卷积神经网络——conv2d

# coding: utf-8

# In[1]:


import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers,Sequential
from tensorflow.examples.tutorials.mnist import input_data
import matplotlib.pyplot as plt
import numpy as np
tf.__version__


# In[2]:


#图片数量，图片长，宽，输入通道数
x=tf.random.normal([1,5,5,3])


# In[3]:


#过滤器的高，宽，输入通道数，输出通道数
w=tf.random.normal([3,3,3,6])


# In[4]:


#输入，过滤器，步长，填充（上下左右填充0）
out1=tf.nn.conv2d(input=x,filters=w,strides=[1,1,1,1],padding=[[0,0],[0,0],[0,0],[0,0]])


# In[5]:


out1


# In[6]:


model=tf.keras.Sequential([keras.layers.Conv2D(6,3),
                          keras.layers.MaxPooling2D(pool_size=2,strides=2),
                          keras.layers.ReLU(),
                          keras.layers.Conv2D(16,3),
                          keras.layers.MaxPooling2D(pool_size=2,strides=2),
                          keras.layers.ReLU(),
                          keras.layers.Flatten(),
                          keras.layers.Dense(120,activation="relu"),
                          keras.layers.Dense(84,activation="relu"),
                          keras.layers.Dense(10,activation="softmax")
                    ])
batch=32
model.build(input_shape=(batch,28,28,1))
model.summary()


# In[7]:


loss=keras.losses.CategoricalCrossentropy()


# In[8]:


model.compile(optimizer=keras.optimizers.Adam(),
             loss=keras.losses.CategoricalCrossentropy(),
             metrics=['accuracy'])


# In[30]:


(x_train,y_train),(x_test,y_test)=tf.keras.datasets.mnist.load_data()
type(x_train),x_train[0]
print(y_train[0])
plt.matshow(x_train[0])
plt.show()


# In[32]:


def preprocess(x,y):
    '''
	最简单的预处理函数:
		转numpy为Tensor、分类问题需要处理label为one_hot编码、处理训练数据
	'''
    # 把numpy数据转为Tensor
    x=tf.cast(x,dtype=tf.float32)/255
    x=tf.reshape(x,[-1,28,28,1])
    y=tf.one_hot(y,depth=10)
    return x,y
(x_train,y_train),(x_test,y_test)=tf.keras.datasets.mnist.load_data()
# 使用 tf.data.Dataset.from_tensor_slices 进行加载
train_db=tf.data.Dataset.from_tensor_slices((x_train,y_train))
print(train_db)
train_db=train_db.shuffle(10000)
#设置 batch size 一次喂入64个数据
train_db=train_db.batch(128)
#  预处理 (预处理函数在下面)
train_db=train_db.map(preprocess)

test_db=tf.data.Dataset.from_tensor_slices((x_test,y_test))
test_db=test_db.shuffle(10000)
test_db=test_db.batch(128)
test_db=test_db.map(preprocess)
print("ok")


# In[20]:


model.fit(train_db,epochs=5)


# In[17]:


model.evaluate(test_db)