ValueError：形状（无，1）和（无，2）不兼容

Question

我正在训练一个面部表情（愤怒 vs 快乐）模型。最后一个密集输出层以前是 1，但是当我预测图像时，它的输出始终为 1，准确度为 64%。所以我将其更改为 2 以获取 2 个输出。但现在我收到这个错误::

Epoch 1/15

---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

<ipython-input-54-9c7272c38dcb> in <module>()
     11     epochs=epochs,
     12     validation_data = val_data_gen,
---> 13     validation_steps = validation_steps,
     14 
     15 )

10 frames

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py in wrapper(*args, **kwargs)
    966           except Exception as e:  # pylint:disable=broad-except
    967             if hasattr(e, "ag_error_metadata"):
--> 968               raise e.ag_error_metadata.to_exception(e)
    969             else:
    970               raise

ValueError: in user code:

    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:571 train_function  *
        outputs = self.distribute_strategy.run(
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:951 run  **
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2290 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2649 _call_for_each_replica
        return fn(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:533 train_step  **
        y, y_pred, sample_weight, regularization_losses=self.losses)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/compile_utils.py:205 __call__
        loss_value = loss_obj(y_t, y_p, sample_weight=sw)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/losses.py:143 __call__
        losses = self.call(y_true, y_pred)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/losses.py:246 call
        return self.fn(y_true, y_pred, **self._fn_kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/losses.py:1527 categorical_crossentropy
        return K.categorical_crossentropy(y_true, y_pred, from_logits=from_logits)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/backend.py:4561 categorical_crossentropy
        target.shape.assert_is_compatible_with(output.shape)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py:1117 assert_is_compatible_with
        raise ValueError("Shapes %s and %s are incompatible" % (self, other))

    ValueError: Shapes (None, 1) and (None, 2) are incompatible

相关代码是：

    model = Sequential([
    Conv2D(32,3, activation='relu', input_shape=(48,48,1)),
    BatchNormalization(),
    MaxPooling2D(pool_size=(3, 3)),
  
    Flatten(),
    Dense(512, activation='relu'),
    Dense(2,activation='softmax')
])
model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['accuracy'])


model.summary()

Model: "sequential_4"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_6 (Conv2D)            (None, 46, 46, 32)        320       
_________________________________________________________________
batch_normalization_4 (Batch (None, 46, 46, 32)        128       
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 15, 15, 32)        0         
_________________________________________________________________
flatten_4 (Flatten)          (None, 7200)              0         
_________________________________________________________________
dense_8 (Dense)              (None, 512)               3686912   
_________________________________________________________________
dense_9 (Dense)              (None, 2)                 1026      
=================================================================
Total params: 3,688,386
Trainable params: 3,688,322
Non-trainable params: 64
_________________________________________________________________


epochs = 15
steps_per_epoch = train_data_gen.n//train_data_gen.batch_size
validation_steps = val_data_gen.n//val_data_gen.batch_size



history = model.fit(
    x=train_data_gen,
    steps_per_epoch=steps_per_epoch,
    epochs=epochs,
    validation_data = val_data_gen,
    validation_steps = validation_steps,
    
)

Answer 1

将 Categorical Cross Entropy 更改为 Binary Cross Entropy，因为您的输出标签是二进制的。也将 Softmax 更改为 Sigmoid，因为 Sigmoid 是二进制数据的正确激活函数

Answer 2

如果您的数据集是image_dataset_from_directory使用加载的，请使用label_mode='categorical'

train_ds = tf.keras.preprocessing.image_dataset_from_directory(
  path,
  label_mode='categorical'
)

或加载flow_from_directory，flow_from_dataframe然后使用class_mode='categorical'

train_ds = ImageDataGenerator.flow_from_directory(
  path,
  class_mode='categorical'
)

Answer 3

您可以将标签从二进制值更改为分类值，然后继续使用相同的代码。例如，

from keras.utils import to_categorical
one_hot_label = to_cateorical(input_labels)
# change to [1, 0, 0,..., 0]  --> [[0, 1], [1, 0], ..., [1, 0]]

您可以通过此链接更好地了解Keras API。

如果您想对两个类使用分类交叉熵，请使用 softmax 并进行一种热编码。对于二元分类，您可以使用二元交叉熵，如前面提到的使用 sigmoid 激活函数提到的答案。

1. 分类交叉熵：

model = Sequential([
    Conv2D(32,3, activation='relu', input_shape=(48,48,1)),
    BatchNormalization(),
    MaxPooling2D(pool_size=(3, 3)),

    Flatten(),
    Dense(512, activation='relu'),
    Dense(2,activation='softmax')  # activation change
])
model.compile(optimizer='adam',
              loss='categorical_crossentropy', # Loss
              metrics=['accuracy'])

2. 二元交叉熵

model = Sequential([
    Conv2D(32,3, activation='relu', input_shape=(48,48,1)),
    BatchNormalization(),
    MaxPooling2D(pool_size=(3, 3)),

    Flatten(),
    Dense(512, activation='relu'),
    Dense(1,activation='sigmoid') #activation change
])
model.compile(optimizer='adam',
              loss='binary_crossentropy', # Loss
              metrics=['accuracy'])

Answer 4

我面临着同样的问题，我的形状是

shape of X (271, 64, 64, 3)
shape of y (271,)
shape of trainX (203, 64, 64, 3)
shape of trainY (203, 1)
shape of testX (68, 64, 64, 3)
shape of testY (68, 1)

和

loss="categorical_crossentropy"

我把它改成

loss="sparse_categorical_crossentropy"

它对我来说就像一种魅力

Answer 5

在多标签分类的情况下，从“categorical_crossentropy”更改为“sparse_categorical_crossentropy”对我有用