python深度学习系列Keras 简单入门

入门级别案例mnist手写数字识别

一、用keras实现手写识别

其中也没有什么需要着重注意的点，把几个易错点罗列一下：

1. (train_images, train_labels), (test_images, test_labels) = mnist.load_data()
   左边要用两个元组接
2. from keras.utils import to_categorical
   用来处理样本标签，将其转换为one-hot编码
   train_labels = to_categorical(train_labels)
   test_labels = to_categorical(test_labels)
3. 源码附上

from keras.datasets import mnist
from keras import models
from keras import layers
from keras.utils import to_categorical



# 加载并观察
def load_data():
    """
    导入数据并观察
    :return:
    """
    (train_images, train_labels), (test_images, test_labels) = mnist.load_data()
    print('训练集样本形状：\n', train_images.shape)     #  (60000, 28, 28)
    print('训练集标签形状：\n', train_labels, train_labels.shape)       #  [5 0 4 ... 5 6 8] (60000,)
    print('---------------------------------')
    print('测试集样本形状：\n', test_images.shape)      #  (10000, 28, 28)
    print('测试集标签形状：\n', test_labels, test_labels.shape)     #  [7 2 1 ... 4 5 6] (10000,)

    return train_images, train_labels, test_images, test_labels

def model():
    """
    建立神经网络模型
    :return:
    """

    # 建立顺序模型
    network = models.Sequential()
    network.add(layers.Dense(512, activation='relu', input_shape=(28 * 28,)))
    network.add(layers.Dense(10, activation='softmax'))     # softmax返回10个概率值，总和为1的数组
    network.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

    return network


def data_processing(train_images, train_labels, test_images, test_labels):
    """
    准备图像数据和标签
    :return:
    """
    train_images = train_images.reshape((60000, 28*28))
    train_images = train_images.astype('float32') / 255

    test_images = test_images.reshape((10000, 28*28))
    test_images = test_images.astype('float32') / 255

    # 准备标签,转换成one-hot编码
    train_labels = to_categorical(train_labels)
    test_labels = to_categorical(test_labels)
    print('转换效果为：\n', train_labels)

    return train_images, train_labels, test_images, test_labels



if __name__ == "__main__":

    train_images, train_labels, test_images, test_labels = load_data()
    train_images, train_labels, test_images, test_labels = data_processing(train_images, train_labels, test_images, test_labels)
    network = model()

    # 训练模型
    network.fit(train_images, train_labels, epochs=5, batch_size=128)   # epochs大约是线程的意思

    # 显示精度
    test_loss, test_accuracy = network.evaluate(test_images, test_labels)

    print('测试损失值：\n', test_loss)

    print('测试准确率：\n', test_accuracy)

疑问

1. 自动训练了一万步，并没有自行指定
   以后解决

一、用keras实现手写识别