Keras：ValueError：logits和labels必须具有相同的形状（（None，2）与（None，1））。

Question

我一直在使用著名的狗与猫Kaggle数据集，并试图创建自己的CNN模型。我是第一次使用image_dataset_from_directory方法导入数据集，该方法将数据集配置为两个文件夹，分别包含猫和狗的图像。

以下是该模型的代码：

from keras.models import Sequential
from keras.layers import Conv2D,MaxPooling2D,\
     Dropout,Flatten,Dense,Activation,\
     BatchNormalization
model=Sequential()
model.add(Conv2D(32,(3,3),activation='relu',input_shape=(128,128,3)))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Conv2D(64,(3,3),activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Conv2D(128,(3,3),activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512,activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(2,activation='sigmoid'))
model.compile(loss = 'binary_crossentropy',
  optimizer='adam',metrics=['accuracy'])

下面是数据集的代码：

Dataset = tf.keras.preprocessing.image_dataset_from_directory(
    TRAIN_DIR,
    labels="inferred",
    label_mode="binary",
    class_names=None,
    color_mode="rgb",
    batch_size=32,
    image_size=(128, 128),
    shuffle=True,
    seed=None,
    validation_split=None,
    subset=None,
    interpolation="bilinear",
    follow_links=False,
)

在运行拟合函数以训练我的卷积神经网络后，出现了以下错误：

ValueError: in user code:

    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\training.py:806 train_function  *
        return step_function(self, iterator)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\training.py:796 step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:1211 run
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2585 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2945 _call_for_each_replica
        return fn(*args, **kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\training.py:789 run_step  **
        outputs = model.train_step(data)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\training.py:748 train_step
        loss = self.compiled_loss(
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\compile_utils.py:204 __call__
        loss_value = loss_obj(y_t, y_p, sample_weight=sw)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\losses.py:149 __call__
        losses = ag_call(y_true, y_pred)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\losses.py:253 call  **
        return ag_fn(y_true, y_pred, **self._fn_kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\util\dispatch.py:201 wrapper
        return target(*args, **kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\losses.py:1605 binary_crossentropy
        K.binary_crossentropy(y_true, y_pred, from_logits=from_logits), axis=-1)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\util\dispatch.py:201 wrapper
        return target(*args, **kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\backend.py:4823 binary_crossentropy
        return nn.sigmoid_cross_entropy_with_logits(labels=target, logits=output)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\util\dispatch.py:201 wrapper
        return target(*args, **kwargs)
    C:\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\ops\nn_impl.py:173 sigmoid_cross_entropy_with_logits
        raise ValueError("logits and labels must have the same shape (%s vs %s)" %

    ValueError: logits and labels must have the same shape ((None, 2) vs (None, 1))

我该怎么解决这个问题？

- xEthereal

---

但是 RGB 规范已经解决了这个问题，对吧？我刚刚使用 input_shape = (128,128,3) 运行了代码，但它没有工作，因为该参数需要一个 2D 向量。 - xEthereal

是的，应该这样，但我不知道为什么会发生这种情况。我也有完全相同的错误。 - Jimit Vaghela

尝试将Dense(2, activation='sigmoid')作为最后一层改为Dense(1, activation='sigmoid')。 - Marco Cerliani

二元分类使用sigmoid函数（基本中的基本）......如果大于0.5，则属于类别1，如果小于0.5，则属于类别0 :-) - Marco Cerliani

哦，抱歉。我只是新手对AI不太熟悉。对不起。 - xEthereal

Answer 1

我终于找到了解决我的问题的方法！我把损失函数从binary_crossentropy修改为sparse_categorical_crossentropy后，成功地训练了模型。然后我还把最后一层的激活函数从sigmoid改成了softmax。

如果您感兴趣，这是整个代码：

from keras.models import Sequential
from keras.layers import Conv2D,MaxPooling2D,\
     Dropout,Flatten,Dense,Activation,\
     BatchNormalization
model=Sequential()
model.add(Conv2D(32,(3,3),activation='relu',input_shape=(128,128,3)))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Conv2D(64,(3,3),activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Conv2D(128,(3,3),activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512,activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(2,activation='softmax'))
model.compile(loss = 'sparse_categorical_crossentropy',
  optimizer='adam',metrics=['accuracy'])

- xEthereal