Question

我想从以下层开始训练深度网络：

def cnn_model():
model = Sequential()

model.add(Conv2D(32, (3, 3), padding='same',
                 input_shape=(3,48, 48),
                 activation='relu'))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))

model.add(Conv2D(64, (3, 3), padding='same',
                 activation='relu'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))

model.add(Conv2D(128, (3, 3), padding='same',
                 activation='relu'))
model.add(Conv2D(128, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))

model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(NUM_CLASSES, activation='softmax'))
return model

model = cnn_model()
# let's train the model using SGD + momentum (how original).
lr = 0.01
sgd = SGD(lr=lr, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
          optimizer=sgd,
          metrics=['accuracy'])


def lr_schedule(epoch):
    return lr*(0.1**int(epoch/10))

使用以下生成器：

model = cnn_model()

# let's train the model using SGD + momentum
lr = 0.01
sgd = SGD(lr=lr, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
              optimizer=sgd,
              metrics=['accuracy'])
def lr_schedule(epoch):
    return lr * (0.1 ** int(epoch / 10))

batch_size = 32
epochs = 30

model.fit(X, Y,
          batch_size=batch_size,
          epochs=epochs,
          validation_split=0.2,
          callbacks=[LearningRateScheduler(lr_schedule),
                     ModelCheckpoint('model.h5', save_best_only=True)]
          )

（验证生成器看起来很相似）。

在训练期间，我收到错误：

ValueError: Error when checking input: expected conv2d_175_input to have 4 dimensions, but got array with shape (0, 1)

输入print(model.summary())

后，我得到以下输出

    _________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_175 (Conv2D)          (None, 32, 48, 48)        896       
_________________________________________________________________
conv2d_176 (Conv2D)          (None, 32, 46, 46)        9248      
_________________________________________________________________
max_pooling2d_85 (MaxPooling (None, 32, 23, 23)        0         
_________________________________________________________________
dropout_103 (Dropout)        (None, 32, 23, 23)        0         
_________________________________________________________________
conv2d_177 (Conv2D)          (None, 64, 23, 23)        18496     
_________________________________________________________________
conv2d_178 (Conv2D)          (None, 64, 21, 21)        36928     
_________________________________________________________________
max_pooling2d_86 (MaxPooling (None, 64, 10, 10)        0         
_________________________________________________________________
dropout_104 (Dropout)        (None, 64, 10, 10)        0         
_________________________________________________________________
conv2d_179 (Conv2D)          (None, 128, 10, 10)       73856     
_________________________________________________________________
conv2d_180 (Conv2D)          (None, 128, 8, 8)         147584    
_________________________________________________________________
max_pooling2d_87 (MaxPooling (None, 128, 4, 4)         0         
_________________________________________________________________
dropout_105 (Dropout)        (None, 128, 4, 4)         0         
_________________________________________________________________
flatten_28 (Flatten)         (None, 2048)              0         
_________________________________________________________________
dense_43 (Dense)             (None, 512)               1049088   
_________________________________________________________________
dropout_106 (Dropout)        (None, 512)               0         
_________________________________________________________________
dense_44 (Dense)             (None, 43)                22059     
=================================================================
Total params: 1,358,155
Trainable params: 1,358,155
Non-trainable params: 0

我试过了什么？我正在阅读关于“Convolution2D”/“输入形状”https://keras.io/layers/convolutional/的文档，所以我试图更改input_shape，我知道我的输出应该是4D但我不知道该怎么做。如果你能提供帮助，请分享。

谢谢

Answer 1

正如@nuric所提到的，问题在于输入图像尺寸。

如果输入图像的高度为48，宽度为48，输入通道为3，结果为（3,48,48），那么您对模型的输入为（批量大小，3,48,48）。

这里批量大小只是给予训练模型的单批训练图像数量

伪代码

training_list = []
for i in range(0, batch_size):
    image = read_image(training_set[i])
    training_list.append(image)
training_set = numpy.array(training_list)

所以现在训练集的维度是（批量大小，3,48,48）

检查模型输入时出错：预期conv2d_175_input有4个维度，但得到的形状为数组（0,1）

1 个答案: