tensorflow入门学习及MNIST手写数字识别学习

tensorflow入门学习及MNIST手写数字识别学习

1. tensorflow安装

如果使用pycharm设置来安装会报下载超时的错误
使用命令安装也需要下载很久
//使用这个命令安装就很快
pip install tensorflow==2.0.0rc1 -i https://pypi.tuna.tsinghua.edu.cn/simple

2. 安装tensorflow结果查看

import tensorflow as tf
# 查看安装版本
print(tf.__version__)
# 打印所有数据集
print(dir(tf.keras.datasets))

#打印结果
2.0.0-rc1
['__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '_sys', 'boston_housing', 'cifar10', 'cifar100', 'fashion_mnist', 'imdb', 'mnist', 'reuters']

3. MNIST数据集可视化

# 加载数据集
mnist = tf.keras.datasets.mnist
# 包括两个数据集,一个是训练数据(60000个),一个是测试数据(10000个)
(x_train, y_train), (x_test, y_test) = mnist.load_data()
# 打印数据格式
print(x_train.shape, y_train.shape)
print(x_test.shape, y_test.shape)

#打印结果
(60000, 28, 28) (60000,)
(10000, 28, 28) (10000,)

# 显示图片
# 导入依赖库
import matplotlib.pyplot as plt
# 随机选取一个数并查看label
image_index = 15000  # 范围是[0,59999)
# 图片显示(彩色)
# plt.imshow(x_train[image_index])
# 图片灰度显示(黑白)
plt.imshow(x_train[image_index], cmap='Greys')
plt.show()
# 打印图片所代表的数字
print(y_train[image_index])

4. MNIST数据集格式转换

import numpy as np
# 将图片从28*28扩充为32*32
x_train = np.pad(x_train, ((0, 0), (2, 2), (2, 2)), 'constant', constant_values=0)
print(x_train.shape)
# 数据类型转换
x_train = x_train.astype('float32')
# 数据正则化
x_train /= 255
# 数据维度转换([n,h,w,c])
x_train = x_train.reshape(x_train.shape[0], 32, 32, 1)
print(x_train.shape)

# 打印结果
(60000, 32, 32)
(60000, 32, 32, 1)

5. 构建LeNet模型

# 构建LeNet模型
model = tf.keras.models.Sequential([
    # 第一层:卷积层编码
    # 参数:卷积核个数,卷积核大小,填充方式,激活函数,输入数据格式
    tf.keras.layers.Conv2D(filters=6, kernel_size=(5, 5), padding='valid', activation=tf.nn.relu,
                           input_shape=(32, 32, 1)),
    # 第二层:池化层
    tf.keras.layers.AveragePooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'),
    # 第三层:卷积层
    tf.keras.layers.Conv2D(filters=16, kernel_size=(5, 5), padding='valid', activation=tf.nn.relu),
    # 第四层:池化层
    tf.keras.layers.AveragePooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'),
    # 扁平处理(多维数据转换为一维数据)
    tf.keras.layers.Flatten(),

    # 第五、六、七层:全连接层
    tf.keras.layers.Dense(units=120, activation=tf.nn.relu),
    tf.keras.layers.Dense(units=84, activation=tf.nn.relu),
    tf.keras.layers.Dense(units=10, activation=tf.nn.softmax)
])

print(model.summary())

# 模型训练

# 超参数设置
num_epochs = 10
batch_size = 64
learning_rate = 0.001
adam_optimizer = tf.keras.optimizers.Adam(learning_rate)

model.compile(optimizer=adam_optimizer,
              loss=tf.keras.losses.sparse_categorical_crossentropy,
              metrics=['accuracy'])

# 用于计算学习时间
import datetime

start_time = datetime.datetime.now()

model.fit(x=x_train,
          y=y_train,
          batch_size=batch_size,
          epochs=num_epochs)
end_time = datetime.datetime.now()
time_cost = end_time - start_time
print("time_cost = ", time_cost)  # CPU time cost: 5min, GPU time cost: less than 1min

# 模型保存(保存在当前目录下)
model.save('lenet_model.h5')

6. 使用模型来识别手写的数字

# 使用模型识别图片数字
import tensorflow as tf

# 加载已经训练好的LeNet模型
model = tf.keras.models.load_model('lenet_model.h5')
model.summary()


import cv2
import matplotlib.pyplot as plt

# 第一步:读取图片
img = cv2.imread('numberImages/9.jpg')  # 8.png
print(img.shape)
plt.imshow(img)
plt.show()

# 第二步:将图片转为灰度图
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
print(img.shape)
plt.imshow(img, cmap='Greys')
plt.show()

# 第三步:将图片的底色和字的颜色取反
img = cv2.bitwise_not(img)
plt.imshow(img, cmap='Greys')
plt.show()

# 第四步:将底变成纯白色,将字变成纯黑色
img[img <= 100] = 0
img[img > 140] = 255  # 130
plt.imshow(img)
plt.show()

# 显示图片
plt.imshow(img, cmap='Greys')
plt.show()

# 第五步:将图片尺寸缩放为输入规定尺寸
img = cv2.resize(img, (32, 32))
plt.show()

# 第六步:将数据类型转为float32
img = img.astype('float32')

# 第七步:数据正则化
img /= 255

# 第八步:增加维度为输入的规定格式
img = img.reshape(1, 32, 32, 1)
print(img.shape)

# 第九步:预测
pred = model.predict(img)

# 第十步:输出结果
print('=========================')
print('预测结果为:', pred.argmax())

9.jpg

# 运行结果
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 28, 28, 6)         156       
_________________________________________________________________
average_pooling2d (AveragePo (None, 14, 14, 6)         0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 10, 10, 16)        2416      
_________________________________________________________________
average_pooling2d_1 (Average (None, 5, 5, 16)          0         
_________________________________________________________________
flatten (Flatten)            (None, 400)               0         
_________________________________________________________________
dense (Dense)                (None, 120)               48120     
_________________________________________________________________
dense_1 (Dense)              (None, 84)                10164     
_________________________________________________________________
dense_2 (Dense)              (None, 10)                850       
=================================================================
Total params: 61,706
Trainable params: 61,706
Non-trainable params: 0
_________________________________________________________________
(651, 650, 3)
(651, 650)
(1, 32, 32, 1)
=========================
预测结果为: 9

LetNet模型简介








资料:

链接:https://pan.baidu.com/s/1kJz0IvNIMpYU3-ZLwM20cQ 
提取码:ty40
posted @ 2020-09-11 20:32  曾强  阅读(43)  评论(0编辑  收藏