生成mnist_model_graph.pb

简介

应用于tensorflow生成.pb文件供Android调用。新的手写AndroidTensorFlowMNISTExample-master.rar

代码

from __future__ import print_functionimport shutilimport os.pathimport tensorflow as tffrom tensorflow.examples.tutorials.mnist import input_dataEXPORT_DIR = './model'if os.path.exists(EXPORT_DIR):    shutil.rmtree(EXPORT_DIR)mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)# Parameterslearning_rate = 0.001training_iters = 200000batch_size = 128display_step = 10# Network Parametersn_input = 784  # MNIST data input (img shape: 28*28)n_classes = 10  # MNIST total classes (0-9 digits)dropout = 0.75  # Dropout, probability to keep units# tf Graph inputx = tf.placeholder(tf.float32, [None, n_input])y = tf.placeholder(tf.float32, [None, n_classes])keep_prob = tf.placeholder(tf.float32)  # dropout (keep probability)# Create some wrappers for simplicitydef conv2d(x, W, b, strides=1):    # Conv2D wrapper, with bias and relu activation    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')    x = tf.nn.bias_add(x, b)    return tf.nn.relu(x)def maxpool2d(x, k=2):    # MaxPool2D wrapper    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],                          padding='SAME')# Create Modeldef conv_net(x, weights, biases, dropout):    # Reshape input picture    x = tf.reshape(x, shape=[-1, 28, 28, 1])    # Convolution Layer    conv1 = conv2d(x, weights['wc1'], biases['bc1'])    # Max Pooling (down-sampling)    conv1 = maxpool2d(conv1, k=2)    # Convolution Layer    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])    # Max Pooling (down-sampling)    conv2 = maxpool2d(conv2, k=2)    # Fully connected layer    # Reshape conv2 output to fit fully connected layer input    fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])    fc1 = tf.nn.relu(fc1)    # Apply Dropout    fc1 = tf.nn.dropout(fc1, dropout)    # Output, class prediction    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])    return out# Store layers weight & biasweights = {    # 5x5 conv, 1 input, 32 outputs    'wc1': tf.Variable(tf.random_normal([5, 5, 1, 32])),    # 5x5 conv, 32 inputs, 64 outputs    'wc2': tf.Variable(tf.random_normal([5, 5, 32, 64])),    # fully connected, 7*7*64 inputs, 1024 outputs    'wd1': tf.Variable(tf.random_normal([7 * 7 * 64, 1024])),    # 1024 inputs, 10 outputs (class prediction)    'out': tf.Variable(tf.random_normal([1024, n_classes]))}biases = {    'bc1': tf.Variable(tf.random_normal([32])),    'bc2': tf.Variable(tf.random_normal([64])),    'bd1': tf.Variable(tf.random_normal([1024])),    'out': tf.Variable(tf.random_normal([n_classes]))}# Construct modelpred = conv_net(x, weights, biases, keep_prob)# Define loss and optimizercost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)# Evaluate modelcorrect_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))# Initializing the variablesinit = tf.initialize_all_variables()# Launch the graphwith tf.Session() as sess:    sess.run(init)    step = 1    # Keep training until reach max iterations    while step * batch_size < training_iters:        batch_x, batch_y = mnist.train.next_batch(batch_size)        # Run optimization op (backprop)        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y,                                       keep_prob: dropout})        if step % display_step == 0:            # Calculate batch loss and accuracy            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_x,                                                              y: batch_y,                                                              keep_prob: 1.})            print("Iter " + str(step * batch_size) + ", Minibatch Loss= " + \                  "{:.6f}".format(loss) + ", Training Accuracy= " + \                  "{:.5f}".format(acc))        step += 1    print("Optimization Finished!")    # Calculate accuracy for 256 mnist test images    print("Testing Accuracy:", \          sess.run(accuracy, feed_dict={x: mnist.test.images[:256],                                        y: mnist.test.labels[:256],                                        keep_prob: 1.}))    WC1 = weights['wc1'].eval(sess)    BC1 = biases['bc1'].eval(sess)    WC2 = weights['wc2'].eval(sess)    BC2 = biases['bc2'].eval(sess)    WD1 = weights['wd1'].eval(sess)    BD1 = biases['bd1'].eval(sess)    W_OUT = weights['out'].eval(sess)    B_OUT = biases['out'].eval(sess)# Create new graph for exportingg = tf.Graph()with g.as_default():    x_2 = tf.placeholder("float", shape=[None, 784], name="input")    WC1 = tf.constant(WC1, name="WC1")    BC1 = tf.constant(BC1, name="BC1")    x_image = tf.reshape(x_2, [-1, 28, 28, 1])    CONV1 = conv2d(x_image, WC1, BC1)    MAXPOOL1 = maxpool2d(CONV1, k=2)    WC2 = tf.constant(WC2, name="WC2")    BC2 = tf.constant(BC2, name="BC2")    CONV2 = conv2d(MAXPOOL1, WC2, BC2)    MAXPOOL2 = maxpool2d(CONV2, k=2)    WD1 = tf.constant(WD1, name="WD1")    BD1 = tf.constant(BD1, name="BD1")    FC1 = tf.reshape(MAXPOOL2, [-1, WD1.get_shape().as_list()[0]])    FC1 = tf.add(tf.matmul(FC1, WD1), BD1)    FC1 = tf.nn.relu(FC1)    W_OUT = tf.constant(W_OUT, name="W_OUT")    B_OUT = tf.constant(B_OUT, name="B_OUT")    # skipped dropout for exported graph as there is no need for already calculated weights    OUTPUT = tf.nn.softmax(tf.matmul(FC1, W_OUT) + B_OUT, name="output")    sess = tf.Session()    init = tf.initialize_all_variables()    sess.run(init)    graph_def = g.as_graph_def()    tf.train.write_graph(graph_def, EXPORT_DIR, 'mnist_model_graph.pb', as_text=False)    # Test trained model    y_train = tf.placeholder("float", [None, 10])    correct_prediction = tf.equal(tf.argmax(OUTPUT, 1), tf.argmax(y_train, 1))    accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))    print("check accuracy %g" % accuracy.eval(            {x_2: mnist.test.images, y_train: mnist.test.labels}, sess))

代码的改进

增加可变的学习速率，以及边训练边验证：

'''Created on 2017年10月24日@author: XuTing'''import mathimport shutilimport os.pathimport tensorflow as tffrom tensorflow.examples.tutorials.mnist import input_dataEXPORT_DIR = './model'if os.path.exists(EXPORT_DIR):    shutil.rmtree(EXPORT_DIR)mnist = input_data.read_data_sets("E:/Program Files/MyEclipseCode/PythonCode/MyPythonCode/MyPythonTry/MNIST_data/", one_hot=True)# Parametersmax_learning_rate = 0.001min_learning_rate = 0.00001decay_speed = 2000.0 lr = tf.placeholder(tf.float32)training_iters = 200batch_size = 128display_step = 10# Network Parametersn_input = 784  # MNIST data input (img shape: 28*28)n_classes = 10  # MNIST total classes (0-9 digits)dropout = 0.75  # Dropout, probability to keep units# tf Graph inputx = tf.placeholder(tf.float32, [None, n_input])y = tf.placeholder(tf.float32, [None, n_classes])keep_prob = tf.placeholder(tf.float32)  # dropout (keep probability)# Create some wrappers for simplicitydef conv2d(x, W, b, strides=1):    # Conv2D wrapper, with bias and relu activation    x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')    x = tf.nn.bias_add(x, b)    return tf.nn.relu(x)def maxpool2d(x, k=2):    # MaxPool2D wrapper    return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],                          padding='SAME')# Create Modeldef conv_net(x, weights, biases, dropout):    # Reshape input picture    x = tf.reshape(x, shape=[-1, 28, 28, 1])    # Convolution Layer    conv1 = conv2d(x, weights['wc1'], biases['bc1'])    # Max Pooling (down-sampling)    conv1 = maxpool2d(conv1, k=2)    # Convolution Layer    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])    # Max Pooling (down-sampling)    conv2 = maxpool2d(conv2, k=2)    # Fully connected layer    # Reshape conv2 output to fit fully connected layer input    fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])    fc1 = tf.nn.relu(fc1)    # Apply Dropout    fc1 = tf.nn.dropout(fc1, dropout)    # Output, class prediction    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])    return out# Store layers weight & biasweights = {    # 5x5 conv, 1 input, 32 outputs    'wc1': tf.Variable(tf.random_normal([5, 5, 1, 32])),    # 5x5 conv, 32 inputs, 64 outputs    'wc2': tf.Variable(tf.random_normal([5, 5, 32, 64])),    # fully connected, 7*7*64 inputs, 1024 outputs    'wd1': tf.Variable(tf.random_normal([7 * 7 * 64, 1024])),    # 1024 inputs, 10 outputs (class prediction)    'out': tf.Variable(tf.random_normal([1024, n_classes]))}biases = {    'bc1': tf.Variable(tf.random_normal([32])),    'bc2': tf.Variable(tf.random_normal([64])),    'bd1': tf.Variable(tf.random_normal([1024])),    'out': tf.Variable(tf.random_normal([n_classes]))}# Construct modelpred = conv_net(x, weights, biases, keep_prob)# Define loss and optimizercost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))optimizer = tf.train.AdamOptimizer(learning_rate=lr).minimize(cost)# Evaluate modelcorrect_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))# Initializing the variablesinit = tf.global_variables_initializer()# Launch the graphwith tf.Session() as sess:    sess.run(init)    step = 1    # Keep training until reach max iterations    while step < training_iters:        learningRate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-step / decay_speed)        batch_x, batch_y = mnist.train.next_batch(batch_size)        batch_tx, batch_ty = mnist.test.next_batch(batch_size)        # Run optimization op (backprop)        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y,                                       keep_prob: dropout, lr:learningRate})        if step % display_step == 0:            # Calculate batch loss and accuracy            loss, acc = sess.run([cost, accuracy], feed_dict={x: batch_x,                                                              y: batch_y,                                                              keep_prob: 1.,                                                              lr:learningRate})            print("Iter " + str(step) + ", Minibatch Loss= " + \                  "{:.6f}".format(loss) + ", Training Accuracy= " + \                  "{:.5f}".format(acc) + ",learningRate:{}".format(learningRate))        if step % 50 == 0:            print(" ")            # Calculate accuracy for 256 mnist test images            testAcc = sess.run(accuracy, feed_dict={x: batch_tx,                                        y: batch_ty,                                        keep_prob: 1.})            print("------------------Testing Accuracy:",testAcc)        step += 1    print("Optimization Finished!")    WC1 = weights['wc1'].eval(sess)    BC1 = biases['bc1'].eval(sess)    WC2 = weights['wc2'].eval(sess)    BC2 = biases['bc2'].eval(sess)    WD1 = weights['wd1'].eval(sess)    BD1 = biases['bd1'].eval(sess)    W_OUT = weights['out'].eval(sess)    B_OUT = biases['out'].eval(sess)# Create new graph for exportingg = tf.Graph()with g.as_default():    x_2 = tf.placeholder("float", shape=[None, 784], name="input")    WC1 = tf.constant(WC1, name="WC1")    BC1 = tf.constant(BC1, name="BC1")    x_image = tf.reshape(x_2, [-1, 28, 28, 1])    CONV1 = conv2d(x_image, WC1, BC1)    MAXPOOL1 = maxpool2d(CONV1, k=2)    WC2 = tf.constant(WC2, name="WC2")    BC2 = tf.constant(BC2, name="BC2")    CONV2 = conv2d(MAXPOOL1, WC2, BC2)    MAXPOOL2 = maxpool2d(CONV2, k=2)    WD1 = tf.constant(WD1, name="WD1")    BD1 = tf.constant(BD1, name="BD1")    FC1 = tf.reshape(MAXPOOL2, [-1, WD1.get_shape().as_list()[0]])    FC1 = tf.add(tf.matmul(FC1, WD1), BD1)    FC1 = tf.nn.relu(FC1)    W_OUT = tf.constant(W_OUT, name="W_OUT")    B_OUT = tf.constant(B_OUT, name="B_OUT")    # skipped dropout for exported graph as there is no need for already calculated weights    OUTPUT = tf.nn.softmax(tf.matmul(FC1, W_OUT) + B_OUT, name="softmax")    out_label = tf.argmax(OUTPUT, 1,name="output")    sess = tf.Session()    init = tf.global_variables_initializer()    sess.run(init)    graph_def = g.as_graph_def()    tf.train.write_graph(graph_def, EXPORT_DIR, 'mnist_model_graph.pb', as_text=False)    print("Save model to {}/mnist_model_graph.pb".format(EXPORT_DIR))    # Test trained model    '''    y_train = tf.placeholder("float", [None, 10])    correct_prediction = tf.equal(tf.argmax(OUTPUT, 1), tf.argmax(y_train, 1))    accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))    print("check accuracy %g" % accuracy.eval(            {x_2: mnist.test.images, y_train: mnist.test.labels}, sess))    '''

运行后的参数，测试：

'''Created on 2017年9月9日@author: admin'''import matplotlib.pyplot as pltimport tensorflow as tfimport  numpy as npimport PIL.Image as Imagefrom skimage import transformfrom tensorflow.examples.tutorials.mnist import input_databatch_size =1def recognize(pb_file_path):    with tf.Graph().as_default():        output_graph_def = tf.GraphDef()        with open(pb_file_path, "rb") as f:            output_graph_def.ParseFromString(f.read())  # rb            _ = tf.import_graph_def(output_graph_def, name="")        with tf.Session() as sess:            tf.global_variables_initializer().run()            input_x = sess.graph.get_tensor_by_name("input:0")            print (input_x)            out_softmax = sess.graph.get_tensor_by_name("softmax:0")            print (out_softmax)            out_label = sess.graph.get_tensor_by_name("output:0")            print (out_label)            mnist = input_data.read_data_sets("E:/Program Files/MyEclipseCode"+\            "/PythonCode/MyPythonCode/MyPythonTry/MNIST_data/", one_hot=True)            batch_tx, batch_ty = mnist.test.next_batch(batch_size)            plt.figure("fig1")            plt.imshow(np.reshape(batch_tx, (28,28)),cmap='gray')            img_out_softmax = sess.run(out_softmax, feed_dict={input_x:batch_tx})            print ("img_out_softmax:", img_out_softmax)            prediction_labels = np.argmax(img_out_softmax, axis=1)            print ("prediction_labels:", prediction_labels)            plt.show()recognize("./model/mnist_model_graph.pb")

参考

【1】MindorksOpenSource/AndroidTensorFlowMNISTExample: Android TensorFlow MachineLearning MNIST Example (Building Model with TensorFlow for Android)
https://github.com/MindorksOpenSource/AndroidTensorFlowMNISTExample
【2】模型下载地址
在tensorflow官方教程里https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android/
1、D:\tensorflow\WORKSPACE
（这是As a first step for all build types, clone the TensorFlow repo with:
git clone –recurse-submodules https://github.com/tensorflow/tensorflow.git
Note that –recurse-submodules is necessary to prevent some issues with protobuf compilation.）

2、D:\tensorflow-master\WORKSPACE

“http://download.tensorflow.org/models/stylize_v1.zip“,

“http://download.tensorflow.org/models/mobile_multibox_v1a.zip“,

“http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_android_export.zip“,

“http://download.tensorflow.org/models/speech_commands_v0.01.zip“,