在 tensorflow 计算机视觉建模中在哪里定义批量大小？

编程问答 2022-04-21

问题描述

我有这样的数据，其中我将 batch_size 指定为 32：

# Preparing and preprocessing the data
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_dir = '/content/pizza_steak/train'
test_dir = '/content/pizza_steak/test'

train_data_gen_aug = ImageDataGenerator(rotation_range=0.2,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,vertical_flip=True,rescale=1./255)
test_data_gen = ImageDataGenerator(rescale=1./255)

train_data_aug = train_data_gen_aug.flow_from_directory(train_dir,target_size=(224,224),class_mode='binary',batch_size=32,seed=42)
test_data = test_data_gen.flow_from_directory(test_dir,seed=42)

返回：

Found 1500 images belonging to 2 classes.
Found 500 images belonging to 2 classes

当我进行如下探索时：

# Explore the data
train_images,train_labels = train_data.next()
train_images_aug,train_labels_aug = train_data_aug.next()
test_images_aug,test_labels_aug = test_data.next()
print('train_data:     ',len(train_data),train_images.shape,train_labels.shape)
print('train_data_aug: ',len(train_data_aug),train_images_aug.shape,train_labels_aug.shape)
print('test_data:      ',len(test_data),test_images_aug.shape,test_labels_aug.shape)

它返回：

train_data:      47 (32,224,3) (32,)
train_data_aug:  47 (32,)
test_data:       16 (32,)

然后我构建并编译模型，并在 bacth_size 中将 None 指定为 InputLayer：

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import InputLayer,Conv2D,MaxPool2D,Flatten,Dense
from tensorflow.keras.activations import relu,sigmoid
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.losses import BinaryCrossentropy
from tensorflow.keras.metrics import binary_accuracy

# create the model
model = Sequential()

# Add the input layer
INPUT_SHAPE = (224,3)
model.add(InputLayer(input_shape=INPUT_SHAPE,batch_size=None,# I enetered the batch size here as None
                       ))

# Add the hidden layers
model.add(Conv2D(filters=10,kernel_size=3,strides=1,padding='valid',activation=relu))
model.add(MaxPool2D(pool_size=(2,2),strides=None,padding='valid'))

# Add the flatten layer
model.add(Flatten())

# Add the output layer
model.add(Dense(units=1,activation=sigmoid))

# Compile the model
model.compile(optimizer=Adam(),loss=BinaryCrossentropy(),metrics=[binary_accuracy])

然后拟合模型，我将batch_size指定为None

history = model.fit(train_data_aug,# bacth_size defined as None
                    epochs=5,verbose=1,validation_data=test_data,steps_per_epoch=len(train_data_aug),validation_steps=len(test_data))

该模型运行良好，当仅训练 5 个时期时，val_binary_accuracy 为 81%。

在什么情况下应该使用其他两个中的批量大小并且可以在所有这些中定义 batch_size 否则可能会导致问题？

解决方法

批量大小是每次梯度更新的样本数。 如果它像您在 model.fit() 中那样未指定，则默认为 32。但是，您的数据采用已经有批次的生成器的形式。所以您不必指定批量大小。

来自 Tensorflow 文档：https://www.tensorflow.org/api_docs/python/tf/keras/Model

Do not specify the batch_size if your data is in the form of datasets,generators,or 
keras.utils.Sequence instances (since they generate batches).
batchsizedeep-learningkeraskeraspythontensorflowtensorflowtensorflow