张量形状与测试中的目标值不匹配

Question

我正在创建一个基于 MobileNetV2 的模型：

# UNQ_C2
# GRADED FUNCTION
def alpaca_model(image_shape=IMG_SIZE, data_augmentation=data_augmenter()):
    ''' Define a tf.keras model for binary classification out of the MobileNetV2 model
    Arguments:
        image_shape -- Image width and height
        data_augmentation -- data augmentation function
    Returns:
    Returns:
        tf.keras.model
    '''
    
    
    input_shape = image_shape + (3,)
    
    ### START CODE HERE
    
    base_model = tf.keras.applications.MobileNetV2(input_shape=None,
                                                   include_top=None, # <== Important!!!!
                                                   weights=None) # From imageNet
    
    # Freeze the base model by making it non trainable
    base_model.trainable = False 

    # create the input layer (Same as the imageNetv2 input size)
    inputs = tf.keras.Input(shape=input_shape) 
    
    print("inputs size: ", str(inputs.shape))
    
    # apply data augmentation to the inputs
    x = data_augmentation(inputs)
    print("x size: ", str(x.shape))
    
    # data preprocessing using the same weights the model was trained on
    x = preprocess_input(x) 
    print("x size: ", str(x.shape))
    
    # set training to False to avoid keeping track of statistics in the batch norm layer
    x = base_model(x, training=False) 
    
    # Add the new Binary classification layers
    # use global avg pooling to summarize the info in each channel
    x = tf.keras.layers.GlobalAveragePooling2D()(x) 
    print("x size: ", str(x.shape))
    
    #include dropout with probability of 0.2 to avoid overfitting
    x = tf.keras.layers.Dropout(0.2)(x)
    print("x size: ", str(x.shape))
    
    # create a prediction layer with one neuron (as a classifier only needs one)
    prediction_layer = tf.keras.layers.Dense(2 ,activation='softmax')(x)
    print("prediction_layer size: ", str(prediction_layer.shape))
    
    ### END CODE HERE
    
    outputs = prediction_layer
    model = tf.keras.Model(inputs, outputs)
    
    return model

但是，测试脚本

model2 = alpaca_model(IMG_SIZE, data_augmentation)

from test_utils import summary, comparator

alpaca_summary = [['InputLayer', [(None, 160, 160, 3)], 0],
                    ['Sequential', (None, 160, 160, 3), 0],
                    ['TensorFlowOpLayer', [(None, 160, 160, 3)], 0],
                    ['TensorFlowOpLayer', [(None, 160, 160, 3)], 0],
                    ['Functional', (None, 5, 5, 1280), 2257984],
                    ['GlobalAveragePooling2D', (None, 1280), 0],
                    ['Dropout', (None, 1280), 0, 0.2],
                    ['Dense', (None, 1), 1281, 'linear']] #linear is the default activation

    comparator(summary(model2), alpaca_summary)
    
    for layer in summary(model2):
        print(layer)

给出以下错误：

Test failed 
 Expected value 

 ['Functional', (None, 5, 5, 1280), 2257984] 

 does not match the input value: 

 ['Functional', (None, None, None, 1280), 2257984]
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-67-0346cb4bf847> in <module>
     10                     ['Dense', (None, 1), 1281, 'linear']] #linear is the default activation
     11 
---> 12 comparator(summary(model2), alpaca_summary)
     13 
     14 for layer in summary(model2):

~/work/W2A2/test_utils.py in comparator(learner, instructor)
     19                   "\n\n does not match the input value: \n\n",
     20                   colored(f"{a}", "red"))
---> 21             raise AssertionError("Error in test")
     22     print(colored("All tests passed!", "green"))
     23 

AssertionError: Error in test

我每一步都打印出张量的形状。我看不出有什么问题，因为[none, 160, 160, 3] 图像最终会在被输入二进制分类器之前被压平为[none, 1280]。

我不确定这里发生了什么。我对 python 和 CNN 很陌生。谢谢。

Answer 1

把模型改成这个：

base_model = tf.keras.applications.MobileNetV2(input_shape=input_shape,
                                                   include_top=None, # <== Important!!!!
                                                   weights='imagenet') # From imageNet