CNN手写字母识别-Keras版本

#coding:utf-8

from tensorflow.examples.tutorials.mnist import input_data
from keras.models import Sequential
from keras.layers import Convolution2D,MaxPooling2D,Flatten,Dense,Dropout
from keras.optimizers import Adam
from keras.utils import plot_model

# MNIST数据存放的路径
file = "../data/MNIST"
# 导入数据
mnist = input_data.read_data_sets(file, one_hot=True)
X_train = mnist.train.images
y_train = mnist.train.labels
X_train = X_train.reshape((-1,28,28,1))

X_test = mnist.test.images
y_test = mnist.test.labels
X_test = X_test.reshape((-1,28,28,1))

model = Sequential()
#卷积1，激活
model.add(Convolution2D(input_shape=(28,28,1),filters=30,kernel_size=5,strides=1,padding='same',activation='relu'))
#池化
model.add(MaxPooling2D(pool_size=(2,2),strides=1,padding='same'))

#卷积2，激活
model.add(Convolution2D(filters=30,kernel_size=5,strides=1,padding='same',activation='relu'))
#池化
model.add(MaxPooling2D(pool_size=(2,2),strides=1,padding='same'))

#Flatten , 转成一维向量
model.add(Flatten())

#全连接1
model.add(Dense(500,activation='relu'))

#Dropout 0.5
model.add(Dropout(0.5))

#全连接2
model.add(Dense(10,activation='softmax'))

#模型结构保存为图片
plot_model(model, to_file='mnist_cnn.png', show_shapes=True)

#优化器
adam = Adam(lr=1e-4)

#编译
model.compile(optimizer=adam,loss="categorical_crossentropy",metrics=['accuracy'])

#训练
model.fit(X_train,y_train,batch_size=100,epochs=10)

#评估效果
loss,acc = model.evaluate(X_test,y_test)
print(loss,acc)