transfer4uspVGG16FtrExtr.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from numpy import genfromtxt\n",
    "import pylab as pl\n",
    "from io import StringIO\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n",
    "\n",
    "import urllib\n",
    "import random\n",
    "import sys\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation, Dropout\n",
    "from keras.layers import LSTM\n",
    "from keras.optimizers import RMSprop\n",
    "from keras.utils.data_utils import get_file\n",
    "from keras.layers.wrappers import TimeDistributed\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import load_model\n",
    "from keras.models import model_from_json\n",
    "from keras.models import Model\n",
    "\n",
    "from IPython.display import clear_output\n",
    "import re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.applications.mobilenet import MobileNet\n",
    "from keras.applications.resnet50 import ResNet50\n",
    "from keras.applications.vgg16 import VGG16\n",
    "from keras.applications.vgg19 import VGG19\n",
    "from keras.applications.xception import Xception\n",
    "from keras.applications.inception_v3 import InceptionV3\n",
    "from keras.applications.inception_resnet_v2 import InceptionResNetV2\n",
    "from keras.applications.densenet import DenseNet121\n",
    "from keras.applications.densenet import DenseNet169\n",
    "from keras.applications.densenet import DenseNet201\n",
    "from keras.applications.nasnet import NASNetLarge\n",
    "from keras.applications.nasnet import NASNetMobile\n",
    "\n",
    "from keras.applications import mobilenet\n",
    "# from keras.applications.mobilenet import preprocess_input, decode_predictions\n",
    "from keras.applications.resnet50 import preprocess_input, decode_predictions\n",
    "from keras.preprocessing import image\n",
    "from keras import applications\n",
    "import tensorflow as tf\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "# Imports\n",
    "import os, os.path, gc, time, shutil, matplotlib, glob"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load image list and set image path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Published Image quantity: 5505\n",
      "'4.62406833671_-74.0571218553_-90.jpg'\n",
      "'4.6461951_-74.0536636_180.jpg'\n",
      "'4.6461951_-74.0536636_180.jpg'\n"
     ]
    }
   ],
   "source": [
    "# Image set directory root\n",
    "# chapinero\n",
    "imgRoot = '/Users/SerG1oAC/Documents/dbintroUdacity/fullstack/vagrant/ggStreetView/map/localidades/chapinero/imgs/'\n",
    "# martires\n",
    "# imgRoot = '/Users/SerG1oAC/Documents/dbintroUdacity/fullstack/vagrant/ggStreetView/map/localidades/martires/imgs/'\n",
    "# usaquen\n",
    "# imgRoot = '/Users/SerG1oAC/Documents/dbintroUdacity/fullstack/vagrant/ggStreetView/map/localidades/usaquen/imgs_usaquen/'\n",
    "\n",
    "# Load published images\n",
    "# chapinero\n",
    "pblImgFile = open('mstrPublishedLst', 'r')\n",
    "# martires\n",
    "# pblImgFile = open('mstrPublishedLst_Martires', 'r')\n",
    "# usaquen\n",
    "# pblImgFile = open('mstrPublishedLst_Usaquen', 'r')\n",
    "\n",
    "pblImgSet =  pblImgFile.readlines() \n",
    "print('Published Image quantity: ' + str(len(pblImgSet)))\n",
    "print(pblImgSet[0][:-2])\n",
    "print(pblImgSet[len(pblImgSet) - 1][:-2])\n",
    "print(pblImgSet[-1][:-2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load VGG19 model without output layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "modelVGG19 = VGG19(weights='imagenet', include_top=False, pooling='avg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image Shape:  (1, 224, 224, 3)\n",
      "Image 4 Nas Shape:  (1, 331, 331, 3)\n",
      "VGG19:  (512,)\n"
     ]
    }
   ],
   "source": [
    "img_path = imgRoot + pblImgSet[np.int(0.0)][:-2].replace(\"'\", \"\")\n",
    "# load image setting the image size to 224 x 224\n",
    "img = image.load_img(img_path, target_size=(224, 224))\n",
    "img4nas = image.load_img(img_path, target_size=(331, 331))\n",
    "# convert image to numpy array\n",
    "x = image.img_to_array(img)\n",
    "# the image is now in an array of shape (3, 224, 224) \n",
    "# need to expand it to (1, 3, 224, 224) as it's expecting a list\n",
    "x = np.expand_dims(x, axis=0)\n",
    "x = preprocess_input(x)\n",
    "\n",
    "x4nas = image.img_to_array(img4nas)\n",
    "# the image is now in an array of shape (3, 224, 224) \n",
    "# need to expand it to (1, 3, 224, 224) as it's expecting a list\n",
    "x4nas = np.expand_dims(x4nas, axis=0)\n",
    "x4nas = preprocess_input(x4nas)\n",
    "\n",
    "print(\"Image Shape: \" ,x.shape )\n",
    "print(\"Image 4 Nas Shape: \" ,x4nas.shape )\n",
    "# extract the features\n",
    "\n",
    "featuresmodelVGG19 = modelVGG19.predict(x)[0]\n",
    "print(\"VGG19: \", featuresmodelVGG19.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generate and save VGG19 based image descriptors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "imgFeatureSet : (5505, 512)\n"
     ]
    }
   ],
   "source": [
    "\n",
    "imgFeatureSet = np.zeros((len(pblImgSet), featuresmodelVGG19.shape[0])) \n",
    "print(\"imgFeatureSet :\", imgFeatureSet.shape)\n",
    "\n",
    "def getImgSetFeatures(model, imgRef, masterImgList):\n",
    "    pblImgFile = open(masterImgList, 'r') \n",
    "    pblImgSet =  pblImgFile.readlines() \n",
    "    print('Published Image quantity: ' + str(len(pblImgSet)))\n",
    "   \n",
    "    imgCounter = 0\n",
    "    for v in pblImgSet:\n",
    "        imgPath = imgRoot + v[:-2].replace(\"'\", \"\")\n",
    "        img = image.load_img(imgPath, target_size=(imgRef.shape[1], imgRef.shape[2]))\n",
    "        x = image.img_to_array(img)\n",
    "        x = np.expand_dims(x, axis=0)\n",
    "        x = preprocess_input(x)\n",
    "        imgFeatureSet[imgCounter] = model.predict(x)[0]\n",
    "        imgCounter = imgCounter + 1\n",
    "        if imgCounter % 10 == 0:\n",
    "            clear_output(wait=True)\n",
    "            print (\"Cost after epoch %i: %f\" % (imgCounter, 100*(imgCounter/len(pblImgSet))))\n",
    "            \n",
    "\n",
    "\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cost after epoch 5500: 99.909173\n"
     ]
    }
   ],
   "source": [
    "getImgSetFeatures(modelVGG19, x, 'mstrPublishedLst')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.savetxt(\"VGG19Chapinero_Ftrs.csv\", imgFeatureSet, delimiter=\",\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  7.00259149e-01   0.00000000e+00   0.00000000e+00   5.45773953e-02\n",
      "   1.22547541e+01   4.19673651e-01   0.00000000e+00   4.11977954e-02\n",
      "   1.14436257e+00   3.52895319e-01   0.00000000e+00   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   0.00000000e+00   8.62121284e-01\n",
      "   0.00000000e+00   3.70051920e-01   0.00000000e+00   0.00000000e+00\n",
      "   5.37817001e+00   1.47125988e+01   2.17260733e-01   0.00000000e+00\n",
      "   5.21877825e-01   0.00000000e+00   2.09709620e+00   7.66196728e-01\n",
      "   3.25401497e+00   0.00000000e+00   8.77244771e-01   0.00000000e+00\n",
      "   2.64648527e-01   0.00000000e+00   1.84404194e+00   0.00000000e+00\n",
      "   1.01462014e-01   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   3.76028232e-02   0.00000000e+00   4.91837084e-01   5.96052967e-02\n",
      "   0.00000000e+00   4.82262701e-01   1.37673492e+01   4.40700874e-02\n",
      "   0.00000000e+00   5.38695335e+00   1.14731455e+00   2.42062733e-02\n",
      "   3.11903596e-01   3.96084994e-01   4.99453098e-01   7.05833054e+00\n",
      "   4.86659706e-01   1.15747452e+00   1.36871529e+00   2.45962524e+00\n",
      "   0.00000000e+00   5.86073339e-01   1.56825376e+00   4.99573648e-02\n",
      "   0.00000000e+00   2.01496780e-01   0.00000000e+00   3.56611848e+00\n",
      "   0.00000000e+00   5.66488683e-01   4.46629107e-01   3.86745065e-01\n",
      "   3.69942963e-01   3.58801801e-03   0.00000000e+00   0.00000000e+00\n",
      "   2.07497549e+00   4.86298114e-01   4.21576738e-01   0.00000000e+00\n",
      "   0.00000000e+00   1.16182208e+00   0.00000000e+00   3.18474650e-01\n",
      "   0.00000000e+00   4.96441245e-01   1.00499010e+01   2.00459301e-01\n",
      "   3.35300374e+00   1.58853900e+00   0.00000000e+00   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   2.30619621e+00   1.45579171e+00\n",
      "   7.87888467e-02   0.00000000e+00   0.00000000e+00   2.67928505e+00\n",
      "   7.25281239e-01   3.09513491e-02   8.65531445e-01   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   0.00000000e+00   8.60612988e-01\n",
      "   2.11464897e-01   7.84139574e-01   2.13366675e+00   7.85624459e-02\n",
      "   2.40843073e-01   1.01048207e+00   5.52574813e-01   0.00000000e+00\n",
      "   5.68399131e-01   7.17207479e+00   8.86666107e+00   2.03969502e+00\n",
      "   1.18523037e+00   0.00000000e+00   0.00000000e+00   4.69633102e+00\n",
      "   5.41988134e+00   4.65848178e-01   0.00000000e+00   0.00000000e+00\n",
      "   1.98902115e-02   1.13113247e-01   1.90057015e+00   4.28455472e-01\n",
      "   5.81567764e-01   0.00000000e+00   1.82000361e-02   1.10748760e-01\n",
      "   0.00000000e+00   5.98516107e-01   6.05662286e-01   7.98757505e+00\n",
      "   0.00000000e+00   0.00000000e+00   0.00000000e+00   1.65007782e+00\n",
      "   1.21163130e+01   0.00000000e+00   1.35180044e+00   0.00000000e+00\n",
      "   2.49841118e+00   2.36170244e+00   2.57943049e-02   2.59226054e-01\n",
      "   0.00000000e+00   0.00000000e+00   4.56764065e-02   1.80798507e+00\n",
      "   6.85525715e-01   0.00000000e+00   0.00000000e+00   8.33438098e-01\n",
      "   8.09687555e-01   0.00000000e+00   9.16999221e-01   9.20538068e-01\n",
      "   0.00000000e+00   5.41279256e-01   9.07439888e-01   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   3.52110314e+00   0.00000000e+00\n",
      "   3.38846482e-02   5.46179152e+00   2.17170984e-01   0.00000000e+00\n",
      "   0.00000000e+00   5.48639745e-02   0.00000000e+00   2.31344756e-02\n",
      "   1.16430484e-01   5.52780814e-02   1.61010587e+00   0.00000000e+00\n",
      "   2.49581871e+01   0.00000000e+00   4.20102835e+00   9.37526941e-01\n",
      "   4.29873943e+00   0.00000000e+00   3.44504029e-01   3.41896558e+00\n",
      "   3.81961524e-01   0.00000000e+00   2.15627020e-03   0.00000000e+00\n",
      "   2.07274485e+00   0.00000000e+00   0.00000000e+00   4.26232368e-02\n",
      "   0.00000000e+00   2.89457083e-01   0.00000000e+00   0.00000000e+00\n",
      "   0.00000000e+00   3.35528076e-01   0.00000000e+00   3.54190683e+00\n",
      "   0.00000000e+00   0.00000000e+00   7.51472473e-01   1.34752047e+00\n",
      "   2.00618193e-01   1.83830246e-01   6.90610790e+00   0.00000000e+00\n",
      "   0.00000000e+00   1.27746388e-01   0.00000000e+00   4.78545688e-02\n",
      "   5.80086422e+00   3.81307006e-02   0.00000000e+00   0.00000000e+00\n",
      "   4.15533185e-01   3.82834002e-02   0.00000000e+00   1.85916889e-02\n",
      "   0.00000000e+00   0.00000000e+00   1.86290264e+00   1.00018036e+00\n",
      "   1.96923208e+00   0.00000000e+00   2.29881611e-02   3.36373389e-01\n",
      "   1.36122778e-01   1.20336008e+01   0.00000000e+00   3.83656956e-02\n",
      "   9.41156000e-02   0.00000000e+00   1.15955405e-01   0.00000000e+00\n",
      "   1.45194542e+00   0.00000000e+00   1.09553826e+00   3.26822400e-01\n",
      "   0.00000000e+00   8.39644223e-02   3.39149070e+00   1.47962108e-01\n",
      "   6.46758318e-01   3.48968744e+00   3.73138094e+00   0.00000000e+00\n",
      "   1.08126543e-01   2.22204164e-01   2.91486406e+00   8.63075829e+00\n",
      "   2.20358044e-01   3.72333717e+00   0.00000000e+00   0.00000000e+00\n",
      "   2.77646399e+00   0.00000000e+00   4.39179516e+00   1.26313663e+00\n",
      "   0.00000000e+00   4.92163785e-02   0.00000000e+00   5.69750145e-02\n",
      "   4.55486961e-02   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   6.17332220e+00   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   0.00000000e+00   3.64828140e-01   2.05830503e+00   0.00000000e+00\n",
      "   0.00000000e+00   1.96973145e-01   0.00000000e+00   0.00000000e+00\n",
      "   3.69293004e-01   0.00000000e+00   6.73971117e-01   1.58015680e+00\n",
      "   4.93827772e+00   7.88893104e-01   8.01086128e-01   0.00000000e+00\n",
      "   1.59272361e+00   1.28607568e-03   1.68940330e+00   1.61177993e-01\n",
      "   1.90746441e-01   0.00000000e+00   1.13728583e+00   1.69929489e-02\n",
      "   1.13685715e+00   1.49909675e-01   0.00000000e+00   1.11849713e+00\n",
      "   1.79708749e-01   0.00000000e+00   4.17500556e-01   2.21620500e-01\n",
      "   3.76966691e+00   0.00000000e+00   6.34786040e-02   8.56223488e+00\n",
      "   5.68963945e-01   5.87759260e-03   6.39164209e-01   2.73921546e-02\n",
      "   9.71412957e-01   2.22638547e-02   0.00000000e+00   0.00000000e+00\n",
      "   4.13596690e-01   0.00000000e+00   1.61248004e+00   6.98730722e-02\n",
      "   3.84202898e-02   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   4.38645154e-01   0.00000000e+00   1.37268343e+01   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   1.36903009e+01   0.00000000e+00\n",
      "   3.36623862e-02   0.00000000e+00   0.00000000e+00   2.62287647e-01\n",
      "   2.09221900e-01   7.54788160e-01   0.00000000e+00   3.61474514e+00\n",
      "   0.00000000e+00   0.00000000e+00   1.74828649e+00   0.00000000e+00\n",
      "   1.73015761e+00   7.15876579e-01   2.88391292e-01   0.00000000e+00\n",
      "   0.00000000e+00   3.49902898e-01   4.01899740e-02   0.00000000e+00\n",
      "   0.00000000e+00   0.00000000e+00   1.56776409e+01   8.04626405e-01\n",
      "   0.00000000e+00   0.00000000e+00   3.13824558e+00   1.50861338e-01\n",
      "   2.12083101e+00   1.83139861e+00   9.43666771e-02   0.00000000e+00\n",
      "   2.36492181e+00   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   0.00000000e+00   7.88269788e-02   4.63128299e-01   9.59807754e-01\n",
      "   3.11840385e-01   3.70445895e+00   7.70684063e-01   0.00000000e+00\n",
      "   0.00000000e+00   1.46785364e-01   1.46669626e-01   6.33662128e+00\n",
      "   8.23833227e-01   0.00000000e+00   8.55376339e+00   0.00000000e+00\n",
      "   6.18737862e-02   1.10124975e-01   0.00000000e+00   0.00000000e+00\n",
      "   7.25852489e-01   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   7.85065666e-02   1.67794275e+00   8.49358663e-02   5.51200449e-01\n",
      "   0.00000000e+00   7.34025538e-01   8.39729726e-01   8.38361323e-01\n",
      "   2.19024634e+00   1.67667174e+00   0.00000000e+00   1.50098872e+00\n",
      "   1.26662791e+00   0.00000000e+00   6.15434766e-01   0.00000000e+00\n",
      "   5.09925461e+00   1.48057818e+00   0.00000000e+00   0.00000000e+00\n",
      "   1.19209841e-01   1.49286389e-01   5.62584221e-01   2.81164289e-01\n",
      "   5.16058981e-01   4.64651972e-01   5.10688126e-01   6.75335407e-01\n",
      "   0.00000000e+00   0.00000000e+00   0.00000000e+00   0.00000000e+00\n",
      "   9.04238597e-02   9.37722623e-02   2.18002900e-01   1.86801827e+00\n",
      "   3.26472968e-01   0.00000000e+00   1.16763078e-01   1.11155903e+00\n",
      "   1.28714815e-01   1.53324723e+00   9.60183024e-01   0.00000000e+00\n",
      "   4.23219728e+00   1.70875740e+00   8.33695680e-02   4.25744690e-02\n",
      "   0.00000000e+00   0.00000000e+00   4.57765222e-01   0.00000000e+00\n",
      "   0.00000000e+00   1.77078247e-01   4.69831750e-02   8.42510909e-02\n",
      "   0.00000000e+00   0.00000000e+00   4.27400112e-01   0.00000000e+00\n",
      "   3.06674510e-01   1.88104999e+00   0.00000000e+00   3.56026936e+00\n",
      "   9.69865501e-01   3.09228015e+00   7.71299899e-02   0.00000000e+00\n",
      "   0.00000000e+00   1.24139094e+00   0.00000000e+00   0.00000000e+00\n",
      "   5.84616005e-01   0.00000000e+00   8.17960203e-02   2.64985800e-01\n",
      "   4.36798902e-03   4.11472559e+00   1.17274411e-01   8.21027160e-02\n",
      "   1.51263669e-01   1.17892705e-01   2.75810093e-01   1.15728843e+00\n",
      "   0.00000000e+00   0.00000000e+00   3.21380272e+01   3.36294127e+00\n",
      "   1.19298017e-02   7.20349371e-01   0.00000000e+00   1.94716528e-01\n",
      "   2.93605375e+00   4.54575837e-01   0.00000000e+00   0.00000000e+00\n",
      "   1.53188026e+00   1.21422720e+00   2.58964688e-01   0.00000000e+00\n",
      "   8.09365463e+00   0.00000000e+00   1.12052292e-01   3.20201740e-02\n",
      "   2.34832239e+00   0.00000000e+00   0.00000000e+00   9.58608687e-01\n",
      "   0.00000000e+00   0.00000000e+00   5.33151293e+00   4.81783009e+00]\n"
     ]
    }
   ],
   "source": [
    "print(imgFeatureSet[0,:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building training set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'genfromtxt' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-2-052cacfcbfaa>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mimageFeatures\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgenfromtxt\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'VGG16_Ftrs.csv'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdelimiter\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m','\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'genfromtxt' is not defined"
     ]
    }
   ],
   "source": [
    "imageFeatures = genfromtxt('VGG16_Ftrs.csv', delimiter=',')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5505, 512)\n"
     ]
    }
   ],
   "source": [
    "print(imageFeatures.shape)\n",
    "imgFtrStd = np.expand_dims(np.std(imageFeatures, axis=0), axis=0) \n",
    "imgFtrMean = np.expand_dims(np.mean(imageFeatures, axis=0) , axis=0) \n",
    "imgNrmFeatures = ((imageFeatures - imgFtrMean)/imgFtrStd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 1024)\n"
     ]
    }
   ],
   "source": [
    "concaFtr = np.expand_dims( np.concatenate((imgNrmFeatures[1], imgNrmFeatures[2]), axis=0), axis=0 )\n",
    "print(concaFtr.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(14115, 3)\n",
      "1.0\n",
      "Training percentage  90 %\n",
      "Training mini batch size 128\n",
      "Total amount of training votes 12703\n",
      "Total amount of test votes 1412\n",
      "Number of training mini batches 100\n",
      "Last batch size 31\n",
      "Number of test mini batches 12\n",
      "Last test batch size 4\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# Load vote indexer file\n",
    "votesIndex = genfromtxt('descriptorIndexerNrml_Nov_24.txt', delimiter=',')\n",
    "print(votesIndex.shape)\n",
    "print(votesIndex[4][2])\n",
    "\n",
    "# training percentage\n",
    "perTrain = 90\n",
    "print('Training percentage ' , perTrain,'%')\n",
    "# miniBatch Size\n",
    "minibatch_size = 128\n",
    "print('Training mini batch size' , minibatch_size)\n",
    "# Total amount of training votes\n",
    "ImgVtTrainSize = np.int(np.floor((votesIndex.shape[0]*perTrain)/100))\n",
    "print(\"Total amount of training votes\" , ImgVtTrainSize)\n",
    "# Total amount of test votes\n",
    "ImgVtTestSize = votesIndex.shape[0] - ImgVtTrainSize\n",
    "print(\"Total amount of test votes\" , ImgVtTestSize)\n",
    "num_minibatches = int(ImgVtTrainSize / minibatch_size)\n",
    "LastBatchSize = ImgVtTrainSize - num_minibatches*minibatch_size\n",
    "if LastBatchSize > 0:\n",
    "    num_minibatches = num_minibatches + 1\n",
    "print(\"Number of training mini batches\" , num_minibatches)\n",
    "print(\"Last batch size\", LastBatchSize)\n",
    "#--Test figures\n",
    "iniTestIdx = ImgVtTrainSize\n",
    "num_Testminibatches = int(ImgVtTestSize / minibatch_size)\n",
    "LastTestBatchSize = ImgVtTestSize - num_Testminibatches*minibatch_size\n",
    "if LastTestBatchSize > 0:\n",
    "    num_Testminibatches = num_Testminibatches + 1\n",
    "print(\"Number of test mini batches\" , num_Testminibatches)\n",
    "print(\"Last test batch size\", LastTestBatchSize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def getMiniBatch(minbIdx):\n",
    "    # minbIdx: miniBatch index\n",
    "    # training set permutation array\n",
    "    iniIdx = minbIdx * minibatch_size\n",
    "    endIdx = iniIdx + minibatch_size - 1\n",
    "    if endIdx > (ImgVtTrainSize - 1):\n",
    "        endIdx = ImgVtTrainSize - 1\n",
    "    vsize = (endIdx - iniIdx) + 1  \n",
    "    # build room\n",
    "    x_train_orig = np.zeros((vsize, imageFeatures.shape[1]*2)) \n",
    "    y_train_orig = np.zeros((vsize,1))\n",
    "    idxIte = 0\n",
    "    for i in range(iniIdx,(endIdx+1)):\n",
    "        imgAIdx = votesIndex[i][0] - 1\n",
    "        imgBIdx = votesIndex[i][1] - 1\n",
    "        voteCode = votesIndex[i][2]\n",
    "        \n",
    "        y_train_orig[idxIte,0] = voteCode\n",
    "        x_train_orig[idxIte] = np.expand_dims(np.concatenate((imgNrmFeatures[np.int(imgAIdx)], imgNrmFeatures[np.int(imgBIdx)]), axis=0), axis=0)[0]\n",
    "            \n",
    "        idxIte =  idxIte + 1\n",
    "\n",
    "    return  x_train_orig, y_train_orig     \n",
    "        \n",
    "    \n",
    "\n",
    "def getTestMiniBatch(minTbIdx):\n",
    "    # minbIdx: miniBatch index\n",
    "    # training set permutation array\n",
    "    minbIdx = int(minTbIdx%num_Testminibatches)\n",
    "    \n",
    "    iniIdx = minbIdx * minibatch_size\n",
    "    endIdx = iniIdx + minibatch_size - 1\n",
    "    if endIdx > (ImgVtTestSize - 1):\n",
    "        endIdx = ImgVtTestSize - 1\n",
    "    vsize = (endIdx - iniIdx) + 1  \n",
    "    # build room\n",
    "    x_train_orig = np.zeros((vsize, imageFeatures.shape[1]*2)) \n",
    "    y_train_orig = np.zeros((vsize,1))\n",
    "    idxIte = 0\n",
    "    for i in range(iniIdx,(endIdx+1)):\n",
    "        imgAIdx = votesIndex[i][0] - 1\n",
    "        imgBIdx = votesIndex[i][1] - 1\n",
    "        voteCode = votesIndex[i][2]\n",
    "        \n",
    "        y_train_orig[idxIte,0] = voteCode\n",
    "        x_train_orig[idxIte] = np.expand_dims(np.concatenate((imgNrmFeatures[np.int(imgAIdx)], imgNrmFeatures[np.int(imgBIdx)]), axis=0), axis=0)[0]\n",
    "            \n",
    "        idxIte =  idxIte + 1\n",
    "    \n",
    "    return  x_train_orig, y_train_orig"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_placeholders(n_x, n_y):\n",
    "    X = tf.placeholder(tf.float32, shape=(None,n_x), name = 'X') \n",
    "    Y = tf.placeholder(tf.float32, shape=(None, n_y), name = 'Y')\n",
    "    \n",
    "    return X,Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def forward_propagation(X):\n",
    "    \n",
    "    prtVlmSize = True\n",
    "    \n",
    "    Z4 = tf.contrib.layers.fully_connected(X, 1024)\n",
    "    Z5 = tf.contrib.layers.fully_connected(Z4, 512)\n",
    "    Z6 = tf.contrib.layers.fully_connected(Z5, 1, activation_fn=None)\n",
    "    \n",
    "    return Z6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def compute_cost(Z6, Y):\n",
    "    cost = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits = Z6, labels = Y))\n",
    "    \n",
    "    return cost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model(learning_rate = 0.0001, num_epochs = 50, print_cost = True, minibatch_size = minibatch_size):\n",
    "    \n",
    "    \n",
    "    ops.reset_default_graph()                         # to be able to rerun the model without overwriting tf variables\n",
    "    tf.set_random_seed(1)                             # to keep results consistent (tensorflow seed)\n",
    "    seed = 3                                          # to keep results consistent (numpy seed)\n",
    "                                        \n",
    "    costs = []                                        # To keep track of the cost\n",
    "    Testcosts = []\n",
    "    lastBestTestCost = 100.\n",
    "    # Create Placeholders of the correct shape\n",
    "    X, Y = create_placeholders(imageFeatures.shape[1]*2, 1)\n",
    "     \n",
    "    # Forward propagation: Build the forward propagation in the tensorflow graph\n",
    "    Z6 = forward_propagation(X)\n",
    "    \n",
    "    # Cost function: Add cost function to tensorflow graph\n",
    "    cost = compute_cost(Z6, Y)\n",
    "    \n",
    "    # Backpropagation: Define the tensorflow optimizer. Use an AdamOptimizer that minimizes the cost.\n",
    "    optimizer = tf.train.AdamOptimizer(learning_rate).minimize(cost)\n",
    "   \n",
    "    # SAVER ops\n",
    "    SIGMOID_op = tf.sigmoid(Z6, name = 'SIGMOID_op')\n",
    "    PREDICT_op = tf.round(SIGMOID_op, name = 'PREDICT_op')\n",
    "    OK_PREDICTED_op = tf.equal(PREDICT_op, Y, name = 'OK_PREDICTED_op')\n",
    "    ACCURACY_op = tf.reduce_mean(tf.cast(OK_PREDICTED_op, \"float\"), name = 'ACCURACY_op')\n",
    "    \n",
    "    # Initialize all the variables globally\n",
    "    init = tf.global_variables_initializer()\n",
    "     \n",
    "    # Add ops to save and restore all the variables.\n",
    "    saver = tf.train.Saver()\n",
    "\n",
    "    # Start the session to compute the tensorflow graph\n",
    "    with tf.Session() as sess:\n",
    "        \n",
    "        # Run the initialization\n",
    "        sess.run(init)\n",
    "        \n",
    "        # Do the training loop\n",
    "        for epoch in range(num_epochs):\n",
    "            minibatch_cost = 0.\n",
    "            minibatch_testCost = 0.\n",
    "            seed = seed + 1\n",
    "            np.random.seed(seed)\n",
    "            for it in range(0, num_minibatches):\n",
    "                X_train, Y_train = getMiniBatch(it)\n",
    "                _ , temp_cost = sess.run([optimizer, cost], feed_dict={X: X_train, Y: Y_train})\n",
    "                minibatch_cost += temp_cost / num_minibatches\n",
    "               \n",
    "            for tx in range(0, num_Testminibatches):\n",
    "                X_test, Y_test = getTestMiniBatch(tx)\n",
    "                temp_testcost = cost.eval({X: X_test, Y: Y_test})\n",
    "                minibatch_testCost += temp_testcost/num_Testminibatches\n",
    "                \n",
    "            # Print the cost every epoch\n",
    "            if print_cost == True and epoch % 1 == 0:\n",
    "                print (\"Cost after epoch %i: %f\" % (epoch, minibatch_cost))\n",
    "                print (\"Test Cost after epoch %i: %f\" % (epoch, minibatch_testCost))\n",
    "            # build cost list\n",
    "            if epoch > 1:\n",
    "                costs.append(minibatch_cost)\n",
    "                Testcosts.append(minibatch_testCost)\n",
    "            # update model if a better test cost is detected\n",
    "            if minibatch_testCost < lastBestTestCost:\n",
    "                lastBestTestCost = minibatch_testCost\n",
    "                # remove previuous best model\n",
    "                files = glob.glob(modelPath + '/*')\n",
    "                for f in files:\n",
    "                    os.remove(f)\n",
    "                # Save the variables to disk.\n",
    "                print(\"Saving a new best model\")\n",
    "                saver.save(sess, modelPath + '/model' + str(int(lastBestTestCost*100))) \n",
    "            \n",
    "        # plot the cost\n",
    "        plt.plot(np.squeeze(costs))\n",
    "        plt.ylabel('costs')\n",
    "        plt.xlabel('iterations (per tens)')\n",
    "        plt.title(\"Learning rate =\" + str(learning_rate))\n",
    "        plt.show()\n",
    "        \n",
    "        plt.plot(np.squeeze(Testcosts))\n",
    "        plt.ylabel('test cost')\n",
    "        plt.xlabel('iterations (per tens)')\n",
    "        plt.title(\"Learning rate =\" + str(learning_rate))\n",
    "        plt.show()\n",
    "        \n",
    "                \n",
    "        return costs,Testcosts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cost after epoch 0: 0.596810\n",
      "Test Cost after epoch 0: 0.726519\n",
      "Saving a new best model\n",
      "Cost after epoch 1: 0.511433\n",
      "Test Cost after epoch 1: 0.616777\n",
      "Saving a new best model\n",
      "Cost after epoch 2: 0.447840\n",
      "Test Cost after epoch 2: 0.536026\n",
      "Saving a new best model\n",
      "Cost after epoch 3: 0.394223\n",
      "Test Cost after epoch 3: 0.471511\n",
      "Saving a new best model\n",
      "Cost after epoch 4: 0.347194\n",
      "Test Cost after epoch 4: 0.414838\n",
      "Saving a new best model\n",
      "Cost after epoch 5: 0.304083\n",
      "Test Cost after epoch 5: 0.361082\n",
      "Saving a new best model\n",
      "Cost after epoch 6: 0.263699\n",
      "Test Cost after epoch 6: 0.308975\n",
      "Saving a new best model\n",
      "Cost after epoch 7: 0.226163\n",
      "Test Cost after epoch 7: 0.259226\n",
      "Saving a new best model\n",
      "Cost after epoch 8: 0.191588\n",
      "Test Cost after epoch 8: 0.210946\n",
      "Saving a new best model\n",
      "Cost after epoch 9: 0.160466\n",
      "Test Cost after epoch 9: 0.166384\n",
      "Saving a new best model\n",
      "Cost after epoch 10: 0.133168\n",
      "Test Cost after epoch 10: 0.127186\n",
      "Saving a new best model\n",
      "Cost after epoch 11: 0.110493\n",
      "Test Cost after epoch 11: 0.095746\n",
      "Saving a new best model\n",
      "Cost after epoch 12: 0.092456\n",
      "Test Cost after epoch 12: 0.074971\n",
      "Saving a new best model\n",
      "Cost after epoch 13: 0.078946\n",
      "Test Cost after epoch 13: 0.064746\n",
      "Saving a new best model\n",
      "Cost after epoch 14: 0.069683\n",
      "Test Cost after epoch 14: 0.061501\n",
      "Saving a new best model\n",
      "Cost after epoch 15: 0.063666\n",
      "Test Cost after epoch 15: 0.063330\n",
      "Cost after epoch 16: 0.059911\n",
      "Test Cost after epoch 16: 0.069021\n",
      "Cost after epoch 17: 0.055968\n",
      "Test Cost after epoch 17: 0.082340\n",
      "Cost after epoch 18: 0.051785\n",
      "Test Cost after epoch 18: 0.171004\n",
      "Cost after epoch 19: 0.041292\n",
      "Test Cost after epoch 19: 0.298543\n",
      "Cost after epoch 20: 0.033823\n",
      "Test Cost after epoch 20: 0.114870\n",
      "Cost after epoch 21: 0.022153\n",
      "Test Cost after epoch 21: 0.091156\n",
      "Cost after epoch 22: 0.015956\n",
      "Test Cost after epoch 22: 0.043005\n",
      "Saving a new best model\n",
      "Cost after epoch 23: 0.011900\n",
      "Test Cost after epoch 23: 0.023603\n",
      "Saving a new best model\n",
      "Cost after epoch 24: 0.009779\n",
      "Test Cost after epoch 24: 0.020147\n",
      "Saving a new best model\n",
      "Cost after epoch 25: 0.008507\n",
      "Test Cost after epoch 25: 0.018442\n",
      "Saving a new best model\n",
      "Cost after epoch 26: 0.008198\n",
      "Test Cost after epoch 26: 0.015527\n",
      "Saving a new best model\n",
      "Cost after epoch 27: 0.006567\n",
      "Test Cost after epoch 27: 0.013566\n",
      "Saving a new best model\n",
      "Cost after epoch 28: 0.006123\n",
      "Test Cost after epoch 28: 0.016219\n",
      "Cost after epoch 29: 0.004917\n",
      "Test Cost after epoch 29: 0.013456\n",
      "Saving a new best model\n",
      "Cost after epoch 30: 0.004127\n",
      "Test Cost after epoch 30: 0.017068\n",
      "Cost after epoch 31: 0.003676\n",
      "Test Cost after epoch 31: 0.015214\n",
      "Cost after epoch 32: 0.003253\n",
      "Test Cost after epoch 32: 0.012791\n",
      "Saving a new best model\n",
      "Cost after epoch 33: 0.002967\n",
      "Test Cost after epoch 33: 0.009276\n",
      "Saving a new best model\n",
      "Cost after epoch 34: 0.006433\n",
      "Test Cost after epoch 34: 0.006325\n",
      "Saving a new best model\n",
      "Cost after epoch 35: 0.002545\n",
      "Test Cost after epoch 35: 0.007691\n",
      "Cost after epoch 36: 0.005726\n",
      "Test Cost after epoch 36: 0.007861\n",
      "Cost after epoch 37: 0.005566\n",
      "Test Cost after epoch 37: 0.013778\n",
      "Cost after epoch 38: 0.005781\n",
      "Test Cost after epoch 38: 0.029095\n",
      "Cost after epoch 39: 0.006886\n",
      "Test Cost after epoch 39: 0.038155\n",
      "Cost after epoch 40: 0.005198\n",
      "Test Cost after epoch 40: 0.051792\n",
      "Cost after epoch 41: 0.008580\n",
      "Test Cost after epoch 41: 0.017938\n",
      "Cost after epoch 42: 0.004344\n",
      "Test Cost after epoch 42: 0.015410\n",
      "Cost after epoch 43: 0.006584\n",
      "Test Cost after epoch 43: 0.014600\n",
      "Cost after epoch 44: 0.006956\n",
      "Test Cost after epoch 44: 0.011863\n",
      "Cost after epoch 45: 0.002499\n",
      "Test Cost after epoch 45: 0.023557\n",
      "Cost after epoch 46: 0.008469\n",
      "Test Cost after epoch 46: 0.008501\n",
      "Cost after epoch 47: 0.003298\n",
      "Test Cost after epoch 47: 0.014860\n",
      "Cost after epoch 48: 0.004298\n",
      "Test Cost after epoch 48: 0.016243\n",
      "Cost after epoch 49: 0.004942\n",
      "Test Cost after epoch 49: 0.011522\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUiiIiEin/w9ao3AGjX4nvAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x136162c50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11fc36668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "costs [0.44783995360136009, 0.39422304823994636, 0.34719371393322934, 0.30408342368900759, 0.26369897119700919, 0.22616284400224684, 0.19158794168382876, 0.16046606458723545, 0.13316828344017267, 0.11049343304708602, 0.092456463910639261, 0.078946255128830634, 0.069683356545865507, 0.063666253145784141, 0.059910526480525736, 0.05596847699955107, 0.051785427532158763, 0.041291936049237844, 0.033822857763734629, 0.022152981688850548, 0.015955661874613728, 0.011900134106981573, 0.0097788889192452159, 0.0085067237094335734, 0.0081976068230869732, 0.0065674717416550231, 0.0061226799703581504, 0.0049165396198077339, 0.0041272669294994532, 0.0036757000582656497, 0.003252844064336386, 0.0029669134564755967, 0.0064325615453071821, 0.0025445660726109045, 0.0057256823850184452, 0.0055659772586295723, 0.0057813236699075805, 0.0068856962406425712, 0.0051981220922607471, 0.0085800147746340361, 0.0043442517001312806, 0.0065841147987521238, 0.0069555444340949173, 0.0024985501885385029, 0.0084689810570853265, 0.0032975695768254808, 0.0042979127617218177, 0.0049421129207257771]\n",
      "Testcosts [0.53602559616168333, 0.47151079277197522, 0.41483797381321585, 0.36108242720365524, 0.30897544945279753, 0.25922649105389917, 0.21094553420941034, 0.16638375942905745, 0.12718581594526768, 0.095745923618475601, 0.074970840476453304, 0.064745874454577759, 0.061500935194393001, 0.063329653503994138, 0.069020697226126984, 0.082339500077068806, 0.17100426989297074, 0.2985433905074994, 0.1148701071118315, 0.091156295345475286, 0.043004733122264348, 0.023602752403045695, 0.020147204748354856, 0.018441505962982774, 0.015526538481935859, 0.013565749550859135, 0.016219017289889354, 0.01345635155060639, 0.017068038112483919, 0.015213851157265406, 0.012791055060612658, 0.0092764104677674677, 0.0063251457080089795, 0.0076912988830978675, 0.0078609775518998504, 0.013778076274320483, 0.029095035686623302, 0.03815548195658873, 0.051791985286399715, 0.017938475270057094, 0.015410292495895796, 0.014600144563398015, 0.011863140884088352, 0.023556784261018041, 0.0085012348863529041, 0.014859505152950685, 0.016243409016169608, 0.011521942481825437]\n"
     ]
    }
   ],
   "source": [
    "costs,Testcosts = model()\n",
    "print('costs', costs)\n",
    "print('Testcosts', Testcosts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from ./transfer4uspVGG16/model0\n",
      "Tensor(\"X:0\", shape=(?, 1024), dtype=float32)\n",
      "Confusion matrix, without normalization\n",
      "[[ 99 146]\n",
      " [195 972]]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU8AAAEmCAYAAADiNhJgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x107ac6668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "saver4Pre = tf.train.import_meta_graph('./transfer4uspVGG16/model0.meta')\n",
    "with tf.Session() as sess:\n",
    "    # Step-2: Now let's load the weights saved using the restore method.\n",
    "    saver4Pre.restore(sess, tf.train.latest_checkpoint('./transfer4uspVGG16'))\n",
    "    # Accessing the default graph which we have restored\n",
    "    graph = tf.get_default_graph()\n",
    "    # Now, let's get hold of the op that we can be processed to get the output.\n",
    "    # In the original network y_pred is the tensor that is the prediction of the network\n",
    "    PREDICT_op = graph.get_tensor_by_name(\"PREDICT_op:0\")\n",
    "    X = graph.get_tensor_by_name(\"X:0\") \n",
    "    print(X)\n",
    "    Y = graph.get_tensor_by_name(\"Y:0\")\n",
    "    # for i in range(0, votesIndex.shape[0]):\n",
    "    X_val = np.zeros((1, imageFeatures.shape[1]*2)) \n",
    "    Y_val = np.zeros((1,1))\n",
    "    # for i in range(0, votesIndex.shape[0]):\n",
    "    prediction = []\n",
    "    actulLabel = []\n",
    "   \n",
    "    for i in range(12703, 14115):\n",
    "        # minus 1 because originall indexing list was done in MATLAB.\n",
    "        # MATLAB index stars at 1\n",
    "        imgAIdx = votesIndex[i][0] - 1\n",
    "        imgBIdx = votesIndex[i][1] - 1\n",
    "        voteCode = votesIndex[i][2]\n",
    "        \n",
    "        Y_val[0,0] = voteCode\n",
    "        X_val[0] = np.expand_dims(np.concatenate((imgNrmFeatures[np.int(imgAIdx)], imgNrmFeatures[np.int(imgBIdx)]), axis=0), axis=0)[0]\n",
    "        \n",
    "        result = sess.run(PREDICT_op, feed_dict={X: X_val, Y: Y_val})\n",
    "        prediction.append(result[0,0])\n",
    "        actulLabel.append(Y_val[0,0])\n",
    "        \n",
    "    \n",
    "    \n",
    "    \n",
    "    \n",
    "    # pout = np.concatenate((np.transpose(actulLabelArr), np.transpose(predictionArr)), axis=1)\n",
    "    # print(pout)\n",
    "    # Compute confusion matrix\n",
    "    cnf_matrix = confusion_matrix(np.asarray(actulLabel), np.asarray(prediction))\n",
    "    np.set_printoptions(precision=2)\n",
    "    # Plot non-normalized confusion matrix\n",
    "    plt.figure()\n",
    "    class_names = ['Zer0', 'One']\n",
    "    plot_confusion_matrix(cnf_matrix, classes=class_names,\n",
    "                      title='Confusion matrix, without normalization')\n",
    "    plt.show()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}