0.文件结构如图,data文件夹用来存放训练后的权重,MNIST_data存放训练数据和测试数据
1.搭建线性回归和cnn两种模型,models.py
import tensorflow as tf
# y = wx+b
def regression(x):
W = tf.Variable(tf.zeros([784,10]),name="W")
b = tf.Variable(tf.zeros([10]),name="b")
y = tf.nn.softmax(tf.matmul(x,W)+b)
return y, [W,b]
def convolutional(x , keep_prob): #输入值,和drop_out 比例
def conv2d(x,W):
return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding="SAME")
def max_pool_2x2(x):
return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding="SAME")
def weight_varible(shape):
initial = tf.truncated_normal(shape,stddev=0.1)
return tf.Variable(initial)
def bias_varible(shape):
initial = tf.constant(0.1,shape=shape)
return tf.Variable(initial)
x_image = tf.reshape(x,[-1,28,28,1]) #-1 样本数任意,,28 28 为矩阵大小,,1 为图像深度
W_conv1 = weight_varible([5,5,1,32]) #卷积核为5x5,单通道输入(灰度图像),三十二个特征输出
b_conv1 = bias_varible([32])
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1)+b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_varible([5, 5, 32, 64]) #卷积核为5x5,连接第一层输出32通道输入,64个输出个特征输出
b_conv2 = bias_varible([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2) #池化层
#full connection 密集连接层
W_fc1 = weight_varible([7*7*64,1024])
b_fc1 = bias_varible([1024])
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1) + b_fc1)
#dropout
h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob)
W_fc2 = weight_varible([1024,10])
b_fc2 = bias_varible([10])
y = tf.nn.softmax(tf.matmul(h_fc1_drop,W_fc2) + b_fc2) #网络输出
return y , [W_conv1, b_conv1,W_conv2,b_conv2, W_fc1,b_fc1,W_fc2,b_fc2] #返回网络输出,和此时的权重和偏置
2.regression.py 回归训练
import os
import input_data
import models
import tensorflow as tf
data = input_data.read_data_sets('MNIST_data',one_hot=True) #通过特定方式,自动下载数据集
###被墙了,下载有问题
with tf.variable_scope("regression"):
x = tf.placeholder(tf.float32,[None,784])
y, variables = models.regression(x)
#train
y_ = tf.placeholder("float",[None,10])
cross_entropy = -tf.reduce_sum(y_*tf.log(y)) #计算交叉熵
train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy) #训练
correct_prediction = tf.equal(tf.arg_max(y,1),tf.argmax(y_,1)) #计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32)) #计算准确率
saver = tf.train.Saver(variables) #保存权重
with tf.Session() as sess:
merged_summary_op = tf.summary.merge_all()
summay_writer = tf.summary.FileWriter('/tmp/mnist_log/2', sess.graph)
# summay_writer.add_graph(sess.graph)
sess.run(tf.global_variables_initializer()) #初始化数据
for _ in range(1000):
batch_xs,batch_ys = data.train.next_batch(100)
sess.run(train_step,feed_dict={x:batch_xs,y_:batch_ys})
print((sess.run(accuracy,feed_dict={x:data.test.images,y_:data.test.labels})))
path = saver.save(
# sess,os.path.join(os.path.normpath(os.path.dirname(__file__)),'data','regression.ckpt'),write_meta_graph=False,write_state=False
sess,os.path.join(os.path.dirname(__file__),'data','regression.ckpt'),write_meta_graph=False,write_state=False ###os.path.dirname(__file__)为当前路径,不能含有汉字,否则会报错:路径不存在
)
print("Saved:",path)
3,convolution.py cnn训练
import os
import models
import tensorflow as tf
import input_data
data = input_data.read_data_sets('MNIST_data',one_hot=True)
#model
with tf.variable_scope("convolutional"):
x = tf.placeholder(tf.float32,[None,784],name = 'x')
keep_prob = tf.placeholder(tf.float32)
y ,variables = models.convolutional(x,keep_prob)
#train
y_ = tf.placeholder(tf.float32,[None,10],name='y')
cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)#AdadeltaOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
saver = tf.train.Saver(variables)
with tf.Session() as sess:
merged_summary_op = tf.summary.merge_all()
summay_writer = tf.summary.FileWriter('/tmp/mnist_log/1',sess.graph)
summay_writer.add_graph(sess.graph)
# summay_writer.add_graph(sess.)
sess.run(tf.global_variables_initializer())
for i in range(20000):
batch = data.train.next_batch(50)
if i%100 == 0:
train_accuracy = accuracy.eval(feed_dict={x:batch[0],y_:batch[1],keep_prob:1.0})
print("srep %d, train accuracy %g" % (i,train_accuracy))
sess.run(train_step,feed_dict={x:batch[0],y_:batch[1],keep_prob:0.5})
print(sess.run(accuracy,feed_dict={x:data.test.images,y_:data.test.labels,keep_prob:1.0}))
path = saver.save(sess,os.path.join(os.path.dirname(__file__),'data','convolutional.ckpt'),write_meta_graph=False,write_state=False)
print("Saved:",path)
4.执行regression.py 和 convolution.py 后训练模型会保存在指定的文件夹下,在下面的main.py中使用训练好的数据集
import numpy as np
import tensorflow as tf
from mnist import models
from PIL import Image
from PIL import ImageDraw
from PIL import ImageFont
x = tf.placeholder("float",[None,784])
sess = tf.Session()
##回归
with tf.variable_scope("regression"):
y1, variables = models.regression(x)
saver = tf.train.Saver(variables)
saver.restore(sess,"mnist/data/regression.ckpt")
##回归
##cnn
with tf.variable_scope("convolutional"):
keep_prob = tf.placeholder("float")
y2,variables = models.convolutional(x,keep_prob)
saver = tf.train.Saver(variables)
saver.restore(sess,"mnist/data/convolutional.ckpt")
##cnn
def regression(input): #处理输入图片
return sess.run(y1,feed_dict={x:input}).flatten().tolist()
def convolutional(input):#处理输入图片(卷积)
return sess.run(y2,feed_dict={x:input,keep_prob:1.0}).flatten().tolist()
def input_image(img): #对图片进行处理
img_l = img.convert("L").resize((28,28))
return ((255 - np.array(img_l, dtype=np.uint8)) / 255.0).reshape([1,784])
if __name__ == '__main__':
each = '1'
blank = Image.new("RGB", [56, 56], "white") # 创建背景[269,70]
drawObject = ImageDraw.Draw(blank) # 加载背景
ttf = 'tyc-num-d325fc6b6b.ttf' # 字体文件,在网上找的字体文件,生成字体,画在图片上,供网络输入
Font4 = ImageFont.truetype(ttf, 24) # 加载字体,字体大小
drawObject.text([20, 20], each, font=Font4, fill='black') # 以字体库生成字体图片,起始坐标[5, 5],字体黑色'123456890.' 026974381
#blank.save(each.replace(".","point")+".jpg")
blank.save(each +".jpg")
photos = input_image(blank) #处理图片
out_put1 = regression(photos)
out_put2 = convolutional(photos)
print(out_put1.index(max(out_put1)))
lis = []
for each_p in out_put2:
lis.append('{:.16f}'.format(each_p)) #cnn输出的是指数形式的值,转化为浮点形式
print(lis.index(max(lis)))
print("###############")