find-similar-images.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Identifying Similar Images with Tensorflow\n",
    "\n",
    "Douglas Duhaime wrote an [excellent tutorial](http://douglasduhaime.com/posts/identifying-similar-images-with-tensorflow.html). This notebook is my experiment trying to follow his steps, using a github repo and a jupyter binder.\n",
    "\n",
    "The repo is at [github](https://github.com/shawngraham/bindr-test-Identifying-Similar-Images-with-TensorFlow). The various python bits and pieces are called from the `requirements.txt` file, which saves us from having to `!pip install`.\n",
    "\n",
    "I have a bunch of images in \\images. In Duhaime's tutorial, he had a folder with 2000 images. I'm just going with 25 here because a) I'm impatient and b) I don't know how many I can push into github.\n",
    "\n",
    "Run the modified classify script (which pulls out the second but last layer), and write it to the new `image_vectors` dir. Then cluster, then project. Then run the affinity propagation notebook, which is written in R."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "## Now Let's Identify Similar Images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      ">> Downloading inception-2015-12-05.tgz 100.0%\n",
      "Succesfully downloaded inception-2015-12-05.tgz 88931400 bytes.\n",
      "2018-05-07 09:52:20.550199: W tensorflow/core/framework/op_def_util.cc:346] Op BatchNormWithGlobalNormalization is deprecated. It will cease to work in GraphDef version 9. Use tf.nn.batch_normalization().\n",
      "2018-05-07 09:52:20.802865: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA\n",
      "parsing 0 images/3969575.jpg \n",
      "\n",
      "results for images/3969575.jpg\n",
      "tobacco shop, tobacconist shop, tobacconist (score = 0.27585)\n",
      "\n",
      "\n",
      "results for images/3969575.jpg\n",
      "bookshop, bookstore, bookstall (score = 0.19276)\n",
      "\n",
      "\n",
      "results for images/3969575.jpg\n",
      "entertainment center (score = 0.05562)\n",
      "\n",
      "\n",
      "results for images/3969575.jpg\n",
      "bookcase (score = 0.04556)\n",
      "\n",
      "\n",
      "results for images/3969575.jpg\n",
      "lab coat, laboratory coat (score = 0.03591)\n",
      "\n",
      "\n",
      "parsing 1 images/3969603.jpg \n",
      "\n",
      "results for images/3969603.jpg\n",
      "book jacket, dust cover, dust jacket, dust wrapper (score = 0.76150)\n",
      "\n",
      "\n",
      "results for images/3969603.jpg\n",
      "web site, website, internet site, site (score = 0.09768)\n",
      "\n",
      "\n",
      "results for images/3969603.jpg\n",
      "comic book (score = 0.01847)\n",
      "\n",
      "\n",
      "results for images/3969603.jpg\n",
      "brass, memorial tablet, plaque (score = 0.00194)\n",
      "\n",
      "\n",
      "results for images/3969603.jpg\n",
      "planetarium (score = 0.00190)\n",
      "\n",
      "\n",
      "parsing 2 images/3969615.jpg \n",
      "\n",
      "results for images/3969615.jpg\n",
      "wardrobe, closet, press (score = 0.19771)\n",
      "\n",
      "\n",
      "results for images/3969615.jpg\n",
      "barbershop (score = 0.13588)\n",
      "\n",
      "\n",
      "results for images/3969615.jpg\n",
      "tobacco shop, tobacconist shop, tobacconist (score = 0.09701)\n",
      "\n",
      "\n",
      "results for images/3969615.jpg\n",
      "lab coat, laboratory coat (score = 0.08760)\n",
      "\n",
      "\n",
      "results for images/3969615.jpg\n",
      "bakery, bakeshop, bakehouse (score = 0.07224)\n",
      "\n",
      "\n",
      "parsing 3 images/3969601.jpg \n",
      "\n",
      "results for images/3969601.jpg\n",
      "ox (score = 0.54208)\n",
      "\n",
      "\n",
      "results for images/3969601.jpg\n",
      "oxcart (score = 0.33635)\n",
      "\n",
      "\n",
      "results for images/3969601.jpg\n",
      "water buffalo, water ox, Asiatic buffalo, Bubalus bubalis (score = 0.01660)\n",
      "\n",
      "\n",
      "results for images/3969601.jpg\n",
      "prison, prison house (score = 0.00627)\n",
      "\n",
      "\n",
      "results for images/3969601.jpg\n",
      "horse cart, horse-cart (score = 0.00568)\n",
      "\n",
      "\n",
      "parsing 4 images/3969577.jpg \n",
      "\n",
      "results for images/3969577.jpg\n",
      "web site, website, internet site, site (score = 0.18976)\n",
      "\n",
      "\n",
      "results for images/3969577.jpg\n",
      "wing (score = 0.11993)\n",
      "\n",
      "\n",
      "results for images/3969577.jpg\n",
      "seashore, coast, seacoast, sea-coast (score = 0.06913)\n",
      "\n",
      "\n",
      "results for images/3969577.jpg\n",
      "television, television system (score = 0.02831)\n",
      "\n",
      "\n",
      "results for images/3969577.jpg\n",
      "ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus (score = 0.02484)\n",
      "\n",
      "\n",
      "parsing 5 images/3969589.jpg \n",
      "\n",
      "results for images/3969589.jpg\n",
      "restaurant, eating house, eating place, eatery (score = 0.14967)\n",
      "\n",
      "\n",
      "results for images/3969589.jpg\n",
      "television, television system (score = 0.13588)\n",
      "\n",
      "\n",
      "results for images/3969589.jpg\n",
      "screen, CRT screen (score = 0.08451)\n",
      "\n",
      "\n",
      "results for images/3969589.jpg\n",
      "library (score = 0.05580)\n",
      "\n",
      "\n",
      "results for images/3969589.jpg\n",
      "monitor (score = 0.05216)\n",
      "\n",
      "\n",
      "parsing 6 images/3969599.jpg \n",
      "\n",
      "results for images/3969599.jpg\n",
      "fire screen, fireguard (score = 0.26967)\n",
      "\n",
      "\n",
      "results for images/3969599.jpg\n",
      "desk (score = 0.06362)\n",
      "\n",
      "\n",
      "results for images/3969599.jpg\n",
      "television, television system (score = 0.02327)\n",
      "\n",
      "\n",
      "results for images/3969599.jpg\n",
      "table lamp (score = 0.02222)\n",
      "\n",
      "\n",
      "results for images/3969599.jpg\n",
      "pedestal, plinth, footstall (score = 0.02173)\n",
      "\n",
      "\n",
      "parsing 7 images/3969573.jpg \n",
      "\n",
      "results for images/3969573.jpg\n",
      "flute, transverse flute (score = 0.09582)\n",
      "\n",
      "\n",
      "results for images/3969573.jpg\n",
      "book jacket, dust cover, dust jacket, dust wrapper (score = 0.09494)\n",
      "\n",
      "\n",
      "results for images/3969573.jpg\n",
      "assault rifle, assault gun (score = 0.06679)\n",
      "\n",
      "\n",
      "results for images/3969573.jpg\n",
      "harmonica, mouth organ, harp, mouth harp (score = 0.05129)\n",
      "\n",
      "\n",
      "results for images/3969573.jpg\n",
      "violin, fiddle (score = 0.04440)\n",
      "\n",
      "\n",
      "parsing 8 images/3969613.jpg \n",
      "\n",
      "results for images/3969613.jpg\n",
      "tobacco shop, tobacconist shop, tobacconist (score = 0.82542)\n",
      "\n",
      "\n",
      "results for images/3969613.jpg\n",
      "toyshop (score = 0.04832)\n",
      "\n",
      "\n",
      "results for images/3969613.jpg\n",
      "bookshop, bookstore, bookstall (score = 0.01201)\n",
      "\n",
      "\n",
      "results for images/3969613.jpg\n",
      "comic book (score = 0.00591)\n",
      "\n",
      "\n",
      "results for images/3969613.jpg\n",
      "barbershop (score = 0.00528)\n",
      "\n",
      "\n",
      "parsing 9 images/3969571.jpg \n",
      "\n",
      "results for images/3969571.jpg\n",
      "book jacket, dust cover, dust jacket, dust wrapper (score = 0.76292)\n",
      "\n",
      "\n",
      "results for images/3969571.jpg\n",
      "passenger car, coach, carriage (score = 0.02995)\n",
      "\n",
      "\n",
      "results for images/3969571.jpg\n",
      "prison, prison house (score = 0.01691)\n",
      "\n",
      "\n",
      "results for images/3969571.jpg\n",
      "screen, CRT screen (score = 0.00655)\n",
      "\n",
      "\n",
      "results for images/3969571.jpg\n",
      "analog clock (score = 0.00444)\n",
      "\n",
      "\n",
      "parsing 10 images/1615157.jpg \n",
      "\n",
      "results for images/1615157.jpg\n",
      "shower cap (score = 0.42766)\n",
      "\n",
      "\n",
      "results for images/1615157.jpg\n",
      "hoopskirt, crinoline (score = 0.09817)\n",
      "\n",
      "\n",
      "results for images/1615157.jpg\n",
      "pickelhaube (score = 0.04220)\n",
      "\n",
      "\n",
      "results for images/1615157.jpg\n",
      "cowboy boot (score = 0.03903)\n",
      "\n",
      "\n",
      "results for images/1615157.jpg\n",
      "jinrikisha, ricksha, rickshaw (score = 0.03347)\n",
      "\n",
      "\n",
      "parsing 11 images/3932077.jpg \n",
      "\n",
      "results for images/3932077.jpg\n",
      "airship, dirigible (score = 0.12418)\n",
      "\n",
      "\n",
      "results for images/3932077.jpg\n",
      "gondola (score = 0.05746)\n",
      "\n",
      "\n",
      "results for images/3932077.jpg\n",
      "dock, dockage, docking facility (score = 0.05331)\n",
      "\n",
      "\n",
      "results for images/3932077.jpg\n",
      "submarine, pigboat, sub, U-boat (score = 0.03511)\n",
      "\n",
      "\n",
      "results for images/3932077.jpg\n",
      "dome (score = 0.03289)\n",
      "\n",
      "\n",
      "parsing 12 images/1952272.jpg \n",
      "\n",
      "results for images/1952272.jpg\n",
      "lumbermill, sawmill (score = 0.44609)\n",
      "\n",
      "\n",
      "results for images/1952272.jpg\n",
      "bakery, bakeshop, bakehouse (score = 0.03943)\n",
      "\n",
      "\n",
      "results for images/1952272.jpg\n",
      "broom (score = 0.03028)\n",
      "\n",
      "\n",
      "results for images/1952272.jpg\n",
      "barrow, garden cart, lawn cart, wheelbarrow (score = 0.02897)\n",
      "\n",
      "\n",
      "results for images/1952272.jpg\n",
      "tobacco shop, tobacconist shop, tobacconist (score = 0.02464)\n",
      "\n",
      "\n",
      "parsing 13 images/1952301.jpg \n",
      "\n",
      "results for images/1952301.jpg\n",
      "minibus (score = 0.16549)\n",
      "\n",
      "\n",
      "results for images/1952301.jpg\n",
      "television, television system (score = 0.12829)\n",
      "\n",
      "\n",
      "results for images/1952301.jpg\n",
      "Model T (score = 0.11771)\n",
      "\n",
      "\n",
      "results for images/1952301.jpg\n",
      "pencil sharpener (score = 0.03841)\n",
      "\n",
      "\n",
      "results for images/1952301.jpg\n",
      "passenger car, coach, carriage (score = 0.03528)\n",
      "\n",
      "\n",
      "parsing 14 images/3969569.jpg \n",
      "\n",
      "results for images/3969569.jpg\n",
      "African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus (score = 0.42135)\n",
      "\n",
      "\n",
      "results for images/3969569.jpg\n",
      "hog, pig, grunter, squealer, Sus scrofa (score = 0.12953)\n",
      "\n",
      "\n",
      "results for images/3969569.jpg\n",
      "skunk, polecat, wood pussy (score = 0.07356)\n",
      "\n",
      "\n",
      "results for images/3969569.jpg\n",
      "wild boar, boar, Sus scrofa (score = 0.04827)\n",
      "\n",
      "\n",
      "results for images/3969569.jpg\n",
      "book jacket, dust cover, dust jacket, dust wrapper (score = 0.03369)\n",
      "\n",
      "\n",
      "parsing 15 images/3969583.jpg \n",
      "\n",
      "results for images/3969583.jpg\n",
      "book jacket, dust cover, dust jacket, dust wrapper (score = 0.42377)\n",
      "\n",
      "\n",
      "results for images/3969583.jpg\n",
      "comic book (score = 0.16741)\n",
      "\n",
      "\n",
      "results for images/3969583.jpg\n",
      "guillotine (score = 0.02499)\n",
      "\n",
      "\n",
      "results for images/3969583.jpg\n",
      "pole (score = 0.02278)\n",
      "\n",
      "\n",
      "results for images/3969583.jpg\n",
      "prison, prison house (score = 0.02061)\n",
      "\n",
      "\n",
      "parsing 16 images/3969597.jpg \n",
      "\n",
      "results for images/3969597.jpg\n",
      "television, television system (score = 0.40918)\n",
      "\n",
      "\n",
      "results for images/3969597.jpg\n",
      "barbershop (score = 0.09649)\n",
      "\n",
      "\n",
      "results for images/3969597.jpg\n",
      "sliding door (score = 0.08329)\n",
      "\n",
      "\n",
      "results for images/3969597.jpg\n",
      "screen, CRT screen (score = 0.05706)\n",
      "\n",
      "\n",
      "results for images/3969597.jpg\n",
      "web site, website, internet site, site (score = 0.04899)\n",
      "\n",
      "\n",
      "parsing 17 images/3969581.jpg \n",
      "\n",
      "results for images/3969581.jpg\n",
      "television, television system (score = 0.34591)\n",
      "\n",
      "\n",
      "results for images/3969581.jpg\n",
      "screen, CRT screen (score = 0.05680)\n",
      "\n",
      "\n",
      "results for images/3969581.jpg\n",
      "web site, website, internet site, site (score = 0.03028)\n",
      "\n",
      "\n",
      "results for images/3969581.jpg\n",
      "entertainment center (score = 0.01993)\n",
      "\n",
      "\n",
      "results for images/3969581.jpg\n",
      "home theater, home theatre (score = 0.01982)\n",
      "\n",
      "\n",
      "parsing 18 images/3969595.jpg \n",
      "\n",
      "results for images/3969595.jpg\n",
      "barbershop (score = 0.57052)\n",
      "\n",
      "\n",
      "results for images/3969595.jpg\n",
      "television, television system (score = 0.07383)\n",
      "\n",
      "\n",
      "results for images/3969595.jpg\n",
      "streetcar, tram, tramcar, trolley, trolley car (score = 0.04607)\n",
      "\n",
      "\n",
      "results for images/3969595.jpg\n",
      "web site, website, internet site, site (score = 0.04582)\n",
      "\n",
      "\n",
      "results for images/3969595.jpg\n",
      "barber chair (score = 0.02506)\n",
      "\n",
      "\n",
      "parsing 19 images/3969609.jpg \n",
      "\n",
      "results for images/3969609.jpg\n",
      "stretcher (score = 0.25366)\n",
      "\n",
      "\n",
      "results for images/3969609.jpg\n",
      "apiary, bee house (score = 0.19872)\n",
      "\n",
      "\n",
      "results for images/3969609.jpg\n",
      "mountain tent (score = 0.03312)\n",
      "\n",
      "\n",
      "results for images/3969609.jpg\n",
      "thresher, thrasher, threshing machine (score = 0.02733)\n",
      "\n",
      "\n",
      "results for images/3969609.jpg\n",
      "park bench (score = 0.02342)\n",
      "\n",
      "\n",
      "parsing 20 images/3969591.jpg \n",
      "\n",
      "results for images/3969591.jpg\n",
      "ox (score = 0.86663)\n",
      "\n",
      "\n",
      "results for images/3969591.jpg\n",
      "oxcart (score = 0.02592)\n",
      "\n",
      "\n",
      "results for images/3969591.jpg\n",
      "whippet (score = 0.00860)\n",
      "\n",
      "\n",
      "results for images/3969591.jpg\n",
      "plow, plough (score = 0.00395)\n",
      "\n",
      "\n",
      "results for images/3969591.jpg\n",
      "worm fence, snake fence, snake-rail fence, Virginia fence (score = 0.00303)\n",
      "\n",
      "\n",
      "parsing 21 images/3969585.jpg \n",
      "\n",
      "results for images/3969585.jpg\n",
      "guillotine (score = 0.12392)\n",
      "\n",
      "\n",
      "results for images/3969585.jpg\n",
      "lumbermill, sawmill (score = 0.10771)\n",
      "\n",
      "\n",
      "results for images/3969585.jpg\n",
      "prison, prison house (score = 0.07205)\n",
      "\n",
      "\n",
      "results for images/3969585.jpg\n",
      "butcher shop, meat market (score = 0.04236)\n",
      "\n",
      "\n",
      "results for images/3969585.jpg\n",
      "stretcher (score = 0.03917)\n",
      "\n",
      "\n",
      "parsing 22 images/3969593.jpg \n",
      "\n",
      "results for images/3969593.jpg\n",
      "triumphal arch (score = 0.06874)\n",
      "\n",
      "\n",
      "results for images/3969593.jpg\n",
      "jinrikisha, ricksha, rickshaw (score = 0.05751)\n",
      "\n",
      "\n",
      "results for images/3969593.jpg\n",
      "horse cart, horse-cart (score = 0.03460)\n",
      "\n",
      "\n",
      "results for images/3969593.jpg\n",
      "academic gown, academic robe, judge's robe (score = 0.03368)\n",
      "\n",
      "\n",
      "results for images/3969593.jpg\n",
      "bearskin, busby, shako (score = 0.03326)\n",
      "\n",
      "\n",
      "parsing 23 images/3969587.jpg \n",
      "\n",
      "results for images/3969587.jpg\n",
      "passenger car, coach, carriage (score = 0.27143)\n",
      "\n",
      "\n",
      "results for images/3969587.jpg\n",
      "steam locomotive (score = 0.07181)\n",
      "\n",
      "\n",
      "results for images/3969587.jpg\n",
      "electric locomotive (score = 0.03989)\n",
      "\n",
      "\n",
      "results for images/3969587.jpg\n",
      "freight car (score = 0.03187)\n",
      "\n",
      "\n",
      "results for images/3969587.jpg\n",
      "airship, dirigible (score = 0.01730)\n",
      "\n",
      "\n",
      "parsing 24 images/3969579.jpg \n",
      "\n",
      "results for images/3969579.jpg\n",
      "military uniform (score = 0.86382)\n",
      "\n",
      "\n",
      "results for images/3969579.jpg\n",
      "pickelhaube (score = 0.06729)\n",
      "\n",
      "\n",
      "results for images/3969579.jpg\n",
      "bulletproof vest (score = 0.00916)\n",
      "\n",
      "\n",
      "results for images/3969579.jpg\n",
      "stretcher (score = 0.00525)\n",
      "\n",
      "\n",
      "results for images/3969579.jpg\n",
      "prison, prison house (score = 0.00484)\n",
      "\n",
      "\n",
      "all done\n"
     ]
    }
   ],
   "source": [
    "!python classify_images.py \"images/*\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "now cluster with his nearest neighbours script. \"Each of those outfiles will identify the 30 images most similar to the given image. To search for more or fewer nearest neighbors, one just needs to update the n_nearest_neighbors variable in the nearest neighbors script.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "!python cluster_vectors.py"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "and now we run the t-sne clustering script"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " * loaded 0 of 25 image vectors\n",
      " * loaded 1 of 25 image vectors\n",
      " * loaded 2 of 25 image vectors\n",
      " * loaded 3 of 25 image vectors\n",
      " * loaded 4 of 25 image vectors\n",
      " * loaded 5 of 25 image vectors\n",
      " * loaded 6 of 25 image vectors\n",
      " * loaded 7 of 25 image vectors\n",
      " * loaded 8 of 25 image vectors\n",
      " * loaded 9 of 25 image vectors\n",
      " * loaded 10 of 25 image vectors\n",
      " * loaded 11 of 25 image vectors\n",
      " * loaded 12 of 25 image vectors\n",
      " * loaded 13 of 25 image vectors\n",
      " * loaded 14 of 25 image vectors\n",
      " * loaded 15 of 25 image vectors\n",
      " * loaded 16 of 25 image vectors\n",
      " * loaded 17 of 25 image vectors\n",
      " * loaded 18 of 25 image vectors\n",
      " * loaded 19 of 25 image vectors\n",
      " * loaded 20 of 25 image vectors\n",
      " * loaded 21 of 25 image vectors\n",
      " * loaded 22 of 25 image vectors\n",
      " * loaded 23 of 25 image vectors\n",
      " * loaded 24 of 25 image vectors\n",
      "building tsne model\n"
     ]
    }
   ],
   "source": [
    "# ok let's project this stuff!\n",
    "from sklearn.manifold import TSNE\n",
    "import numpy as np\n",
    "import glob, json, os\n",
    "\n",
    "# create datastores\n",
    "vector_files = []\n",
    "image_vectors = []\n",
    "chart_data = []\n",
    "# note the maximum here - change up for total number you've got!\n",
    "maximum_imgs = 25\n",
    "\n",
    "# build a list of image vectors\n",
    "vector_files = glob.glob('image_vectors/*.npz')[:maximum_imgs]\n",
    "for c, i in enumerate(vector_files):\n",
    "  image_vectors.append(np.loadtxt(i))\n",
    "  print(' * loaded', c, 'of', len(vector_files), 'image vectors')\n",
    "\n",
    "# build the tsne model on the image vectors\n",
    "print('building tsne model')\n",
    "model = TSNE(n_components=2, random_state=0)\n",
    "np.set_printoptions(suppress=True)\n",
    "fit_model = model.fit_transform( np.array(image_vectors) )\n",
    " \n",
    "# store the coordinates of each image in the chart data\n",
    "for c, i in enumerate(fit_model):\n",
    "  image_name = os.path.basename(vector_files[c]).replace('.npz', '') \n",
    "  chart_data.append({\n",
    "    'image': os.path.join('images', image_name),\n",
    "    'x': i[0],\n",
    "    'y': i[1]\n",
    "  })"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "then we write the `chart_data` to file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "thefile = open('image_tsne_projections.json', 'w')\n",
    "for item in chart_data:\n",
    "  thefile.write(\"%s\\n\" % item)\n",
    "thefile.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Now go view your work by clicking on the 'jupyter' button above! Outputs, folders, etc, all visible there.\n",
    "\n",
    "Before you can work with that json file though, you need to add a `[` and a `]` to the front and end of the file, and a `,` at the end of each line (except the last one). Easily done in Regex; doing it in the code chunk is a bit beyond me still.\n",
    "\n",
    "If you don't do this, the import json in the affinity propagation notebook will throw and error."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "duhaime",
   "language": "python",
   "name": "duhaime"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}