test_final_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "from model import FaceRec\n",
    "from sklearn.datasets import fetch_lfw_people\n",
    "from PIL import Image, ImageFilter\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "CKPT_LOCATION = 'checkpoints/c1s_9_c2n_64_rs_1_c2s_6_c2d_0.7_lr_0.0001_c2vl_32_c1n_256_c1vl_16_c1s_5_c1nf_16--FaceRec--1522700434.769402'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ModelBase::Loading ckpt ...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:The saved meta_graph is possibly from an older release:\n",
      "'model_variables' collection should be of type 'byte_list', but instead is of type 'node_list'.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "The saved meta_graph is possibly from an older release:\n",
      "'model_variables' collection should be of type 'byte_list', but instead is of type 'node_list'.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from checkpoints/c1s_9_c2n_64_rs_1_c2s_6_c2d_0.7_lr_0.0001_c2vl_32_c1n_256_c1vl_16_c1s_5_c1nf_16--FaceRec--1522700434.769402\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Restoring parameters from checkpoints/c1s_9_c2n_64_rs_1_c2s_6_c2d_0.7_lr_0.0001_c2vl_32_c1n_256_c1vl_16_c1s_5_c1nf_16--FaceRec--1522700434.769402\n",
      "ModelBase::Ckpt ready\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Error encountered when serializing model_variables.\n",
      "Type is unsupported, or the types of the items don't match field type in CollectionDef.\n",
      "'Tensor' object has no attribute 'to_proto'\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error encountered when serializing model_variables.\n",
      "Type is unsupported, or the types of the items don't match field type in CollectionDef.\n",
      "'Tensor' object has no attribute 'to_proto'\n"
     ]
    }
   ],
   "source": [
    "model = FaceRec(\"FaceRec\", output_folder='/tmp')\n",
    "model.load(CKPT_LOCATION)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "people = fetch_lfw_people(color=True, min_faces_per_person=25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def downsample_images(img):\n",
    "    img = Image.fromarray(img.astype('uint8'), 'RGB')\n",
    "    img = img.resize((32,32), Image.ANTIALIAS)\n",
    "    return np.array(img)\n",
    "\n",
    "def rotate_images(img, degree=180):\n",
    "    img = img * 255\n",
    "    img = Image.fromarray(img.astype('uint8'), 'RGB')\n",
    "    img = img.rotate(degree)\n",
    "    return np.array(img)/255\n",
    "\n",
    "def blur_image(img, sigma=0.1):\n",
    "    img = img * 255\n",
    "    img = Image.fromarray(img.astype('uint8'), 'RGB')\n",
    "    img = img.filter(ImageFilter.GaussianBlur(sigma))\n",
    "    return np.array(img)/255\n",
    "\n",
    "X_faces = people.images\n",
    "Y_faces = people.target\n",
    "target_names = people.target_names\n",
    "\n",
    "X_faces = np.array([downsample_images(ab) for ab in X_faces])\n",
    "X_faces = X_faces / 255\n",
    "X_faces_rotated = np.array([rotate_images(ab) for ab in X_faces])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_top_5_prediction(image):\n",
    "    softmax = model.predict(np.array([image]))\n",
    "    top_5 = softmax.argsort()[0][-5:][::-1]\n",
    "    top_5_names = [(target_names[i], softmax[0][i]) for i in top_5]\n",
    "    return top_5_names\n",
    "\n",
    "def print_top_5_prediction(image):\n",
    "    top_5_names = get_top_5_prediction(image)\n",
    "    print(\"\\nPredicted Names and softmax score: \")\n",
    "    for name in top_5_names:\n",
    "        print(\"%s: %s\" %(name[0], name[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x12e1c2470>"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12e590860>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "index = 144\n",
    "image = X_faces[index]\n",
    "plt.imshow(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Jennifer Capriati', 0.029991947),\n",
       " ('Laura Bush', 0.027784284),\n",
       " ('Guillermo Coria', 0.027079802),\n",
       " ('Gerhard Schroeder', 0.0268241),\n",
       " ('Arnold Schwarzenegger', 0.02634984)]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_top_5_prediction(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Guillermo Coria'"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target_names[Y_faces[index]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x12e2127f0>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAHwhJREFUeJztnVuMXNeVnv9V966+kH1js3mnNBxLHGtEC4TiwIrhzGAGijEYWUFg2A+GHozhIBgDMTB5EBwgdoA8eILYhp8c0JEwmsDxJSMbFgaeiyMY0VwysihFomRRtiSSktgku5tsNtm36qo6tfJQpYSk9r+7yGZXU9r/BxCs3qv2Ofvsc1adqv2ftZa5O4QQ6ZHb7AEIITYHOb8QiSLnFyJR5PxCJIqcX4hEkfMLkShyfiESRc4vRKLI+YVIlMJ6OpvZgwC+CSAP4L+6+1dj78/lil4olIO24a1bab+tW7cE20sFPvxGfYXaFi5fojZkTW7zVrDZYk9JRmzFIh9/sVjk24TdsM0iXcz4PcAiHR382NiTo7HnSXOFPLXV6g1qW66t8m3mw3N8+coi7ZPFzmd0PiLdYueMdIzNLzvPWdZAq9WMXSD/fws3+3ivmeUB/ArA7wA4A+A5AJ9191dZn1JpwCe23RO0Pfz7/5Lu6+Hf/2Swfc/2Ydrn/KmXqe1//eWT1JZdvkBt+Wb4IsuRdgDwOrdNbp+gth2Tk9RmuchntocduVjifUrFErXlIx9CjSZ3yHoj/CGa5fh12Tca/pAHgNenpqjthddOU1upfzTY/pc/+3vaZ36lTm0o8/nInH945Y33syzsgw2P3IjIB/alC6fQaKx05fzr+dp/P4A33P2ku9cBfA/AQ+vYnhCih6zH+XcCeOeqv8902oQQ7wPW9Zu/G8zsCIAjAJDP86+XQojesp47/xSA3Vf9vavTdg3uftTdD7v74VwutoglhOgl63H+5wAcMLP9ZlYC8BkAT92aYQkhNpqb/trv7k0z+wKAv0Zb6nvc3X8R65PLGaqV8Ff/vXv4csH2kbAMuDxznvb5h7/+C2pbnjlLbYVsidoMYalvaGCA9jlw8G5q27VjB7WF99RmaXmZ2szDC739/VXap1wOy69r0coyams0wyvVC3U+v81VLr/d86E7qK1cHaS2N9+aCbbfvX93sB0A/vEVrhQBfdxk/GdtRs4LwKXiSBeuOHa1zt9mXb/53f0nAH6ynm0IITYHPeEnRKLI+YVIFDm/EIki5xciUeT8QiTKhj/hd83O8jmMjoZlsV2TI7xjcz7Y/L//lst5F8/+ktomtnBJplrgkliBfFbu3s5lyrGRIWpDiwfGxDSbvhIff6VUCbeXuUSVy/N9ZZEox3qDj7/MIi5zfByXl7gMmC3wAKkP7d7D+zXCx9aKRDJeXORRn7+MBBhZJEqz2YpFi5Jm44FC7Pq4kTA93fmFSBQ5vxCJIucXIlHk/EIkipxfiETp7Wp/oYCx0XCQzugwD86YOvl6sP3UiRdpnx1EVQCAiRG+om8tnsJpYnx7sH2wj48dTb7+uhIJ0MkikT2tyJLu7PRseF8rNdqnGMmFyEN3gEZktZ9l67ICv98Uy2GlAgBKkfkoRCZrYih8rrNWOL0XAPyTe8Op5gDg3Ew4UAgA5ht8jgt5fs21yBq9R9KCtZgadAPL/brzC5Eocn4hEkXOL0SiyPmFSBQ5vxCJIucXIlF6HNiTx9hgWOobzPPMvm+eeivYXlzlUtOuvbziTX3lMrXFKtvUlsJSzpY+Xjkocy6WFSMBOvVlLhudn56mtlVSRSdWmKmvwnP4WZ7LTbG4EyYDxuTB1WUeUNOXcTl1tLiNb5NU3ym0+HnZv5PLgB/+9b3U9vcv8hSW+XzkPkuqLLGqPACQQ3jybyCFn+78QqSKnF+IRJHzC5Eocn4hEkXOL0SiyPmFSJR1SX1mdhrAAtrBX013Pxx7fytrYOlKWKZamj9D+517+5Vge3+R50XLZzznWy4S+jQ+NkZtM9PhiK5KpPpwOZI7r1bjY2w2uRSVo7WagC39YUmsEMkvVyxxmTUfifijEhUAy4WlqPnlK7RPY5GX61pa4NLn0ACfx9pKeJurkbmvRsqv3Xv3AWo7dYbn93tnjh+b58LRjHnnEaZGpL54obdruRU6/z939wu3YDtCiB6ir/1CJMp6nd8B/I2ZPW9mR27FgIQQvWG9X/sfcPcpM9sG4Kdm9pq7P3P1GzofCkcAoFqJlDcWQvSUdd353X2q8/8MgB8BuD/wnqPuftjdD5dLN1cHXghx67lp5zezfjMbfPc1gN8FEF6WF0Lcdqzna/8EgB9ZW3YqAPjv7v5XsQ6Neg3nzrwatB37R9713FsvBdvvipT4Kha4VNZscIkwZ7zfxEQ42mvqzFnaZ3goUoYs8tlba0Skvkg04NiOcDRjpY//5JqdnaO2i8u8hFahyBNuToyHJdPBCk9k+auTx6htbDwcDQoAWTNWCis8j4MVfunHIvBGKlwWfeDwIWp76m//gdoWV8OyY67AJV2n5b+6z+B5087v7icB3Huz/YUQm4ukPiESRc4vRKLI+YVIFDm/EIki5xciUXqawLOVNbA0F47qe/Wln9N+d2wLR1mNjm2hfQYGuAzVavGIrqVFntxzYmJHsL1a4fuavXiR2pDjstHrb4WTlgKAlfj+6r96Ldg+fYGPY3qa2xZrvJ5gJZL4cwt5mnNilCc7ZVGTANDfz4/5wK/vobZKXzj6bXiEj2P78BC1zc/z6MIDk/x6/OhdkcSfL4Sl7FaTRwJmGZH0PCJ7Xofu/EIkipxfiESR8wuRKHJ+IRJFzi9EovR0tb+vXMa9B+4I2g7sHKf9xgbCwyxHSmsVIyvRq3N8RXQ1kjtvO6lPZZEiScUiD8JZWuGlq3It/rnsDZ6nrVwMKwh3ToaVCgD4jTv2UVulygNxCiRPHwA0SbBKvcGVlt2TE9Rm4OdseXme2rwZXhVfuBTJPBcJ/Oqv9PN+Ta4EHNrDr+/c6r5g+8h2fs7GhsMBY0ef/Bnt8579dv1OIcQHCjm/EIki5xciUeT8QiSKnF+IRJHzC5EoPZb6SvjwnbuDtsECzz22dTAslxW40oRGxksdVQd4PrtWxqW+jAQEVfu5nHdpnufH2xGRcoYGucSWj8iHE9vCOfxiud2aDS45Li3zwJ65SzwgqDocDnIZGua5+GorK9SWAz8vtRoPxpq/PBtsP3nqFO2zPMePqxxO4wgAGCxH8gJu4dJz9eCHgu3DEztpnxbJ8ViOlVe7Dt35hUgUOb8QiSLnFyJR5PxCJIqcX4hEkfMLkShr6gJm9jiA3wMw4+4f7rSNAPg+gH0ATgP4tLtfWmtb+ZxhuEoi9HJcyqmUSDQdD6aDtbi0NbKV51rLaBkkoNEIy15jozznW2OVy1dzF89T29j4dmrLIseWNcPltZoZjwRcXY1FzHGpb3qaj3/uYljiHI7k8Bsb4TLgWEQiLOf4PewSyQs43B/OCwkA+3byua81eOTeciRicXiYH3erFT5np98M52MEQPM/xqIm37OJLt7zpwAevK7tUQBPu/sBAE93/hZCvI9Y0/nd/RkA13+MPwTgic7rJwB86haPSwixwdzsb/4Jdz/XeX0e7Yq9Qoj3Eete8HN3R+TZUTM7YmbHzOzY4kr3v0eEEBvLzTr/tJlNAkDnf1ptwd2Puvthdz880MefbxZC9Jabdf6nADzSef0IgB/fmuEIIXpFN1LfdwF8AsCYmZ0B8GUAXwXwAzP7PIC3AHy6m52ZAQXycVPt4+WYnMhUThJqAkAuYjOPaIQZt5XK4W8ulRKPwNu/dz+1eYuX5Lo4Ey5rBgCTkzzaa2UpLBst17hEtbjMbReIZAcArUji0rHxsWB7gV0AAJoNHolZb3Lb5UvhyD0AqK2G+9173yHaZ1sk2emlpQVqW57n0YClSELZXCV8zkqDXNIdGg/L1QUiiwffu9Yb3P2zxPTbXe9FCHHboSf8hEgUOb8QiSLnFyJR5PxCJIqcX4hE6WkCz3wuh4GBsCxmziPLMlI7rWzhyCYAMOMyyeLyIrXFEl0WSR28rBirncclzDvv4DLgyZMnqe31116ltv6hsATUyPhxzc1doba+fh6xuGtfOBkrAOwkdff6ylyKunyF19w7e3aK2mZnuMR2112/EWzfvXsf7YM8vycORepDztd4BORqnV/fhVJYBswTaRkAKv3hKEfLKYGnEGIN5PxCJIqcX4hEkfMLkShyfiESRc4vRKL0VOqDATmSbLFR4/XivBmWqQpVLvV5JOKMRXoBQF8k+qrVCicZddIOAIU8H0elxGsGHjx4kNqq1X5qmzoXTqp58O67aJ9cntf+Q4tfIgP9g9RWrYQlzmKBy6KliMR25RLPD3tgP5dMDx68O7yvKp/75cj1YXV+rvvKPLpzsR6O3GsTPu5inm/PLHydWuS6726vQogPPHJ+IRJFzi9Eosj5hUgUOb8QidLb1X53GqRTyPOV+0IxvLKZNfnKcd15mvCh2Cp1P18Fnr8UDjxpRVaASwWuHpRLPOhncJCv6P/mPbz008DQ28H2lTpfwS44H3+MepMrNPlG+L7ikRpr5y9eoLaVjO/rrnvuobaBreHApMx5oFMlx9WPWqRUWrXKg6CySMBNbTmsBORy/Nqp18PXfmR4791+928VQnyQkPMLkShyfiESRc4vRKLI+YVIFDm/EInSTbmuxwH8HoAZd/9wp+0rAP4AwLt1kr7k7j9Ze3eGvIV3WY0EWmSNcDmpfCSIoVrlQRF9VS6xrayuUNvAYFhiG4rkufNIabCZCzxnHVo8kKVaHaC2Pbv3Btvnr/A8fZcv8xJURgKxAKDA1VmU+sLGpUj+xIVFngNv545dfBx5LomdPRMue7ZU4+fZCjzP4Piu7dRWHebXQXmRz3HzYvhc1yJz5ei+LBejmzv/nwJ4MND+DXc/1PnXheMLIW4n1nR+d38GAK/WKIR4X7Ke3/xfMLPjZva4mfFHzoQQtyU36/zfAnAngEMAzgH4GnujmR0xs2NmduxKpBS0EKK33JTzu/u0u2fu3gLwbQD3R9571N0Pu/vhochCmxCit9yU85vZ5FV/PgzglVszHCFEr+hG6vsugE8AGDOzMwC+DOATZnYI7dpWpwH8YXe7MxiJ3svluG7UyrOoMx7ClEUOrVThUX2DW8eoLW9heWWOSDUAcHmO25YXuLTVynikXSMSobd9cluwfXLHDtpnIBLJOHuBR9q1WjxyslINR8ZdunyZby8i3WYtfp86/spr1GaF8DYHtnK5tBCJwDv19hlqK0XmcXySS4TFxfA8Zgv8Z3KRyJtm3d/P13R+d/9soPmxrvcghLgt0RN+QiSKnF+IRJHzC5Eocn4hEkXOL0Si9DSBpwNokMSJtSaXrxhFUhIKAAaGtlDb9h07qW3+Eo9+e4vJPB6JfOvjsuL24bAsBwDmPDnp5Utcfjs/E45iW1nlstzo6Ci19fXxObZIMs6Fy+F5bEWOqxyRypYj8ubgtnFq698SjrQrliMJNRt8rrDCowGXSFJNAMBFfl31V8PXaqnIj/nMmZlge73OE51ej+78QiSKnF+IRJHzC5Eocn4hEkXOL0SiyPmFSJTe1uoDwGLVohFdWVgeHB/kct7YtglqW17hEsrFOR51VmGJOiORVLHaacsReTPn4ZqGAIAiT97YN0SkxcjH/IVIjbxqP0+EOj7KIyBzJAKyVufHtbDCoxwHtvLahX393LZIJM4Tb5ykfUbHR6htz5491DYzw+fxzNnz1LadRFwODnKZuD4VlnQjJQjfg+78QiSKnF+IRJHzC5Eocn4hEkXOL0Si9Dywp0lW9Vcjy+J9xXA+uFhAzfw8D6S4eJHXIJm7yEtoDZPV7b7IquzPn3uO2uo1vrr9iX/2MWqrRrIg//JEeFU5FwnCaTV4MEj9PLeVizzv4t0fOhhsn4nM/cvHX6Y2j+xrbHKS2s5euBhsf+nECdpn3/791HZfjasVx55/gdoGI+XoFkmw0MAYV1Oq/eFrIJfj5/k97+36nUKIDxRyfiESRc4vRKLI+YVIFDm/EIki5xciUbop17UbwJ8BmEBbrTvq7t80sxEA3wewD+2SXZ92d16bCkC92cTUdDj32O7tPBCnMhAO3Gg1uQw1O8sDKWZmZqltNSLlzF0JB/2cnQvLSQDw6olXqa0QCWYqkjJTAFCO2PIkR95CpEzWxZnwOQGAvXv3Udv8PN/m9HQ48CSiOGKgj5fQWljisujs1Dneb3Ex2L5jlOdP9DovlfaL47ws5c7tvCTa0BAPPpq9FJ7/0hYuDzab4fnwSI7E6+nmzt8E8MfufhDARwH8kZkdBPAogKfd/QCApzt/CyHeJ6zp/O5+zt1f6LxeAHACwE4ADwF4ovO2JwB8aqMGKYS49dzQb34z2wfgIwCeBTDh7u9+3zqP9s8CIcT7hK6d38wGADwJ4Ivufs2zs+7uIPWyzeyImR0zs2NLK5F86EKIntKV85tZEW3H/467/7DTPG1mkx37JIDgqoW7H3X3w+5+uL8vXFNcCNF71nR+a5dleQzACXf/+lWmpwA80nn9CIAf3/rhCSE2im6i+j4G4HMAXjazFzttXwLwVQA/MLPPA3gLwKfX3FLL0ayHo/darUjUGYn4q63WaJ98JM9dqcQjxFZXeV690fFwlNXwBI++2rN7F7XVl/n4lxYXqK2/yqMIB0jJK4/M745JLlFNTmyntqGINLe0uBRsr5R4ROLHH/g4tV0m5b8AYC4iYy6uhiPmViPJ7kp9XJYbjpQ2y3JcZnv5xHFqmyay9MAYzyXY3xfOrZjPdb+Mt6bzu/vfAVSQ/u2u9ySEuK3QE35CJIqcX4hEkfMLkShyfiESRc4vRKL0NIFny4FVEjG1XOMS20qFJPAscTkvA4/OK/SR7QHIk/JOADBzLizJzF8JR44BQJVIMgAwPjJObXu28aSUzTqPZmw2wuPfu+sO2mfLEClDhniy0MEyf2irQc5nLJFosczPS2Ubn6vRES6/XVoMy4BXlnmU4DtneZTj22+/Q23LzbC8CQBnzp2htjJJ7jkQkRzzWfi+nbsBqU93fiESRc4vRKLI+YVIFDm/EIki5xciUeT8QiRKj6U+xxKRqZqRKKs8kZTyXOlD3rjUN9jPJZRcxsdRKYbHsUIi2ADg8twFapsl0iEAFHL84MoRSay/LywbDQ1yOW98ZJjaqkxmBdDK+Bw3GuFoOp4aEyiV+TFX+/n4mxmPprs4H57/C+d40s9GJNoSTS4FD5a49Hnw1w5Q2xCJFBwa4Me8OBvOlesRP7oe3fmFSBQ5vxCJIucXIlHk/EIkipxfiETp6Wp/M8swuzAftA3M8VXlnRPh1ehiiffps8ihNfmac3WYr3wX8+FtTo7xwJJaZHXYcrHp5wEwC5EyWbWl8Cr71kGeb28gEui0usxzCbYaPBiLBTRlkXJS9Xp47ADQjCgLIOcFAEbHtwbbW5HbXhaRJFgQDgDELrnFFa4ILTfCx1Zr8mNeImXIWhHl43p05xciUeT8QiSKnF+IRJHzC5Eocn4hEkXOL0SirCn1mdluAH+GdgluB3DU3b9pZl8B8AcAZjtv/ZK7/yS6rZyhRAJPLlwKByoAwKnTp4Ltd+3lpbAqkVJSiMhNaHHbajMclFSIlAbb2s/HUSjwsmHliIxZ2clLaOUQHotH5M2M5FUEgMYKDxQZHNhCbf0k999KJEeiR+TN2Fwt1nggTm0lLEdu3cLPS6PBcyQuLfN8jTnn99KC8euqSUqKkcp2AICsGT6uGwns6UbnbwL4Y3d/wcwGATxvZj/t2L7h7v+5670JIW4buqnVdw7Auc7rBTM7AWDnRg9MCLGx3NBvfjPbB+AjAJ7tNH3BzI6b2eNmxh+NE0LcdnTt/GY2AOBJAF909ysAvgXgTgCH0P5m8DXS74iZHTOzY7VIvnkhRG/pyvnNrIi243/H3X8IAO4+7e6Zu7cAfBvA/aG+7n7U3Q+7++FKiS/aCCF6y5rOb2YG4DEAJ9z961e1X11S5mEAr9z64QkhNopuVvs/BuBzAF42sxc7bV8C8FkzO4S2/HcawB+utSHL5VEuh/PnFTMuAS0vh6Ob5ud5pFTJuPxWynNJySMhXUxGK0Zyz8Wkw+WFK9RW40PElcg2WTRgmeQfBIByntuGBnm+w1Ikl+ASKYfVIBFsbfhBZwUuYbFoSwBY9fB1tXQlHF0KAMUy316rxWXF5UVuKxb5XFkrPCeNVT5XlUJ4e7FyaNfTzWr/3yF8VqKavhDi9kZP+AmRKHJ+IRJFzi9Eosj5hUgUOb8QidLTBJ7ecqzUwvJFiZTkAoBCPpwMsplFJLZINsUsUjSqUOD9iiThZi4iUXkkCWPJ+GdvPiJH5ou8n5HP80okSjCf5w9fNUj0GACsLnBpK0/mqlTm+2o2uZxXi0QD1iOJRPOl8DjKxdgDZ5GSbZEH1VrOxxGTOFmQadbgkm5/NSzB5nLd38915xciUeT8QiSKnF+IRJHzC5Eocn4hEkXOL0Si9Fbqc0eLJJKsDA3SfqVKOBlkISJfZZHINydRVACQj0hALDKrWeM15qwViRKMSH0Wic7K5fgYmWxHco8CALJIfTcm2QFApcLHUciHj60eqV1YjCTpjMmR+TyXfDMi2xUi59lbXOqLlQwkqiIAoJ5xWbTJtL5IJKCTqD7cQFSf7vxCJIqcX4hEkfMLkShyfiESRc4vRKLI+YVIlB5LfS006mFZrJgfof0qlXDEnzG5A4BF5J9GpLZbJR9LtBhuz0clO2qCRZKMRm2R09bKwmMpRJJcxqSymOSYRSSxJtEWc7lIJOYNRKRdTTGSSDSfhcfYioy9ZbF6d5HrqsXnqsEVX4BEkkYuASwsh/0ou4FafbrzC5Eocn4hEkXOL0SiyPmFSBQ5vxCJsuZqv5lVADwDoNx5/5+7+5fNbD+A7wEYBfA8gM+5R5KYoZ3HbGr6fHg/kQCYfC68gjm4ZYiPuxQOBgKASmR1PiOBRwBQr4cPrxDZXiynWqzMVKx0VT6SnzBP9hdb7c8iJco8FmBEgnfaNrK/yPY8F1MWIsvlsXGQwJ5GgwcYZRGFJhfJ4WdMDgKiEUElooBkKzxgbGr6XLC9EclneD3d3PlXAfyWu9+LdjnuB83sowD+BMA33P3XAFwC8Pmu9yqE2HTWdH5vs9j5s9j55wB+C8Cfd9qfAPCpDRmhEGJD6Oo3v5nlOxV6ZwD8FMCbAObd/d3vMmcA7NyYIQohNoKunN/dM3c/BGAXgPsB3NXtDszsiJkdM7Nj9chvSyFEb7mh1X53nwfwMwD/FMBWs/+38rQLwBTpc9TdD7v74VLkMVIhRG9Z0/nNbNzMtnZe9wH4HQAn0P4Q+Fedtz0C4McbNUghxK2nm8CeSQBPWDvSJAfgB+7+F2b2KoDvmdl/BPB/ADy21oaaWRNzV+aDNm9xiaKYD8s11T5e4muoupvatlbC5b8AIF/gye6YohT9RrMaSfoWUYZimdhiJcBaJBokIwEuQDz4yCOBIk2PSX3hccTkvFgwUyxcJRZ8tEoCjOosbx4QDTBajcx9MxIQlO/j0nO+GLZdnr1A+8wvLATbY/kYr2dN53f34wA+Emg/ifbvfyHE+xA94SdEosj5hUgUOb8QiSLnFyJR5PxCJIrFpJxbvjOzWQBvdf4cA8C1jN6hcVyLxnEt77dx7HX38W422FPnv2bHZsfc/fCm7Fzj0Dg0Dn3tFyJV5PxCJMpmOv/RTdz31Wgc16JxXMsHdhyb9ptfCLG56Gu/EImyKc5vZg+a2S/N7A0ze3QzxtAZx2kze9nMXjSzYz3c7+NmNmNmr1zVNmJmPzWz1zv/D2/SOL5iZlOdOXnRzD7Zg3HsNrOfmdmrZvYLM/s3nfaezklkHD2dEzOrmNnPzeylzjj+Q6d9v5k92/Gb75sZr1PWDe7e039oFzt7E8AdAEoAXgJwsNfj6IzlNICxTdjvxwHcB+CVq9r+E4BHO68fBfAnmzSOrwD4tz2ej0kA93VeDwL4FYCDvZ6TyDh6OidoR3QPdF4XATwL4KMAfgDgM532/wLgX69nP5tx578fwBvuftLbqb6/B+ChTRjHpuHuzwCYu675IbQToQI9SohKxtFz3P2cu7/Qeb2AdrKYnejxnETG0VO8zYYnzd0M598J4J2r/t7M5J8O4G/M7HkzO7JJY3iXCXd/Nxn7eQATmziWL5jZ8c7Pgg3/+XE1ZrYP7fwRz2IT5+S6cQA9npNeJM1NfcHvAXe/D8C/APBHZvbxzR4Q0P7kRzx5zUbyLQB3ol2j4RyAr/Vqx2Y2AOBJAF909ytX23o5J4Fx9HxOfB1Jc7tlM5x/CsDVObZo8s+Nxt2nOv/PAPgRNjcz0bSZTQJA5/+ZzRiEu093LrwWgG+jR3NiZkW0He477v7DTnPP5yQ0js2ak86+bzhpbrdshvM/B+BAZ+WyBOAzAJ7q9SDMrN/MBt99DeB3AbwS77WhPIV2IlRgExOivutsHR5GD+bE2kn4HgNwwt2/fpWpp3PCxtHrOelZ0txerWBet5r5SbRXUt8E8O82aQx3oK00vATgF70cB4Dvov31sYH2b7fPo13z8GkArwP4nwBGNmkc/w3AywCOo+18kz0YxwNof6U/DuDFzr9P9npOIuPo6ZwA+E20k+IeR/uD5t9fdc3+HMAbAP4HgPJ69qMn/IRIlNQX/IRIFjm/EIki5xciUeT8QiSKnF+IRJHzC5Eocn4hEkXOL0Si/F/uhfbBQ9hLEAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12e166e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "image_rotated = rotate_images(image)\n",
    "plt.imshow(image_rotated)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1302cf7b8>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12e4f9e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "image_blurred = blur_image(image,3)\n",
    "plt.imshow(image_blurred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "from data_handler import get_face_data\n",
    "\n",
    "x_train, y_train, x_test, y_test = get_face_data()\n",
    "x_test = x_test / 255\n",
    "x_train = x_train / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_loss_acc_for_rotated_images(degree):\n",
    "    x_rotated_test = np.array([rotate_images(ab, degree) for ab in x_test])\n",
    "    loss, acc, _ = model.evaluate_dataset(x_rotated_test, y_test)\n",
    "    print(\"Rotated Test dataset (%s degrees), loss = %s, acc = %s\" % (degree, loss, acc))\n",
    "    return loss, acc\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_loss_acc_for_blurred_images(sigma):\n",
    "    x_blurred_test = np.array([blur_image(ab, sigma) for ab in x_test])\n",
    "    loss, acc, _ = model.evaluate_dataset(x_blurred_test, y_test)\n",
    "    print(\"Blurred Test dataset (%s sigma), loss = %s, acc = %s\" % (sigma, loss, acc))\n",
    "    return loss, acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Rotated Test dataset (180 degrees), loss = 0.8253456, acc = 0.052941176470588235\n"
     ]
    }
   ],
   "source": [
    "loss, acc = get_loss_acc_for_rotated_images(180)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Rotated Test dataset (10 degrees), loss = 0.2792917, acc = 0.7803921568627451\n"
     ]
    }
   ],
   "source": [
    "loss, acc = get_loss_acc_for_rotated_images(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Blurred Test dataset (5 sigma), loss = 0.6148513, acc = 0.1588235294117647\n"
     ]
    }
   ],
   "source": [
    "loss, acc = get_loss_acc_for_blurred_images(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Rotated Test dataset (5 degrees), loss = 0.1536024, acc = 0.9313725490196079\n",
      "Rotated Test dataset (10 degrees), loss = 0.2792917, acc = 0.7803921568627451\n",
      "Rotated Test dataset (30 degrees), loss = 0.73775244, acc = 0.17647058823529413\n",
      "Rotated Test dataset (45 degrees), loss = 0.8269824, acc = 0.054901960784313725\n",
      "Rotated Test dataset (90 degrees), loss = 0.7646246, acc = 0.049019607843137254\n",
      "Rotated Test dataset (115 degrees), loss = 0.79466015, acc = 0.041176470588235294\n",
      "Rotated Test dataset (180 degrees), loss = 0.8253456, acc = 0.052941176470588235\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbfb1980128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "degrees = [5, 10, 30, 45, 90, 115, 180]\n",
    "total_acc = []\n",
    "total_loss = []\n",
    "\n",
    "for degree in degrees:\n",
    "    loss, acc = get_loss_acc_for_rotated_images(degree)\n",
    "    total_loss.append(loss)\n",
    "    total_acc.append(acc)\n",
    "    \n",
    "plt.figure(figsize=(10,8))\n",
    "plt.title('Degrees Rotated vs. Accuracy')\n",
    "plt.xlabel('Degrees Rotated')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.plot(degrees, total_acc, label='Accuracy')\n",
    "plt.plot(degrees, total_loss, label='Loss')\n",
    "plt.legend(loc='best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Blurred Test dataset (0.1 sigma), loss = 0.08788459, acc = 0.9372549019607843\n",
      "Blurred Test dataset (0.5 sigma), loss = 0.0929313, acc = 0.9372549019607843\n",
      "Blurred Test dataset (1 sigma), loss = 0.1345572, acc = 0.9313725490196079\n",
      "Blurred Test dataset (1.5 sigma), loss = 0.22201838, acc = 0.8627450980392157\n",
      "Blurred Test dataset (2 sigma), loss = 0.31600922, acc = 0.703921568627451\n",
      "Blurred Test dataset (2.5 sigma), loss = 0.40148932, acc = 0.5254901960784314\n",
      "Blurred Test dataset (3 sigma), loss = 0.46428537, acc = 0.39019607843137255\n",
      "Blurred Test dataset (4 sigma), loss = 0.55547607, acc = 0.24705882352941178\n",
      "Blurred Test dataset (5 sigma), loss = 0.6148513, acc = 0.1588235294117647\n",
      "Blurred Test dataset (6 sigma), loss = 0.6564907, acc = 0.11176470588235295\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12e5efe48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sigmas = [0.1, 0.5, 1, 1.5, 2, 2.5, 3, 4, 5, 6]\n",
    "total_acc = []\n",
    "total_loss = []\n",
    "\n",
    "for sigma in sigmas:\n",
    "    loss, acc = get_loss_acc_for_blurred_images(sigma)\n",
    "    total_loss.append(loss)\n",
    "    total_acc.append(acc)\n",
    "    \n",
    "plt.figure(figsize=(10,8))\n",
    "plt.title('Gaussian Blur vs. Accuracy')\n",
    "plt.xlabel('Sigma')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.plot(sigmas, total_acc, label='Accuracy')\n",
    "plt.plot(sigmas, total_loss, label='Loss')\n",
    "plt.legend(loc='best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sigma = 0.100000\n",
      "Sigma = 0.500000\n",
      "Sigma = 1.000000\n",
      "Sigma = 1.500000\n",
      "Sigma = 2.000000\n",
      "Sigma = 2.500000\n",
      "Sigma = 3.000000\n",
      "Sigma = 4.000000\n",
      "Sigma = 5.000000\n",
      "Sigma = 6.000000\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12e004518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for sigma in sigmas:\n",
    "    image_blurred = blur_image(image,sigma)\n",
    "    print(\"Sigma = %f\" % sigma)\n",
    "    plt.imshow(image_blurred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "STREAM b'IHDR' 16 13\n",
      "STREAM b'IDAT' 41 8192\n"
     ]
    }
   ],
   "source": [
    "img = Image.open('george_bush.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "image = img.convert('RGB').resize((32, 32), Image.ANTIALIAS)\n",
    "image = np.array(image) / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}