tensorflow在android上的demo加入mnist手写数字识别

1.下载tensorflow源代码

1
git clone https://github.com/tensorflow/tensorflow.git

假设tensorflow的根目录为 TENSORFLOW_ROOT

2.训练模型

训练的模型采用是最简单的模型，训练脚本如下：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
import tensorflow as tf
import os.path
# Import data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("data/", one_hot=True)
g = tf.Graph()
with g.as_default():
    # Create the model
    x = tf.placeholder(tf.float32, [None, 784])
    W = tf.Variable(tf.zeros([784, 10]), name="vaiable_W")
    b = tf.Variable(tf.zeros([10]), name="variable_b")
    y = tf.nn.softmax(tf.matmul(x, W) + b)
    
    # Define loss and optimizer
    y_ = tf.placeholder(tf.float32, [None, 10])
    cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
    train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)
    
    sess = tf.Session()
    
    # Train
    init = tf.global_variables_initializer()
    sess.run(init)
    
    for i in range(1000):
        batch_xs, batch_ys = mnist.train.next_batch(100)
        train_step.run({x: batch_xs, y_: batch_ys}, sess)
    
    # Test trained model
    correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    
    print(accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}, sess))
    
# Store variable
_W = W.eval(sess)
_b = b.eval(sess)
sess.close()
# Create new graph for exporting
g_2 = tf.Graph()
with g_2.as_default():
    # Reconstruct graph
    x_2 = tf.placeholder(tf.float32, [None, 784], name="input")
    W_2 = tf.constant(_W, name="constant_W")
    b_2 = tf.constant(_b, name="constant_b")
    y_2 = tf.nn.softmax(tf.matmul(x_2, W_2) + b_2, name="output")
    
    sess_2 = tf.Session()
    
    init_2 = tf.global_variables_initializer()
    sess_2.run(init_2)
    
    graph_def = g_2.as_graph_def()
    
    tf.train.write_graph(graph_def, './model/beginner-export',
			                         'beginner-graph.pb', as_text=False)

训练完成之后，得到了模型文件 beginner-graph.pb

3.替换模型

android例子的位置在TENSORFLOW_ROOT/tensorflow/examples/android目录下面
拷贝上一步生成的模型文件到 assets目录下面
新建一个标签文件 mnist_labels.txt 放入 assets目录下面
文件内容如下：

1
2
3
4
5
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9

4.修改源码

首先修改 org.tensorflow.demo 包下面的 ClassifierActivity.java 文件
修改前:

1
2
3
4
5
6
7
8
9
private static final int NUM_CLASSES = 1001;
private static final int INPUT_SIZE = 224;
private static final int IMAGE_MEAN = 117;
private static final float IMAGE_STD = 1;
private static final String INPUT_NAME = "input:0";
private static final String OUTPUT_NAME = "output:0";
private static final String MODEL_FILE = "file:///android_asset/tensorflow_inception_graph.pb";
private static final String LABEL_FILE = "file:///android_asset/imagenet_comp_graph_label_strings.txt";

修改后:

1
2
3
4
5
6
7
8
9
private static final int NUM_CLASSES = 10; // 10类[0-9]
private static final int INPUT_SIZE = 28;  // 图像尺寸28*28
private static final int IMAGE_MEAN = 117;
private static final float IMAGE_STD = 1;
private static final String INPUT_NAME = "input";	// 输入节点名称
private static final String OUTPUT_NAME = "output"; // 输出节点名称
private static final String MODEL_FILE = "file:///android_asset/beginner-graph.pb";
private static final String LABEL_FILE = "file:///android_asset/mnist_labels.txt";

接着修改 TensorFlowImageClassifier.java 文件中的
函数1:

1
public int initializeTensorFlow(){}

修改前:

1
2
3
4
5
// Pre-allocate buffers.
outputNames = new String[] {outputName};
intValues = new int[inputSize * inputSize];
floatValues = new float[inputSize * inputSize * 3];
outputs = new float[numClasses];

修改后:

1
2
3
4
5
// Pre-allocate buffers.
outputNames = new String[] {outputName};
intValues = new int[inputSize * inputSize];
floatValues = new float[inputSize * inputSize];
outputs = new float[numClasses];

函数2:

1
public List<Recognition> recognizeImage(final Bitmap bitmap) {}

修改前:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Trace.beginSection("preprocessBitmap");
// Preprocess the image data from 0-255 int to normalized float based
// on the provided parameters.
bitmap.getPixels(intValues, 0, bitmap.getWidth(), 0, 0, bitmap.getWidth(), bitmap.getHeight());
for (int i = 0; i < intValues.length; ++i) {
	final int val = intValues[i];
	floatValues[i * 3 + 0] = (((val >> 16) & 0xFF) - imageMean) / imageStd;
	floatValues[i * 3 + 1] = (((val >> 8) & 0xFF) - imageMean) / imageStd;
	floatValues[i * 3 + 2] = ((val & 0xFF) - imageMean) / imageStd;
}
Trace.endSection();
// Copy the input data into TensorFlow.
Trace.beginSection("fillNodeFloat");
inferenceInterface.fillNodeFloat(inputName, new int[] {1, inputSize, inputSize, 3}, floatValues);
Trace.endSection();

修改后:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Trace.beginSection("preprocessBitmap");
// Preprocess the image data from 0-255 int to normalized float based
// on the provided parameters.
bitmap.getPixels(intValues, 0, bitmap.getWidth(), 0, 0, bitmap.getWidth(), bitmap.getHeight());
for (int i = 0; i < intValues.length; ++i) {
	final int val = intValues[i];
	int R = (val >> 16) & 0xFF;
	int G = (val >> 8) & 0xFF;
	int B = val & 0xFF;
	float Y = (float)(1-(0.3*R + 0.59*G + 0.11*B)/255);
	floatValues[i] = Y>0.2?Y:0;
}
Trace.endSection();
// Copy the input data into TensorFlow.
Trace.beginSection("fillNodeFloat");
inferenceInterface.fillNodeFloat(inputName, new int[] {1, inputSize*inputSize}, floatValues);
Trace.endSection();

5.编译运行

编译apk

1
bazel build //tensorflow/examples/android:tensorflow_demo

安装apk

1
adb install -r bazel-bin/tensorflow/examples/android/tensorflow_demo.apk

运行TF Classify程序