Tensorflow：-不支持的操作数类型：“Sequential”和“Sequential”

Question

我正在遵循Siamese Network教程，使用tensorflow构建人脸识别模型

def build_net(img_shape):
    """
    :type img_shape: tuple. Shape of input image. Here is(1,height, width). 1 because pgm file only has one channel.
    :rtype:tensorflow Sequential
    """
    model = tf.keras.Sequential()
    # convolution layer 1
    model.add(tf.keras.layers.Conv2D(filters = 16, kernel_size = 3, strides = 1, activation = "relu", input_shape = img_shape, data_format = "channels_first"))
    model.add(tf.keras.layers.MaxPool2D(pool_size = 2))
    model.add(tf.keras.layers.Dropout(0.1))
    # convolution layer 2
    model.add(tf.keras.layers.Conv2D(filters = 32, kernel_size = 3, strides = 1))
    model.add(tf.keras.layers.MaxPool2D(pool_size = 2))
    model.add(tf.keras.layers.Dropout(0.1))

    model.add(tf.keras.layers.Flatten())
    model.add(tf.keras.layers.Dense(1024))
    model.add(tf.keras.layers.Dropout(0.25))
    model.add(tf.keras.layers.Dense(512, activation='relu'))
    # deep face mentioned that there are 67 points to detect on a human face, so use 70 features.
    model.add(tf.keras.layers.Dense(70, activation='relu'))
    print(model.summary())
    return model

并定义一个dist来计算两个输出向量之间的距离。

im1_features = build_net(input_dim)
im2_features = build_net(input_dim)
dist = tf.keras.layers.Lambda(lambda tensors: tf.keras.backend.abs[tensors[0] - tensors[1]])([im1_features, im2_features])

dist中发生错误

  File "e:\School\AIAS\proj\build_model.py", line 102, in <lambda>
    dist = tf.keras.layers.Lambda(lambda tensors: tf.keras.backend.abs[tensors[0] - tensors[1]])([im1_features, im2_features])
TypeError: unsupported operand type(s) for -: 'Sequential' and 'Sequential'

如何让函数build_net返回向量而不是顺序对象？

更新

我将代码更改为：

def build_net(img_shape):
    """
    :type img_shape: tuple. Shape of input image. Here is(1,height, width). 1 because pgm file only has one channel.
    :rtype:tensorflow Sequential
    """
    model = tf.keras.Sequential()
    # convolution layer 1
    model.add(tf.keras.layers.Conv2D(filters = 16, kernel_size = 3, strides = 1, activation = "relu", input_shape = img_shape, data_format = "channels_first"))
    model.add(tf.keras.layers.MaxPool2D(pool_size = 2))
    model.add(tf.keras.layers.Dropout(0.1))
    # convolution layer 2
    model.add(tf.keras.layers.Conv2D(filters = 32, kernel_size = 3, strides = 1))
    model.add(tf.keras.layers.MaxPool2D(pool_size = 2))
    model.add(tf.keras.layers.Dropout(0.1))

    model.add(tf.keras.layers.Flatten())
    model.add(tf.keras.layers.Dense(1024))
    model.add(tf.keras.layers.Dropout(0.25))
    model.add(tf.keras.layers.Dense(512, activation='relu'))
    # deep face mentioned that there are 67 points to detect on a human face, so use 70 features.
    model.add(tf.keras.layers.Dense(70, activation='relu'))
    img = tf.keras.Input(shape = img_shape)
    res = model(img)
    return res

现在，它返回以下内容

​

错误变成：function' object is not subscriptable

我还是找不到矢量

Answer 1

尝试像这样调用tf.keras.abs：

tf.keras.backend.abs(
    x
)

不

tf.keras.backend.abs[
    x
]

它是一个函数，而不是一个数组。这解决你的问题了吗？