TensorFlow中的Fashion MNIST图像识别实战

1.导入相应的库：
关于Fashion MNIST数据集的介绍：看这位博主：
https://blog.youkuaiyun.com/qq_28869927/article/details/85079808

import os 
import cv2 
import numpy as np
import tensorflow as tf

import keras
import functools
from keras.preprocessing.image import ImageDataGenerator

2.加载数据集：

(x_train,y_train),(x_test,y_test)=tf.keras.datasets.fashion_mnist.load_data()

print(np.shape(x_train))
print(np.shape(y_train))
print(np.shape(x_test))
print(np.shape(y_test))

3.转换为独热编码：

# y_train=tf.keras.utils.to_categorical(y_train,num_classes=10)
# y_test=tf.keras.utils.to_categorical(y_test,num_classes=10)
y_train=tf.one_hot(y_train,depth=10).numpy()
y_test=tf.one_hot(y_test,depth=10).numpy()
print(tf.shape(y_train))
print(tf.shape(y_test))

4.归一化处理：

x_train,x_test=x_train.astype('float')/255.0,x_test.astype('float')/255.0

x_train=tf.reshape(x_train,[-1,28,28,1])
x_test=tf.reshape(x_test,[-1,28,28,1])

print(np.shape(x_train))
print(np.shape(y_train))
print(np.shape(x_test))
print(np.shape(y_test))

5.设置全局参数：

batch_Size=32
num_classes=10
EPOCHES=5
num_predictions=20

6.搭建传统的CNN模型：

model=tf.keras.Sequential([
    tf.keras.layers.InputLayer(input_shape=(28,28,1)),
    
    tf.keras.layers.Conv2D(32,kernel_size=[3,3],strides=[1,1],padding='same'),
    tf.keras.layers.Activation('relu'),
    
    tf.keras.layers.Conv2D(32,kernel_size=[3,3],strides=[1,1]),
    tf.keras.layers.Activation('relu'),
    tf.keras.layers.MaxPooling2D(pool_size=[2,2]),
    tf.keras.layers.Dropout(0.25),
    
    tf.keras.layers.Conv2D(64,kernel_size=[3,3],strides=[1,1],padding='same'),
    tf.keras.layers.Activation('relu'),
    
    tf.keras.layers.Conv2D(64,kernel_size=[3,3],strides=[1,1]),
    tf.keras.layers.Activation('relu'),
    tf.keras.layers.MaxPooling2D(pool_size=[2,2]),
    tf.keras.layers.Dropout(0.25),
    
    tf.keras.layers.Flatten(),
    
    tf.keras.layers.Dense(512),
    
    tf.keras.layers.Activation('relu'),
    tf.keras.layers.Dropout(0.5),
    
    tf.keras.layers.Dense(num_classes),
    tf.keras.layers.Activation('softmax')
])

model.summary()

7.数据增强：

data=ImageDataGenerator(
    featurewise_center=False,  
    samplewise_center=False,
    featurewise_std_normalization=False,
    samplewise_std_normalization=False,
    zca_whitening=False,
    zca_epsilon=1e-06,
    rotation_range=0,
    width_shift_range=0.1,
    height_shift_range=0.1,
    shear_range=0.,
    zoom_range=0.,
    channel_shift_range=0.,
    fill_mode='nearest',
    cval=0.,
    horizontal_flip=True,
    vertical_flip=False,
    rescale=None,
    preprocessing_function=None, 
    data_format=None,
    validation_split=0.0
)

了解更多关于ImageDataGenerator：可以看这位博主的文章：
https://blog.youkuaiyun.com/qq_27825451/article/details/90056896

train_gen=data.flow(x_train,y_train,batch_size=batch_Size)
test_gen=data.flow(x_test,y_test,batch_size=batch_Size)

8.模型优化器的选择和编译：

optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001)
model.compile(optimizer=optimizer,
              loss='categorical_crossentropy',
             metrics=['accuracy'] )

9.训练：

history=model.fit(
    train_gen,
    epochs=EPOCHES,
    verbose=1,
    validation_data=test_gen,
    workers=5,
    batch_size=batch_Size
)
model.save('fashion_Mnist_FM.h5')

训练结果：

10.查看history中的情况：

print(history.history)

11.画出accuracy,val_accuracy,loss,val_loss: