TensorFlow 之手写数字识别MNIST

官方文档：
MNIST For ML Beginners - https://www.tensorflow.org/get_started/mnist/beginners
Deep MNIST for Experts - https://www.tensorflow.org/get_started/mnist/pros

版本：
TensorFlow 1.2.0 + Flask 0.12 + Gunicorn 19.6

相关文章：
TensorFlow 之入门体验
TensorFlow 之手写数字识别MNIST
TensorFlow 之物体检测
TensorFlow 之构建人物识别系统

MNIST相当于机器学习界的Hello World。

这里在页面通过 Canvas 画一个数字，然后传给TensorFlow识别，分别给出Softmax回归模型、多层卷积网络的识别结果。

（1）文件结构

│ main.py
│ requirements.txt
│ runtime.txt
├─mnist
│ │ convolutional.py
│ │ model.py
│ │ regression.py
│ │ __init__.py
│ └─data
│          convolutional.ckpt.data-00000-of-00001
│          convolutional.ckpt.index
│          regression.ckpt.data-00000-of-00001
│          regression.ckpt.index
├─src
│ └─js
│          main.js
├─static
│ ├─css
│ │      bootstrap.min.css
│ └─js
│          jquery.min.js
│          main.js
└─templates
        index.html

（2）训练数据

下载以下文件放入/tmp/data/，不用解压，训练代码会自动解压。

引用

http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz

执行命令训练数据（Softmax回归模型、多层卷积网络）

# python regression.py
# python convolutional.py

执行完成后在 mnist/data/ 里会生成以下几个文件，重新训练前需要把这几个文件先删掉。

引用

convolutional.ckpt.data-00000-of-00001
convolutional.ckpt.index
regression.ckpt.data-00000-of-00001
regression.ckpt.index

（3）启动Web服务测试

# cd /usr/local/tensorflow2/tensorflow-models/tf-mnist
# pip install -r requirements.txt
# gunicorn main:app --log-file=- --bind=localhost:8000

浏览器中访问：http://localhost:8000

*** 运行的TensorFlow版本、数据训练的模型、还有这里Canvas的转换都对识别率有一定的影响~！

（4）源代码

Web部分比较简单，页面上放置一个Canvas，鼠标抬起时将Canvas的图像通过Ajax传给后台API，然后显示API结果。

引用

src/js/main.js -> static/js/main.js
templates/index.html

main.py

import numpy as np
import tensorflow as tf
from flask import Flask, jsonify, render_template, request

from mnist import model

x = tf.placeholder("float", [None, 784])
sess = tf.Session()

# restore trained data
with tf.variable_scope("regression"):
    y1, variables = model.regression(x)
saver = tf.train.Saver(variables)
saver.restore(sess, "mnist/data/regression.ckpt")

with tf.variable_scope("convolutional"):
    keep_prob = tf.placeholder("float")
    y2, variables = model.convolutional(x, keep_prob)
saver = tf.train.Saver(variables)
saver.restore(sess, "mnist/data/convolutional.ckpt")

def regression(input):
    return sess.run(y1, feed_dict={x: input}).flatten().tolist()

def convolutional(input):
    return sess.run(y2, feed_dict={x: input, keep_prob: 1.0}).flatten().tolist()

# webapp
app = Flask(__name__)

@app.route('/api/mnist', methods=['POST'])
def mnist():
    input = ((255 - np.array(request.json, dtype=np.uint8)) / 255.0).reshape(1, 784)
    output1 = regression(input)
    output2 = convolutional(input)
    print(output1)
    print(output2)
    return jsonify(results=[output1, output2])

@app.route('/')
def main():
    return render_template('index.html')

if __name__ == '__main__':
    app.run()

mnist/model.py

import tensorflow as tf


# Softmax Regression Model
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]


# Multilayer Convolutional Network
def convolutional(x, keep_prob):
    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_variable(shape):
        initial = tf.truncated_normal(shape, stddev=0.1)
        return tf.Variable(initial)

    def bias_variable(shape):
        initial = tf.constant(0.1, shape=shape)
        return tf.Variable(initial)

    # First Convolutional Layer
    x_image = tf.reshape(x, [-1, 28, 28, 1])
    W_conv1 = weight_variable([5, 5, 1, 32])
    b_conv1 = bias_variable([32])
    h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
    h_pool1 = max_pool_2x2(h_conv1)
    # Second Convolutional Layer
    W_conv2 = weight_variable([5, 5, 32, 64])
    b_conv2 = bias_variable([64])
    h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
    h_pool2 = max_pool_2x2(h_conv2)
    # Densely Connected Layer
    W_fc1 = weight_variable([7 * 7 * 64, 1024])
    b_fc1 = bias_variable([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)
    # Readout Layer
    W_fc2 = weight_variable([1024, 10])
    b_fc2 = bias_variable([10])
    y = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)
    return y, [W_conv1, b_conv1, W_conv2, b_conv2, W_fc1, b_fc1, W_fc2, b_fc2]

mnist/convolutional.py

import os
import model
import tensorflow as tf

from tensorflow.examples.tutorials.mnist import input_data
data = input_data.read_data_sets("/tmp/data/", one_hot=True)

# model
with tf.variable_scope("convolutional"):
    x = tf.placeholder(tf.float32, [None, 784])
    keep_prob = tf.placeholder(tf.float32)
    y, variables = model.convolutional(x, keep_prob)

# train
y_ = tf.placeholder(tf.float32, [None, 10])
cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
train_step = tf.train.AdamOptimizer(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:
    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("step %d, training 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)

mnist/regression.py

import os
import model
import tensorflow as tf

from tensorflow.examples.tutorials.mnist import input_data
data = input_data.read_data_sets("/tmp/data/", one_hot=True)

# model
with tf.variable_scope("regression"):
    x = tf.placeholder(tf.float32, [None, 784])
    y, variables = model.regression(x)

# train
y_ = tf.placeholder("float", [None, 10])
cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
train_step = tf.train.GradientDescentOptimizer(0.01).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:
    sess.run(tf.global_variables_initializer())
    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.dirname(__file__), 'data', 'regression.ckpt'),
        write_meta_graph=False, write_state=False)
    print("Saved:", path)

参考：
http://memo.sugyan.com/entry/20151124/1448292129

TensorFlow 之 手写数字识别MNIST

猜你喜欢

TensorFlow 之手写数字识别MNIST