Простой тест нейронной сети

1-1. Только ввод и вывод, без скрытого слоя

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据集
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)
#每个批次的大小
batch_size=100
#计算一共有多少批次
n_batch = mnist.train.num_examples // batch_size
#定义两个placehold
x = tf.placeholder(tf.float32,[None,784])
y = tf.placeholder(tf.float32,[None,10])

#创建一个简单的神经网络
W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))
prediction = tf.nn.softmax(tf.matmul(x,W)+b)
#二次代价函数
loss = tf.reduce_mean(tf.square(y-prediction))
#使用梯度下降法
train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)

#初始化变量
init = tf.global_variables_initializer()
#存放结果到一个布尔型变量中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(20):
        for batch in range(n_batch):
            batch_xs,batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys})

        acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels})
        print("Iter "+str(epoch) + ",Testing Accuracy "+str(acc))

Результаты тренировки 20 раз таковы:

Iter 0,Testing Accuracy 0.8303
Iter 1,Testing Accuracy 0.8712
Iter 2,Testing Accuracy 0.8808
Iter 3,Testing Accuracy 0.8887
Iter 4,Testing Accuracy 0.8944
Iter 5,Testing Accuracy 0.8973
Iter 6,Testing Accuracy 0.8992
Iter 7,Testing Accuracy 0.9032
Iter 8,Testing Accuracy 0.9034
Iter 9,Testing Accuracy 0.905
Iter 10,Testing Accuracy 0.9068
Iter 11,Testing Accuracy 0.9074
Iter 12,Testing Accuracy 0.9083
Iter 13,Testing Accuracy 0.9096
Iter 14,Testing Accuracy 0.9101
Iter 15,Testing Accuracy 0.9109
Iter 16,Testing Accuracy 0.9118
Iter 17,Testing Accuracy 0.9121
Iter 18,Testing Accuracy 0.9127
Iter 19,Testing Accuracy 0.9133

1-2.Использование функции перекрестной энтропии

# #二次代价函数
# loss = tf.reduce_mean(tf.square(y-prediction))
#交叉熵定义二次代价函数作用是加快模型收敛速度
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))

Результаты обучения следующие:

Iter 0,Testing Accuracy 0.8239
Iter 1,Testing Accuracy 0.892
Iter 2,Testing Accuracy 0.9011
Iter 3,Testing Accuracy 0.9052
Iter 4,Testing Accuracy 0.9086
Iter 5,Testing Accuracy 0.9098
Iter 6,Testing Accuracy 0.912
……
Iter 15,Testing Accuracy 0.921
Iter 16,Testing Accuracy 0.9207
Iter 17,Testing Accuracy 0.9221
Iter 18,Testing Accuracy 0.9213
Iter 19,Testing Accuracy 0.922

2-1 Нейронная сеть с тремя скрытыми слоями

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据集
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)
#每个批次的大小
batch_size=100
#计算一共有多少批次
n_batch = mnist.train.num_examples // batch_size
#定义两个placehold
x = tf.placeholder(tf.float32,[None,784])
y = tf.placeholder(tf.float32,[None,10])
keep_prob = tf.placeholder(tf.float32)
#创建一个简单的神经网络
W1 = tf.Variable(tf.truncated_normal([784,500],stddev=0.1))
b1 = tf.Variable(tf.zeros([500])+0.1)
L1 = tf.nn.tanh(tf.matmul(x,W1)+b1)
L1_drop = tf.nn.dropout(L1,keep_prob)

W2 = tf.Variable(tf.truncated_normal([500,300],stddev=0.1))
b2 = tf.Variable(tf.zeros([300])+0.1)
L2 = tf.nn.tanh(tf.matmul(L1_drop,W2)+b2)
L2_drop = tf.nn.dropout(L2,keep_prob)

W3 = tf.Variable(tf.truncated_normal([300,10],stddev=0.1))
b3 = tf.Variable(tf.zeros([10])+0.1)
prediction = tf.nn.softmax(tf.matmul(L2_drop,W3)+b3)

#二次代价函数
loss = tf.reduce_mean(tf.square(y-prediction))
#使用梯度下降法
train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)

#初始化变量
init = tf.global_variables_initializer()
#存放结果到一个布尔型变量中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(51):
        for batch in range(n_batch):
            batch_xs,batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys,keep_prob:1.0})

        test_acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels,keep_prob:1.0})
        train_acc = sess.run(accuracy,feed_dict={x:mnist.train.images,y:mnist.train.labels,keep_prob:1.0})
        print("Iter "+str(epoch) + ",Testing Accuracy "+str(test_acc))

Результаты тренировки 50 раз таковы:

Iter 0,Testing Accuracy 0.8852
Iter 1,Testing Accuracy 0.9103
Iter 2,Testing Accuracy 0.9179
Iter 3,Testing Accuracy 0.9257
……
Iter 13,Testing Accuracy 0.9502
Iter 14,Testing Accuracy 0.9504
……
Iter 24,Testing Accuracy 0.9605
Iter 25,Testing Accuracy 0.9603
……
Iter 49,Testing Accuracy 0.9699
Iter 50,Testing Accuracy 0.9698

2-2 Вместо этого используйте функцию перекрестной энтропии

loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))

После пятидесяти тренировок:

Iter 0,Testing Accuracy 0.9239
……
Iter 4,Testing Accuracy 0.9529
……
Iter 7,Testing Accuracy 0.9619
……
Iter 14,Testing Accuracy 0.9714
……
Iter 49,Testing Accuracy 0.9769
Iter 50,Testing Accuracy 0.9765

2-3 Дальнейшие изменения в модели оптимизации оптимизатора

• Использование AdamOptimizer

#train_step = tf.train.AdamOptimizer(1e-2).minimize(loss)

Iter 0,Testing Accuracy 0.9255
……
Iter 50,Testing Accuracy 0.9255

• Использование MomentumOptimizer

#train_step = tf.train.MomentumOptimizer(learning_rate=0.1,momentum=0.9).minimize(loss)

Iter 0,Testing Accuracy 0.9422
Iter 1,Testing Accuracy 0.9573
……
Iter 5,Testing Accuracy 0.9708
……
Iter 50,Testing Accuracy 0.9793

2-4 Оптимизировано до 98%

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据集
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)

#每个批次的大小
batch_size=100
#计算一共有多少批次
n_batch = mnist.train.num_examples // batch_size

#定义两个placehold
x = tf.placeholder(tf.float32,[None,784])
y = tf.placeholder(tf.float32,[None,10])
keep_prob = tf.placeholder(tf.float32)
lr = tf.Variable(0.001,dtype=tf.float32) #学习率

#创建一个简单的神经网络
W1 = tf.Variable(tf.truncated_normal([784,500],stddev=0.1))
b1 = tf.Variable(tf.zeros([500])+0.1)
L1 = tf.nn.tanh(tf.matmul(x,W1)+b1)
L1_drop = tf.nn.dropout(L1,keep_prob)

W2 = tf.Variable(tf.truncated_normal([500,300],stddev=0.1))
b2 = tf.Variable(tf.zeros([300])+0.1)
L2 = tf.nn.tanh(tf.matmul(L1_drop,W2)+b2)
L2_drop = tf.nn.dropout(L2,keep_prob)

W3 = tf.Variable(tf.truncated_normal([300,10],stddev=0.1))
b3 = tf.Variable(tf.zeros([10])+0.1)
prediction = tf.nn.softmax(tf.matmul(L2_drop,W3)+b3)

#交叉熵代价函数
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=prediction))
#训练
train_step = tf.train.AdamOptimizer(lr).minimize(loss)

#初始化变量
init = tf.global_variables_initializer()
#存放结果到一个布尔型变量中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(51):
        sess.run(tf.assign(lr,0.001 * (0.95 ** epoch)))
        for batch in range(n_batch):
            batch_xs,batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys,keep_prob:1.0}) 
        learning_rate = sess.run(lr)
        acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels,keep_prob:1.0})
        print("Iter "+str(epoch) + ",Testing Accuracy= "+str(acc)+" Learning Rate=  "+str(learning_rate))

Результаты обучения следующие:

Iter 0,Testing Accuracy= 0.9497 Learning Rate=  0.001
Iter 1,Testing Accuracy= 0.9629 Learning Rate=  0.00095
Iter 2,Testing Accuracy= 0.971 Learning Rate=  0.0009025
Iter 3,Testing Accuracy= 0.9664 Learning Rate=  0.000857375
Iter 4,Testing Accuracy= 0.973 Learning Rate=  0.00081450626
Iter 5,Testing Accuracy= 0.9766 Learning Rate=  0.0007737809
Iter 6,Testing Accuracy= 0.9756 Learning Rate=  0.0007350919
Iter 7,Testing Accuracy= 0.9753 Learning Rate=  0.0006983373
Iter 8,Testing Accuracy= 0.9767 Learning Rate=  0.0006634204
Iter 9,Testing Accuracy= 0.9786 Learning Rate=  0.0006302494
Iter 10,Testing Accuracy= 0.9782 Learning Rate=  0.0005987369
Iter 11,Testing Accuracy= 0.9781 Learning Rate=  0.0005688001
Iter 12,Testing Accuracy= 0.9767 Learning Rate=  0.0005403601
Iter 13,Testing Accuracy= 0.9796 Learning Rate=  0.0005133421
Iter 14,Testing Accuracy= 0.9807 Learning Rate=  0.000487675
Iter 15,Testing Accuracy= 0.9809 Learning Rate=  0.00046329122
……
Iter 49,Testing Accuracy= 0.9818 Learning Rate=  8.099471e-05
Iter 50,Testing Accuracy= 0.9825 Learning Rate=  7.6944976e-05

3 Распознавание изображений с помощью Inception-v3

daima
import tensorflow as tf
import os
import numpy as np
import re
from PIL import Image
import matplotlib.pyplot as plt



class NodeLookup(object):
    def __init__(self):  
        label_lookup_path = 'inception_model/imagenet_2012_challenge_label_map_proto.pbtxt'   
        uid_lookup_path = 'inception_model/imagenet_synset_to_human_label_map.txt'
        self.node_lookup = self.load(label_lookup_path, uid_lookup_path)

    def load(self, label_lookup_path, uid_lookup_path):

        proto_as_ascii_lines = tf.gfile.GFile(uid_lookup_path).readlines()
        uid_to_human = {}
        #一行一行读取数据
        for line in proto_as_ascii_lines :
            #去掉换行符
            line=line.strip('\n')
            #按照'\t'分割
            parsed_items = line.split('\t')
            #获取分类编号
            uid = parsed_items[0]
            #获取分类名称
            human_string = parsed_items[1]
            #保存编号字符串n********与分类名称映射关系
            uid_to_human[uid] = human_string

        # 加载分类字符串n********对应分类编号1-1000的文件
        proto_as_ascii = tf.gfile.GFile(label_lookup_path).readlines()
        node_id_to_uid = {}
        for line in proto_as_ascii:
            if line.startswith('  target_class:'):
                #获取分类编号1-1000
                target_class = int(line.split(': ')[1])
            if line.startswith('  target_class_string:'):
                #获取编号字符串n********
                target_class_string = line.split(': ')[1]
                #保存分类编号1-1000与编号字符串n********映射关系
                node_id_to_uid[target_class] = target_class_string[1:-2]

        #建立分类编号1-1000对应分类名称的映射关系
        node_id_to_name = {}
        for key, val in node_id_to_uid.items():
            #获取分类名称
            name = uid_to_human[val]
            #建立分类编号1-1000到分类名称的映射关系
            node_id_to_name[key] = name
        return node_id_to_name

    #传入分类编号1-1000返回分类名称
    def id_to_string(self, node_id):
        if node_id not in self.node_lookup:
            return ''
        return self.node_lookup[node_id]


#创建一个图来存放google训练好的模型
with tf.gfile.FastGFile('inception_model/classify_image_graph_def.pb', 'rb') as f:
    graph_def = tf.GraphDef()
    graph_def.ParseFromString(f.read())
    tf.import_graph_def(graph_def, name='')


with tf.Session() as sess:
    softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
    #遍历目录
    for root,dirs,files in os.walk('images/'):
        for file in files:
            #载入图片
            image_data = tf.gfile.FastGFile(os.path.join(root,file), 'rb').read()
            predictions = sess.run(softmax_tensor,{'DecodeJpeg/contents:0': image_data})#图片格式是jpg格式
            predictions = np.squeeze(predictions)#把结果转为1维数据

            #打印图片路径及名称
            image_path = os.path.join(root,file)
            print(image_path)
            #显示图片
            img=Image.open(image_path)
            plt.imshow(img)
            plt.axis('off')
            plt.show()

            #排序
            top_k = predictions.argsort()[-5:][::-1]
            node_lookup = NodeLookup()
            for node_id in top_k:     
                #获取分类名称
                human_string = node_lookup.id_to_string(node_id)
                #获取该分类的置信度
                score = predictions[node_id]
                print('%s (score = %.5f)' % (human_string, score))
            print()

Результат распознавания

images/man.jpg
Windsor tie (score = 0.73132)
suit, suit of clothes (score = 0.19451)
stole (score = 0.01027)
jean, blue jean, denim (score = 0.00458)
wool, woolen, woollen (score = 0.00434)

airliner (score = 0.80073)
space shuttle (score = 0.05414)
wing (score = 0.02091)
airship, dirigible (score = 0.00550)
warplane, military plane (score = 0.00169)

lakeside, lakeshore (score = 0.93102)
breakwater, groin, groyne, mole, bulwark, seawall, jetty (score = 0.01073)
goose (score = 0.00414)
boathouse (score = 0.00362)
valley, vale (score = 0.00271)

acoustic guitar (score = 0.97403)
banjo (score = 0.00399)
electric guitar (score = 0.00208)
pick, plectrum, plectron (score = 0.00061)
garter snake, grass snake (score = 0.00035)

sports car, sport car (score = 0.62812)
convertible (score = 0.14632)
racer, race car, racing car (score = 0.13355)
car wheel (score = 0.00704)
limousine, limo (score = 0.00574)

golden retriever (score = 0.94793)
Labrador retriever (score = 0.00335)
flat-coated retriever (score = 0.00066)
Tibetan mastiff (score = 0.00063)
Leonberg (score = 0.00061)