Tutorial_3 - Convolutional network for MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  CNN做MNIST分类\n",
    "refer: http://wiki.jikexueyuan.com/project/tensorflow-zh/tutorials/mnist_pros.html<br/>\n",
    "@author: huangyongye\n",
    "<br/>@date: 2017-02-24\n",
    "<br/><br/>在测试的过程中出现了ResourceExhaustedError ，就是资源耗尽了！！！\n",
    "<br/>之前在keras中用同样的网络和同样的数据集来做这个例子的时候。keras占用了 5647M 的显存（训练过程中设了 validation_split = 0.2， 也就是1.2万张图）。\n",
    "\n",
    "但是我用 tensorflow 自己写的 batch=50 来测试发现呀只有529的占用显存！！！只是在最后做测试的时候因为是对10000多张图片一次性做预测才占用了 8721M 的显存<font color=\"red\">这里的测试集是 1 万张</font>。如果把预测时候的 batch 也设得比较小的话，那么整个网络只需要很小的显存了。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "\n",
    "# 设置按需使用GPU\n",
    "config = tf.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "sess = tf.InteractiveSession(config=config)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.导入数据，用 tensorflow 导入"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
      "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
      "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
      "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
      "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
      "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
      "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
      "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n",
      "(10000, 10)\n",
      "(55000, 10)\n"
     ]
    }
   ],
   "source": [
    "# 用tensorflow 导入数据\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets('MNIST_data', one_hot=True)\n",
    "# 看看咱们样本的数量\n",
    "print mnist.test.labels.shape\n",
    "print mnist.train.labels.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 构建网络"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 权值初始化\n",
    "def weight_variable(shape):\n",
    "    # 用正态分布来初始化权值\n",
    "    initial = tf.truncated_normal(shape, stddev=0.1)\n",
    "    return tf.Variable(initial)\n",
    "\n",
    "def bias_variable(shape):\n",
    "    # 本例中用relu激活函数，所以用一个很小的正偏置较好\n",
    "    initial = tf.constant(0.1, shape=shape)\n",
    "    return tf.Variable(initial)\n",
    "\n",
    "# 定义卷积层\n",
    "def conv2d(x, W):\n",
    "    # 默认 strides[0]=strides[3]=1, strides[1]为x方向步长，strides[2]为y方向步长\n",
    "    return tf.nn.conv2d(x, W, strides=[1,1,1,1], padding='SAME')\n",
    "\n",
    "# pooling 层\n",
    "def max_pool_2x2(x):\n",
    "    return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')\n",
    "\n",
    "\n",
    "X_ = tf.placeholder(tf.float32, [None, 784])\n",
    "y_ = tf.placeholder(tf.float32, [None, 10])\n",
    "\n",
    "# 把X转为卷积所需要的形式\n",
    "X = tf.reshape(X_, [-1, 28, 28, 1])\n",
    "# 第一层卷积：5×5×1卷积核32个 [5，5，1，32]\n",
    "W_conv1 = weight_variable([5,5,1,32])\n",
    "b_conv1 = bias_variable([32])\n",
    "h_conv1 = tf.nn.relu(conv2d(X, W_conv1) + b_conv1)\n",
    "\n",
    "# 第一个pooling 层\n",
    "h_pool1 = max_pool_2x2(h_conv1)\n",
    "\n",
    "# 第二层卷积：5×5×32卷积核64个 [5，5，32，64]\n",
    "W_conv2 = weight_variable([5,5,32,64])\n",
    "b_conv2 = bias_variable([64])\n",
    "h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)\n",
    "\n",
    "# 第二个pooling 层,输出[None, 7, 7, 64] ? \n",
    "h_pool2 = max_pool_2x2(h_conv2)\n",
    "\n",
    "# flatten\n",
    "h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])\n",
    "\n",
    "# fc1\n",
    "W_fc1 = weight_variable([7*7*64, 1024])\n",
    "b_fc1 = bias_variable([1024])\n",
    "h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)\n",
    "\n",
    "# dropout: 输出的维度和h_fc1一样，只是随机部分值被值为零\n",
    "keep_prob = tf.placeholder(tf.float32)\n",
    "h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)\n",
    "\n",
    "# 输出层\n",
    "W_fc2 = weight_variable([1024, 10])\n",
    "b_fc2 = bias_variable([10])\n",
    "y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "看看各层输出 tensor 的维度"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"Placeholder:0\", shape=(?, 784), dtype=float32)\n",
      "Tensor(\"Placeholder_1:0\", shape=(?, 10), dtype=float32)\n",
      "Tensor(\"Relu:0\", shape=(?, 28, 28, 32), dtype=float32)\n",
      "Tensor(\"MaxPool:0\", shape=(?, 14, 14, 32), dtype=float32)\n",
      "Tensor(\"Relu_1:0\", shape=(?, 14, 14, 64), dtype=float32)\n",
      "Tensor(\"MaxPool_1:0\", shape=(?, 7, 7, 64), dtype=float32)\n",
      "Tensor(\"Relu_2:0\", shape=(?, 1024), dtype=float32)\n",
      "Tensor(\"Softmax:0\", shape=(?, 10), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "print(X_)\n",
    "print(y_)\n",
    "print(h_conv1)\n",
    "print(h_pool1)\n",
    "print(h_conv2)\n",
    "print(h_pool2)\n",
    "print(h_fc1)\n",
    "print(y_conv)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.训练和评估"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b> 在测试的时候不使用 mini_batch， 那么测试的时候会占用较多的GPU（4497M），这在 notebook 交互式编程中是不推荐的。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step 0, training accuracy 0.04\n",
      "step 1000, training accuracy 1\n",
      "step 2000, training accuracy 1\n",
      "step 3000, training accuracy 0.98\n",
      "step 4000, training accuracy 0.98\n",
      "step 5000, training accuracy 0.98\n",
      "step 6000, training accuracy 1\n",
      "step 7000, training accuracy 1\n",
      "step 8000, training accuracy 1\n",
      "step 9000, training accuracy 1\n",
      "test accuracy 0.9908\n"
     ]
    }
   ],
   "source": [
    "cross_entropy = -tf.reduce_sum(y_*tf.log(y_conv))\n",
    "train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)\n",
    "\n",
    "correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, \"float\"))\n",
    "sess.run(tf.global_variables_initializer())\n",
    "for i in range(10000):\n",
    "    batch = mnist.train.next_batch(50)\n",
    "    if i%1000 == 0:\n",
    "        train_accuracy = accuracy.eval(feed_dict={\n",
    "            X_:batch[0], y_: batch[1], keep_prob: 1.0})\n",
    "        print \"step %d, training accuracy %g\"%(i, train_accuracy)\n",
    "    train_step.run(feed_dict={X_: batch[0], y_: batch[1], keep_prob: 0.5})\n",
    "\n",
    "print \"test accuracy %g\"%accuracy.eval(feed_dict={\n",
    "    X_: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b> 下面改成了 test 也用 mini_batch 的形式， 显存只用了 529M,所以还是很成功的。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step 0, training acc 0.04\n",
      "step 500, training acc 0.9\n",
      "step 1000, training acc 0.96\n",
      "step 1500, training acc 1\n",
      "step 2000, training acc 0.96\n",
      "step 2500, training acc 1\n",
      "step 3000, training acc 0.98\n",
      "step 3500, training acc 0.96\n",
      "step 4000, training acc 0.98\n",
      "step 4500, training acc 0.98\n",
      "testing step 20, test_acc_sum 19.74\n",
      "testing step 40, test_acc_sum 39.47\n",
      "testing step 60, test_acc_sum 59.14\n",
      "testing step 80, test_acc_sum 78.83\n",
      "testing step 100, test_acc_sum 98.61\n",
      " test_accuracy 0.9861\n"
     ]
    }
   ],
   "source": [
    "# 题外话：在做这个例子的过程中遇到过：资源耗尽的错误，为什么？\n",
    "# -> 因为之前每次做 train_acc  的时候用了全部的 55000 张图，显存爆了.\n",
    "\n",
    "# 1.损失函数：cross_entropy\n",
    "cross_entropy = -tf.reduce_sum(y_ * tf.log(y_conv))\n",
    "# 2.优化函数：AdamOptimizer\n",
    "train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)\n",
    "\n",
    "# 3.预测准确结果统计\n",
    "#　预测值中最大值（１）即分类结果，是否等于原始标签中的（１）的位置。argmax()取最大值所在的下标\n",
    "correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.arg_max(y_, 1))  \n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "\n",
    "\n",
    "# 如果一次性来做测试的话，可能占用的显存会比较多，所以测试的时候也可以设置较小的batch来看准确率\n",
    "test_acc_sum = tf.Variable(0.0)\n",
    "batch_acc = tf.placeholder(tf.float32)\n",
    "new_test_acc_sum = tf.add(test_acc_sum, batch_acc)\n",
    "update = tf.assign(test_acc_sum, new_test_acc_sum)\n",
    "\n",
    "# 定义了变量必须要初始化，或者下面形式\n",
    "sess.run(tf.global_variables_initializer())\n",
    "# 或者某个变量单独初始化 如：\n",
    "# x.initializer.run()\n",
    "\n",
    "# 训练\n",
    "for i in range(5000):\n",
    "    X_batch, y_batch = mnist.train.next_batch(batch_size=50)\n",
    "    if i % 500 == 0:\n",
    "        train_accuracy = accuracy.eval(feed_dict={X_: X_batch, y_: y_batch, keep_prob: 1.0})\n",
    "        print \"step %d, training acc %g\" % (i, train_accuracy)\n",
    "    train_step.run(feed_dict={X_: X_batch, y_: y_batch, keep_prob: 0.5})  \n",
    "\n",
    "# 全部训练完了再做测试，batch_size=100\n",
    "for i in range(100): \n",
    "    X_batch, y_batch = mnist.test.next_batch(batch_size=100)\n",
    "    test_acc = accuracy.eval(feed_dict={X_: X_batch, y_: y_batch, keep_prob: 1.0})\n",
    "    update.eval(feed_dict={batch_acc: test_acc})\n",
    "    if (i+1) % 20 == 0:\n",
    "        print \"testing step %d, test_acc_sum %g\" % (i+1, test_acc_sum.eval())\n",
    "print \" test_accuracy %g\" % (test_acc_sum.eval() / 100.0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. 查看网络中间结果"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b> 在学习 CNN 的过程中，老是看到他们用图片的形式展示了中间层卷积的输出。好吧，这下我必须得自己实现以下看看呀！！！</b>\n",
    "<br/> 关于 python 图片操作主要有 matplotlib 和 PIL 两个库（refer to: http://www.cnblogs.com/yinxiangnan-charles/p/5928689.html）。\n",
    "<br/>我们使用 matplotlib 来完成这个任务。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.1 图像操作基础"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.  0.  0.  0.  0.  0.  1.  0.  0.  0.]\n",
      "img_data shape = (784,)\n"
     ]
    }
   ],
   "source": [
    "# 我们先来看看数据是什么样的\n",
    "img1 = mnist.train.images[1]\n",
    "label1 = mnist.train.labels[1]\n",
    "print label1  \n",
    "print 'img_data shape =', img1.shape  # 我们需要把它转为 28 * 28 的矩阵\n",
    "img1.shape = [28, 28]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(28, 28)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAB4NJREFUeJzt3V+o13cdx/Hf8Ryd6PTUlGnq0tDONhi1FmyWjFbDcHXa\nqmXI1kUsGhHZ2iBi4U0QUWGM2B/6t7GbnS5GtSztj5CMYJsIy6RyM5bMpbk0nM4d/5zj+XUR0U3f\n99Fzjkd/5/V43L7O93x/ok+/F5/z+52udrvdAvJMu9AvALgwxA+hxA+hxA+hxA+hxA+hxA+hxA+h\nxA+heibzZqunrfXjhHCebR15sutsvs6TH0KJH0KJH0KJH0KJH0KJH0KJH0KJH0KJH0KJH0KJH0KJ\nH0KJH0KJH0KJH0JN6vv5OT+O3rmycfvk/b8tr53fc6zcB65ZVu7t4eFy5+LlyQ+hxA+hxA+hxA+h\nxA+hxA+hHPV1gJ6FC8r9rg2bGre7ew+M694D3cvrL3DU17E8+SGU+CGU+CGU+CGU+CGU+CGU+CGU\nc/6LQM+SxeV+zS/2l3t1lr9690fKa9ct2lHuTF2e/BBK/BBK/BBK/BBK/BBK/BBK/BDKOf9F4JWH\n5pb75gWby71/zy2N24xPDZXXfuMrHy33Fae3lzudy5MfQokfQokfQokfQokfQokfQokfQjnnnwTT\nZs4s99VXvDiu7//ylrc1bov+8Ux57YovHRzXvelcnvwQSvwQSvwQSvwQSvwQSvwQylHfJOhavrTc\nP9D783L/+uGryn3JQzsbt5HySpJ58kMo8UMo8UMo8UMo8UMo8UMo8UMo5/yTYPFjfy/3D886We73\nPHVTuS8ffO5cXxJ48kMq8UMo8UMo8UMo8UMo8UMo8UMo5/wToPtNveV++7z6HP7h164o9ys37i33\n4XKF/8+TH0KJH0KJH0KJH0KJH0KJH0KJH0I5558Ax26uP1d/zayn633fO8q9fXD/Ob+mTtB+7zvL\n/dUbZpd7z2C73Od//9lzfk1JPPkhlPghlPghlPghlPghlPghlPghlHP+CXBifu7/oe331Gf1tz36\nu8bt7t7Hymund3WP6TX91xP3zWveVtY/W3HmyJFx3bsT5P6rhXDih1Dih1Dih1Dih1Dih1CO+ibA\n0bfXby0dzYsvLSr3vtZ5fEvvtPo47eBP+8p967sfKffLu6u35db3PtMeKffurvrZdeecfzVuGzYu\nL6/t++wfyr01cqbeO4AnP4QSP4QSP4QSP4QSP4QSP4QSP4Ryzj8BFj47yjn/HfV85fID5T6enyI4\n8/7ryr3v238p998s/vEod6g/XvvxY5c3bt8a+ER57WUv1Of8x99SP7sG7v1O47bjg98tr133vvXl\n3r3t+XLvBJ78EEr8EEr8EEr8EEr8EEr8EEr8EMo5/wSY89Lr5b5n6I3zev/2qmsbt689+qPy2lUz\n6///j4+cLPcbtn+m3Jd+7p+N21sPPVNeO5pLR9n7r7qncfvZmgfLa2ccrP9OO//d/J78EEv8EEr8\nEEr8EEr8EEr8EEr8EMo5/wQY2Vm/J/75k0vK/avLNpf7F9d/vty/d2/zmfXKmeP7Nde33vWFcl+2\n65VyHz50aFz3H4/5O5r/7Nfedkl57etXX1bus3aP6SVdVDz5IZT4IZT4IZT4IZT4IZT4IZT4IZRz\n/klwYOjN5b5uzpFy33n/I6PcYexn+TtPnSr3Gdt2lfvw0Okx33u8unrqf76Hr29+1/2fT58or710\nz9Fyr3+jQGfw5IdQ4odQ4odQ4odQ4odQ4odQjvomwZb1N5X7fU/87bzd+9eD9VtXH7xxTbm3hw5O\n5MuZUCdXv6vc9976g8btjr0fKq8d+dMLY3pNncSTH0KJH0KJH0KJH0KJH0KJH0KJH0I5558E3U//\nsdz799xS7r/s+9WY7/3amVnlPrKg/ojq1sFX673dPsdX9D+jvSV3sP+6ct/4wMOj3GF647Lvgb7y\nytmt7aN8787nyQ+hxA+hxA+hxA+hxA+hxA+hxA+hnPNPhpHmj5ButVqtwz9cWu/ffKPc53fPbtxG\n+1jwdVsGyn3Vro+X++CmheXeu3eocXv5Y+Wlrb39ze/H/4/mc/xWq/4sg7nb/lpeW/+NTQ2e/BBK\n/BBK/BBK/BBK/BBK/BBK/BCqqz2O92Ofq9XT1k7ezaaQk/3Xl3vvl/c1bj9Zsbm8dnrX2H+999k4\n1W4+57+kqz6n3zd8vNw37K8/e//QzcON28jgYHltJ9s68mTX2XydJz+EEj+EEj+EEj+EEj+EEj+E\nctQ3xR1fe0O5H7r9RLnvvvHxcd3/6t9/eszXzttUf+z43IHnxvy9pzJHfUBJ/BBK/BBK/BBK/BBK\n/BBK/BDKOT9MMc75gZL4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4\nIZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4\nIZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IZT4IVRXu92+0K8BuAA8+SGU+CGU+CGU+CGU\n+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CGU+CHU\nvwF95B6z7gsSEwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f99a2a75210>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "# import matplotlib.image as mpimg  # 用于读取图片，这里用不上\n",
    "\n",
    "print img1.shape\n",
    "plt.imshow(img1)\n",
    "plt.axis('off') # 不显示坐标轴\n",
    "plt.show()   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "好吧，是显示了图片，但是结果是热度图像。我们想显示的是灰度图像。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f999c5967d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 我们可以通过设置 cmap 参数来显示灰度图\n",
    "plt.imshow(img1, cmap='gray') # 'hot' 是热度图\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "我们想看 Conv1 层的32个卷积滤波后的结果，显示在同一张图上。 python 中也有 plt.subplot(121) 这样的方法来帮我们解决这个问题。如下：先看两个试试"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAHQAAABGCAYAAAAdKLCBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLVQYWqgwtFBhaKHC0EKF8R+9ru9etu56YgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f999c540110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplot(4,8,1)\n",
    "plt.imshow(img1, cmap='gray')\n",
    "plt.axis('off')\n",
    "plt.subplot(4,8,2)\n",
    "plt.imshow(img1, cmap='gray')\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.2 显示网络中间结果"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "好了，有了前面的图像操作基础，我们就该试试吧！！！"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 28, 28, 32)\n",
      "<type 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "# 首先应该把 img1 转为正确的shape (None, 784)\n",
    "X_img = img1.reshape([-1, 784])\n",
    "y_img = mnist.train.labels[1].reshape([-1, 10])\n",
    "# 我们要看 Conv1 的结果，即 h_conv1\n",
    "result = h_conv1.eval(feed_dict={X_: X_img, y_: y_img, keep_prob: 1.0})\n",
    "print result.shape\n",
    "print type(result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "好的，我们成功的计算得到了 h_conv1，那么赶紧 imshow() 看看吧！！！"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAADwCAYAAAAQNZ8pAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nloaf3J1SAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f999c482e90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for _ in xrange(32):\n",
    "    show_img = result[:,:,:,_]\n",
    "    show_img.shape = [28, 28]\n",
    "    plt.subplot(4, 8, _ + 1)\n",
    "    plt.imshow(show_img, cmap='gray')\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "哈哈，成功啦！从上面的结果中，我们可以看到不同的滤波器（卷积核）学习到了不同的特征。比如第一行中，第一个滤波器学习到了边缘信息，第5个卷积核，则学习到了骨干的信息。感觉好有趣，不由自主的想对另一个数字看看。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0.  0.  0.  0.  0.  0.  0.  0.  1.  0.]\n",
      " [ 0.  0.  0.  0.  0.  0.  1.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  0.  1.  0.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  0.  0.  1.  0.  0.  0.]\n",
      " [ 0.  0.  0.  1.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  0.  1.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  0.  0.  0.  1.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]]\n"
     ]
    }
   ],
   "source": [
    "# 输出前10个看看，我选择数字 8 来试试\n",
    "print mnist.train.labels[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAADwCAYAAAAQNZ8pAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngTt0DofDSRG4Q+dwOJwU4X+5OO9fyObn9wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f999c4b0210>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 首先应该把 img1 转为正确的shape (None, 784)\n",
    "X_img = mnist.train.images[0].reshape([-1, 784])\n",
    "result = h_conv1.eval(feed_dict={X_: X_img, keep_prob: 1.0})\n",
    "\n",
    "for _ in xrange(32):\n",
    "    show_img = result[:,:,:,_]\n",
    "    show_img.shape = [28, 28]\n",
    "    plt.subplot(4, 8, _ + 1)\n",
    "    plt.imshow(show_img, cmap='gray')\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "第一个核还是主要学习到了边缘特征，第五个核还是学到了骨干特征（当然在某种程度上）。好吧，本次就到这啦！"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}