图像识别实例(CNN)：CIFAR-10-基于keras的python学习笔记（十二--五）

版权声明：本文为博主原创文章，未经博主允许不得转载。https://blog.csdn.net/weixin_44474718/article/details/88421795

一、问题描述

数据集由60000张32x32的彩色图像，50000张用于训练，10000张用于评估。训练数据集被均匀的分为10个类别，每个类别5000张。

二、导入数据

from keras.datasets import cifar10
from matplotlib import pyplot as plt
from scipy.misc import toimage
import numpy as np

# 导入数据
(X_train, y_train), (X_validation, y_validation) = cifar10.load_data()

for i in range(0, 9):
    plt.subplot(331 + i)
    plt.imshow(toimage(X_train[i]))

# 显示图片
plt.show()

# 设定随机种子
seed = 7
np.random.seed(seed)

三、简单卷积神经网络

预处理：归一化（将数据调整到0-1）、one-hot编码（便于模型的输出）

1. 卷积层：具有32个特征图，感受野大小：3*3
2. dropout层，20%
3. 卷积层：具有32个特征图，感受野大小：3*3
4. dropout层，20%
5. 池化层：采样因子（pool_size）为2*2
6. flatten层
7. 全连接层：512个神经元和rule激活函数
8. dropout层，50%
9. 输出层：10个神经元

import numpy as np
from keras.datasets import cifar10
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Dropout
from keras.layers import Flatten
from keras.layers.convolutional import Conv2D
from keras.layers.convolutional import MaxPooling2D
from keras.optimizers import SGD
from keras.constraints import maxnorm
from keras.utils import np_utils
from keras import backend
backend.set_image_data_format('channels_first')

# 设定随机种子
seed = 7
np.random.seed(seed=seed)

# 导入数据
(X_train, y_train), (X_validation, y_validation) = cifar10.load_data()

# 格式化数据到0-1之前
X_train = X_train.astype('float32')
X_validation = X_validation.astype('float32')
X_train = X_train / 255.0
X_validation = X_validation / 255.0

# one-hot编码
y_train = np_utils.to_categorical(y_train)
y_validation = np_utils.to_categorical(y_validation)
num_classes = y_train.shape[1]

def create_model(epochs=25):
    model = Sequential()
    model.add(Conv2D(32, (3, 3), input_shape=(3, 32, 32), padding='same', activation='relu', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Conv2D(32, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Flatten())
    model.add(Dense(512, activation='relu', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.5))
    model.add(Dense(10, activation='softmax'))
    lrate = 0.01
    decay = lrate / epochs
    sgd = SGD(lr=lrate, momentum=0.9, decay=decay, nesterov=False)
    model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
    return model

epochs = 25
model = create_model(epochs)
model.fit(x=X_train, y=y_train, epochs=epochs, batch_size=32, verbose=2)
scores = model.evaluate(x=X_validation, y=y_validation, verbose=0)
print('Accuracy: %.2f%%' % (scores[1] * 100))

四、大型卷积神经网络

复杂就复杂在多了几个层。

import numpy as np
from keras.datasets import cifar10
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Dropout
from keras.layers import Flatten
from keras.layers.convolutional import Conv2D
from keras.layers.convolutional import MaxPooling2D
from keras.optimizers import SGD
from keras.constraints import maxnorm
from keras.utils import np_utils
from keras import backend
backend.set_image_data_format('channels_first')

# 设定随机种子
seed = 7
np.random.seed(seed=seed)

# 导入数据
(X_train, y_train), (X_validation, y_validation) = cifar10.load_data()

# 格式化数据到0-1之前
X_train = X_train.astype('float32')
X_validation = X_validation.astype('float32')
X_train = X_train / 255.0
X_validation = X_validation / 255.0

# one-hot编码
y_train = np_utils.to_categorical(y_train)
y_validation = np_utils.to_categorical(y_validation)
num_classes = y_train.shape[1]

def create_model(epochs=25):
    model = Sequential()
    model.add(Conv2D(32, (3, 3), input_shape=(3, 32, 32), padding='same', activation='relu', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Conv2D(32, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Conv2D(128, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Conv2D(128, (3, 3), activation='relu', padding='same', kernel_constraint=maxnorm(3)))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Flatten())
    model.add(Dropout(0.2))
    model.add(Dense(1024, activation='relu', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Dense(512, activation='relu', kernel_constraint=maxnorm(3)))
    model.add(Dropout(0.2))
    model.add(Dense(10, activation='softmax'))
    lrate = 0.01
    decay = lrate / epochs
    sgd = SGD(lr=lrate, momentum=0.9, decay=decay, nesterov=False)
    model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
    return model

epochs = 25
model = create_model(epochs)
model.fit(x=X_train, y=y_train, epochs=epochs, batch_size=32, verbose=2)
scores = model.evaluate(x=X_validation, y=y_validation, verbose=0)
print('Accuracy: %.2f%%' % (scores[1] * 100))

五、改进模型

参照network in network 这篇论文（http://arxiv.org/pdf/1312.4400.pdf）来实现一个改进的模型，池化层采用了GlobalAveragePooling（作为最后一个池化层）。 在阿里云的CPU执行了5天？？？？，准确率87.96%。

import keras
import numpy as np
from keras.datasets import cifar10
from keras.models import Sequential
from keras.layers import Dropout, Activation
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.initializers import RandomNormal
from keras import optimizers
from keras.callbacks import LearningRateScheduler, TensorBoard

batch_size = 128
epochs = 200
iterations = 391
num_classes = 10
dropout = 0.5
log_filepath = './nin'


def normalize_preprocessing(x_train, x_validation):
    x_train = x_train.astype('float32')
    x_validation = x_validation.astype('float32')
    mean = [125.307, 122.95, 113.865]
    std = [62.9932, 62.0887, 66.7048]
    for i in range(3):
        x_train[:, :, :, i] = (x_train[:, :, :, i] - mean[i]) / std[i]
        x_validation[:, :, :, i] = (x_validation[:, :, :, i] - mean[i]) / std[i]

    return x_train, x_validation


def scheduler(epoch):
    if epoch <= 60:
        return 0.05
    if epoch <= 120:
        return 0.01
    if epoch <= 160:
        return 0.002
    return 0.0004


def build_model():
    model = Sequential()

    model.add(Conv2D(192, (5, 5), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.01), input_shape=x_train.shape[1:],
                     activation='relu'))
    model.add(Conv2D(160, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(Conv2D(96, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2), padding='same'))

    model.add(Dropout(dropout))

    model.add(Conv2D(192, (5, 5), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(Conv2D(192, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(Conv2D(192, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(MaxPooling2D(pool_size=(3, 3), strides=(2, 2), padding='same'))

    model.add(Dropout(dropout))

    model.add(Conv2D(192, (3, 3), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(Conv2D(192, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))
    model.add(Conv2D(10, (1, 1), padding='same', kernel_regularizer=keras.regularizers.l2(0.0001),
                     kernel_initializer=RandomNormal(stddev=0.05), activation='relu'))

    model.add(GlobalAveragePooling2D())
    model.add(Activation('softmax'))

    sgd = optimizers.SGD(lr=0.1, momentum=0.9, nesterov=True)
    model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
    return model


if __name__ == '__main__':
    np.random.seed(seed=7)
    # load data
    (x_train, y_train), (x_validation, y_validation) = cifar10.load_data()
    y_train = keras.utils.to_categorical(y_train, num_classes)
    y_validation = keras.utils.to_categorical(y_validation, num_classes)

    x_train, x_validation = normalize_preprocessing(x_train, x_validation)

    # build network
    model = build_model()
    print(model.summary())

    # set callback
    tb_cb = TensorBoard(log_dir=log_filepath, histogram_freq=0)
    change_lr = LearningRateScheduler(scheduler)
    cbks = [change_lr, tb_cb]

    '''
    # set data augmentation
    print('Using real-time data augmentation.')
    from keras.preprocessing.image import ImageDataGenerator
    datagen = ImageDataGenerator(horizontal_flip=True, width_shift_range=0.125, height_shift_range=0.125,
                                 fill_mode='constant', cval=0.)
    datagen.fit(x_train)

    # start training
    model.fit_generator(datagen.flow(x_train, y_train, batch_size=batch_size), steps_per_epoch=iterations,
                        epochs=epochs, callbacks=cbks, validation_data=(x_validation, y_validation), verbose=2)
    '''
    model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, callbacks=cbks,
              validation_data=(x_validation, y_validation), verbose=2)
    model.save('nin.h5')