用cnn模型,以cifar-10-batches-py内的训练数据,训练一个图片分类模型

import keras
import numpy as np
import os
import pickle
from keras import Sequential
from keras.callbacks import ModelCheckpoint
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout
from keras.optimizers import Adam
# 读取数据
def load_data(file):
    with open(file, 'rb') as fo:
        dict = pickle.load(fo, encoding='bytes')
    return dict
# 转换数据格式
def convert_data(data):
    images = np.reshape(data[b'data'], (len(data[b'data']), 3, 32, 32)).transpose(0, 2, 3, 1)
    labels = np.array(data[b'labels'])
    return images, labels
# 加载数据集
def load_dataset():
    train_files = ['data_batch_1', 'data_batch_2', 'data_batch_3', 'data_batch_4', 'data_batch_5']
    test_file = 'test_batch'
    X_train = []
    y_train = []
    for file in train_files:
        data = load_data(os.path.join('cifar-10-batches-py', file))
        images, labels = convert_data(data)
        X_train.append(images)
        y_train.append(labels)
    X_train = np.vstack(X_train)
    y_train = np.hstack(y_train)
    data = load_data(os.path.join('cifar-10-batches-py', test_file))
    X_test, y_test = convert_data(data)
    return X_train, y_train, X_test, y_test
# 构建CNN模型
def build_model(input_shape, num_classes):
    model = Sequential()
    model.add(Conv2D(32, (3, 3), activation='relu', input_shape=input_shape))
    model.add(Conv2D(32, (3, 3), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))
    model.add(Conv2D(64, (3, 3), activation='relu'))
    model.add(Conv2D(64, (3, 3), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))
    model.add(Flatten())
    model.add(Dense(512, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(num_classes, activation='softmax'))
    return model
# 主函数
if __name__ == '__main__':
    # 加载数据
    X_train, y_train, X_test, y_test = load_dataset()
    # 定义超参数
    input_shape = (32, 32, 3)
    num_classes = 10
    batch_size = 128
    epochs = 50
    # 构建CNN模型
    model = build_model(input_shape, num_classes)
    model.summary()
    # 编译模型
    model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])
    # 设置回调函数
    checkpoint = ModelCheckpoint(filepath='weights.hdf5', monitor='val_acc', verbose=1, save_best_only=True)
    # 训练模型
    history = model.fit(X_train, keras.utils.to_categorical(y_train, num_classes), batch_size=batch_size, epochs=epochs, verbose=1, validation_data=(X_test, keras.utils.to_categorical(y_test, num_classes)), callbacks=[checkpoint])
    # 评估模型
    score = model.evaluate(X_test, keras.utils.to_categorical(y_test, num_classes), verbose=0)
    print('Test loss:', score[0])
    print('Test accuracy:', score[1])

问题 

        找不到保存的模型文件

 分析

        可能是因为checkpoint函数是临时保存,或者是只有在botebook环境下的才行

解决方法

model.save('my_model.h5') # 将模型保存到 my_model.h5 文件中

使用这个保存