在keras中使用cnn有困难
Difficulty with cnn in keras
我正在尝试构建一个 dcnn，但出现此错误：
ValueError: ('The specified size contains a dimension with value <= 0', (-192, 1024))
真的，我不知道这个错误的原因，这是我的代码：
数据：
c_X = open("C:/Users/PC/Desktop/Notebooks/Isabelle/mfcc_train_I_C_I_C_2.dat", "r")
c_y = open("C:/Users/PC/Desktop/Notebooks/Isabelle/phoneme_train_I_C_I_C_2.dat", "r")
c_X = np.fromfile(c_X, np.dtype('float32'))
c_y = np.fromfile(c_y, np.dtype('int8'))
c_X = c_X.reshape(886887,1120)
c_X = c_X.reshape(c_X.shape[0], 1, 20, 56)
c_y = one_hot(c_y)
#c_y = np.append(c_y, np.zeros((374975,1)), axis=1)
X_3 = apendice(Colere_X, c_X)
y_3 = apendice(Colere_y, c_y)
#print(c_X.shape, c_y.shape)
print(X_3.shape, y_3.shape)
(1123867, 1, 20, 56) (1123867, 38)
这是我的神经网络实现（我认为问题出在这里）：
model = Sequential()
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu',input_shape=(1, 20, 56)))
model.add(Dropout(0.25))
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu'))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
# Compile the model
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
# Train the model
start = time.time()
model_info = model.fit(X_3, y_3, batch_size=100, \
epochs=20, verbose=2, validation_data=(X_test, y_test))
end = time.time()
模型总结如下：
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_21 (Conv2D) (None, -1, 18, 32) 16160
_________________________________________________________________
dropout_16 (Dropout) (None, -1, 18, 32) 0
_________________________________________________________________
conv2d_22 (Conv2D) (None, -3, 16, 32) 9248
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, -2, 8, 32) 0
_________________________________________________________________
dropout_17 (Dropout) (None, -2, 8, 32) 0
_________________________________________________________________
conv2d_23 (Conv2D) (None, -4, 6, 32) 9248
_________________________________________________________________
conv2d_24 (Conv2D) (None, -6, 4, 32) 9248
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, -3, 2, 32) 0
_________________________________________________________________
dropout_18 (Dropout) (None, -3, 2, 32) 0
_________________________________________________________________
flatten_6 (Flatten) (None, -192) 0
=================================================================
Total params: 43,904
Trainable params: 43,904
Non-trainable params: 0
_________________________________________________________________
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-17-589407073ff5> in <module>()
13 model.add(Flatten())
14 model.summary()
---> 15 model.add(Dense(256, activation='relu'))
16 model.add(Dropout(0.5))
17 model.add(Dense(num_classes, activation='softmax'))
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\models.py in add(self, layer)
467 output_shapes=[self.outputs[0]._keras_shape])
468 else:
--> 469 output_tensor = layer(self.outputs[0])
470 if isinstance(output_tensor, list):
471 raise TypeError('All layers in a Sequential model '
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\engine\topology.py in __call__(self, inputs, **kwargs)
567 '`layer.build(batch_input_shape)`')
568 if len(input_shapes) == 1:
--> 569 self.build(input_shapes[0])
570 else:
571 self.build(input_shapes)
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\layers\core.py in build(self, input_shape)
823 name='kernel',
824 regularizer=self.kernel_regularizer,
--> 825 constraint=self.kernel_constraint)
826 if self.use_bias:
827 self.bias = self.add_weight(shape=(self.units,),
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\legacy\interfaces.py in wrapper(*args, **kwargs)
85 warnings.warn('Update your `' + object_name +
86 '` call to the Keras 2 API: ' + signature, stacklevel=2)
---> 87 return func(*args, **kwargs)
88 wrapper._original_function = func
89 return wrapper
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\engine\topology.py in add_weight(self, name, shape, dtype, initializer, regularizer, trainable, constraint)
389 if dtype is None:
390 dtype = K.floatx()
--> 391 weight = K.variable(initializer(shape), dtype=dtype, name=name)
392 if regularizer is not None:
393 self.add_loss(regularizer(weight))
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\initializers.py in __call__(self, shape, dtype)
206 limit = np.sqrt(3. * scale)
207 return K.random_uniform(shape, -limit, limit,
--> 208 dtype=dtype, seed=self.seed)
209
210 def get_config(self):
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\backend\theano_backend.py in random_uniform(shape, minval, maxval, dtype, seed)
2189 seed = np.random.randint(1, 10e6)
2190 rng = RandomStreams(seed=seed)
-> 2191 return rng.uniform(shape, low=minval, high=maxval, dtype=dtype)
2192
2193
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\theano\sandbox\rng_mrg.py in uniform(self, size, low, high, ndim, dtype, nstreams)
854 raise ValueError(
855 "The specified size contains a dimension with value <= 0",
--> 856 size)
857
858 else:
ValueError: ('The specified size contains a dimension with value <= 0', (-192, 256))
非常感谢您的帮助。
提前致谢。
这是一个与 Keras 中的频道排序相关的问题。您可能已将 ~/.keras/keras.json 中的 image_dim_ordering 或 image_data_format parameter 设置为 tf 或 channels_last，具体取决于您使用的 Keras 版本。问题是您提供的 input_shape 是 th 或 channels first 顺序，这会扰乱维度的解释方式并产生负维度大小。
如果您先将 Keras 与 TensorFlow 一起使用，然后切换到 Theano，就会发生这种情况。
解决方案很简单，您可以将 keras.json 配置文件中的相应参数设置为 th 或 channels_first 以匹配您的文件，或者更改 input_shape 的数据到 (20, 56, 1)。两种解决方案都应该有效，您应该更喜欢您正在使用的后端的本机通道排序（因为它稍微快一点）。
我正在尝试构建一个 dcnn，但出现此错误：
ValueError: ('The specified size contains a dimension with value <= 0', (-192, 1024))
真的，我不知道这个错误的原因，这是我的代码：
数据：
c_X = open("C:/Users/PC/Desktop/Notebooks/Isabelle/mfcc_train_I_C_I_C_2.dat", "r")
c_y = open("C:/Users/PC/Desktop/Notebooks/Isabelle/phoneme_train_I_C_I_C_2.dat", "r")
c_X = np.fromfile(c_X, np.dtype('float32'))
c_y = np.fromfile(c_y, np.dtype('int8'))
c_X = c_X.reshape(886887,1120)
c_X = c_X.reshape(c_X.shape[0], 1, 20, 56)
c_y = one_hot(c_y)
#c_y = np.append(c_y, np.zeros((374975,1)), axis=1)
X_3 = apendice(Colere_X, c_X)
y_3 = apendice(Colere_y, c_y)
#print(c_X.shape, c_y.shape)
print(X_3.shape, y_3.shape)
(1123867, 1, 20, 56) (1123867, 38)
这是我的神经网络实现（我认为问题出在这里）：
model = Sequential()
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu',input_shape=(1, 20, 56)))
model.add(Dropout(0.25))
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(32, (3, 3), border_mode='valid', activation='relu'))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
# Compile the model
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
# Train the model
start = time.time()
model_info = model.fit(X_3, y_3, batch_size=100, \
epochs=20, verbose=2, validation_data=(X_test, y_test))
end = time.time()
模型总结如下：
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_21 (Conv2D) (None, -1, 18, 32) 16160
_________________________________________________________________
dropout_16 (Dropout) (None, -1, 18, 32) 0
_________________________________________________________________
conv2d_22 (Conv2D) (None, -3, 16, 32) 9248
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, -2, 8, 32) 0
_________________________________________________________________
dropout_17 (Dropout) (None, -2, 8, 32) 0
_________________________________________________________________
conv2d_23 (Conv2D) (None, -4, 6, 32) 9248
_________________________________________________________________
conv2d_24 (Conv2D) (None, -6, 4, 32) 9248
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, -3, 2, 32) 0
_________________________________________________________________
dropout_18 (Dropout) (None, -3, 2, 32) 0
_________________________________________________________________
flatten_6 (Flatten) (None, -192) 0
=================================================================
Total params: 43,904
Trainable params: 43,904
Non-trainable params: 0
_________________________________________________________________
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-17-589407073ff5> in <module>()
13 model.add(Flatten())
14 model.summary()
---> 15 model.add(Dense(256, activation='relu'))
16 model.add(Dropout(0.5))
17 model.add(Dense(num_classes, activation='softmax'))
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\models.py in add(self, layer)
467 output_shapes=[self.outputs[0]._keras_shape])
468 else:
--> 469 output_tensor = layer(self.outputs[0])
470 if isinstance(output_tensor, list):
471 raise TypeError('All layers in a Sequential model '
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\engine\topology.py in __call__(self, inputs, **kwargs)
567 '`layer.build(batch_input_shape)`')
568 if len(input_shapes) == 1:
--> 569 self.build(input_shapes[0])
570 else:
571 self.build(input_shapes)
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\layers\core.py in build(self, input_shape)
823 name='kernel',
824 regularizer=self.kernel_regularizer,
--> 825 constraint=self.kernel_constraint)
826 if self.use_bias:
827 self.bias = self.add_weight(shape=(self.units,),
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\legacy\interfaces.py in wrapper(*args, **kwargs)
85 warnings.warn('Update your `' + object_name +
86 '` call to the Keras 2 API: ' + signature, stacklevel=2)
---> 87 return func(*args, **kwargs)
88 wrapper._original_function = func
89 return wrapper
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\engine\topology.py in add_weight(self, name, shape, dtype, initializer, regularizer, trainable, constraint)
389 if dtype is None:
390 dtype = K.floatx()
--> 391 weight = K.variable(initializer(shape), dtype=dtype, name=name)
392 if regularizer is not None:
393 self.add_loss(regularizer(weight))
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\initializers.py in __call__(self, shape, dtype)
206 limit = np.sqrt(3. * scale)
207 return K.random_uniform(shape, -limit, limit,
--> 208 dtype=dtype, seed=self.seed)
209
210 def get_config(self):
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\keras\backend\theano_backend.py in random_uniform(shape, minval, maxval, dtype, seed)
2189 seed = np.random.randint(1, 10e6)
2190 rng = RandomStreams(seed=seed)
-> 2191 return rng.uniform(shape, low=minval, high=maxval, dtype=dtype)
2192
2193
~\Anaconda3\envs\tensorflow-gpu\lib\site-packages\theano\sandbox\rng_mrg.py in uniform(self, size, low, high, ndim, dtype, nstreams)
854 raise ValueError(
855 "The specified size contains a dimension with value <= 0",
--> 856 size)
857
858 else:
ValueError: ('The specified size contains a dimension with value <= 0', (-192, 256))
非常感谢您的帮助。 提前致谢。
这是一个与 Keras 中的频道排序相关的问题。您可能已将 ~/.keras/keras.json 中的 image_dim_ordering 或 image_data_format parameter 设置为 tf 或 channels_last，具体取决于您使用的 Keras 版本。问题是您提供的 input_shape 是 th 或 channels first 顺序，这会扰乱维度的解释方式并产生负维度大小。
如果您先将 Keras 与 TensorFlow 一起使用，然后切换到 Theano，就会发生这种情况。
解决方案很简单，您可以将 keras.json 配置文件中的相应参数设置为 th 或 channels_first 以匹配您的文件，或者更改 input_shape 的数据到 (20, 56, 1)。两种解决方案都应该有效，您应该更喜欢您正在使用的后端的本机通道排序（因为它稍微快一点）。