tensorflow2.0 手写数字识别demotensorflow2.0 手写数字识别demo

95 阅读 0 评论 63 点赞

我是靠谱客的博主拼搏战斗机，这篇文章主要介绍tensorflow2.0 手写数字识别demotensorflow2.0 手写数字识别demo，现在分享给大家，希望可以做个参考。

tensorflow2.0 手写数字识别demo

tensorflow2.0 使用keras高级API，可以摆脱繁琐的tf.nn了。

tf.keras 用法和 keras基本相同。

通过手写数字识别CNN demo，熟悉tf.keras的基本用法。

demo 代码

复制代码

#%%
import tensorflow as tf
import time
#%%
#自动加载mnist dataset
mnist = tf.keras.datasets.mnist
#%%
#training set 60000 samples, test set 10000 samples
#labels格式0-9数字
(x_train, y_train), (x_test, y_test) = mnist.load_data()
#%%
#normalization
x_train, x_test = x_train/255.0, x_test/255.0
#%%
print(y_train[0:20])
#%%
#building model
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Reshape((28, 28, 1))) # 1 是 channels，model.fit()需要channels
model.add(tf.keras.layers.Conv2D(64, (5,5), activation=tf.keras.activations.relu))
model.add(tf.keras.layers.MaxPool2D(2,2))
model.add(tf.keras.layers.Conv2D(128, (3,3),activation=tf.keras.activations.relu))
model.add(tf.keras.layers.AveragePooling2D((2,2)))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(512, activation=tf.keras.activations.relu))
model.add(tf.keras.layers.Dense(10, activation=tf.keras.activations.softmax))
model.compile(
optimizer='SGD',
loss='sparse_categorical_crossentropy',
#如果想使用categorical_crossentropy, 需要把labels转成onehot格式
metrics=['accuracy']
)
#%%
#training
now = time.time()
model.fit(x_train, y_train, epochs=20) #跑20次就结束吧
#testing
model.evaluate(x_test, y_test)
print(time.time()-now)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
#%%
import tensorflow as tf
import time
#%%
#自动加载mnist dataset
mnist = tf.keras.datasets.mnist
#%%
#training set 60000 samples, test set 10000 samples
#labels格式0-9数字
(x_train, y_train), (x_test, y_test) = mnist.load_data()
#%%
#normalization
x_train, x_test = x_train/255.0, x_test/255.0
#%%
print(y_train[0:20])
#%%
#building model
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Reshape((28, 28, 1))) # 1 是 channels，model.fit()需要channels
model.add(tf.keras.layers.Conv2D(64, (5,5), activation=tf.keras.activations.relu))
model.add(tf.keras.layers.MaxPool2D(2,2))
model.add(tf.keras.layers.Conv2D(128, (3,3),activation=tf.keras.activations.relu))
model.add(tf.keras.layers.AveragePooling2D((2,2)))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(512, activation=tf.keras.activations.relu))
model.add(tf.keras.layers.Dense(10, activation=tf.keras.activations.softmax))
model.compile(
optimizer='SGD',
loss='sparse_categorical_crossentropy',
#如果想使用categorical_crossentropy, 需要把labels转成onehot格式
metrics=['accuracy']
)
#%%
#training
now = time.time()
model.fit(x_train, y_train, epochs=20) #跑20次就结束吧
#testing
model.evaluate(x_test, y_test)
print(time.time()-now)