Tensorflow2.0+Tensorflow2.1+mnist手写体识别双版本+预处理单个图片测试

最近才开始学机器学习，看Tensorflow实战Google看的真是头疼，全是老代码，都得自己手动解决。现在终于把书上手写体识别的一整套代码都改完了，2.1版本实测能跑，献给报错的同学！

1.首先是书上出现的第一个完全体代码

没有注释看不懂的看这个链接！ 不同的地方是为了能在2.1跑，以及文件路径要改成我自己的
Tensorflow学习–最佳的深度学习实践案例
要想自己手动升级代码到2.0版本的同学可以输入

tf_upgrade_v2 --infile ***.py --outfile ***_update.py

星号改成文件名，在控制台输入，控制台的位置要在你要升级的文件的位置，这个命令会把文件名.py修改后
，生成一个文件名_update.py的文件，就是修改完的文件。
然后打开修改完的开头加入下面代码，禁用动态就行了。

tf.compat.v1.disable_v2_behavior()

对于找不到tutorials的人，看这个链接！
Tensorflow 2.0 !!! No module named ‘tensorflow.examples.tutorials’
其实就是点进去，找到他的微云，下载tutorials，然后拷贝到\Lib\site-packages\tensorflow_core\examples
这个Lib的位置在哪，要看你怎么装的环境，比如我用的Anaconda3，还装了虚拟环境起名叫tsfgpu就在
Anaconda3\envs\tsfgpu\Lib\site-packages\tensorflow_core\examples

mnist数据集不知道放哪，不知道在哪的
如何导入MNIST数据集

下面是改完的全代码，你想跑的话把涉及到的路径改了就行

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

tf.compat.v1.disable_v2_behavior()

INPUT_NODE = 784
OUTPUT_NODE = 10
LAYER1_NODE = 500
BATCH_SIZE = 100
LEARNING_RATE_BASE = 0.8
LEARNING_RATE_DECAY = 0.99
REGULARIZATION_RATE = 0.0001
TRAINING_STEPS = 30000
MOVING_AVERAGE_DECAY = 0.99

def inference(input_tensor, avg_class, weights1, biases1, weights2, biases2):

    if avg_class == None:
        layer1 = tf.nn.relu(tf.matmul(input_tensor, weights1) + biases1)
        return tf.matmul(layer1, weights2) + biases2

    else:
        layer1 = tf.nn.relu(tf.matmul(input_tensor, avg_class.average(weights1)) + avg_class.average(biases1))
        return tf.matmul(layer1, avg_class.average(weights2)) + avg_class.average(biases2)

def train(mnist):

    x = tf.compat.v1.placeholder(dtype=tf.float32, shape=[None, INPUT_NODE], name='x-input')
    y_ = tf.compat.v1.placeholder(dtype=tf.float32, shape=[None, OUTPUT_NODE], name='y-input')

    weights1 = tf.Variable(tf.random.truncated_normal([INPUT_NODE, LAYER1_NODE], stddev=0.1))
    biases1 = tf.Variable(tf.constant(0.1, shape=[1, LAYER1_NODE]))

    weights2 = tf.Variable(tf.random.truncated_normal([LAYER1_NODE, OUTPUT_NODE], stddev=0.1))
    biases2 = tf.Variable(tf.constant(0.1, shape=[1, OUTPUT_NODE]))

    y = inference(x, None, weights1, biases1, weights2, biases2)

    global_step = tf.Variable(0, trainable=False)

    variable_averages = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)

    variables_averages_op = variable_averages.apply(tf.compat.v1.trainable_variables())

    average_y = inference(x, variable_averages, weights1, biases1, weights2, biases2)

    cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(input=y_, axis=1))

    cross_entropy_mean = tf.reduce_mean(input_tensor=cross_entropy)

    regularizer = tf.keras.regularizers.l2(0.5 * (REGULARIZATION_RATE))

    regularization = regularizer(weights1) + regularizer(weights2)

    loss = cross_entropy_mean + regularization

    learning_rate = tf.compat.v1.train.exponential_decay(
        LEARNING_RATE_BASE,
        global_step,
        mnist.train.num_examples / BATCH_SIZE,
        LEARNING_RATE_DECAY
    )

    train_step = tf.compat.v1.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)

    with tf.control_dependencies([train_step, variables_averages_op]):
        train_op = tf.no_op(name='train')

    correct_prrediction = tf.equal(tf.argmax(input=average_y, axis=1), tf.argmax(input=y_, axis=1))

    accuracy = tf.reduce_mean(input_tensor=tf.cast(correct_prrediction, tf.float32))

    with tf.compat.v1.Session() as sess:
        tf.compat.v1.initialize_all_variables().run()

        validate_feed = {x: mnist.validation.images, y_: mnist.validation.labels}

        test_feed = {x: mnist.test.images, y_: mnist.test.labels}

        for i in range(TRAINING_STEPS):

            if i % 1000 == 0:

                validate_acc = sess.run(accuracy, feed_dict=validate_feed)
                print("Aftrr %d training step(s),validation accuancy" 
                      "using average model is %g " % (i, validate_acc))

            xs, ys = mnist.train.next_batch(BATCH_SIZE)
            sess.run(train_op, feed_dict={x:xs, y_:ys})

        test_acc = sess.run(accuracy, feed_dict = test_feed)
        print("After %d training strep(s),test accuracy using average model is %g" % (TRAINING_STEPS, test_acc))

def main(argv=None):

    mnist = input_data.read_data_sets("./traindata/", one_hot=True)
    train(mnist)

if __name__ == '__main__':
        tf.compat.v1.app.run()

2.重构之后变成3个部分了！

看不懂的，需要注释的看这个链接！
TensorFlow学习之实现MNIST识别（实现断点重训）-----详细注解版

下面是我自己改完路径，升级完能跑Tensorflow2.1的代码，一样请你改路径使用！

mnist_inference_update

import tensorflow as tf

tf.compat.v1.disable_v2_behavior()

INPUT_NODE = 784
OUTPUT_NODE = 10
LAYER1_NODE = 500

def get_weight_variable(shape, regularizer):

    weights = tf.compat.v1.get_variable("weights", shape, initializer=tf.compat.v1.truncated_normal_initializer(stddev=0.1))

    if regularizer != None:
        tf.compat.v1.add_to_collection('losses', regularizer(weights))
    return weights

def inference(input_tensor, regularizer):
    with tf.compat.v1.variable_scope('layer1'):
        weights = get_weight_variable([INPUT_NODE, LAYER1_NODE], regularizer)
        biases = tf.compat.v1.get_variable("biases", [LAYER1_NODE], initializer=tf.compat.v1.constant_initializer(0.0))
        layer1 = tf.nn.relu(tf.matmul(input_tensor, weights) + biases)

    with tf.compat.v1.variable_scope('layer2'):
        weights = get_weight_variable([LAYER1_NODE, OUTPUT_NODE], regularizer)
        biases = tf.compat.v1.get_variable("biases", [OUTPUT_NODE], initializer=tf.compat.v1.constant_initializer(0.0))
        layer2 = tf.matmul(layer1, weights) + biases

    return layer2

mnist_train_update

import time
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

import mnist_inference_update
import mnist_train_update

tf.compat.v1.disable_v2_behavior()

EVAL_INTERVAL_SECS = 10
MODEL_SAVE_PATH = "MNIST_model/"
MODEL_NAME = "mnist_model"

def evaluate(mnist):
    with tf.Graph().as_default() as g:

        x = tf.compat.v1.placeholder(tf.float32, [None, mnist_inference_update.INPUT_NODE], name='x-input')
        y_ = tf.compat.v1.placeholder(tf.float32, [None, mnist_inference_update.OUTPUT_NODE], name='y-input')

        validate_feed = {x: mnist.validation.images, y_: mnist.validation.labels}
        y = mnist_inference_update.inference(x, None)

        correct_prediction = tf.equal(tf.argmax(input=y, axis=1), tf.argmax(input=y_, axis=1))
        accuracy = tf.reduce_mean(input_tensor=tf.cast(correct_prediction, tf.float32))

        variable_averages = tf.train.ExponentialMovingAverage(mnist_train_update.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.compat.v1.train.Saver(variables_to_restore)

        while True:
            with tf.compat.v1.Session() as sess:
                ckpt = tf.train.get_checkpoint_state(mnist_train_update.MODEL_SAVE_PATH)
                sess.run(tf.compat.v1.initialize_all_variables())
                if ckpt and ckpt.model_checkpoint_path:

                    saver.restore(sess, ckpt.model_checkpoint_path)

                    global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]

                    accuracy_score = sess.run(accuracy, feed_dict=validate_feed)
                    print("After %s training step(s), validation accuracy = %g" % (global_step, accuracy_score))
                else:
                    print('no checkpoint file found')
                    time.sleep(EVAL_INTERVAL_SECS)

def main(argv=None):
    mnist = input_data.read_data_sets("../traindata/", one_hot=True)
    evaluate(mnist)

if __name__ == '__main__':
    tf.compat.v1.app.run()

mnist_eval_update

import time
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

import mnist_inference_update
import mnist_train_update

tf.compat.v1.disable_v2_behavior()

EVAL_INTERVAL_SECS = 10
MODEL_SAVE_PATH = "MNIST_model/"
MODEL_NAME = "mnist_model"

def evaluate(mnist):
    with tf.Graph().as_default() as g:

        x = tf.compat.v1.placeholder(tf.float32, [None, mnist_inference_update.INPUT_NODE], name='x-input')
        y_ = tf.compat.v1.placeholder(tf.float32, [None, mnist_inference_update.OUTPUT_NODE], name='y-input')

        validate_feed = {x: mnist.validation.images, y_: mnist.validation.labels}
        y = mnist_inference_update.inference(x, None)

        correct_prediction = tf.equal(tf.argmax(input=y, axis=1), tf.argmax(input=y_, axis=1))
        accuracy = tf.reduce_mean(input_tensor=tf.cast(correct_prediction, tf.float32))

        variable_averages = tf.train.ExponentialMovingAverage(mnist_train_update.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.compat.v1.train.Saver(variables_to_restore)

        while True:
            with tf.compat.v1.Session() as sess:
                ckpt = tf.train.get_checkpoint_state(mnist_train_update.MODEL_SAVE_PATH)
                sess.run(tf.compat.v1.initialize_all_variables())
                if ckpt and ckpt.model_checkpoint_path:

                    saver.restore(sess, ckpt.model_checkpoint_path)

                    global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]

                    accuracy_score = sess.run(accuracy, feed_dict=validate_feed)
                    print("After %s training step(s), validation accuracy = %g" % (global_step, accuracy_score))
                else:
                    print('no checkpoint file found')
                    time.sleep(EVAL_INTERVAL_SECS)

def main(argv=None):
    mnist = input_data.read_data_sets("../traindata/", one_hot=True)
    evaluate(mnist)

if __name__ == '__main__':
    tf.compat.v1.app.run()

3.我想自己弄个图片试试，不用mnist的数据怎么办

首先我们要弄明白，mnist图片的是啥样的。
他的图片就是黑白的，28*28像素，并且是黑底白字！没错就是黑底白字!
这是我用画图做的黑底白字，

如果你的图片是白底黑字，那么把图片变成一行784列的矩阵之后，比如图片a变成矩阵b了，要用1-b再用。

图片处理的translate函数看不懂的看这个链接！
【手写数字图片预处理】用skimage将自己的手写数字转换为mnist数据集
skimage包没有的要自己安装，我用Anaconda，进入我的虚拟环境tsfgpu后，控制台输入

pip install -U scikit-image -i https://pypi.tuna.tsinghua.edu.cn/simple

记得用清华源
pypi 镜像使用帮助

最后我的全部代码如下，修改自mnist_eval_update，改了路径就能用了，是单个图片的测试
最后输出的是预测结果的矩阵，是个一行10列的。每一列代表0-9，其中最大的说明就是那个数
比如结果
[[18.982655 -3.8378751 1.0982409 -4.460092 -5.644824 -1.3264737
-2.2123477 1.0031121 -8.763886 4.256794 ]]
第一个数是18，10个里面最大，所以是数字0！

import time
import tensorflow as tf

from skimage import io,data,transform,color
import numpy as np

import mnist_inference_update
import mnist_train_update

tf.compat.v1.disable_v2_behavior()

MNIST_SIZE = 28
IMAGE_PATH = "./image/w0.PNG"

EVAL_INTERVAL_SECS = 10
MODEL_SAVE_PATH = "MNIST_model/"
MODEL_NAME = "mnist_model"

def translate(image_path):

    img = io.imread(image_path)
    img_gray = color.rgb2gray(img)

    translated_img = transform.resize(img_gray, (MNIST_SIZE, MNIST_SIZE))

    flatten_img = np.reshape(translated_img, 784)

    imgarray = np.array([flatten_img])

    result = np.around(imgarray, decimals=3)

    print(result)

    return result


def evaluate():
    with tf.Graph().as_default() as g:

        x = translate(image_path=IMAGE_PATH)
        y = mnist_inference_update.inference(tf.cast(x, tf.float32), None)

        variable_averages = tf.train.ExponentialMovingAverage(mnist_train_update.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.compat.v1.train.Saver(variables_to_restore)

        with tf.compat.v1.Session() as sess:
            ckpt = tf.train.get_checkpoint_state(mnist_train_update.MODEL_SAVE_PATH)
            sess.run(tf.compat.v1.initialize_all_variables())
            if ckpt and ckpt.model_checkpoint_path:

                saver.restore(sess, ckpt.model_checkpoint_path)

                outresult = sess.run(y)

                print(outresult)
            else:
                print('no checkpoint file found')
                time.sleep(EVAL_INTERVAL_SECS)

def main(argv=None):
    evaluate()

if __name__ == '__main__':
    tf.compat.v1.app.run()