在 MNIST 数据集上重塑 3DCNN 的目标数组大小 - ValueError

Question

我正在尝试在 MNIST 数据集上运行 3DCNN。我的代码抛出了这样的 ValueError：

ValueError：输入数组的样本数应与目标数组相同。找到 15000 个输入样本和 60000 个目标样本。

这是我的代码：

import tensorflow
import keras
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv3D, MaxPooling3D
from keras import backend as K

batch_size = 128
num_classes = 10
epochs = 12

img_rows, img_cols, img_dep = 28, 28, 4

(x_train, y_train), (x_test, y_test) = mnist.load_data()

print y_train.shape[0]

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(15000, 1, img_dep, img_rows, img_cols)
    x_test = x_test.reshape(2500, 1, img_dep, img_rows, img_cols)
    input_shape = (1, img_dep, img_rows, img_cols)
else:
    x_train = x_train.reshape(15000, img_dep, img_rows, img_cols, 1)
    x_test = x_test.reshape(2500, img_dep, img_rows, img_cols, 1)
    input_shape = (img_dep, img_rows, img_cols, 1)

x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255
print('x_train shape:', x_train.shape)
print(x_train.shape[0], 'train samples')
print(x_test.shape[0], 'test samples')

y_train.reshape(15000, img_dep)
y_test.reshape(2500, img_dep)

print('x_train shape', x_train.shape)
print('x_test shape', x_test.shape)
print('y_train shape', y_train.shape)
print('y_test shape', y_test.shape)

y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)

model = Sequential()
model.add(Conv3D(32, kernel_size=(3, 3, 3),
         activation='relu',
         input_shape=input_shape))
model.add(Conv3D(64, (2, 2, 2), activation='relu'))
#model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))

model.compile(loss=keras.losses.categorical_crossentropy,
          optimizer=keras.optimizers.Adadelta(),
          metrics=['accuracy'])

model.fit(x_train, y_train,
      batch_size=batch_size,
      epochs=epochs,
      verbose=1,
      validation_data=(x_test, y_test))

score = model.evaluate(x_test, y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])

知道我哪里出错了吗？我对 ML 很陌生，因此这可能是一个简单修复的简单错误。我在网上找不到任何使用 MNIST 数据运行 3DCNN 的示例。谢谢！

回溯消息是：

Traceback (most recent call last):
  File "3dconvnet.py", line 67, in <module>
    validation_data=(x_test, y_test))
  File "/home/jackson/anaconda2/lib/python2.7/site-packages/keras/models.py", line 853, in fit
    initial_epoch=initial_epoch)
  File "/home/jackson/anaconda2/lib/python2.7/site-packages/keras/engine/training.py", line 1406, in fit
    batch_size=batch_size)
  File "/home/jackson/anaconda2/lib/python2.7/site-packages/keras/engine/training.py", line 1308, in _standardize_user_data
    _check_array_lengths(x, y, sample_weights)
  File "/home/jackson/anaconda2/lib/python2.7/site-packages/keras/engine/training.py", line 229, in _check_array_lengths
    'and ' + str(list(set_y)[0]) + ' target samples.')
ValueError: Input arrays should have the same number of samples as target arrays. Found 15000 input samples and 60000 target samples.

打印形状的输出是：

('x_train shape', (15000, 4, 28, 28, 1))
('x_test shape', (2500, 4, 28, 28, 1))
('y_train shape', (60000,))
('y_test shape', (10000,))

编辑* 1：添加回溯

编辑** 2：添加打印形状的输出

Answer 1

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train.shape
Out[3]: (60000, 28, 28)

y_train.shape
Out[4]: (60000,)

MNIST 数据集包含 60000 张 28x28 图像，具有一个（不是四个）通道。而不是这个代码片段

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(15000, 1, img_dep, img_rows, img_cols)
    x_test = x_test.reshape(2500, 1, img_dep, img_rows, img_cols)
    input_shape = (1, img_dep, img_rows, img_cols)
else:
    x_train = x_train.reshape(15000, img_dep, img_rows, img_cols, 1)
    x_test = x_test.reshape(2500, img_dep, img_rows, img_cols, 1)
    input_shape = (img_dep, img_rows, img_cols, 1)

你需要

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(-1, 1, img_rows, img_cols)
    x_test = x_test.reshape(-1, 1, img_rows, img_cols)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(-1, img_rows, img_cols, 1)
    x_test = x_test.reshape(-1, img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1)

因为图像深度（颜色通道数）= 1。另外，您不需要

y_train.reshape(15000, img_dep)
y_test.reshape(2500, img_dep)

Answer 2

据我所知，3d CNN 用于体积数据或时间图像数据。 要在 mnist 数据集中使用，您需要以某种方式将其转换为 5d 张量。您的主要错误是 x train 的大小为 25000，而 y train 的大小为 60000。 这就是问题。希望它解决。