.keras.Sequential（）。这将起作用。

model = tensorflow.keras.Sequential()
model.add(vgg_model.layers[0])

Answer 2

您无需创建InputLayer，只需以与Conv2D /其他层相同的方式导入BatchNormalization层，例如：

<section>
      <div class="user">Need some help with HTML</div>
      <div class="user">And CSS maybe</div>
      <div class="user">Want it to work with long-lenth messages as well, you know. And in all the browsers, even IE...</div>
      <div>Sure</div>
      <div>Lets test this one</div>
      <div>Quite a good in terms of browser support in 2019</div>
      <div class="user">Awsome!</div>
      <div class="user">Thank you so much</div>
      <div>You are welcome</div>
      <div class="user">Goodbye</div>
</section>

<footer>
<p>Sticky Footer</p>
</footer>

代替将其导入为独立的Keras图层，即：

from tensorflow.keras.layers import Conv2D, Flatten, MaxPooling2D, Dropout, BatchNormalization

Answer 3

要添加到@Manoj Mohan的答案中，您可以使用input_layer model中的input_layer将Keras添加到layers中，如下所示：

import keras
from keras.models import Sequential
from keras.layers import InputLayer

model = Sequential()
model.add(InputLayer(input_shape=shape, name=name))
....

如果您使用内置的TensorFlow Keras，则导入是不同的，其他情况仍然相同

import tensorflow as tf
import tensorflow.keras as keras
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import InputLayer

model = Sequential()
model.add(InputLayer(input_shape=shape, name=name))
....

主要部分，如果要将图层导入到顺序模型中，可以使用以下语法。

import keras
from keras.models import Sequential, load_model
from keras import optimizers
from keras.applications.vgg16 import VGG16
from keras.applications.vgg19 import VGG19

# For VGG16 loading to sequential model  
model = Sequential(VGG16().layers)
# For VGG19 loading to sequential model  
model = Sequential(VGG19().layers)

添加的层必须是类Layer的实例。找到：<tensorflow.python.keras.engine.input_layer.inputlayer object =“” at =“” 0x000001fa104cbb70 =“”>

3 个答案: