codes/super_resolution/super_resolution.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 이미지 Super-resolution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting data/MNIST/train-images-idx3-ubyte.gz\n",
      "Extracting data/MNIST/train-labels-idx1-ubyte.gz\n",
      "Extracting data/MNIST/t10k-images-idx3-ubyte.gz\n",
      "Extracting data/MNIST/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import scipy.misc as misc\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import tensorflow.contrib.slim as slim\n",
    "\n",
    "# TensorFlow 내장 MNIST 데이터셋 모듈\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets(\"data/MNIST\", one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def batch_norm_params(is_training):\n",
    "    return {\n",
    "        \"decay\": 0.9,\n",
    "        \"epsilon\": 1e-5,\n",
    "        \"scale\": True,\n",
    "        \"updates_collections\": None,\n",
    "        \"is_training\": is_training\n",
    "    }\n",
    "\n",
    "\n",
    "def arg_scope(is_training):\n",
    "    with slim.arg_scope([slim.conv2d],\n",
    "        weights_initializer=slim.variance_scaling_initializer(),\n",
    "        activation_fn=tf.nn.relu,\n",
    "        normalizer_fn=slim.batch_norm,\n",
    "        normalizer_params=batch_norm_params(is_training),\n",
    "        stride=1, padding=\"SAME\"):\n",
    "        with slim.arg_scope([slim.batch_norm],\n",
    "                            **batch_norm_params(is_training)) as arg_scp:\n",
    "            return arg_scp\n",
    "\n",
    "\n",
    "def srnet(inputs, scope=\"srnet\"):\n",
    "    net = inputs\n",
    "    with tf.variable_scope(scope) as scp:\n",
    "        end_pts_collection = scp.name+\"end_pts\"\n",
    "        with slim.arg_scope([slim.conv2d],\n",
    "                            outputs_collections=end_pts_collection):\n",
    "            net = slim.conv2d(net, 64, [3, 3],\n",
    "                              normalizer_fn=None,\n",
    "                              normalizer_params=None,\n",
    "                              scope=\"conv1\")\n",
    "\n",
    "            net = slim.conv2d(net, 64, [3, 3], scope=\"conv2\")\n",
    "            net = slim.conv2d(net, 64, [3, 3], scope=\"conv3\")\n",
    "            \n",
    "            net = slim.conv2d(net, 64, [3, 3], scope=\"conv4\")\n",
    "            net = slim.conv2d(net, 64, [3, 3], scope=\"conv5\")\n",
    "            \n",
    "            net = slim.conv2d(net, 64, [3, 3], scope=\"conv6\")\n",
    "            net = slim.conv2d(net, 1, [3, 3],\n",
    "                              activation_fn=tf.nn.tanh,\n",
    "                              normalizer_fn=None,\n",
    "                              normalizer_params=None,\n",
    "                              scope=\"logit\")\n",
    "\n",
    "            end_pts = slim.utils.convert_collection_to_dict(end_pts_collection)\n",
    "\n",
    "    return net, end_pts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "inputs = tf.placeholder(tf.float32, [None, 28, 28, 1])\n",
    "labels = tf.placeholder(tf.float32, [None, 28, 28, 1])\n",
    "is_training = tf.placeholder(tf.bool)\n",
    "\n",
    "with slim.arg_scope(arg_scope(is_training)):\n",
    "     logits, model = srnet(inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500 0.0142021\n",
      "1000 0.0120737\n",
      "1500 0.0105551\n",
      "2000 0.00944535\n",
      "2500 0.00964404\n",
      "3000 0.00911137\n"
     ]
    }
   ],
   "source": [
    "learning_rate = 0.01\n",
    "\n",
    "loss_op = tf.losses.mean_squared_error(labels, logits)\n",
    "opt = tf.train.AdamOptimizer(learning_rate).minimize(loss_op)\n",
    "\n",
    "sess_config = tf.ConfigProto(gpu_options=tf.GPUOptions(allow_growth=True))\n",
    "sess = tf.Session(config=sess_config)\n",
    "sess.run(tf.global_variables_initializer())\n",
    "\n",
    "for step in range(3000):\n",
    "    high_res, _ = mnist.train.next_batch(128)\n",
    "    high_res    = high_res.reshape((-1, 28, 28))\n",
    "    \n",
    "    low_res = np.empty((high_res.shape[0], 28, 28))\n",
    "    for i, image in enumerate(high_res):\n",
    "        low        = misc.imresize(image, (7, 7)) / 255.0\n",
    "        low_res[i] = misc.imresize(low, (28, 28)) / 255.0\n",
    "        \n",
    "    high_res = high_res.reshape((-1, 28, 28, 1))\n",
    "    low_res  = low_res.reshape((-1, 28, 28, 1))\n",
    "        \n",
    "    _, loss = sess.run([opt, loss_op], \n",
    "        feed_dict={inputs: low_res, labels: high_res, is_training: True})\n",
    "    \n",
    "    if (step+1) % 500 == 0:\n",
    "        print(step+1, loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 테스트"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_high = mnist.test.images[:5].reshape(-1, 28, 28)\n",
    "test_low  = np.empty((test_high.shape[0], 28, 28))\n",
    "for i, image in enumerate(test_high):\n",
    "    low         = misc.imresize(image, (7, 7)) / 255.0\n",
    "    test_low[i] = misc.imresize(low, (28, 28)) / 255.0\n",
    "    \n",
    "test_high  = test_high.reshape((-1, 28, 28, 1))\n",
    "test_low   = test_low.reshape((-1, 28, 28, 1))    \n",
    "test_super = sess.run(logits, feed_dict={inputs:test_low, is_training:False})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAACRCAYAAADTnUPWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAGWxJREFUeJztnXv0VlWZx7+PgoCC3FG5iIpUVAsxh7IRjZUzA3lJ12RM\nTWgXRWnKbGFeanQwM3F0Rgsns8lpTK3QVUtHjUm7KKOiKbjwQiQCGnciLgrITXjmj71/x+c8vO/m\nPS/v7/d7f4fvZy0We7/Puexz9tnPb5/vefbeoqoghBDS8TmgvQtACCGkMdChE0JISaBDJ4SQkkCH\nTgghJYEOnRBCSgIdOiGElAQ69FZCRMaKyPJ92P8bInJHI8vUTIjIfBEZW+O2r4vI37RykQohIp8T\nkSf3Yf/bReTqRpaJtA778vyJyJEisllEDmx0uSrRFA69GRtsW1LJ+avq9ap6QXuVaV+oVJ/eAarq\n+1T18TYvXDtQyfmr6mRV/VZ7lalWRGSMiMwWkTdEZL2IPCUio9u7XM2Kf/ZVdamqdlfVXW1x/k5t\ncZJmRkQ6qerb7V0O0jqwfutHRA4F8DCALwK4D8BBAE4GsL2NyyEARFV3J7ZhPaNJeugpRGSSiCyK\nvYMHRWRg/P2bInJrTHcWkS0iclPMdxORbSLSp8LxxorIchG5QkRWA/jv+PsZIjJPRDbGHslIs88V\nIrJCRDaJyCsicmr8vYuIfEdEVsZ/3xGRLlWuQ0XkWJO/U0SuE5FDAPwvgIHx1WyziAwUkWtE5B6z\n/cejTLFRRB4XkRHG9rqIfE1EXow9qXtFpOs+3fhWxvZkYn39WEQ2iMgCEbm8glw1qpbri73hp0Tk\nFhFZB+Ca+PsX4rE3iMgjIjI0/i5x2z+LyJsi8pKIvD/aeorIXSKyVkT+JCJXicgebUZEjor128n8\n9riIXBDr6XYAH451uzHa7xSR68z2FZ/zaFMRmSwir8b6/150cq3NuwBAVX+mqrtUdauqPqqqL8Zy\n+Wc0dx/iPZgmIs/Ge/s/tk2KyImxrW0UkRfESHBx32+LyFMA3gJwjC9cfIauEJEXAWwRkU6x7fwi\n1tlrIvIVs/0HRWROLMsaEbnZ2Kq2L3dOX2/Z27WI3A3gSAAPxbq+vMI9GRjrd32s70nmWNeIyH3x\nmdsUy/NXtVYW0OQOXUQ+CmAagAkAjgDwJwAzonkWgLExPRrAagCnxPyHAbyiquurHPpwAH0ADAVw\noYgcD+BHAC4C0BfADwA8KMFhvxvAlwGMVtUeAMYBeD0e558BnAhgFIDjAHwQwFVFrlFVtwD4GICV\n8dWsu6qudPfhXQB+BuCrAPoDmInw0BxkNpsAYDyAowGMBPC5IuVoZ6YCOAqh0f4tgIkVtilyfR8C\nsATAYQC+LSJnAfgGgL9HuH9PINxPAPg7hOfmXQB6xvOsi7Zb42/HAPgIgPMAfL7IhanqAgCTATwd\n67aX32Yvz3kLZyA85yPjduOKlKNOFgLYFf/YfkxEetdxjPMAfAHhut4GMB0ARGQQgF8CuA6hLX4N\nwC9EpL/Z91wAFwLogXBPKvFpAKcD6AVgN4CHALwAYBCAUwF8VURa7tV3AXxXVQ8FMAzhraPW9rVX\nVPVcAEsBnBnr+sYKm80AsBzAQADnALg+1n8LH4/b9ALwIID/KFKGpnboAD4D4Eeq+ryqbgfwdYSe\nzlEAngYwXET6IjTI/wIwSES6IzS+WYnj7gYwVVW3q+pWhIfmB6r6+9gT+THCa+WJAHYB6ALgvSLS\nWVVfV9XFpnzXquqfVXUtgG8iPISN5h8A/FJVf62qOwH8G4BuAP7abDNdVVfGP2IPIfyRaU8eiL2d\njbFXelti2wkArlfVDaq6HLHRO4pc30pVvVVV3471OxnANFVdEF/Lr0fo8Q8FsBPBYbwH4bV+gaqu\nkvAR61MAvq6qm1T1dQD/jtap39Rz3sINqrpRVZcCeAxtUL+q+iaAMQAUwA8BrI29y8MKHOZuVX05\ndlyuBjAh3tuJAGaq6kxV3a2qvwYwB8BpZt87VXV+rMedVY4/XVWXxXoeDaC/ql6rqjtUdUks96fi\ntjsBHCsi/VR1s6o+E3+vpX3tMyIyBMBJAK5Q1W2qOg/AHQh/9Fp4Mt6TXQDuRugo1kyzO/SBMH+Z\nVXUzQu9pUKzAOQjO+xQEBz4b4YbtzaGvVdVtJj8UwKXOAQ0BMFBVFyH85b4GwJ9FZIZ5Hc6VL6YH\novH4+7AbwDKEXkgLq036LQDdW6EcRThbVXu1/APwT4ltByJcTwvLKmxT5Pr8/kMBfNfU7XoAgvAc\n/Q6hF/Q9hPr9TwnacT8AnbFn/Q5C46n6nJtt2qV+4x+4z6nqYADvj2X9ToFD2Lr4E8I97YdQJ590\nbW4MQk++0r61HH8ognRpj/kNhDc1ADgf4U3sjyLynIicEX+vpX01goEA1qvqJvObf6Z8PXcVI+Xt\njWZ36CsRKgkAIEFv7gtgRfxpFoCPAjgewHMxPw5B+vi/xHH9FJPLAHzbOiBVPVhVfwYAqvpTVR0T\ny6IA/rVS+RD0s5xcYngLwMEmf3iiPB5/HwThD86Kqnt0LFYBGGzyQ/bxeJXq9yJXv91UdTYAqOp0\nVT0BwHsRGvxlAP6C0KPz9Vvpnm+J/zeqfv1z3hSo6h8B3Ing2IFw3dWuuQVbl0ci3NO/INTJ3a5O\nDlHVG+wpaymWSS8D8Jo7Zg9VPS2W/1VV/TSAAQht+OfxXhdpX3u75lSZVwLoIyI9zG/Vnqm6aCaH\n3llEupp/nRB0rc+LyCgJHxuvB/D7+PoLBAd+HoA/qOoOAI8DuAChUtcWOPcPAUwWkQ9J4BAROV1E\neojIu0Xko/H82wBsRZBsEMt3lYj0F5F+AP4FwD2VT4F5AP5RRA4UkfEIbxEtrAHQV0R6Vtn3PgCn\ni8ipItIZwKUIktDsAtfYzNwH4Osi0jtqq19u8PFvj8d/H5B97PxkTI+O9d4ZobFuA7A7vvLeh6DB\n94jyzBRUqN/4rK0AMDHW7xcQNNoW1gAYnNBk9/actwsi8h4RuVREBsf8EATNukWqmAfgFAmx1j0R\npCLPRBF5r4gcDOBaAD+P9/YeAGeKyLh4z7pK+MA4uMIxauVZAJskfCjtFo/7folhliIyUUT6xx74\nxrjPbhRrX/MAnCYifUTkcIS3d8saVPiACwCquiwec1q83pEIbw3VfEZhmsmhz0Rwli3/rlHV3yDo\nbr9A6MUNwzt6GBBuTje80xv/A0KDTPXO90BV5wCYhPDqvQHAIrzz0a0LgBsQehWrEf66tzy41yHI\nPi8CeAnA8/G3SlwC4EyEB+kzAB4w5/8jQqNeEl8Vc7KNqr6CoDneGstxJsKHlx1FrrOJuRbhQ9Fr\nAH4D4OdoYGicqt6P0CObISJvAngZ4UM0AByK8Ad9A8Lr7zoAN0XbxQhOfgmAJwH8FOHjeSUmIfTs\n1wF4H/LO4HcA5gNYLSJ/qVC+vT3n7cUmhA/MvxeRLQiO/GUEh4eoe9+L8PzPRQhx9NyN0KtfDaAr\ngK/EfZcBaPlYvRahd30Z9sEnxT8UZyB8X3gNoa3cgfBhGwgf1eeLyGaED6Sf0hC5U6R93Y3w0fV1\nAI/G67dMQ+jkbRSRr1XY/9MIAQArAdyP8C3vN3VdcAVEucAFaTJE5IsIje0je92YNC0i8jiAe1S1\ntCOem41m6qGT/RQROUJEThKRAySEiV6K0HshhBRgvx8pSpqCgxBi/49GkKRmIB3mSAipACUXQggp\nCZRcCCGkJLSp5CIifB1oElS1YXOBsF6bh0bWa9++fRtSrykVQNpkSpp38GWx56+3LG2hcqxbt66m\nwrGHTgghJYEOnRBCSgIdOiGElASGLRJCKuI15Vq14rbQzOs9TuqaimjhtWrv/pitfW/YQyeEkJJA\nh04IISWhaSQX/7ph8wcccEBVm9/Pbps6pif16lVvWFJqv9Q5ithqPR8hRSnyPNltU+2sXhkndb56\npZJ9odbzp/xTqp3XW0720AkhpCTQoRNCSEmgQyeEkJLQtBp6p07vFK1z58512Q488MCqNs+uXbsq\npn2+Xl3R599+++2czeb9+Xfv3l0x7Y+ZshGyr9jnK/Vdy9OIb0Cp72H7EjZY6znr/S7QGjp5CvbQ\nCSGkJNChE0JISWgaycW/wlkppWvXrjnbQQe9s9Zuly5dcja7rZdY/HEsO3fuzNJeDrE2L4ekSEkn\n9pg+v2NHfinDlBxTpDyE7Au+jVqsfOClv1qlBf8sF5E59nW7RmLPmZrdsch+tcIeOiGElAQ6dEII\nKQl06IQQUhKaRkNPhS16nfzggw+umPb51H6ebdu2ZWmvb1ub19dTeB3M7rt9+/acbevWrTUft9o5\nGLZIWpNUyF2toYn+GbVt1IcZp47pj1MvjQh/bITW36j92EMnhJCSQIdOCCEloUNILn40qH1N8zJK\njx49svQhhxxS1eZDsGyooA8btPm9hWTZ4/pzWCln8+bNOdumTZuy9JYtW3K2t956K0tb+QfISzW+\nbI16LSX7J0VC7iz+ubfPoQ8dHj58eJb2bWL+/PlZ2ssx9hwp/+BJhRL7dm9t/hy2PCmf0NZhk+yh\nE0JISaBDJ4SQkkCHTgghJaFpNXSrkfkh/KlpAaym3rNnz5ytV69eWdpOHwCkZzRMzTLntTVbHm+z\nGvr69etztnXr1mXpDRs25Gw2v3HjxpzN5n24ZZEQy2bjnHPOyeUnTZqUpVeuXJmz2e8KP/nJT3K2\n1atXZ+lFixY1soj7HfWG6tk2c/nll+ds5513Xpb2obuzZs3K0k8++WTO9tRTT2XpNWvW5GxW3/Zt\noFu3brm8ba/eJ9ht/feEFStWVD2HzReZlbIRsIdOCCElgQ6dEEJKgrTlaEIRqXoyH2pk5ZGUdNKn\nT5+crW/fvhXTPu9DGmtdlNq/lnXv3j2Xt8f157CjQ1etWpWzWWnApvdmW7t2bVWbD3+0qGrD3v1S\n9VovS5YsyeWPOuqouo5jw0FtGFxbsXz58ix944035mxz5sxp+PkaWa99+vTJ1Wu9MxzacMDbb789\nZ/PSmsVKFz5sMTW7o93Py5BFFoix7cfLvlbmueGGG3K2Rx55pOp+qRkrU754/fr1Nd189tAJIaQk\n0KETQkhJoEMnhJCS0CHCFlOhgqkZFb32bvV2b7PH9Dq5tXld3Or5/rjeZsOyli5dWnU/H1pl743X\nBO1waa8zdmRsmCIAjBw5MksvWLAgZxsxYkSW/sAHPpCzjR07NkufeOKJOduyZcuy9JAhQ2oumw9T\ns98xjjjiiKr7+TpvDQ29vbDPqL8/9rmcOnVqzmZDE4855picbdiwYVn6uOOOy9n69++fpX3osvUJ\nvi15ndpOBeBnQE2FRw8cODBLT5kyJWd74oknKh7f0xrfL9lDJ4SQkkCHTgghJaFpJJfUBPY+nMjK\nDv41yY4atLMUAnnJI7WAdJFZ5rw8Y2d09GGTtqxvvvlmzmZHjnoZyb76eVnF5ttjQdzW4re//W0y\nb/nVr35V1da7d+8sPWrUqJxt7ty5WXr06NE1l83PeLlw4cIs7eUgK/MtXry45nM0A0Xagd3Wj5C2\nEowf5XvvvfdW3c+2Fy+dpKRV2yZ8WLFvd7ZsVsYBgAkTJmTps846K2ezbdvXqz2/n8Gx1oW2uUg0\nIYTs59ChE0JISaBDJ4SQktA0GrrXya1+lhrCntrPr4BidWqvrVnd2ocoWZ3c6+Jep7d43c9q/3ZI\nOpCfNdHPxGhnW/T72evlCkV7Yu/dY489VnW7lEa/Nz7xiU9kaavZA8BLL72Upa1e3BHwmnmt32j8\nc2h1Yx/2a21eb7ZatA/XtbOT2rTfz5c5NZPqyy+/nLN99rOfzdK+Xu3Q/5tvvjlns9/qisy22Igw\nRvbQCSGkJNChE0JISegQkot/FbGhRj6EzMosfqEI+7rnww2tzOLDBq1twIABlS8gYsOp/EhRiw+f\nqlVy8fKTvf7UqDTSOPwzcNttt2Vp/4p97bXXZmlfrx2N1AIX1ubba2pGw1plBn8+336rbevbROo4\ngwcPztnGjRtX9Ry33HJLln711VdzNus/Ugtmp+5hvbCHTgghJYEOnRBCSgIdOiGElISm1dBtCJOf\nvc3qxqmh8N5mh/v7of9WS0uFLQ4aNChn89va8CY/857V03xIpdXQfRiW1dD9fgxbbHu+9KUv5fJ2\nyLj/bvPKK6+0SZnagloXifakhrs3oixFzp0q9wUXXJDLH3744Vnahp8CwF133VX1mNbvcLZFQggh\ndUGHTgghJaFpJJfUxPOtEY7nZ3azozr9KFIbiujDJH3ejhz18ojFhy2+8cYbWdrKL97mwxbtCDpK\nLq3HSSedlKWvvPLKqtudffbZubwffUgaT2qmVi+5eF/Sr1+/LH3++edXPc5VV12Vs9l27sOcU1JK\nvbJVrbCHTgghJYEOnRBCSgIdOiGElISm0dDbAqtZ+bBFq5vbVWaA/BB+G8oE7Dl7nCWld6d0cpsG\n8lq81+xteCc19NbjtNNOy9L++4udqfHpp59uszK1NakVdWrVjVOktiuiN6eG/vspAyZPnpyl7cLP\nQH4Ba78qlv2u5n1JaqqDauUsYkvBHjohhJQEOnRCCCkJdOiEEFISSqeh16o9FdHQ7RD+ww47rOp+\n/vw+Dt3GrqY0dB+jbo9jh/oD+WkRWmMo8f6KX21q/PjxWdqvrDN16tQs7VfWKSuN0rtr3W9vKw9Z\nrG7u49B79uyZy1900UVZ2tfdTTfdlKVTKy2lpgSud9WnemEPnRBCSgIdOiGElITSSS4p7OuODz2z\nYUgpycXOrAfsKblYvORiF3iuN2zRh2HVGiJFinHZZZfl8scff3yW9iFss2fPbpMytTf1hh/WKjOk\n9vO2lORhQxP9TK124Wcg337tDIoA8Oyzz2ZpL8fUumJSvTJovfuxh04IISWBDp0QQkoCHTohhJQE\nactQNxFp+MlSq4WkbHbaTAA49thjs/Tw4cNzNpv3+rrX72ze2+xKRH6l8IULF2bpRYsW5Wx2tfhG\nDLGO2zYsfqo16rWtOf3003P5Bx54IJe30zjYEEYAeOaZZ1qvYAVpZL326dOnIfVaq6bcqLBFew4/\nnH/u3Lm5vJ1KY9SoUTmbXX3Kr35mSa3IVGTofyrccd26dTXVK3vohBBSEujQCSGkJHT4sEX/umPD\nEX1ool1ZxM6gCOQlGD8a1IYt9ujRI2fziwLbcERvW7NmTZZeu3ZtzmZDE32IFEeAtg59+/bN0tOn\nT8/Z/Cv2zJkzs3QzSSzNQiNklSLHtDZ/TDvK94477sjZfLihDU/1I7TtM1BQzmzodkVgD50QQkoC\nHTohhJQEOnRCCCkJHV5D9/qZ1c3tcH4gP2NaEQ190KBBFY8P5IfzA/kZFVetWpWzrVy5MkvbEEYg\nraFbqKfXj9fF7RD+o48+OmdbvHhxLn/11Ve3XsFKRpFwvBSpaS3stzP/He2SSy7J0qNHj87ZXnjh\nhVz+/vvvr1o2e1w/5Ua9qzdVO0ajYA+dEEJKAh06IYSUhA4vufjXLRuW1LVr15zNzqzWu3fvnM1K\nLgMGDMjZ/Ggzi5VRAGDr1q1ZevXq1TnbsmXLsnRqEQs/QxxllsYwbNiwXP6EE06ouu2UKVNyeS/B\nkOqkJJciI51t3stltt2fcsopOZudUdGOsgaAK6+8Mpe3YcbeX6QWyrAUmeWUC1wQQgipCTp0Qggp\nCXTohBBSEkqtofvVhOwCsT5s0YY0+nBHO2WADyn0erfVwr1+Z/N2ljef58pDjWPo0KFZ+tFHH626\nnV+h6OGHH261MnVUUvpvrQsj13oMIN8OfDu30zZceOGFOZtt5zNmzMjZnn/++VzeLhaf0vB9ua2+\n3qgwzWrHKAJ76IQQUhLo0AkhpCTsV5KLDVU89NBDq+7nZRQbYmjDEoE9F3S2I0ftoggAsH379iy9\nY8eOnM2+wjFMsXHY1/Ejjzyy6nazZs3K5VkH9c+MWOviE3s7n22Tvp1PnDgxS/vwU9tGH3zwwZzN\ntkF/3NRC1EWod1HsRsAeOiGElAQ6dEIIKQl06IQQUhI6vIbuhwTb4bs2FBHIhyp6Dd3Oouj1bauh\n27BEID90GMjr5l5vt6GJXqenht4YxowZk8tffPHF7VSSjk9qlaAUqTC+IvvZEOERI0bkbOeee26W\n9m3Zrgz23HPP5Wx+xaJUu6t1xaQi1PoNgWGLhBCyn0OHTgghJaHDSy4+tMhKLkXCFq3kkgpb9BKL\nnzXRhi3axS6AfMiUD+2qdUY6kubkk0/O5f0zYLEzKHopjRQjNTqyVpvHSqbjx4/P2exIUY+tVy+f\nelp79sN6z1dvudhDJ4SQkkCHTgghJYEOnRBCSkKH19C91mRnT7OzJAJAt27dsrRfncSGP3p922rf\nqVkSgbxml5qZscjKLaQx+AWCTz311CztZ8Yke1JrWF2REL9aj+kXVbergfnvH9///veztG+DqZkg\nfbuv1VZvmGajQiEt7KETQkhJoEMnhJCS0OElF0Is06ZNS+ZJ7RRZ7LleOcbiZQ0rpTz00EM527x5\n87K0lz2XLl1asVzAnrMt2nDleheqKCKRNmLxixTsoRNCSEmgQyeEkJJAh04IISVBGCJHCCHlgD10\nQggpCXTohBBSEujQCSGkJNChE0JISaBDJ4SQkkCHTgghJYEOnRBCSgIdOiGElAQ6dEIIKQl06IQQ\nUhLo0AkhpCTQoRNCSEmgQyeEkJJAh04IISWBDp0QQkoCHTohhJQEOnRCCCkJdOiEEFIS6NAJIaQk\n0KETQkhJoEMnhJCSQIdOCCElgQ6dEEJKwv8DgknsfDZewZIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7f8ad53198>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAACRCAYAAADTnUPWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7f8837d9b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAACRCAYAAADTnUPWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7f840a8c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAACRCAYAAADTnUPWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7f83b59748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAACRCAYAAADTnUPWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7f7f437d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_high  = test_high.reshape((-1, 28, 28))\n",
    "test_low   = test_low.reshape((-1, 28, 28))\n",
    "test_super = test_super.reshape((-1, 28, 28))\n",
    "\n",
    "for i, (low, high, sup) in enumerate(zip(test_low, test_high, test_super)):\n",
    "    fig = plt.figure()\n",
    "    a = fig.add_subplot(1, 3, 1)\n",
    "    a.axis(\"off\")\n",
    "    a.set_title(\"Low resolution\")\n",
    "    plt.imshow(low, cmap=\"gray\")\n",
    "    \n",
    "    a = fig.add_subplot(1, 3, 2)\n",
    "    a.axis(\"off\")\n",
    "    a.set_title(\"High resolution\")\n",
    "    plt.imshow(high, cmap=\"gray\")\n",
    "    \n",
    "    a = fig.add_subplot(1, 3, 3)\n",
    "    a.axis(\"off\")\n",
    "    a.set_title(\"Super resolution\")\n",
    "    plt.imshow(sup, cmap=\"gray\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}