robbbit
diff --git a/‎.gitignore‎
Lines changed: 4 additions & 0 deletions b/‎.gitignore‎
Lines changed: 4 additions & 0 deletions
diff --git a/‎Deep_Learning_with_TensorFlow/1.4.0/Chapter07/.ipynb_checkpoints/1. TFRecord样例程序-checkpoint.ipynb‎
Lines changed: 32 additions & 15 deletions b/‎Deep_Learning_with_TensorFlow/1.4.0/Chapter07/.ipynb_checkpoints/1. TFRecord样例程序-checkpoint.ipynb‎
Lines changed: 32 additions & 15 deletions
diff --git a/‎Deep_Learning_with_TensorFlow/1.4.0/Chapter07/.ipynb_checkpoints/7. 数据集基本使用方法-checkpoint.ipynb‎
Lines changed: 224 additions & 0 deletions b/‎Deep_Learning_with_TensorFlow/1.4.0/Chapter07/.ipynb_checkpoints/7. 数据集基本使用方法-checkpoint.ipynb‎
Lines changed: 224 additions & 0 deletions
@@ -7,6 +7,10 @@ Deep_Learning_with_TensorFlow/datasets/flower_processed_data.npy
 Deep_Learning_with_TensorFlow/1.4.0/Chapter05/5. MNIST\346\234\200\344\275\263\345\256\236\350\267\265/MNIST_model/*
 Deep_Learning_with_TensorFlow/1.4.0/Chapter07/output.tfrecords
 Deep_Learning_with_TensorFlow/1.4.0/Chapter07/data.tfrecords*
+Deep_Learning_with_TensorFlow/1.4.0/Chapter07/output_test.tfrecords
+Deep_Learning_with_TensorFlow/1.4.0/Chapter07/test1.txt
+Deep_Learning_with_TensorFlow/1.4.0/Chapter07/test2.txt
+Deep_Learning_with_TensorFlow/1.4.0/Chapter08/sin.png
 Deep_Learning_with_TensorFlow/1.4.0/Chapter10/log/*
 Deep_Learning_with_TensorFlow/1.4.0/Chapter11/log/*
 .DS_Store
 
@@ -33,7 +33,8 @@
       "Extracting ../../datasets/MNIST_data/train-labels-idx1-ubyte.gz\n",
       "Extracting ../../datasets/MNIST_data/t10k-images-idx3-ubyte.gz\n",
       "Extracting ../../datasets/MNIST_data/t10k-labels-idx1-ubyte.gz\n",
-      "TFRecord文件已保存。\n"
+      "TFRecord训练文件已保存。\n",
+      "TFRecord测试文件已保存。\n"
      ]
     }
    ],
@@ -45,27 +46,43 @@
     "def _bytes_feature(value):\n",
     "    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))\n",
     "\n",
-    "# 读取mnist数据。\n",
+    "# 将数据转化为tf.train.Example格式。\n",
+    "def _make_example(pixels, label, image):\n",
+    "    image_raw = image.tostring()\n",
+    "    example = tf.train.Example(features=tf.train.Features(feature={\n",
+    "        'pixels': _int64_feature(pixels),\n",
+    "        'label': _int64_feature(np.argmax(label)),\n",
+    "        'image_raw': _bytes_feature(image_raw)\n",
+    "    }))\n",
+    "    return example\n",
+    "\n",
+    "# 读取mnist训练数据。\n",
     "mnist = input_data.read_data_sets(\"../../datasets/MNIST_data\",dtype=tf.uint8, one_hot=True)\n",
     "images = mnist.train.images\n",
     "labels = mnist.train.labels\n",
     "pixels = images.shape[1]\n",
     "num_examples = mnist.train.num_examples\n",
     "\n",
-    "# 输出TFRecord文件的地址。\n",
-    "filename = \"output.tfrecords\"\n",
-    "writer = tf.python_io.TFRecordWriter(filename)\n",
-    "for index in range(num_examples):\n",
-    "    image_raw = images[index].tostring()\n",
+    "# 输出包含训练数据的TFRecord文件。\n",
+    "with tf.python_io.TFRecordWriter(\"output.tfrecords\") as writer:\n",
+    "    for index in range(num_examples):\n",
+    "        example = _make_example(pixels, labels[index], images[index])\n",
+    "        writer.write(example.SerializeToString())\n",
+    "print(\"TFRecord训练文件已保存。\")\n",
     "\n",
-    "    example = tf.train.Example(features=tf.train.Features(feature={\n",
-    "        'pixels': _int64_feature(pixels),\n",
-    "        'label': _int64_feature(np.argmax(labels[index])),\n",
-    "        'image_raw': _bytes_feature(image_raw)\n",
-    "    }))\n",
-    "    writer.write(example.SerializeToString())\n",
-    "writer.close()\n",
-    "print \"TFRecord文件已保存。\""
+    "# 读取mnist测试数据。\n",
+    "images_test = mnist.test.images\n",
+    "labels_test = mnist.test.labels\n",
+    "pixels_test = images_test.shape[1]\n",
+    "num_examples_test = mnist.test.num_examples\n",
+    "\n",
+    "# 输出包含测试数据的TFRecord文件。\n",
+    "with tf.python_io.TFRecordWriter(\"output_test.tfrecords\") as writer:\n",
+    "    for index in range(num_examples_test):\n",
+    "        example = _make_example(\n",
+    "            pixels_test, labels_test[index], images_test[index])\n",
+    "        writer.write(example.SerializeToString())\n",
+    "print(\"TFRecord测试文件已保存。\")"
    ]
   },
   {
 
@@ -0,0 +1,224 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": [
+    "import tempfile\n",
+    "import tensorflow as tf"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 1. 从数组创建数据集。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "1\n",
+      "4\n",
+      "9\n",
+      "25\n",
+      "64\n"
+     ]
+    }
+   ],
+   "source": [
+    "input_data = [1, 2, 3, 5, 8]\n",
+    "dataset = tf.data.Dataset.from_tensor_slices(input_data)\n",
+    "\n",
+    "# 定义迭代器。\n",
+    "iterator = dataset.make_one_shot_iterator()\n",
+    "\n",
+    "# get_next() 返回代表一个输入数据的张量。\n",
+    "x = iterator.get_next()\n",
+    "y = x * x\n",
+    "\n",
+    "with tf.Session() as sess:\n",
+    "    for i in range(len(input_data)):\n",
+    "        print(sess.run(y))\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 2. 读取文本文件里的数据。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "File1, line1.\n",
+      "File1, line2.\n",
+      "File2, line1.\n",
+      "File2, line2.\n"
+     ]
+    }
+   ],
+   "source": [
+    "# 创建文本文件作为本例的输入。\n",
+    "with open(\"./test1.txt\", \"w\") as file:\n",
+    "    file.write(\"File1, line1.\\n\") \n",
+    "    file.write(\"File1, line2.\\n\")\n",
+    "with open(\"./test2.txt\", \"w\") as file:\n",
+    "    file.write(\"File2, line1.\\n\") \n",
+    "    file.write(\"File2, line2.\\n\")\n",
+    "\n",
+    "# 从文本文件创建数据集。这里可以提供多个文件。\n",
+    "input_files = [\"./test1.txt\", \"./test2.txt\"]\n",
+    "dataset = tf.data.TextLineDataset(input_files)\n",
+    "\n",
+    "# 定义迭代器。\n",
+    "iterator = dataset.make_one_shot_iterator()\n",
+    "\n",
+    "# 这里get_next()返回一个字符串类型的张量，代表文件中的一行。\n",
+    "x = iterator.get_next()  \n",
+    "with tf.Session() as sess:\n",
+    "    for i in range(4):\n",
+    "        print(sess.run(x))\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 3. 解析TFRecord文件里的数据。读取文件为本章第一节创建的文件。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "7\n",
+      "3\n",
+      "4\n",
+      "6\n",
+      "1\n",
+      "8\n",
+      "1\n",
+      "0\n",
+      "9\n",
+      "8\n"
+     ]
+    }
+   ],
+   "source": [
+    "# 解析一个TFRecord的方法。\n",
+    "def parser(record):\n",
+    "    features = tf.parse_single_example(\n",
+    "        record,\n",
+    "        features={\n",
+    "            'image_raw':tf.FixedLenFeature([],tf.string),\n",
+    "            'pixels':tf.FixedLenFeature([],tf.int64),\n",
+    "            'label':tf.FixedLenFeature([],tf.int64)\n",
+    "        })\n",
+    "    decoded_images = tf.decode_raw(features['image_raw'],tf.uint8)\n",
+    "    retyped_images = tf.cast(decoded_images, tf.float32)\n",
+    "    images = tf.reshape(retyped_images, [784])\n",
+    "    labels = tf.cast(features['label'],tf.int32)\n",
+    "    #pixels = tf.cast(features['pixels'],tf.int32)\n",
+    "    return images, labels\n",
+    "\n",
+    "# 从TFRecord文件创建数据集。这里可以提供多个文件。\n",
+    "input_files = [\"output.tfrecords\"]\n",
+    "dataset = tf.data.TFRecordDataset(input_files)\n",
+    "\n",
+    "# map()函数表示对数据集中的每一条数据进行调用解析方法。\n",
+    "dataset = dataset.map(parser)\n",
+    "\n",
+    "# 定义遍历数据集的迭代器。\n",
+    "iterator = dataset.make_one_shot_iterator()\n",
+    "\n",
+    "# 读取数据，可用于进一步计算\n",
+    "image, label = iterator.get_next()\n",
+    "\n",
+    "with tf.Session() as sess:\n",
+    "    for i in range(10):\n",
+    "        x, y = sess.run([image, label]) \n",
+    "        print(y)\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 4. 使用initializable_iterator来动态初始化数据集。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": [
+    "# 从TFRecord文件创建数据集，具体文件路径是一个placeholder，稍后再提供具体路径。\n",
+    "input_files = tf.placeholder(tf.string)\n",
+    "dataset = tf.data.TFRecordDataset(input_files)\n",
+    "dataset = dataset.map(parser)\n",
+    "\n",
+    "# 定义遍历dataset的initializable_iterator。\n",
+    "iterator = dataset.make_initializable_iterator()\n",
+    "image, label = iterator.get_next()\n",
+    "\n",
+    "with tf.Session() as sess:\n",
+    "    # 首先初始化iterator，并给出input_files的值。\n",
+    "    sess.run(iterator.initializer,\n",
+    "             feed_dict={input_files: [\"output.tfrecords\"]})\n",
+    "    # 遍历所有数据一个epoch。当遍历结束时，程序会抛出OutOfRangeError。\n",
+    "    while True:\n",
+    "        try:\n",
+    "            x, y = sess.run([image, label])\n",
+    "        except tf.errors.OutOfRangeError:\n",
+    "            break \n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 2",
+   "language": "python",
+   "name": "python2"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 2
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython2",
+   "version": "2.7.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 1
+}