Ejercicio_CNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Convolutional Neural Networks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Realizaremos un ejercicio descrito en el blog www.aprendemachinelearning.com para hacer reconocimiento de imágenes."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Importar Librerías"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:12:59.231970Z",
     "start_time": "2018-11-08T00:12:51.800950Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/jbagnato/anaconda3/envs/python36/lib/python3.6/site-packages/sklearn/cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "import re\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:13:12.550878Z",
     "start_time": "2018-11-08T00:12:59.234748Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import Sequential,Input,Model\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.advanced_activations import LeakyReLU"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Cargar set de Imágenes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:45.248080Z",
     "start_time": "2018-11-08T00:13:12.553292Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "leyendo imagenes de  /Users/jbagnato/github-folders/fastai/sportimages/\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/golf 1\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/basket 9768\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/tenis 8823\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/natacion 8937\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/ciclismo 5172\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/beisball 7533\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/futbol 7752\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/americano 7617\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/f1 9348\n",
      "/Users/jbagnato/github-folders/fastai/sportimages/boxeo 5053\n",
      "Directorios leidos: 10\n",
      "Imagenes en cada directorio [9769, 8823, 8937, 5172, 7533, 7752, 7617, 9348, 5053, 7124]\n",
      "suma Total de imagenes en subdirs: 77128\n"
     ]
    }
   ],
   "source": [
    "dirname = os.path.join(os.getcwd(), 'sportimages')\n",
    "imgpath = dirname + os.sep \n",
    "\n",
    "images = []\n",
    "directories = []\n",
    "dircount = []\n",
    "prevRoot=''\n",
    "cant=0\n",
    "\n",
    "print(\"leyendo imagenes de \",imgpath)\n",
    "\n",
    "for root, dirnames, filenames in os.walk(imgpath):\n",
    "    for filename in filenames:\n",
    "        if re.search(\"\\.(jpg|jpeg|png|bmp|tiff)$\", filename):\n",
    "            cant=cant+1\n",
    "            filepath = os.path.join(root, filename)\n",
    "            image = plt.imread(filepath)\n",
    "            images.append(image)\n",
    "            b = \"Leyendo...\" + str(cant)\n",
    "            print (b, end=\"\\r\")\n",
    "            if prevRoot !=root:\n",
    "                print(root, cant)\n",
    "                prevRoot=root\n",
    "                directories.append(root)\n",
    "                dircount.append(cant)\n",
    "                cant=0\n",
    "dircount.append(cant)\n",
    "\n",
    "dircount = dircount[1:]\n",
    "dircount[0]=dircount[0]+1\n",
    "print('Directorios leidos:',len(directories))\n",
    "print(\"Imagenes en cada directorio\", dircount)\n",
    "print('suma Total de imagenes en subdirs:',sum(dircount))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creamos las etiquetas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:45.269861Z",
     "start_time": "2018-11-08T00:16:45.251786Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cantidad etiquetas creadas:  77128\n"
     ]
    }
   ],
   "source": [
    "labels=[]\n",
    "indice=0\n",
    "for cantidad in dircount:\n",
    "    for i in range(cantidad):\n",
    "        labels.append(indice)\n",
    "    indice=indice+1\n",
    "print(\"Cantidad etiquetas creadas: \",len(labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:45.285925Z",
     "start_time": "2018-11-08T00:16:45.273489Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 golf\n",
      "1 basket\n",
      "2 tenis\n",
      "3 natacion\n",
      "4 ciclismo\n",
      "5 beisball\n",
      "6 futbol\n",
      "7 americano\n",
      "8 f1\n",
      "9 boxeo\n"
     ]
    }
   ],
   "source": [
    "deportes=[]\n",
    "indice=0\n",
    "for directorio in directories:\n",
    "    name = directorio.split(os.sep)\n",
    "    print(indice , name[len(name)-1])\n",
    "    deportes.append(name[len(name)-1])\n",
    "    indice=indice+1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:45.498672Z",
     "start_time": "2018-11-08T00:16:45.290061Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of outputs :  10\n",
      "Output classes :  [0 1 2 3 4 5 6 7 8 9]\n"
     ]
    }
   ],
   "source": [
    "y = np.array(labels)\n",
    "X = np.array(images, dtype=np.uint8) #convierto de lista a numpy\n",
    "\n",
    "# Find the unique numbers from the train labels\n",
    "classes = np.unique(y)\n",
    "nClasses = len(classes)\n",
    "print('Total number of outputs : ', nClasses)\n",
    "print('Output classes : ', classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creamos Sets de Entrenamiento y Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:45.669596Z",
     "start_time": "2018-11-08T00:16:45.502716Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training data shape :  (61702, 21, 28, 3) (61702,)\n",
      "Testing data shape :  (15426, 21, 28, 3) (15426,)\n"
     ]
    }
   ],
   "source": [
    "train_X,test_X,train_Y,test_Y = train_test_split(X,y,test_size=0.2)\n",
    "print('Training data shape : ', train_X.shape, train_Y.shape)\n",
    "print('Testing data shape : ', test_X.shape, test_Y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:49.319746Z",
     "start_time": "2018-11-08T00:16:45.673944Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Ground Truth : 9')"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAATwAAACOCAYAAACscLTMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAH8hJREFUeJztnXuUZFd13r99b1V1VXf1dE1190z39Mz0zAihpyVkTSQgIIxeCJwEWEY2sOzgxESL2GaFFRwj23kQYhxsVuL4EZKwHFkQsDCLhAgTWIYosMTLoAeSRqORRtI8e6af0+9HPe/OH1Uytc8+M139qu6eu39r9eo+t/e999xz9z11a599vkPMDMMwjDgQbHYFDMMwWoV1eIZhxAbr8AzDiA3W4RmGERuswzMMIzZYh2cYRmywDm+dIKJTRHTnJp5/iIh+ZrPOb2ws5l/rw7bp8Ijo3UT0QyJaIKKx+t+/SkS02XW7FET0dSKar/+UiajUUP6vqzzm54joo+tcVfcc/4KIRohohoj+jIhSG3m+zcb8SxxzQ/2LiNJE9EdEdJ6IpojoT4gosVHna2RbdHhE9GEAfwTgkwD6AOwG8AEAfxeA90EkorBlFbwEzPxWZs4ycxbA5wH8wStlZv6Aa9+qG38piOhnAXwYwJsBHARwFYB/vamV2kDMv1rO7wC4EcB1qPnWawH8VkvOzMxb+gdAF4AFAD+3jN2DAP4LgK/V7e+s7/tZAOMATgP4lwCCuv1HAXyuYf8DABhAol7+NoB/B+B7AOYAfANAT4P9L9WPeQG1G3gKwJ1N1PF3nW131vf9bQAjAP4cwPsBfLvBJlGv2wEAvwqgDKAEYB7Al+s2QwD+OYAjAGYAPASgbZVt/kUAH2sovwXA0Gb7gvnXZeNfTwF4Z0P5HwI42Yr7vR3e8F4HoA3Aw03YvhfAxwF0AvgugD9BzSkPAXgTag37j1Zw7vfW7Xeh9kn/GwBARNei5vy/BGAPgG4Ae1dwXJe9ALIA9qPmcBeFmT8F4C8B/B7XPsXf2fDvnwdwF2rXe3O9fgoiOkhE00S05yKnuQ7A0w3lpwEMEFFXMxezzTD/aqBF/kX1n8byASLKNnMxa2E7dHg9ACaYufLKBiL6fr1Bl4jotgbbh5n5e8wcofYp9QsAfouZ55j5FID/gIvcpIvw58x8nJmXUHvreU19+7sAfJWZH2XmIoB/BSBa9RUCFQAfZeZS/Vyr5T8x8wgzXwDw1Yb6Cpj5JDPnmPn8RY6TRe1T/BVe+btzDXXbqph/Nc96+dfXAXyIiHqIqB/AB+vbM2uoW1Nshw7vAoCextgDM7+emXP1/zVew9mGv3tQ+9Q83bDtNICBFZx7pOHvRdQ6AqD2qfu352LmhXpdVssoM5fWsP8rXKy+K2UewI6G8o6G7Zcb5l/Ns17+9TEAR1H75vBdAF8GUAAwsabaNcF26PB+AKAI4O1N2DZKv0yg9ik82LBtP4Bz9b8XALQ3/K9vBXUaBrDvlQIRtaP2tWO1uJI1y9VtoyVujqIWVH6FGwGcY+bpDT7vZmD+1WL/YuZFZv6nzDzAzFcAmALwONcDehvJlu/w6g/ZvwXwKSJ6FxFliSggotcA6LjEflXUviZ8nIg6iWgQtaDr5+omTwG4jYj212NTKxkl+hKAv0dEb6ina3wM69uWTwO4gYh+iogyAP6N8/9R1OIoG8VnAfwTIrqaiPKoBeMf3MDzbRrmX633LyLaS0T99XZ+PWqDMh/dqPM1suU7PABg5j9AzZl+E8AYajfkvwH4CIDvX2LXD6L2aXYCtVfnvwDwQP2Y30QtOPsMgCdQi0k0W5+jAH6tfrxh1D6hhlZyTcsc/zkAv4faSN4LAB51TP4MwI31HKYvrfT4RHSonqflDSoz81cB/GH9vKcAvIjaQ3dZYv7VWv8CcCWAv0EtRPIAgN9g5kdWep7VQC14izQMw9gSbIs3PMMwjPXAOjzDMGKDdXiGYcSGNXV4RHQPEb1ARC8R0f3rVSnDAMy/jPVn1YMW9cnTx1GbajIE4DEA76mPABnGmjD/MjaCtSgn3ALgJWY+AQBE9AXUkjcv6pCZTJI7O9POVvcl06fGEyxrAeLlbZx8yqb6+sBzJHdHn0kkbQKfypDn/KzsfJVcvuK6ivr8rvJRKtXmOc7yM5rc2lSrep+RkQsTzNy77MF+wor9KwgCDgPpK247+FpOtY3nmpMJKY4SVaueAzm+7Lu/7PqgNko456pGuj7ufn4VK33s0LlWX3uEqaQoF4vFZQ/t+6qYTMjuJRFqgZmK045BqLukRJsUrCkW9aSRpeJSU/61lg5vAHKqzRCAWy+1Q2dnGu+697DYRo76DsHtEAGCfBADT+uG5DQcaSchp1MsVbTTsuO0iZRuolrOaeOB9bnKzk1pC7TKELG+kMhRHYrI92A5N5y025ZLsk7JQHdmCUfi7uDBg8pmqbAgyoHPaZ3TT0/NKpt///ufPa02XpoV+1cYBMh3yum+Raf5qpHuGBKBfMBRWlQ2/d07RXlxdkbZhKFsY450W7kPa7VcVjb5vNRomHfuAQAslWQn1JZIKpvI01F2Ow9P2fMwde3ZLcovvHxC2bgul4n0cfbme0R5Z+cOZTM9PyfK6VxO2fQOHhDlE6fPKJsfv/B0U/61lg6vqY8UIroPwH0AkM3qh84wLsKK/cv7Fm0YDaxl0GIIDfP9UJOgUeoIzPxpZj7MzIczGf0pZBgXYcX+Ffhe/Q2jgbW84T0G4EoiOojahOl3o6bvdQkChMEyMTxPTINIvr6TJ/IQOXEXX+jJFalNJbUaTdWJKXBVn6tYKohywtOKqYS8rkpZxx1SCc8br/vMeq6jWnXeZHxf8UP54eKLjbjvQ0tLWjmo7MbjKrpC84uyPZLJdflgW7F/MUcoR9JXIuf7dme7Vrga3CMFTubnppRNCOkXlG1XNlWnbYpF/XV1oSS/7vflddhpfmZSnsvz4trnfPWbWfAI2Xg+AMpt8t4slnR8rjA6KsodKR1myiSduNrcnLIpOM/J1Jx+loKEvLipC2PKJpGW52pPrv6DbdUdHjNXiOjXAfw1gBDAA/U5gIaxZsy/jI1gTfr2zPw11CSvDWPdMf8y1hsLehiGERtavIIRgSFjCKQSxjypIqg4ZX82lTTypZPI/SpLOq4WOmkXHOn6tDt5SiUnVgEA5aqM3wSetJTCvI6fcEbul/AFCJ1YZODmfwEqVSaCvo5SSV7/yNiwsnFTTjzNgaobP3XTdloFAeT4z8AumRrRHup41E3XvlqUi0WdBnL0+WOiPDOpBYjzORmPm5/VAr69XTL2lknr1JXFOekDu5z0DkDHXxc9j0TBEzdezEo/ZM9zUi7I5y3pSeXJOOlaYVrHw8uRvA72pIEF8lQ4eEjL8EXsxN7D1b+n2RueYRixwTo8wzBig3V4hmHEBuvwDMOIDS0etABYBUDdRFYd2KRg+RngavK0Z9I/NSNCEMjzJzwfCe7gR0i6GUNnIn5U1cHpdEYnHhcDmXQahp7rcK6NWdtU3MnlkQ5gB0m5X5n14IuqN+nrcMdVCoX1WBFw5QQgpJ3BoXxWzt8MPInThwblGtfznkGoG26+SZSfffqYsvnZu+4R5a89/H+UzdHjUvvATc4FgH379olye0IPeHFJRvt7PQnM4xN60OSUMyCT9PiOO++7zfOkpJ25u26iPQC0d8o1kDo79ZpIQ8682ExK27z5Z+4U5Weee0HZ4MnH9DYP9oZnGEZssA7PMIzYYB2eYRixoeUxPHJjdE6CrE/IMHB0xbxih462nCsUAACRG/xLeDTzHJvQMxG+7OQLtyX1hPS33P0PRHn/Hq0191df+bradnr8b0SZq3oCOkjWySfSGTjJ2hT6hCZlO/qSitlJ5PWJOLrabGFqc2SawiBEPi215FCW172woAUSpublth899WNlUyrLdrjjjW9WNrf8ndeL8rmXtW5buSrvy/i0jrPlcvIarjt0pbLJOn554vnnlU1/u46HdTnP25xHu5BKjq94Eobz+bwot3dqMYVMh0zynvDEFHfkpPYehVoz7x33/mNRvv70OWXzqf/9ebXNh73hGYYRG6zDMwwjNliHZxhGbLAOzzCM2NDSQQsiIFAx7+X7XHZs+CJrkkkjX3ayDBgHbkIzgKIToE2QDsaWyzLwm/YEWsOqDDxfcfCnlc0Hfvkqte3jf/x+UZ5f0gvKsJMczZ5FhIKE22baxl3kpeKOxkArVfhU1APn2EsFzypXLSCgAJk2eb+cMQIkO7Jqv/EpqdbbsWOXsjn5nEwYfuGEHpAouj7nWVhnZ6c8f1TRg1KDA4OiPD+nFwxacPzip67Qg2JX33FAbTs2LZWRP//QF5TNdEEmQ1NSJz7PVWVy+ey0HgxaOifPVVzSSdbpNvmctGW1MszJUZksPV3UijfNYm94hmHEBuvwDMOIDdbhGYYRG1qbeMxA4CQIL7+uvS/RePkYnhIc8OBbqNhd7WtkdFLZvPycXC3w19//YWVz2+1vkxsWKsomndCxiG4nEbNUHFE2ZThKspG+VnZXNvN8tukwp8cdWNbbt/qaez+Ki5sTw4uiCIvOCmodTkJssajrtm+PnKyfzep41KGDMt564003KJu5sjx3lNT+tX9ArpDWRTrON5iTSb1nzp1SNtOT46Lc1qMXsC5NaQXrmwevFuXjV79K2ZyakjHDeY9ax2JJ+uDSoo41V6qyHXd4VnpLO9t69/Ypm1kn9see56ZZ7A3PMIzYYB2eYRixwTo8wzBig3V4hmHEhparpURVGQCNyFHm9Q02kKuE4lFUcbZpTQ+AnKznkrsGIYBkQqoQu0m1AHB+SAaD//D3P6lsnvvO46J8x+ter2yixXm1jZ1bUnIzZ6FFh9mj7hw4RlWPEkrkqP968pcRhrI+ZY9yxuKiHMhIJLSScytgZpScRN6BnEwK7+jQCiLj56T6xtFnjyubZLtUxLn7nruUTdlRqDk/fFbZ9JNsm8GsTlrnMbkEZHlkXNmUpqdE+cTzR5RN5gqtsrLkOMI1hwaUTdu0vNauvQeUTdVR4p6e0HUMHZXtakknWXd1y0G6vkGdQD0x8rI8TlUvCdks9oZnGEZssA7PMIzYYB2eYRixwTo8wzBiQ4vVUgJk0jJLmp1Bi4pnmUY3mu4ukwgA5Cw3R2o5SADOzIo2j3x7JZKB53y3Vtd4y1veKMrVKT37oD2Qmeez53UgnBd0dvoIyVkcruoJ4Fkp0SPxTq6siWdWSdK9fjU7Qw8i6WU2gXJJZtTncjrrvxVQECCVlpn7587JAaadO/T9vPVGqWQTXnutsjk57Mx4Ye2nI+fluZ59+hll8/yYnMVweFDPdFiYkbLrQ+Na0jxok/ehu0sPfkwv6QGvJUd55ct/pZcZKDmKM7l+PWPj+ZdeEuW5Wa3ocv3VctBkbFwfpyMrB0iuePVpZZNISkWVJ584qmyaxd7wDMOIDdbhGYYRG6zDMwwjNrQ48ThC1VFKdUJviAKPMq8TsnMVV3wH8gkew1F9qJR1ImTBUQNJt+lYTbJN1nFxUSddBl0yVpRircBRLuol8jizvJpxxWkQ1jLScENt7EnoTqTkfpWSJ8nZUUJZWtLXkW2XMZ/Qs5RjK6hWGbNzUlkjFcprWgqlujEAtGVkMnB3X17ZHLj+elEeG9VLDnampTJwh2f5zmJC1i/do9WVewcPyA3nupTNE0eeFuVcqG0mE/o6smmZtLvnmpuUTcFR9K54nrc7brtblI8dfU7ZTE1Oyzru0EoormLR2ZM6hteWknWeHdc2zWJveIZhxAbr8AzDiA3W4RmGERuW7fCI6AEiGiOiZxu25Ynom0T0Yv33zo2tpnE5Yz5mtIpmBi0eBPCnAD7bsO1+AI8w8yeI6P56+SPLH4pB5A4UyKB44NE5IaeaKqkWgJO/DPLIUpOrIOJRQsmkncTUqj5OpzMgMVrRS/bNzEjFi6kOHcDubtOqD0nnjkR6HAFhKJO3CxVtVHaUUMKEvtUVlveiCj2IU3YUZcoVLZG+Myfbw11GsgkexDr4GEeMwpI8d0dWDkgU5vSgxbf+3zdEOXJ9AMDeK2UycuW4TiS/57Y3OfXRie3IyAThxbSWK8/tlgMZvR3aT/isTOLl3YeUzWnWkuoD7VKdhPM6+b0wPibKSc/AWcEZ4LrhRj34cd5JxD5zRj8ng4Oy3ouLC8pm4oJM+q7Qiv3rb1n2DY+ZHwXgLuzwdgCfqf/9GQDvWHUNjNhjPma0itWmpexm5mEAYOZhItJj63WI6D4A9wFAZ+fqF98wYkdTPtboX4GFpI1l2HAPYeZPM/NhZj7cntErmBvGWmj0L+vwjOVY7RveKBH11z95+wGMLbsHakrFCWcyPDtOWlUz4wF24npEzTi2/p7Pzn4BPMq8ThJvIvR00iUZd+zdpRMqx0dkYupcvlvZDOzSk+x3Z6Ui79mzOsHVVUWOdOgNcJSK2TPZfd5ZWq/dE0+am5bfNH2xQErI9qhWfEs5rpgV+xiDUXXimQvTMia0K639qyOSfnHjzbcom+8fOSbKVU+bF+dlUvGCUwaAPmdJyOEFbXPqmWdFeedOPV6T6ZFKxaMlfV2nPEuMjjvh3olRPel/bkTG3gbyOqY5OiqTf2+55bCy2XugX5THxkaVzdCQFMuY88RYx6ecZyC1+g+21e75FQDvq//9PgAPr7oGhuHHfMxYd5pJS3kIwA8AXEVEQ0T0KwA+AeAuInoRwF31smGsCvMxo1Us+5WWmd9zkX/dsc51MWKK+ZjRKizKaxhGbGj5Mo1QSwo6fa5vlUa4isee9QQdqj7lZOfYZZ+4slOuVLQ6SOgsN5cMdDJuqkvWcYnPK5uO3n617a23XiPKX/zS15TNxLQMNIcJrXZLkMmqkUepuL1NJkNXynqwoVyWUe7sTp1AHTkN66pYt4p8Vw4/98a3i23BnAzcT5+WSr0A0FmVCcK70/oar3aWKjzkWQKxPSWP86pXvVrZLDnteWFyStlkks5A3qK+LzsyciDhwqIe/Mjs7FXbzp49Jcp7erXKyuyIrGMqrZ+3u2+Xy1Rec+1VyiZyFLRD0gnyY6Py+n/0+FPKpup0EWHCk9DdJPaGZxhGbLAOzzCM2GAdnmEYsaGlMbyIgWLFje/I7/XsSiD7j+TZIo/rW7TMTXKOVF2AhHOcNiV2AKQSMjly39U6OXnvTpmMvHhBx/Dy+7RS8o6sjBm+9+f1ClonzsiE4R8+OaRslpy41OyivtZkKJOcl+Z1vLItJW3aPIIHZTfzOdycz9EoqqKwIGN2uVBet2/eeXtSTrL/0Xce18d2Vjubyemk3rMnXhblc8N6snyhIh3TVfwFgK5umWgcekKi+/tl/Hd+TCcQF0taGCCbkufry3Uom2vvul2UuaDjjO7KgeeH9Mpqs3My6btnd4+ySWbk+QcmLyibxIQ8f8mTRI9p/Xz5sDc8wzBig3V4hmHEBuvwDMOIDdbhGYYRG1o6aMFgf8CxgXD5nGItbwyA4S7TqPtyN2E54YlgJ6oy+HvL4QPKJpd21DVe7VGzILkEY3lJJ3juHNAJy0ielDY5vd/AwH5R7t2l1W5PnJXt8egPTiqbalkOtkxPaaWK7l1S5SXyfUayo97iCcS3gtm5GXzjWzJRO+v4RY8nqTiZl1J7p4f1IFDvoGzzHx95Rtlks7I9Jxf0QMLQsBzsmPQkHu/rlQnD+3drpZ0oOCXKJyfmlc2h17xWbXv++BFRTrEe2Lj9F39RlIfP6sGXp554QpR3dunk96//X6kkne7QqitBRir0jHvao+IuG1lZvX/ZG55hGLHBOjzDMGKDdXiGYcQG6/AMw4gNLVdLcWdAuBMifEobATlLOXpmY0Tuco/kuTRnt0xSn2t8WEpXz07qYPDt90gVjCS/rGyy/c6MhFkdwAZ5tqUdJfOiXmqPIQc7dvcNKJuevutE+ZnndCD+tCMfn/FIvLclpAx+taoVL9y2L1c3Z9AiCAJ0pGW7pxyp+ytvuFnvl5IDGb1dekZAbpfc9uRTP1A2O5xFqhaKWsEkSMuBjfyePcomzMg2nyt7BrfcfTxKJBNn9UBV4AzUdWS0fz3y7e+J8szErLIZOil9Jww9szECOdjStVv7acURPlnwnCuRlG224BkMahZ7wzMMIzZYh2cYRmywDs8wjNjQ0hgeMSNRkjGEKsuYi7skIwAkIZMjg4SOjZSS8jhLSR2bIJYxlsqcVisJSjKocOyIVsjd2y8/JzrT+jiHCvL82Xbd1FGgr3V26lZR3r3bkxxdlsoUOyp6+btMt4wDXdGn2+yJZ06J8q791ymbImQycshaUcWNqIaBZ/nLFrArn8cH3/kLYtvMuLw3+U6dyE1t8l6dmdQxs5KjDn3r6+5WNm1O7C2qetR4nNgylbSaceAoT+/K66TeoZERUW7v26dsRuf0PZ+aeZMo5zq06ktUeFGU+3q0D/b2yGTtMNytbApFqYQyPL6gbMqhbLMDg/paF4sy/tyZ0zH86WNarcWHveEZhhEbrMMzDCM2WIdnGEZssA7PMIzY0Fq1FAI4kEmpSSduH/rkwZ2cyijy2IQy8NzRtV+Z7OiUKhR//Zf/Q9ncdMUBUZ6c0wHbbzwiA8Z9PTo4ffRZmZg5MqwTKkOd54uDB+UAxL33vkHv1y0DxDznkbd22iwM9dJ2He1SvYI9uvhuIrhPgp+dJTFd+e9WMTk9jYceflhsq8zKdt+R0hL1Bw7JpTG79l+jbM6el9Lj56Z0om1/n7wvB/Z6EsJzMoG5PakHeKYvSN85PTKmbM6OyIG8dNGjMpLSA3fFipTj9y1n2tEhE33PDz2vbGbm5X79/TrZnCkvyt09etnIAsmHINXpUTmCHMjoyd+kbI4dO6q2+bA3PMMwYoN1eIZhxAbr8AzDiA0tjeEFREilZHyg4sQQqpGeBA3IuIM7WR0A5hbktuk5PcG4VJEJpbldVymbKCGDiuyJfY1OymTcxXkdv9i/R8bHMl06wRShjp8kEzI+GIY6DoOk/Jyam9UJpidOypjGqZNazTiTlkm4QaCvNXLEAnyLaLohu8izjGYrKFQqOD4mY6CdTl3SpJfUTDlJ6omEVuadLU6LcjnQCbIIZHsWF3RrnRmTCbILC1pxeMbZNlf0KWPLpN6pCe1LZdL7Fcsy1jc7ptWM2+m4PFWo489VR9DjxNkjysaN4fXu1ontmR17RTlI6vuTScu4Jy+jmn4p7A3PMIzYYB2eYRixwTo8wzBig3V4hmHEhtYqHjOAqgx2Vp2IN7EetAggbQLyqXHIhNJSVQc/l4oygN3vUWAlyADt1KIOPJ8/JYPKiUgn2p48IQcJcnqlPfT06yUDo4IMav/F576kbHLtso1mLygTLBZkMLpYOKBsslkZeJ8ulJVNGLrX5gsYR05pcwYtEASopB2VmjbpB+0Jne0dkvzcPz+s1W9KkRwk6Nt7UNnksvLcC1M6Ybg8I/0r8gwUtbVLZ1lK6jaPnKTiqKrfXZKRVuPJtMvryLd7FKwXh2V90vr8c4tyoGx6QSc+Hzz0KlH+6cM6YXhmSd6PXHe/srnuenmcky89p2yaxd7wDMOIDdbhGYYRG5bt8IhoHxF9i4iOEdFRIvpn9e15IvomEb1Y/71z46trXG6YfxmtpJkYXgXAh5n5SSLqBPAEEX0TwC8DeISZP0FE9wO4H8BHLnUgBoEjGbMglvGeauRJ0K3IOINPFdnVE0indJwvlZbnrpZ18KvCUpW1J68VcucnZfxibEhP3p+5IONAU7NaKfjMiI6NDDkhpsgTDmMnn7TTI0KwZ69ss91X6uu4MCsbrVzW9QkCGcP0hCsVK4zhrZt/deVyuOetf19s66nKuOT5x55U+81MS9Xf0Yq+yPQ+Ge/ds+eQshnIyfjYqYkRZVOal7HdgqdBi23y3s15FKQTTmg5l8srG/KIQVyYln45zzpu2+YkFRdL2mbHThn/7cjpeHQuJ30u0abj6tmEtHn5xClls1SQbTY7rWOszbLsGx4zDzPzk/W/5wAcAzAA4O0APlM3+wyAd6y6FkZsMf8yWsmKYnhEdADATQB+CGA3Mw8DNacFsOviexrG8ph/GRtN0x0eEWUB/E8AH2JmPbnu4vvdR0SPE9HjS0uer6uGgfXxr0JBzyk2jEaa6vCIKImaM36emf9XffMoEfXX/98PQCcdAWDmTzPzYWY+nMno7/CGsV7+lU57gpmG0QAtp05LRIRaDGWSmT/UsP2TAC40BJXzzPybyxxrHMBpAD0AJi5luwXZjnUGtk69B5lZSd6afwm2Y723Sp29/uXSTIf3BgDfAXAEP0mp/23U4ixfBLAfwBkA9zKzXuTSf8zHmflwM7Zbhe1YZ2Dr19v86ydsx3pvtzovm5bCzN+FXwYNAO5Y3+oYccP8y2glNtPCMIzYsFkd3qc36bxrYTvWGdi+9V4L2/Wat2O9t1Wdl43hGYZhXC7YV1rDMGJDyzs8IrqHiF4gopfq6QZbDiJ6gIjGiOjZhm1bejK7TcKvYf61MVwu/tXSDo+IQgD/GcBbAVwL4D1EdG0r69AkDwK4x9l2P2qT2a8E8Ei9vJV4ZRL+NQBeC+DX6m271eu9bph/bSiXhX+1+g3vFgAvMfMJZi4B+AJqk8S3FMz8KAA352tLT2a3SfgAzL82jMvFv1rd4Q0AONtQHqpv2w5sm8nsMZ6Eb/7VArazf7W6w/Ou49ziOlzWrHYS/mWC+dcGs939q9Ud3hCAfQ3lvQC0eubWpKnJ7JvJWibhXyaYf20gl4N/tbrDewzAlUR0kIhSAN4N4CstrsNq+QqA99X/fh+AhzexLor6JPz/DuAYM//Hhn9t6XqvM+ZfG8Rl41/M3NIfAG8DcBzAywB+p9Xnb7KODwEYBlBG7a3hVwB0ozYK9WL9d36z6+nU+Q2ofX17BsBT9Z+3bfV6m3+Zf7Xyx2ZaGIYRG2ymhWEYscE6PMMwYoN1eIZhxAbr8AzDiA3W4RmGERuswzMMIzZYh2cYRmywDs8wjNjw/wHbeHyBl/zxHwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 360x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=[5,5])\n",
    "\n",
    "# Display the first image in training data\n",
    "plt.subplot(121)\n",
    "plt.imshow(train_X[0,:,:], cmap='gray')\n",
    "plt.title(\"Ground Truth : {}\".format(train_Y[0]))\n",
    "\n",
    "# Display the first image in testing data\n",
    "plt.subplot(122)\n",
    "plt.imshow(test_X[0,:,:], cmap='gray')\n",
    "plt.title(\"Ground Truth : {}\".format(test_Y[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preprocesamos las imagenes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:50.798162Z",
     "start_time": "2018-11-08T00:16:49.322999Z"
    }
   },
   "outputs": [],
   "source": [
    "train_X = train_X.astype('float32')\n",
    "test_X = test_X.astype('float32')\n",
    "train_X = train_X / 255.\n",
    "test_X = test_X / 255."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hacemos el One-hot Encoding para la red"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:50.815482Z",
     "start_time": "2018-11-08T00:16:50.800831Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original label: 0\n",
      "After conversion to one-hot: [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n"
     ]
    }
   ],
   "source": [
    "# Change the labels from categorical to one-hot encoding\n",
    "train_Y_one_hot = to_categorical(train_Y)\n",
    "test_Y_one_hot = to_categorical(test_Y)\n",
    "\n",
    "# Display the change for category label using one-hot encoding\n",
    "print('Original label:', train_Y[0])\n",
    "print('After conversion to one-hot:', train_Y_one_hot[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creamos el Set de Entrenamiento y Validación"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.218520Z",
     "start_time": "2018-11-08T00:16:50.818992Z"
    }
   },
   "outputs": [],
   "source": [
    "#Mezclar todo y crear los grupos de entrenamiento y testing\n",
    "train_X,valid_X,train_label,valid_label = train_test_split(train_X, train_Y_one_hot, test_size=0.2, random_state=13)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.228116Z",
     "start_time": "2018-11-08T00:16:51.222460Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(49361, 21, 28, 3) (12341, 21, 28, 3) (49361, 10) (12341, 10)\n"
     ]
    }
   ],
   "source": [
    "print(train_X.shape,valid_X.shape,train_label.shape,valid_label.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creamos el modelo de CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.244776Z",
     "start_time": "2018-11-08T00:16:51.238704Z"
    }
   },
   "outputs": [],
   "source": [
    "#declaramos variables con los parámetros de configuración de la red\n",
    "INIT_LR = 1e-3 # Valor inicial de learning rate. El valor 1e-3 corresponde con 0.001\n",
    "epochs = 6 # Cantidad de iteraciones completas al conjunto de imagenes de entrenamiento\n",
    "batch_size = 64 # cantidad de imágenes que se toman a la vez en memoria"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.384131Z",
     "start_time": "2018-11-08T00:16:51.252188Z"
    }
   },
   "outputs": [],
   "source": [
    "sport_model = Sequential()\n",
    "sport_model.add(Conv2D(32, kernel_size=(3, 3),activation='linear',padding='same',input_shape=(21,28,3)))\n",
    "sport_model.add(LeakyReLU(alpha=0.1))\n",
    "sport_model.add(MaxPooling2D((2, 2),padding='same'))\n",
    "sport_model.add(Dropout(0.5))\n",
    "\n",
    "sport_model.add(Flatten())\n",
    "sport_model.add(Dense(32, activation='linear'))\n",
    "sport_model.add(LeakyReLU(alpha=0.1))\n",
    "sport_model.add(Dropout(0.5))\n",
    "sport_model.add(Dense(nClasses, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.401674Z",
     "start_time": "2018-11-08T00:16:51.386676Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 21, 28, 32)        896       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_1 (LeakyReLU)    (None, 21, 28, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 11, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 11, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 4928)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                157728    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_2 (LeakyReLU)    (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 158,954\n",
      "Trainable params: 158,954\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "sport_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:16:51.472349Z",
     "start_time": "2018-11-08T00:16:51.406817Z"
    }
   },
   "outputs": [],
   "source": [
    "sport_model.compile(loss=keras.losses.categorical_crossentropy, optimizer=keras.optimizers.Adagrad(lr=INIT_LR, decay=INIT_LR / 100),metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Entrenamos el modelo: Aprende a clasificar imágenes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:49.562522Z",
     "start_time": "2018-11-08T00:16:51.474807Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Variable *= will be deprecated. Use `var.assign(var * other)` if you want assignment to the variable value or `x = x * y` if you want a new python Tensor object.\n",
      "Train on 49361 samples, validate on 12341 samples\n",
      "Epoch 1/6\n",
      "49361/49361 [==============================] - 40s 814us/step - loss: 1.5198 - acc: 0.4897 - val_loss: 1.0611 - val_acc: 0.7136\n",
      "Epoch 2/6\n",
      "49361/49361 [==============================] - 38s 775us/step - loss: 1.2002 - acc: 0.6063 - val_loss: 0.8987 - val_acc: 0.7717\n",
      "Epoch 3/6\n",
      "49361/49361 [==============================] - 43s 864us/step - loss: 1.0886 - acc: 0.6469 - val_loss: 0.8078 - val_acc: 0.7977\n",
      "Epoch 4/6\n",
      "49361/49361 [==============================] - 41s 832us/step - loss: 1.0166 - acc: 0.6720 - val_loss: 0.7512 - val_acc: 0.8180\n",
      "Epoch 5/6\n",
      "49361/49361 [==============================] - 36s 725us/step - loss: 0.9647 - acc: 0.6894 - val_loss: 0.7033 - val_acc: 0.8323\n",
      "Epoch 6/6\n",
      "49361/49361 [==============================] - 40s 802us/step - loss: 0.9258 - acc: 0.7032 - val_loss: 0.6717 - val_acc: 0.8379\n"
     ]
    }
   ],
   "source": [
    "# este paso puede tomar varios minutos, dependiendo de tu ordenador, cpu y memoria ram libre\n",
    "# como ejemplo, en mi Macbook pro tarda 4 minutos\n",
    "sport_train = sport_model.fit(train_X, train_label, batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(valid_X, valid_label))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:49.676566Z",
     "start_time": "2018-11-08T00:20:49.566203Z"
    }
   },
   "outputs": [],
   "source": [
    "# guardamos la red, para reutilizarla en el futuro, sin tener que volver a entrenar\n",
    "sport_model.save(\"sports_mnist.h5py\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluamos la red"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:54.462929Z",
     "start_time": "2018-11-08T00:20:49.678643Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15426/15426 [==============================] - 5s 310us/step\n"
     ]
    }
   ],
   "source": [
    "test_eval = sport_model.evaluate(test_X, test_Y_one_hot, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:54.474683Z",
     "start_time": "2018-11-08T00:20:54.465378Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.6687967825782881\n",
      "Test accuracy: 0.8409179307662388\n"
     ]
    }
   ],
   "source": [
    "print('Test loss:', test_eval[0])\n",
    "print('Test accuracy:', test_eval[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:55.014647Z",
     "start_time": "2018-11-08T00:20:54.479693Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "accuracy = sport_train.history['acc']\n",
    "val_accuracy = sport_train.history['val_acc']\n",
    "loss = sport_train.history['loss']\n",
    "val_loss = sport_train.history['val_loss']\n",
    "epochs = range(len(accuracy))\n",
    "plt.plot(epochs, accuracy, 'bo', label='Training accuracy')\n",
    "plt.plot(epochs, val_accuracy, 'b', label='Validation accuracy')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "plt.figure()\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:58.050602Z",
     "start_time": "2018-11-08T00:20:55.021862Z"
    }
   },
   "outputs": [],
   "source": [
    "predicted_classes2 = sport_model.predict(test_X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:58.262575Z",
     "start_time": "2018-11-08T00:20:58.052878Z"
    }
   },
   "outputs": [],
   "source": [
    "predicted_classes=[]\n",
    "for predicted_sport in predicted_classes2:\n",
    "    predicted_classes.append(predicted_sport.tolist().index(max(predicted_sport)))\n",
    "predicted_classes=np.array(predicted_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:58.272559Z",
     "start_time": "2018-11-08T00:20:58.264703Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((15426,), (15426,))"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predicted_classes.shape, test_Y.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Aprendamos de los errores: Qué mejorar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:20:59.822110Z",
     "start_time": "2018-11-08T00:20:58.275464Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 12972 correct labels\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "correct = np.where(predicted_classes==test_Y)[0]\n",
    "print(\"Found %d correct labels\" % len(correct))\n",
    "for i, correct in enumerate(correct[0:9]):\n",
    "    plt.subplot(3,3,i+1)\n",
    "    plt.imshow(test_X[correct].reshape(21,28,3), cmap='gray', interpolation='none')\n",
    "    plt.title(\"{}, {}\".format(deportes[predicted_classes[correct]],\n",
    "                                                    deportes[test_Y[correct]]))\n",
    "\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:21:00.942267Z",
     "start_time": "2018-11-08T00:20:59.829572Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2454 incorrect labels\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEYCAYAAAC3LjroAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzsvXeYZVlZL/z77X1Spc5hOk3qmSGKiCQVL3MFBEa5oFcuoIMzZD7hA+/HRRDRCzrIiAF8DBfxIiAZBCVKEBwyM+Q4TO6ZTtOxcjhh7/X98b5rr/eE6q6u6q5TVb1+z1NPnbPX3muvs9e713rzS+ccIiIiIiIiLJJ+DyAiIiIiYuUhbg4REREREV2Im0NERERERBfi5hARERER0YW4OUREREREdCFuDhERERERXTjjzYHkPpKPPReDWcsg+e8kr+n3OM43kPxFkrf0exynAsnrSB4neW+/x7IaQdKRvEw/D5D8GMlxkh9cwLU3kHzuIu+7oHfajm8R9yjWW5KvIfmuxfSzGJSW60arESRfA+Ay59zVS+3LOffEpY8o4kzhnPsSgPv0exzzgeQeAC8DcJFz7qgeewuARwO4HMCznXNv798IVx1+A8B2AJudc61zeaO1/k5HtVLEmgXJ1cD8XATghN8YFN8D8DsAvt2fIa1qXATg1nO9MZwPWOzm8DCSPyY5SvJtJGu+geTzSN5O8iTJj5Lcqcd/XkXnPfr9p0mOkbyvft9J8kMkj5G8i+RLTJ9Vkm8ieUj/3kSyupCBknw7yb8j+QmSkyRvJLnXtP81yf0kJ0h+i+Qv6vEnAHgVgKeRnCL5PT3+LJI3a193knxBx/2eTPK72t8d2k+b+EoyIflqkneTPEryn0mu17aLVQy9huQ9+sz+4MynaOWA5Cv1WUwq3fyaabuW5FdIvlHp4U6llWt1Xo5a0V1p4S/02Rwh+WaSA9p2JckDJF+hKpq3+WPm+j0kP6x0doLk3+rxvSQ/r8eOk3w3yQ3mun0k/xfJ76vK4v0LofvTPJfHAvgsgJ1KY28HAOfc3znnPgdgbgmPfVWB5ENIfkdp5IP6fK8z7ad9viRfC+CPEN7Z5yzw9ntJ3qTz+hGSm0yfjyT5VaXN75G80rTZd/oykl/QPo6TfH/HPa5S2j5O8s9JJnrdKemur3DOndEfgH0AfghgD4BNAL4C4Dpt+yUAxwE8BEAVwN8A+KK59nUAPg9gAMD3AbxYjycAvgWZ2AqASwHcCeDx2v7HAL4OYBuArQC+CuBPFjjetwM4CeDhEDXauwG8z7RfDWCztr0MwL0Aatr2GgDv6ujvVwDsBUCI6D8D4CHa9nAA4wAep79pF4D7atsNAJ6rn58N4Hb9ncMAPgzgndp2MQAH4B/1Of00gDqA+53pXK2UPwBPBbBTn8nTAEwD2KFt1wJoAXgWgBTAdQDuAfB3SkO/DGASwLCe/yYAH1XaGwHwMQCv17Yrta8/02sH9NgBbU8hXPkbAQwBqAF4lLZdpvNWVRr7IoA3ddD9Tfo7NgG4GcALF0L3p3k2xfh6tH0ZwLX9nr9loI8KgLsBvBRAGcCvA2hg4euKg6h/gR7v7GnufQOAgwAeqDTxIX+9vr8nAFyltPs4/b7VXOvf6fcC+AM9r6ArM77/VLq5EMCt5rqF0N1jF/Pbljwvi5jIff6l0O9XAbhDP78VwBtM2zCAJoCL9XsZsgn8AMCnAFCPPwLAPR33+X0Ab9PPdwC4yrQ9HsC+BY737QD+b8d4f3KK80cB/PRCJwPAvwF4qX7+BwBvPAUReoL4HIDfMW330edUQtgcdpv2mwA8fbmI4pwTHfBdAE/Wz9cCuM20/ZT+/u3m2AkAD4ZsyNMA9pq2nwNwl36+ErKo1Ez7lQibw88BOAagtIAxPgXAdzro/mrz/Q0A3rwQuj/NfYrx9Wg7XzaH/wJZoNnx2/3mcLp1Zambw/Xm+/2VhlIAr4Aybab90wCuMdf6d/qfAbzFvrfmGgfgCeb77wD43BnQXV82h8Wqlfabz3dDuCno/7t9g3NuCvJi79LvTchi/UAAf+n0F0P0hDtVdBsjOQZR6Wzv1W/HPRcC6wUyAyEuAADJl6maaFzvux7Alvk6IvlEkl9X8XYMstn48/dANrLTodfvKSH83lOOebWB5G+rqs3P7QPR/oyPmM+zAOCc6zw2DOGsBgF8y/T1KT3uccw5N586Zg+Au10PfTTJbSTfR/IgyQkA70I3Hcw3J6ek+4jTYieAg2Y9ANrXmHP9fDvXszJk7i8C8NSOdelRAHb06OP3IMzLTSR/RPLZp7mHV7cvhO76gsVuDnvM5wsBHNLPhyAPFABAcgiisjmo33cB+N8A3gbgLxnsBvsh3N8G8zfinLuqV78d91w0KPaFVwD4HwA2Ouc2QNRC1FNcx/lViNj5FxDOdgOAT5rz90NUTqdDr9/TQvsiuSZA8iKIiuzFEA+SDRC1JE95YW8ch2wUDzB0st45ZzfOU6UZ3g/gQvY2VL9er32Qc24dRN240DGeku4jTovDAHaRtM/brjHn+vl2rmdNCK3th0gOdl0acs5d39mBc+5e59zznHM7AbwAwN+z3X11vjVzKXR3TrHYzeFFJHer4eZVALzx5T0AnkXywbqQ/imAG51z+3Ti3w4REZ8DIYg/0etuAjChhsQBkinJB5J8mLa/F8CrSW4luQVimyj8fSkG3CsX8TtGIIvyMQAlkn8EYJ1pPwLgYm88guhGq3p+i+QTITpxj7fq738Mxei8i2pw78B7AfxPkpeQHIY8p/f34mjXAIYgxH8MEIM+RHI4YzjncshG80aS27S/XSQfv8AuboLQ3fUkh0jWSP6Cto0AmAIwpkzMy89gaPPSvY7xBopb9IJAskIxdhNAWce5lj0LvwYgA/BikiWST4bY7zxO+XxPBQYHj4tPcdrVJO9PchBi3/wX51wGWWOeRPLxuibVKA4Ou3vc56nm+CiE5jNzystJbqQ45LwUYc1cCt2dUyyW4N4D4DMQo/GdECMinHhY/CGEuz4M4aKfrte8BKI2+UMVH58FmfBf1Il4EkSvfBdk1/6/EBUPtP9vQozYP4C4+F0HADohU3r8TPFpAP8OMRDdDfEOseKfD6I5QfLbzrlJ/R0fgBDAb0KMo9Dff5P+rjdCJJAvoF1C8PgnAO+EGJ/u0vv+v4sY/4qHc+7HAP4SsgAcgdgUvrKELl8BMeZ/XcXw/8AC4xgMnV0GMXofgBjIAeC1EIPnOIBPQJwEFoTT0D0gXOOZ/ObPQCSkn4fosWchevk1CedcA2KEfg6AMQj3/HGII8ZCnu+psAfybp9KyngnhHG9F2JMfonedz+AJ0MY4GOQteHl6L1uPgzAjSSnIGvCS51zd5n2j0Dsrd+F0Ndb9fii6e5cwxuEVy1IXg1RM/x+v8cSEdEJZV4+6Jz7uX6PZTWB5I0Qg//bltjPqyF2qH84OyM7f7DqN4eIiIjVD5KPBnALRGvwWwDeDOBS59zhvg7sPMZqiCCNiIhY+7gPRF07DPH4+424MfQXSzJykXwCyVsokYuvPFuDiljdiHQRMR/mow3n3Fucc9vVG+hBzrlP9HOcEUtQK5FMIYbcx0EMe98A8Aw1QEacp4h0ETEfIm2sLixFrfRwALc75+4EAJLvg1j2551ocSlL4Npc0dWlVw+V07RoGRwcBABMz84AAFqtvGhL00rbdWDo03tL587fQQ7YjVC8IsOxxHgKhvF1b5zFEb1JkoTxlrSvqhMPtpr1Vs7ES/UY3HHnnA3aWms4Y7qoVEpuoFaBdXP3n/0hJqEtOYVXp/f4DOf0YH58pzpfli4SvU/ei2di+4c8C56KaVrS31KzXbedXyqleo9AM5nSxd333LPW6QI4Q9oolSuuUq2BSZhvTxe5f8DmQdvzAMDlYb0o6fxkfs56XOfpINfrEtNfqVzWNuudKmi2Wm3XtaGjL5eZc9gRzmBpXM+fm57sG10sZXPYhXa3zwOQNBjzgkxQKdXQtAu5arbYlAezbXhj0fYzP/sgAMCN3/8OAOD4WL1oWz+iMSWZTmgS2koaWtdo6svoZGKb9UZxTiublcvrMrHVaqVoy52f7KYcCO8yirskcpOBoZGibXsu/V/anAIAXF4KhODGTwAA3pzVbWT0WsQZ08VArYKff/h9kKbhZSzry1guy8OvDYT5qVXk2RcvnAsvVaUi51UrA9DGos2//KWSXNfSlzrLmmEsg9L3dEMWAbthpQXjItdPTU4XbSPDmwEAF114hfTdtMyK/JZNmySfm2d6AGB8fBwA8LwXvWCt0wVwhrRRqdZw2YMeiVqtyG+ItCxLVr0lc2YXZD/3Ho3Z8L5v2iDrytT4BAAga4aQIt9/Q/ucnpZ5HRgeKs65YKcERU/OTAJoZzmOnTgufU/Je5+YxtasrBjDNaHHxvRs0eZpyyk9slou2gZGJK7z5q98tm90sZTNoVcUXxe/RfL5AJ4fjuRI7K6tHxPnOcVuCaDXDXONL6kMyEOvmw255X+Vcgl5M9PBBYLwC0IyLIQx3bDxZ3Knakna6rOToaUmBFiqyEKxrhIWtG0leel/dpsQ0m8+4meKtttuFDf3N3/jRqxxnDFd1GoV5EiQtAUu+27k+VqGraUMQUqd9DzcstmUefUbQaUcFhaPel3nXjcXv3gDwOSEvLx5OtB13awuHnByXZaFxWh2Ru535x3iTn/0yImibU4XqQ0bJb5y/foQZ9nKwgJ2HuC0tGHpolwdQDIwhJZhznK/KJTl2dOsJQ3n6ULOqQyHTXhiTrQPqKjEb5lBL0gqc1LV99hvRAAwNScbxvphYQaP6oYAAFBpYPS4zLndpPzmNTUTGAnzWwEAwzWhh9Q8Cr/R9BNLMUgfQHtI+G70SGmhhqaHOuceys7VPmIt4ozpolKOTnPnCU5LG5YuSpUFZeWPOEdYyubwDQCXawqICiRi8aOnuSZi7SPSRcR8iLSxirBols051yL5YkgKihTAPznnfnS66xLkwSgEIC1UCd56HNrYYVm2YhcoaqCG6hua5qdkc5KUM2mKiqCmW2CjORPOUX1zq1BJGGNoRcTK+oxcXxksan+g1dK+vb56LvTZUoPjiVH5fuc9IWJ//8E1l1OvJxZFFyRKaa0wGgJAop+ZlPyBcHoqHGVJ25JSt8GysA842ybHyuV2nsgaHoeG1PCofVtp19OtV1XlRp01Nyv04HK5x4YNoV5LulmOzc0JPR0/EYq+zRqV5VrHmdJGDodZ1wIzY6j1qiM/v3Z+cpkDqtoxMbrItKTqJLQbnwGg6URdyFRV2/oezzaDHXPsuNgqjqhTzOxssB1US0IzWzbKnA8PhzyQiUrFh+6VkI3c3Hf9OlEnlZXWZyaDKmnO9N8vLEmed859EpKVNCKiQKSLiPkQaWP1YHmVvc7BZS0wN5ybd01TA3NuEhm21N00Vy6dJmlpOZHzGrns7mnJeCo0hLuvZeIJUlEj5ebNIU36f33iUwAAH/vcfwAAjh4fL9pyNTyyJhJEoxGMhlQuoa7Hxow0Ulbutj4qXMaddwfJIW+G/iPaUasO4Ior7odqNeiYPXde0nktpaGtVBKyTdMyOuG9nPKmzE/b3KkbiT/H9XBl9Z4rXoixnkwzMzN6TB0cjJdUtZy0tVUqoW12dlrvK2MhAx1baSmiHQSQJO2OKYODMj+eW6/XA3dfV7dgr4xIjGtoU+fFz31qnrtz6ubqvRPVwGymtzBOz7WEBsrGe66sdJjpujFoDNINHdOQ0vbUjFkvlOxSdZAYMVLupqpIHyfRP6zlNMAREREREYvEsrMtzjnkCH6nibIFmW7TTesjrG5g3h89NT7NDdXVMhXuPp+ZKNp2rhMO/tE/I3ESF2yWzN/JulA87jH//bcAAL/2354MAHjT3/5d0faZz9+gY5W9s7w+6I+b6rfs/d9aRmI52hLuZLQlbfuc0ZMnarfIl1yjaM3BOYfWXI7UmSBH/ViYoAwbk6sEkLNRXO9Rqai0qVKm5R49pz83K5JlL++5CbUzuWxMxmGCMoOkIYPydAkAjab06V0XfXAbAMw1pM+6xtnU54I0MjsbON+IdjjkyPI6KobL37h+SP/LOzl+crRoO3ZMnmWriGAMdOGDa4v5sbSmokZCfZfpzwnz5Fpq61CptT5tJAAdXqq2jnWV4AZ97NgxAMB6dYsfrAZC3qC2zURjahIjcQzpuvJD9A9RcoiIiIiI6ELcHCIiIiIiurDsaiWJaDQhj4mIT5lXIZlzO9VKFhV1X6yrQbjmgiH75+57PwDAlQ+8BIC4xAHAEQQXs5vvkCj++jGJdLx8T6hVPvkgKf166123AQDGp01deWrErYq6uXF389Js0+f1KVm1hXfL7fopEc6h1Wi2PZtMo9oTdSn14jwAlDU63afbcCZfQdaQzz7SPnPdrtE+/9K0qnSmpoKKwKuONo2059mRtnY3WavOmtX8X974bK/zOcEaDVFTNBthTM1md66eCAFJlMrlNvfPGWPQBYDpmeDymcEV1wEhkh4Aapq+wjs9JMaZwa8vTXVeyNUltmT9ZtThIB8Xd9NBF+ixpLrwobLcozUZxrR1WNRfPuK5buhx86Cou30w/txUiKJ2PZN7LS+i5BARERER0YXllxzYnpgq8dx1yyc4MeyjupY5dUOjMUzVlStwur9tWR+MQBes15xIY2IMqkOkk/q6HcU5t98qbqb1Q3cCAC7aHvLdXPkQKUn8s5fKrv+hT3yqaBtVV7Y5n8DJhvi3PLfrh2+4wjTuw/PBuRytbAbOSJSZcvepSlxlZxPsaOI8pZXMuD97d1U/B7NzwZW1WhUaWb9+o7apoXh2rDjHuzqOTso823xPmQZjDQ6IUXT37gvDfTXAbv9+yZN29GgIdKuogXKwLHl5yhvaE8QJvtbj2PkNByJjqY2DHZ8Ww//EjEh9tEGOmnHT6XvYMHmrSiq95bmX9IPE7yVAem4975Yac03Ul07JmrTVuMWnSmzecF5XhwcA2L1TsoX45Hyjo8aAPiZ0NzIidJGYXFDWEaJfiCtWREREREQXlt+VlUlg6wAU2RHUfcymyEi85AB1KUuCnrBUFe4tb8p1FROUkqqycE6DS2aUu5iuBU6grl1tXCd6v0N3/qRo2wTh+rakwl0843GPLNre+9mvAgCOqP44N0FWXupJfNZQ87sT/RYdF7uRuSYmGofbguBqatuh6oZbDE/TZ1X1KQzm5gKn5rlAn77CSwsAkCtHOa5BiuNj8r9RN6kKVCecarr1gQEjGRZSrRBtq2maCtdZadu27YKizWfp9C6smcnp37MGQITAAcjYXv9CJQCvTbAMdiEB6DwNDwRO3Nf3yPXZWzfmwjRYuD1raveKSeei2ZgH0G1vqni61evLpXDf/SePal9Cx7VN68OAvc1M/zeM1iRJ+s+3938EERERERErDnFziIiIiIjowjKrlQiwWqiSgJC7xttrE+PC5cv8ebdEZ6JdfQ6bVksLqawLUcyzGq06rtqGExNiXDx84vZw/XatwnRY3FTziRC57GoSfZ2lct2mDaFK3/0vvQgAcOwnYnjM07ZKJPJPpcPE+GZmWf9d01YsEgcONJHUAmGkFZ1zDZGeMzocH3xc1SIpIyPbirYNG6QiW02NwLaKmHchvfdemfN1Q0Izm9YH46JXF9SduBU6k3nVeyNPaaWwH/zo+6GpyNvUzW951VFTjZrWfbXZiHQxH0pJii0DI20qls4Snr1KxBSuxsZRIVW1ks+15YxLqXdl9aqmJPVu8kFl7N/fdFBVQeZ+Tba7I+dmfQvqK1WRV4zaXM/zY3FttWX7r26MkkNERERERBeWVXJI0jIGh7ejlQQDYitTl7RM85ukoW5rQ4+11AiUmt00rwtXv04NRbsu2Fy0HTkmAW5bLxQ3Mr8DDrRC7vyTh38AABgak926NRHKOpZ3qHE7lf9j02G8w8ptViH1GfIsGCxzX3i+Z7ybfnMTiOiAE+/BlqnzUVeJ0gciOcPHOH3OdeXAGyYQcWpGDIDlpKOgPGBcFOVYtaZSn6kxm3suju3cJADkmebaUkPl4FAwLvrTPPc3Vzd1PppeolRpKDf5mkoxCG4+lNMUFwyPYGwiZDT2WViHtNRq2ZT79CViEw2YHKyFtcQ7L8xOzrWdC4T6DT7br6eZXvWpp5Jujp7u1N8tEkNPnqLpRVJbr7xnRdXlRZQcIiIiIiK6sLySQ1JCbWgL0krglrJcs1k2ZZ8aqAXbQVYRXb8r6TGTJbGmer4rdom+OM0Dd39cOY2J+nYAQFWLkW8yOsi9O0TSyNRd9UgeOJBMOYiW5oyfNpWoBtdJdtVyIhJDIw/utTl8H7rn0nIZPgVIlBw6kSQlDFa2IU1txS/VKfvvpq2kvodUiTIxTsOlUnuAWb1uU694lk4D7LSfZlt6FtVpe7dklrrb1PZl6zl4O4iXDlqNQBc+XQe9m6xhyTIaV+iINuStFqbHRlG1bp3K5bc0gNHa9aq+lofO3fiJ7moIWzfK+zs4GNxN6ypFeBuDr8Ewa1ykvaThS9H04qrTHhJDZ+bfxEoEer4PCrZt/ZcbouQQEREREdEDcXOIiIiIiOjCsqqVnCNarRIyU+6zoSqbUknziwztLtrSdWJQrozcAQCoH7uraNumuZAu2Smqp8NH7y7a7j0haqULNksmxIu3Sk6l4XIQQYephvCNqlrgSNE2mYt4OTopLov1zOR08nZlX1LQqDS8COlzBKXOionqeouITuQ5UZ8hyuXwLCtq9C0yqSbGcKtqAyb6NI0r4eyczFlDo6Fhcu94o6LPktrM2rN4AsEoWYJXC5nry15lpa6SxrBcUXUWKdcNDZi51+hu7zJp1VE2ujuiHXnuMDfVwObNwdkkz0SddPKEOJBs2hrckPdethdAyI911+23FW333HMPAGAKo9p3eBO94dlHMZeLSH2jJ1K10pD3jDZup9547NVCPaObk+62wi2343r53H+shDFERERERKwwLHtupTwHUhMFl+qOWm8Jd3V8MhidyxOaiwayXdtygT/1gMsBAFPTktnw0NHjRdtMQwNVyuLKdlRLCe7ZFjKv+rzwzudIKgdOINfrDx8Vg9bISDCSD6+TjJ7rR6Tv6WlTStBbltTwSJMrRVP1RMmhBxIQA6UKSmZ+SxrIVFYDpGXGvEG4CFoyjXWVJoYHhZMvl4OrsZcKvIfiwIAEytl6AZ6TX1fz8xty7IfcT3K/rGXKUOp9fClQW4PEB995F0cTy3lKt8fzHQmJaqWE8bFgWPYuqTWllaQeDPp333IzgMCRT00E549NmnutCH7LA8341E2+Poh3jChXLe2IATullpE1NOPpzwffWXpsD2xDOyH7Q11HAJfFILiIiIiIiBWI5a/nUHJtYe3UndUXZM+aQQfrq2uVVRe3eWOwCwxoGPrBE8IdzJmf0tLIqWFNqbFOc6HmRoc4Pioc4cysBMZNzJl8qWWRVI4fF71mrRyCadbvEA5iaEiOtSaDxFKIB3obu/kXeu3IKXaBCVCuJbBhg/4xZS3vdhq4uFSz83pXVhrdf81XAVNu3QegAUC5JnOXaQqF8UmfSiFMis/iOjmnmXxN4bF6xWcJ9ukOTJH6VOjWc7btleDag6pKpqrd1PQUInojITBYzVE2VduyofZzyCChpT5wUt//2vrA+7a0tkNJOffM2By8NFFK2yv9JUlYE5qqYajWNJDSzL13VS4l3cup8xKkuji3ubIqPTilWSPMtEmz/UKUHCIiIiIiuhA3h4iIiIiILiyvKysdcraQGlfWirov5mpInMmCEWniuBzLZ0R1c8mlIZdN3hR10OScnD9qCo2nGtXqoBlbR+T75GgoB9lQY+JUXUS6uZYpKKIi4+SUliQ0hq1crxvQrJ8JQkRuXpSy9BkjTVZWLxpHi3QXsjzD+NQYUmOQLqsaqaoqAmei472qyGfrTWyGTRX/vTujdWU9cfwwAGB8XGjHlwvdsC7Qlc8K3FDF1kB5uGjLVW3hVYQ2orulGVcrGjlvC8V4I7dXK1WqRgVajxHS84FsoVIeR8kU+0l9lLl3WEh7GJY1M67Nq5WrytLpOmNdWcuVDkNyoZ4yy2NBR9KnqfMUvApS73US7ut8WVJVZbKHnTlRWknK5n49svsuN/o/goiIiIiIFYdlNkg7uFKO5mzgloqi3olw4DXDPbIlxroqhPPavX1X0TY5LplXT4yKVNF0IW//YK2qfct9pseEU5ydCdZFV/JZHcVoPTwc9snDxySz5+icbPMjU+E6H9A0XBPjeIpQSB4a0Oc8J2GCndyKyJayMpHnDrNzdZSD3REO3u1UOK9SM3B6JeX+ahqUlpbChWXNtJppptZaxZQeXSfSXmtWnBGqGlg3VLaBjEIHSSrXDQ2F66emhY5SbasY6WBSAya9cdG65c6qs4VTbjU3RvK81X/D40pFqUxs2Vpue3O81ObdkEvGL7ihxt9Ec2Ylxolhzmd/1rbcSPyp114ox+/zcaUmWVKqUkUz02BHK5WkPvhV/9dN7RF1dKFKjXa+i1ojGnyXDRktRNUWGe4PouQQEREREdGF5XdlzVqA2dFnWn5/kt1zJAk7a2nyTgDA3l1y/qzJvHrTvXLeePlBAIDcBbezLevUlVW5xrkpsTVUaiEIrqE/fVsudRmONDcWbXcckb6mKGH7J5phFx+bFElh93Y5NnhrqBExq9xIA8LV0EQ4+eyRMXt/NxJWMJBehCF1IwUAKhdH9Qe+YHuoxteqCx1MTYstaGBDyLDpA9XGRtUNOQ20NrxO/CBrSXtqjqFamN8ZzfY5XNYaAsYksK7SHnznpQwAqCppebuCdY8dqWqmVv1NTeOuPVSOEuV8cMyQVU+2ZVD1doBGQwLjnBE3kyIoUqVHw90nHdXayibbbqb0lKnk4VsqpspjUyVRNyeainUm+69TW+O4akFmGcZU2yCZoRMNojtpPN+zxAdhasVJBrtpfeIY+o0oOUREREREdCFuDhERERERXVh2tRJcgkrFFMjxeYhUum61ghw/oCLjrj2SnfWugwdNP9LmCrfC0OegGna8qyHL3WJm1hQRsCgXaOw/3vWwqe6TjYYpVJ62lxS00Ztz6FBgctorAAAgAElEQVQ7tP1uRMwDwiFhHTS8yvCwqgRnhB5mp0N+HV+kp6LqmrRkjNXqDrhtu7inlo3qp1KS83fslIIvXvVjs7Ju3CKqpzzRnFsm0tnTj//fyruL1Pvz7XWexnxxe1/q0h6L6Eaapli3bl1XwRwguCqnRvXjn2Wiqmmr2vMZef28NEwRKD93wUU57bq+WDs0Oj83tJqqGitVNTKbNk+brCWTR8VhYd8dgY7v1fD7TbtV3bk3qLar0ZU1IiIiImIlYlklB0ICTWzGSs/YtdTgU04Dx7Vzj9R2SHS3PrD/SNHWWn8RACD35SRNYN3goLiIldWAOKS1IpwRD+ot6XNABYdWHozVBUcJX2siPKbh4eG2cxaKXtxPhEeOFHVcsXdvceTSSy4BAOy/S2p4TIwGjmtQk+pnEOkiNYFQaapGRT3kMhM85x0atCSt5yLbDJdqrG72KDKfZa7t/GbWzX36/7ZPz9H6/3M9yk9GdMM5N+/zCU4Bxi24mCs5ZqUyPx/+XbZrkG/z0oh/t+05hUSoa1GWh/e/pXQwNauOEnPhuq16vx3DIslyh3G7nhJnlmRA7j83G/JszU6N9/zdy4koOUREREREdGGZbQ5EghTOuJX5zJY+jml4MLgz7t69EwBwZP+tAIC5euDiGnXd3VUYKBndY+H6pkFHiUojzqQ99G6mPpimNR52e59Z06FbJ+25C8/9WZ2n0wAXf51rMzREo8N8SAjUysTBe0Klv7FjkuqiMSO62rIJOGtpcKTTXAQDA8FdtdnSDLwdVbYAwKmNIGsoXfTIrT+r9oCs1W2n8lxsYUNozd9mr/O1Qzwna7nhM5VAzyd4TUPZuKt2cvVWsvPz6Y+NjYV0Of6YryrXMnPXKdUXNqVWj1w3KjmgZCr9qbs1NVi3bOZ0QOlvm2ocNtSCS/YOtY9NJuN6/WjRNj3W/wqBp5UcSO4h+Z8kbyb5I5Iv1eObSH6W5G36f+Pp+opYO4h0ETEfIm2sDSxErdQC8DLn3P0APBLAi0jeH8ArAXzOOXc5gM/p94jzB5EuIuZDpI01gNOqlZxzhwEc1s+TJG8GsAvAkwFcqae9A8ANAF5xqr6yLMPY2ASqJiK10VL1TNlnUg37lc+Geu+9EpU83QrieGNKRftEjX0M+Y9OnBBR7rZxuW5AIxBbzdB3MxOj5mAu9zjqNhVtrSJSUsY0PhmMQ75QuS8fOdsIUY2OPpKzW61kC8OsBZxVumhlGD9xEnMmR9JksCgDAGpGtZBpnDnVZTE3ceetXOjCq5PaynX6DKgd7qZ1o+YpVEZ1tJ0DdLuyZmZO/WlZD0N2p5HatqU9CsSsdpwt2sjyHNPT0yEHkYGfJ6sS8sZmH7nu1cOAKf+6ThxPzBR0ua72dlSQPuua9TcxubN8zixfhGqoEta3RM+f1qJgjSxk+R3T6Pi5moytUg0uzlaV1i+ckUGa5MUAfgbAjQC2KxF4Ytg2zzXPJ/lNkt90JrVyxNrBUunCv3ARaw9nShuWLuZmI130E1yoQYzkMIAvAHidc+7DJMeccxtM+6hz7pQ6RJLHAEwDOH6q81YotuDsjPsi59zW05+2OhDpAsDZoY01RRfA0mlD6eJunL13b7mxquliQTItyTKADwF4t3Puw3r4CMkdzrnDJHcANnd1bzjntooE4R66+CH3B6t13OcSkS4Eq3ns5wpngzb8orhan+9qHbfHQryVCOCtAG52zv2VafoogGv08zUAPnL2hxexUhHpImI+RNpYG1iI5PALAJ4J4Ackv6vHXgXgegAfIPkcAPcAeOq5GWLECkWki4j5EGljDWAh3kpfBuYtY/aYRdzzLYu4ZiVgtY77nCDSRRtW89jPOiJtFFit4wZwBgbpiIiIiIjzBzG3UkREREREF+LmEBERERHRhWXbHEg+geQtJG8nuaLD5mNumOXFaqGNSBfLi0gX/cWy2BxIpgBuBfA4AAcAfAPAM5xzPz7nN18E1Ad7h3Pu2yRHAHwLwFMAXAvgpHPueiXWjc65U6aGiDg1VhNtRLpYPkS66D+WS3J4OIDbnXN3OucaAN4HybOyIuGcO+yc+7Z+ngRgc8O8Q097B4QAIpaGVUMbkS6WFZEu+ozl2hx2Adhvvh/QYysei8kbFHFGWJW0EeninCPSRZ+xXJtDL5/nFe9Dq7lhPgTgd51zE/0ezxrFqqONSBfLgkgXfcZybQ4HAOwx33cDOLRM914UTpUbRtsXlDco4rRYVbQR6WLZEOmiz1iuzeEbAC4neQnJCoCnQ/KsrEjE3DDLilVDG5EulhWRLvqMZYuQJnkVgDcBSAH8k3Pudcty40WA5KMAfAnADwD4siCvgugRPwDgQmhuGOfcyb4Mcg1htdBGpIvlRaSL/iKmz4iIiIiI6EKMkI6IiIiI6ELcHCIiIiIiuhA3h4iIiIiILsTNISIiIiKiC3FziIiIiIjoQtwcIiIiIiK6EDeHiIiIiIguxM0hIiIiIqILcXOIiIiIiOhC3BwiIiIiIroQN4eIiIiIiC7EzSEiIiIiogt92xxI7iP52LPc57Ukv3w2+zzFvQZIfozkOMkPLsc91zpIOpKX6edV83xJXkhySuseRywzlkI3JG8g+dxF3vffSV6zgPOK8S3iHsU6SfI1JN+1mH4Wg9Jy3Wilg6QDcLlz7vYFXvIbALYD2Oyca2kxj38A8FAAOwBc4pzbd04Ge36g7fn2ezCngnPuHgDD/R5HBIBlpBvn3BPPZf/9RlQrLR4XAbjVEGAO4FMA/nv/hrSm0Pl8VyRIRgZrZWFV0M1qQL83h4eR/DHJUZJvI1kDAJIbSX6c5DFt+zjJ3f4iVR/dSXKS5F0kf6tX5yT/nOSXSa7X788mebP2+WmSF+nxL+ol31P1wNNONWiSrwXwRwCepuc/xzl3xDn395AKVuctSD6E5Hd0bj5I8v0krzPtzyN5O8mTJD9KcmePPrqe7wLu+yyd20mljReYtitJHiD5eySPkjxM8ikkryJ5q47lVeb8hOQrSd5B8gTJD5DcpG0Xq5rgOSTvAfB5c6yk52xSej6ktPZvevx0dH0DyT8h+RX9HZ8hucW0/zeSPyI5pufe7wynZ8WiX3Sj2EvyJlVFfcTPtfb5SJJf1Wf+PZJXmrZCJUXyMpJf0D6Ok3x/xz2uUro8rutSotftJfl5pbPjJN9NcsOCH9y5hHOuL38A9gH4IaRO7CYAXwFwnbZthnDggwBGAHwQwL9p2xCACQD30e87ADxAP18L4MuQTe8fAXwawKC2PQXA7QDuB1GnvRrAV814HIDLzmD8rwHwrh7HS9rXxf16tn2c0wqAuwG8FEAZwK8DaJh5/SUAxwE8BEAVwN8A+GKvOZjv+Z7i3r8CYC+kMP2jAcwAeIi2XQmgBVk4ygCeB+AYgPcofT0AwByAS/X83wXwdUjd4ipEXfhebbtYx/nPSosD5lhJz/kEgPcD2Kj3e/Tp6FrbbwBwB4ArtN8bAFyvbVcAmAbwOO3z95SeK/2e91VONzcAOAjggTqfH/LXA9gF4ASAqyBryuP0+1Zz7XP183sB/IGeVwPwqI7x/SdknbsQwK3musu03yqArQC+COBN5tp9AB67mN+25HnpI0HsA/BC8/0qAHfMc+6DAYzq5yEAY/qSDXScdy2kNN/7dZIrpu3fATzHfE8gC8hFnQS2wPH3nCic35vDf9EXjebYl81L/lYAbzBtwwCa/lkt5SXvMZZ/A/BS/XwlgFkAqX4f0Xs9wpz/LQBP0c83A3iMaduh4ywhbASXmnZ/rKTn5gA2LmCMBV3r9xsAvNp8/x0An9LPfwjgAx30exDAlf2e99VMNzAbsH6/P2RjSgG8AsA7O87/NIBrzLV+kf9nAG8BsLvHPRyAJ3TM6+fmGc9TAHzHfN+HPm0O/VYr7Tef7wawEwBIDpL8B5J3k5yA7KYbSKbOuWkATwPwQgCHSX6C5H1NP5cBeDKA1zrnGub4RQD+WsXDMQAnIVzmrnP2684/7ARw0CklK/Z3tN/tvzjnpiCc2JLngOQTSX5d1Q5jEGZjiznlhHMu08+z+v+IaZ9FMCpfBOBfDa3cDCCDGDo97O+y2APgpHNutMcY56Vrc9q95vOMGVPns8t1DGuBfvtGNz3udTdEetkCoYOnejpQWngUhAHoxO9B1pObVPX37NPcw69120i+j+RBpYl3oZ1u+4Z+bw57zOcLARzSzy8DcB8IZ7cOwlkA8vDhnPu0c+5xkEn6CUSF5HEzgGcB+HeS9zHH9wN4gXNug/kbcM599az/qvMXhwHsIklzzM7xIcgLBwAgOQRRtRxcyk1JViGS4l8A2O6c2wDgk1B6WQT2A3hiB63UnHN2nPMVX98PYNM8euNT0vVp0PnsCHm2S3p2KwR9oZt57nUhRCo5DpnLd3bQwZBz7vrODpxz9zrnnuec2wngBQD+nu3uq/Otda+H0NKDlCauxuLp9qyi35vDi0juVgPQqyDqIEDE/lkAY9r2v/0FJLerYW4IQB3AFISrK+Cce6/29x8k9+rhNwP4fZIP0H7Wk3yquewIgEttP2pkvHKhP4ZiUK/q16p+P5/wNchcvJhkieSTATzctL8HwLNIPlgX9D8FcKNbgMuvMfpe3KO5AnnuxwC0SD4RwC8v4Xe8GcDrGBwWtupvOS2cc4chKsy/VwN0maTfBOal6wXgAwB+heRjSJYhG00dwFpgbvpFNx5Xk7w/yUEAfwzgX1TKfBeAJ5F8PMmUZI3i3LC7swOSTzXHRyELvl2XXq70sAdiW7Fr3RSEJnYBePnpftNyod+bw3sAfAbAnfrnvRPeBDHIHYcYBj9lrkkgL8YhiGro0RAdXhucc++ATPTnSV7snPtXAH8G4H0qvv0QgPVTfg2Ad6j4+D90oqcA/OAMfs+sXgOIRDN7inPXHFSN9+sAngOxC10N4OOQRQzOuc9BdOcfgnCLewE8fYHd74GI413conNuEsBLIAvoKIDfBPDRJfyUv9brP0NyEkKDjziD658J4T5/AuAoxMANnJquTwnn3C2Q5/k3ev2TADypQ3W6KtEvujF4J4C3Q1R6NQgtwTm3H6KifhWE8dgPWbx7rZsPA3AjySkI7bzUOXeXaf8IxK71XYjDwlv1+GshhvZxPf7hBf6ucw62q/kiPEheDfGC+v1+j2U1g+SNAN7snHvbEvt5NYBjzrl/ODsji1jJiHTTf8TNIeKsguSjAdwC4W5/C6KiuVTVLRERPRHpZuUhRndGnG3cB6LeGYb47P9GfMEjFoBINysMS7I5kHwCyVsokYuvPFuDili9cM69BcA1AA5Agr1+qr8jilhJmG/NcM69xTm3Xb2BHuSc+0Q/xxmxBLWS+mbfConuOwBJG/EM59yPz97wIlYbIl1EzIdIG6sLS1ErPRzA7c65OwGA5Psglv15J5pMXJKkbU68Tt3FqUetp7MrrmPbfwAIm1p3W7hf+/UwG2GW535MXfft6qdNvnLaVd7jxI7vPTbeRj077pzbOv/dVj3OmC5KlZIr18oop4EcK2XxCG7MzgEA8iw878zPAZ2/PrRlkm/N01NiJs/PcZrKsSSR2DM7TXkmX1rO9xOQJtTzXVdbqVTSY0pPSbhvpn1mWabXh9/iaXNibHKt0wVwhrRRKldcpTrQNkH+2Tu0v//Sn/+v85Sbd7Q45jovQ1KsD+33T9MQm+jnjMy7zs31Pomen5bCdVmrKceUdmDGlCiNFGuZodUkEXo6OTbWN7pYyuawC+1RfwfQw92P5PMBPF+/oVIeQKVSMWe0vzilcnhAfnL8ZCcmAWajIR581Wp3KEGrJS92ohPi+0lMHKqfmGpFAlCtBNVq1eWYEsLwSLhHpSqdzDVnZBzN4K2apnK/XBeW3BCC/30Hbh8vIj3XKM6YLqoDVTz0cT+Locpg0b6xtg4AMHr4OACgal+4RJ5loyL/K+vDdXMNmbuNgyMAgOa0mZ+SzPmGzZJXrTog101NzRXnTI6KJ/J4fRIAUDP33aB00JqblnG0ghfp+nUbAQBl/xuSatE2MSnnz83J+c1ms2jzm8pH3vXJtU4XwAJow9JFqVTGrgv2Fu8OEBgDvyZUauE5D6+TuEPq5m+fsz/m30m78Ptj5XJZ7qH3Y96tdc/njsr15bAmTDfl/MFNuoZXw5iOjwr9UtcEu/KhJdfV9XqY9a1Skj5OfvsLfaOLpdgcevHaXayy6hIf6px7aC/uPmLN4YzpolyNfhHnCU5LG5Yu0jTSRT+xlM3hANpDwncjhIRHnL+IdBExHyJtrCIsZWv+BoDLSV4CiT58OiQydV4kSYLaYK1NpCsG4rKuY17S8OdbdVSuDEdS8vtbYEpKXr/nVVYqNnp9MgCUE6939uJmqA1SV3HUq5fSUug7VbVBqVTRc8x1c6Ke8KpDKyn1+s1rFGdMFyRRSVKUzTOay+RZDm8eAABsNKqjZktURXXKs28YtcP6jaJa2LF1GwDgth//pGhzELXB8aMi6o9sFPXSEf0OAK2mzPnQkKiljDmjsIlUajKmVhZ4q2Yu183OKA2UgmqhoVkUXFnoLy0ZZpnnVU2aM6QNAmmpwyakKh9933PD365bL6o9r2qiUQk2VPffUk3v3FxQJXqVslcnNTKZy8HhoeKc6WlRDY6Pz2rfYVBTTtaCLdvl/rUtIT/j4LbL5b66TpSs+npO+irpupS7sF4US9W3v4B+YdGbg5PSmC+GpLBNAfyTc+5HZ21kEasSkS4i5kOkjdWFJSn1nHOfhGS/XNjNymVs335Bm6HIG5bzXHbtJOnWdAUJIgx340bZ1bOW67rOc+nBs6FbKim4hab8HxgIJYCrNeFSW1lD72+uy5V7rAj3WBoaCONc5yWVpvYdDJatwnh5smssaw1nShdwQJI5kIarojxDVoXVayXBsFzPJwCEeS5bTwOV9k4cE8Ph9NRU0eQ5y3IqnF6z2U0XA0My96l6p6wbXFe01arUPqf1SLfnif8N5YFAj5NzE3pO9+vmXL/Tmy0vzoQ2MgATGcHEaAz0cRWePqXQlgzLO1wekHfSz6U9f2JC5mLnnpA7b3JSnA8y5e5rNTE2W+liZL06OJw8AQCYNqSTp3I/V1sPAOC6bUVbbUjG5Ne8xGgaKl48UMeXlnGuskb4fuH8osyIiIiIiAVhed0BHOByoloJ3HZFXcLm6uIaandM7+bn/+dmMy2XanpMOMU8t/p9OT8pbA/eR9nGSaik0sMW0Gp5dzfRXZbL4Rz683XTbzbCoAYHRZpJKFJCYrkaBmkpogMuB7I6stw854rYB/KyzNmcC8/P25ASnYOSkSg9h9aYk3nZsjXofzONlch17nOV7IaHgm7ZY6vaLlwe7jszIxxmQ+9RKgd6amh8RVM5Q1cKY/Kukpmnubaf3iNeJkKRtL1DAEB4/bz+N8/Pqc7eSwklMwdNdXH28QYto72Ym53Wu6kdQ20VWStIDtT5bdRFgm3CuMJSJVEfi2PpUdelRNe52kC4ruU1Cz5+JrP2yzr6jSg5RERERER0IW4OERERERFdWFa1knMOzWbWpt7xUYk1iPGo2QriVJE2Az5SurvPalVUVFYd1Zluo9XyrmrBQFxEX6f+utB5wnYjtzVcthoijg6o0StNQqTk1KS6uen1TOyAY0DPfEiSBNVqBS4Nz6vl1TmZPDerlEs1zjTzboEmtUZZI96n6jLXc61g5B4a1ghnVWFevGcnAODokVDueWpS6G/XTnHHv+32m4u248fFmWBwSO5fSspFW6ZqCmqUawmhLXGqIlO1h1WBRsyP1GUYaUy2HXOdj848y7kjBwAAs+KLgGP2PH2XvbF52jgxeBd5v174KOyt60Ol14MHpVbQjm0auV8Pa8KhWZn7PJP3PzGOFQ1VMzY9qbqgJpudVWccVU/WjZtrfQVoG6PkEBERERHRhWVlZ5OEqNUqbUYk/7miKRQqJpWCdzf1Rt96Pbgz+t3eJ0+z0ojvs3Bl7WG8KpJzFQFrNumV/K9Wy13XJR3ujzYIzrsqFgbIZmjrmagvAoAkUWu5HCUzh34+KhXh9EqGj3GJcPfeYDhjjHeVTKWDlndmCJzapXulnPiRI5KuplqRPo8cPFCck83KGG657XYAwNhY4FzL6kixbkRdFhMjURYSivSZGVGnoq6O3qHCGc4yYn4MVRI8dPcQMtf9brdynyAxvFeVktBDS+eg3gjvn3eHrU+I1Fgy+Y/crJw/qUkeh4ZFOsjNmjB3XFxYByoysZXZQHMD6pK6vSY0N9AMbYm6qWYqZabG6OyTB+bs5tG9o8x4V8vyIUoOERERERFdWFbJIc9zTE9PF/p6wKTE8IEghhP3rmjlQdnlbQqKSqWq56tbY4/gOd+Xv8fAQLXrHKo22wbYeUalWtMsjUZvjTRv69Pet6wpEzLn03UE9jGPLovzwjkicwkqZg4yHxTpuW1npU35n5Y1EDG3NiylFZUo7fwcPXovAGByUvixUirzWjW2jtKw9Dmj2VzTNNgOBgbkc0XtXN4FEgAmxmfaftPwcAie81mefZbPPI+Sw0KQJA5Dw622OfCfvf0mM7ZCTyLe3bxupDefOscNyfUNm7GVcv6QdjVQFaK7YCSsN/URva/y01UGW+OWdfL58gtEorznREikOjfh6UhTuhtbR9UH56rbdslIlN4t9170D1FyiIiIiIjoQtwcIiIiIiK6sLxqJQc0XIapsZBfaL2TnCXexcy6f+b6eXidRh7PBfHSW5KbTrNgmkIZULfWqhq3vfF5yEbCqqyfqPrC5lFJfGS2qo7SchD3Sr7CmI/CrNiIx1zPkbZBLQADtBceiWiHA9DMiNQEhWZaCKUyKM931uRI8kWV6vq8aXIWOVUTOo2cbxhj9WxJ3KWpEfM7tkn09BCDmvPQ3eKyeHJGaNSqQL1qYm5WVAX33nu067cMacZWGtWRj+CebXmXR6M+6OohwsOVU+RbR2AVsllHSQia3FS+xUfOD5iI+6LIj3eLN0ufV0e5Qu2nGXbH9hXnXLBR20qiLkzM9ZWSzPlgU1yiq2lwnLl4naqYW6KCNBpMNNW13hvXm4YY8kxODDmFlx9RcoiIiIiI6MKySg5pKcHIpnU4dixwXMfHhUMrTctQtm3bUrQNj0hA07jmtKnXAxdYUw6tqXmMbN6loiyoSgANdXlszoQguGpVpALX8HV9w7ZNzaDqXQ7TSpBYCjdVvc6XHwSAPGk3qjvj9ppl0SB9KjgCZWN4rCq3nalVsZSYHDr0Lso+E6oJYITWYxiQ69ZvC4bDTVoe9NARMUgfPHAEADA7OVGcM7JRDNobhtS5wHD5U9PSd7Mu8zs4GPr2LrPBbTpcl5R8UKavfW1rOET+bD5kJMbLZVhfk5LPxKv0US2Vu9pSX0rUzIFTqc1TSmpfR59ry9eV99f1cBzI/P2dcY7R+1U41fYfAGqpz9Omzg+lHi73vm8jRyZd0X7Lj0iZERERERFdWFbJgSTSahlpLdw2Rbtd4ORkSGXQYjtXbzfyrOG5dPlvdfrefpEoV+/dIqdmw44+UhL7Q2PGu7KGfXJwUHTTToc5MxdcFhtNkUJKmvZjwATT5D5nuzIzOYOk03LR5jA/HByaaNWDZOdrNCSqpLUZUKdVgsy0tkbZtA1q/Ydd22QSLtgccvqPjgttHbpHgt6SktiEduwI+fcv2C42sLHR78l9B4KdqqzVv06O+iyeRmpUfbWPcbIZQYtf6dOq2OzAMSBuXjjnUM9aqKYhkDGtanbTarcruefEm5rt1Fl1gkqe/nG7JDz3tJiX9uqQNl2Pn7Ok3J0hupm1X18eDLVhSj67sNrQ5oyrvicWLzFY7UWSRckhIiIiImIFIm4OERERERFdWN6srABaeY5Nm7cWx7zxuFYTMXF0NKiVprVYu1fzbNu+xVwn4toll1wCALjnnnuKNq9imp4WQ/bclKghBoaCAXFOi39cuEeuv/DCC4u2detFLDypJQEPHNpftE1NifHSZ48dnwrZT0ZGRCVRVVWTVR/Ux/tfvGOlwrkcWasOVoMKJysio4U+bOR86t2JfUS6KcdaUnXBOpXsW80jRdvJ41N6jqoNVP83axwdRqfk+qkJoaeduy4p2oZHRA01qdHTzZaJjle351bqx2+MofT/NSurMTzGYOn5USmXsHv7ljbVkf/s1TS+iA8QSg77DM1WTeORu1bbOXKsXd0XMjZ33zfVqOayyYdU0gjryVzuXzcq7mZRxMxnhja5oCrtTgwWRP8dWKLkEBERERHRhWU3SJfTSpHtFAj1HLxBeLAWuMCNGySfemFgNoFux45KtvYZLfZe5GgyqGppvg3rJefJxMRY0bZjxw4AwBWXXQ4A+Ma3vlm0lcoyls3q+njsWMgM76WRLVulrVQy3GPZZwLtLj1a7pH7KULhAGQtuLw7i6aXMsfHgwRQb2geI+9WbLjAuTnPvWk21yAshmy7WvbVM/62sPv4hEirm9SN2gZHtmZFSvSSqQ3Ioloo/VBaplNPI6FcaBQXFoJSqYSt2za2HSsyLWfdc+9a7RKDM+6gE2PjeqzSdV3eYRD2ubtaJqtypvVFfG2Xpuk79XUgtM+WyaPWVKOzz9hcMrmVGkpbvqsMYUxcASJlXLEiIiIiIrqwrJJDKU2xdfMG5IZD3LhROANfqyFrBbfRdetEhz+r6QoOHgjZDotzRpQTaAU937Zt4pq4ZYtw916XaDl6v8tPzYhdYc8eaweR82sD0veGTfcp2ioVeWQ1zRRr9YVNzePuOUvbNrtB7Cbf/VLXTzjvQQIpS2gau4KXKKvD6lY8Hkh1Wm1ISUk4tKrJtusT6P7olsMAgG0bgyvr5vVS+c0X8WpBUiFs2hxcWX02hvEj+wAAM2NBkm34+aSOyQRC1VXf7SuTtWcJVp22123bDLPov8viSkXmckw15trqovj31tudUvOcU7UD+Oecm9oJI0MSNOttV1C1edwAACAASURBVE3jNl1IEd5dVSW9pGbETkWubrWJ9WXNvau99G21Cf5+fmwlkx7I20ha9Okzut1y+4koOUREREREdCFuDhERERERXVhmg3SOajqDZh5UQJnK+GoDxs5tJrpQRzei6p0Ltuwt2jqLglt4V7RKom5kej+aLI2+sEZSk2PVqolaVbE00TShmXEr82Li6Jh3oTPGJ1UneTWYbfMiZEQ3iASlpNZW0jNRgjiqzgCj4yH/kS+c4qPUExON3FDRfG5OVJIzB4Kr4/ETx+WYFnYfHJR5mjOurBNTcmz0sBgwB9Q9GQDKqmaoDgmN5i7MaVF21peFbJk2LUdKtNrOlWNRrTQfWs0mjh26t63IV6Fu9MW2ehSI8u9oZtSU/pg39No5SNWRhOoe79us+7RXPZVKWkDMOtWkPoebZlsoh7ZE+ygXKq/w+5zPAadrX2JdnK3aqk+IkkNERERERBeWVXJIEqBWy1Aqhx25WtOdVd0L0zS0eaPMkGbItPmT0rTdla1mjEeNpgQ7NVvzc2W+Lx9MZ3PhNBpauF6NVtbAVLgjai72Ni5Qa1qmpXpXW7MVcrxHtMM5Is9SWBtcVbmv6Tl5br78JwCUa+o62MMl1CViePScV2LcA9NijkWC8Ak9E4ZzZrRuxODQbjnH+MI21T11MBVpIimF0qAu85KC0G+1FrjdVPunBuiZtD5RcjgVcgdXb4KmjAuVn235TMuGBgoHENeeIwkABgdE2mspXTQNZ+5XHB/XVgQrGgcWX1vGeS2ECXJs6LHZhvwfKIe1JFci0yTObeVJm5nWpNG8T802V9YoOURERERErEAsq+SA3CGfzTBoqmtVNdjJ52UfHgmuh04zmfp0FFZh5/WQnQFG8ln68LYGz0VaTr6lnN54U7i/Vh6CnbJUXVH1+npm+87b+rbuqqFN+raSTiuzOfwj2pDkcJUGTLYClHQ+vMtgnpjnp4+8UtR/MPYiTWuSaEW43DDmDYgU0IS4TzecuDpnLtg6putSX6RaUe7RuB5C03XMabBVyaRQ8IFU3h6S2FoCyvoO1IQuZ2YDra0El8WVC8LlZbSa4Tk3G+3vUSntlry8q7yV0BpqZyrSYBg7RpGFVbPueoHBvts+7UVa8Sk2zCiVDobUzTXPgg0r8wGTer3VUFRK7bx5uWyk1BVQOTJKDhERERERXYibQ0REREREF5ZVrZTnQKPu2gy8WzR/0eSkz1sTCvJkRV4SiZq2Rmfv0ub/txXKKERNry7oFukSVSVsGtrcdb1XGfXK7ph35PNxrttd1bvAWffVVuHaGHIERQick+fZS8VSuCAalWKnq2GSWqcAkfd79eVFdd/my87m9W4346RHLix/XfHfuDr6z7nSWpu2Qwu3eDdKSxc2x09EN0j2dFe37R6d2U3bysfqfPpzes2BLwBUlBk2NFC4t2ft3+19vLHcGrs9rfg+LVmGMbWXKe0cX78QJYeIiIiIiC4sq+SQpiWMDG/G5VeEYLbLLrsMAPDDH34fADA2djKcX9ai3GqQTtryqJT0vxolDXORsPeeZzMieg7PNdVFzUoe+t+bO50JTmmpVOENzJbzK2c+0EZ3fVMaNHHRID0vnEOWZb3z7+de6uue06J0o+XyPRfXg9v0fRU1AYosqeEcXzvEGyVpLI/+tMKYaXIr0X8ugqzQdV2v8UaD9PxIkgQDAwNtxmOPIl+alRrZ/i5bh4HOfGf2uRdBc2iX4toN0tI2M9du2Lbn+Qy+c6a+SLUizjc+aNdqRrx0EGg8LMdz9eAm3S9EySEiIiIiogvLKjmUy2Xs3LkLhw4GvfsPf/BjAKFmQrUWIl48J+Alh1Yz7PbDwxLUUql4qSL8FL8De47D/y+bsHa/2w9Wu6sxdV7nrJusujF6E0KzaTmQvO1/r7aIbjg4tFqtNm6uqJmgXJ3Nze/pYnhYsqq2DOtf77D7tOuGU+1bOcVcC8JXa+Yc6Xv9eunbGe7TF4f30kVzNoy3rkGO3o5hUyEkqdCoz8xp0zKsBJfFlQoyQaVSaZMc2CEZ9pIcPKzkMDQ01HZOL6mg1VElzs5NQY+aNNqOyc9nQb4urEU+eLNWG9DvQTpRMiru1xZQy25pablxWsmB5B6S/0nyZpI/IvlSPb6J5GdJ3qb/N56ur4i1g0gXEfMh0sbawELUSi0AL3PO3Q/AIwG8iOT9AbwSwOecc5cD+Jx+jzh/EOkiYj5E2lgDOK1ayTl3GMBh/TxJ8mYAuwA8GcCVeto7ANwA4BWn6qvVauH48eNtxhxvlPFi2vR0KPbjRb/6XLO43mN6Wgw2Rc4dUyijiHhU0SwU++k2BCboNhR7EdT/T3tc5/+32jI/thc4b4uQbq0tg/TZpAuXOzQaDTSM2g/q3ucjpBuNbpfPLVukSM/4RDDyjY6O6vndua88L+SjbMtauGmTybw6qX1NTEgWWKtWauicVyoyzrZSkx30YNVKWUPa1L+iUD3Zca4lnC3ayLIMExMTPdVKhWspu93MPaxaqVMd1Qss+ejntOtc/3nDel8eOCydfiy+rW4Kj9Xr8rnR6FYfdq4lvdyn+4kzGg3JiwH8DIAbAWxXIvDEsG2ea55P8pskv2lfioi1g6XSRasZff3XKs6UNtrWi7m1t3GuJrCX+2DPE8lhAF8A8Drn3IdJjjnnNpj2UefcKXWIJI8BmIZPi7m6sAVnZ9wXOee2nv601YFIFwDODm2sKboAlk4bShd34+y9e8uNVU0XC/JWIlkG8CEA73bOfVgPHyG5wzl3mOQOAEdP149zbivJbzrnHrr4IfcHq3Xc5xKRLgSreeznCmeDNvyiuFqf72odt8dCvJUI4K0AbnbO/ZVp+iiAa/TzNQA+cvaHF7FSEekiYj5E2lgbWIjk8AsAngngByS/q8deBeB6AB8g+RwA9wB46rkZYsQKRaSLiPkQaWMNYCHeSl8G5i1X9ZhF3PMti7hmJWC1jvucINJFG1bz2M86Im0UWK3jBnAGBumIiIiIiPMHK8uxNiIiIiJiRSBuDhERERERXVi2zYHkE0jeQvJ2kis6bD7mhllerBbaiHSxvIh00V8si82BksfiVgCPA3AAwDcAPMM59+NzfvNFQH2wdzjnvk1yBMC3ADwFwLUATjrnrldi3eicO2VqiIhTYzXRRqSL5UOki/5juSSHhwO43Tl3p3OuAeB9kDwrKxLOucPOuW/r50kANjfMO/S0d0AIIGJpWDW0EeliWRHpos9Yrs1hF4D95vsBPbbisZi8QRFnhFVJG5EuzjkiXfQZy7U59PJ5XvE+tJob5kMAftc5N9Hv8axRrDraiHSxLIh00Wcs1+ZwAMAe8303gEPLdO9F4VS5YbR9QXmDIk6LVUUbkS6WDZEu+ozl2hy+AeBykpeQrAB4OiTPyopEzA2zrFg1tBHpYlkR6aLPWLYIaZJXAXgTgBTAPznnXrcsN14ESD4KwJcA/ACAryDyKoge8QMALoTmhnHOnezLINcQVgttRLpYXkS66C9i+oyIiIiIiC7ECOmIiIiIiC7EzSEiIiIiogtxc4iIiIiI6ELcHCIiIiIiuhA3h4iIiIiILsTNISIiIiKiC3FziIiIiIjoQtwcIiIiIiK6EDeHiIiIiIguxM0hIiIiIqILcXOIiIiIiOhC3BwiIiIiIrpwzjYHkvtIPnaR1/6I5JWnOediko5kaZH3cCQv089vJ3ndYvo5w3veQPK55vt1JI+TvPdc3zsiYi2C5C+SvKXf45gPJK8l+eVz0O+i19eFYlEL67mGc+4B/R7DuQbJPQBeBuAi59yqKgISEbFS4Jz7EoD79HscqwUkXwPgMufc1ac7N6qV+oeLAJyIG0NExOKwWK1BxMJwrjeHh5H8MclRkm8jWfMNJH+V5HdJjpH8KskHmbZCZCL5cJLfJDlB8gjJv+q4x7NJHiJ5mOTLTB8PJ/k17f8wyb/VilJLAsnNJD+m4/mGqoa+bNp/Xo+P6/+f79HHYwF8FsBOklMk377UcUVE9AskX0nyDpKT+r7/mmm7luRXSL5R38U79R25luR+kkdJXmPOr5L8C5L36Pv+ZpID2nYlyQMkX6Gq2Lf5Y+b6PSQ/TPIYyRMk/1aP7yX5eT12nOS7SW4w1+0j+b9Ifl/f3fd3rFfPI3k7yZMkP0py55k9Iv6N9vsTko8xDc8iebM+uztJvsC0bSH5cX1uJ0l+iWTXmk3yviTvIvl0/b6T5If0GdxF8iV6/AmQIkRP03Xne6cctXPunPwB2Afgh5A6sJsAfAXAddr2EEg91UdAqjxdo+dXzbWP1c9fA/BM/TwM4JH6+WJIwfH3AhgC8FMAjpnrfhbAIyGqs4sB3Awp/O3H5yDiFQC83Y9tAb/rffo3COD+APYD+LK2bQIwCuCZet9n6PfN2n4DgOfq5ysBHDhXzz/+xb/l+gPwVAA7Iczm0wBMA9ihbdcCaAF4lr7r10Gqov0dgCqAXwYwCWBYz38TpLzmJgAjAD4G4PXadqX29Wd67YB9j7T/7wF4o64JNQCP0rbLADxOr9sK4IsA3mR+wz4AN+nv2KTrxQu17ZcAHNd1qwrgbwB8cYHPxv/+/wmgrM9nHMAmbf8VAHsBEMCjAcwAeIi2vR7Am/W6MoBfRCjQtg/AY3VM9wD4VT2eAPgWgD8CUAFwKYA7ATxe218D4F0LGvs5JJh9/uHq96sA3KGf/w+AP+k4/xYAj7Y/XD9/EcBrAWzpOP9iyAJ/X3PsDQDeOs94fhfAv5rvZ7w5KPE1AdzHHLsOYXN4JoCbOq75GoBr9fMNiJtD/FvjfwC+C+DJ+vlaALeZtp/Sd2+7OXYCwIN1gZwGsNe0/RyAu/TzlQAaAGqmvXiP9NxjAEoLGONTAHzHfN8H4Grz/Q0A3qyf3wrgDaZtWNeBixdwn2sBHIIu6nrsJijD2+P8fwPwUv38x5C605f1OG+frosHAPxXc/wRAO7pOPf3AbxNP78GC9wczrVaab/5fDdkVwZE3/4yFZfGSI5BJIxeotpzAFwB4CeqpvnVhdyD5BUqkt1LcgLAnwLYssTfsxUiEdh72s87dQwWdwPYtcT7RkSsWJD8baMiHgPwQLS/a0fM51kAcM51HhuGvF+DAL5l+vqUHvc45pybm2coewDc7Zxr9RjjNpLvI3lQ14N3oXs9sF6DMzomoOO9ds5NQTa0hb7XB52uzAq7Tj2R5NdVbTQGYaL9uP4cwO0APqMqp1d29PtCAF91zv2nOXYRRF1t19ZXAdi+wLEWONebwx7z+ULIDgrIgvo659wG8zfonHtvZwfOuducc88AsA0iTv4LyaEF3OP/APgJgMudc+sgD4hL/D3HICLi7nnufwgyORYXAji4xPtGRKxIkLwIwD8CeDFEfboBok5ezLt2HLJRPMCsC+udc8PmnFMVvd8P4EL2NlS/Xq99kK4HV5/BGNvea11/NmPh7/UukvZeFwI4RLIK4EMA/gIiSW0A8Ek/LufcpHPuZc65SwE8CcD/Z+0VkM3hQpJvNMf2QyQtu7aOOOeu0vZTPb82nOvN4UUkd5PcBFmc36/H/xHAC0k+goIhkr9CcqSzA5JXk9zqnMsBjOnhzJzyhyQHST4Aotf09xgBMAFgiuR9Afw/Cx00JQbiys7jzrkMwIcBvEbveV8Av21O+SSAK0j+JskSyadB7BIfX+i9IyJWGYYgC84xQAysEMnhjKHv+D8CeCPJbdrfLpKPX2AXNwE4DOB6XVNqJH9B20YATAEYI7kLwMvPYGjvAfAskg/WBf1PAdzonNunY7yB4iI6H7YBeAnJMsmnArgfZK2oQGwYxwC0SD4RYoOB9vurJC/TjWUCsu7ZtW8SwBMA/BeS15tnMEEx2g+QTEk+kOTDtP0IgIt7GbY7ca43h/cA+AzEIHInRD8P59w3ATwPwP/f3pnGSHZddfx/36u9qtfpbVb3jD2xsziZxAuO4hArixQQUfIBghOBAgL5CxEggQhEQghQJCMRIBJCyIRIYREhURCgEEiIhe04iuMttmfG49mnZ3p632t7Va/eu3w457572+VM9yyumuo5vy/V8+5bbk33O/ee/a9BDtszINvcG/FRAMeVUhUAXwLw8OvUyif5+scB/LnW+rt8/HcBfBr0H/h3sIvGFVFK7QP9ER39Cad8FsAASAX9R5BDvMHfaxnAz4HyF5YB/B7IUbS0nWcLQq+htX4VwBdBvrV5kE/hB9dxy8+B3udn2PzzPWwzj4E3bx8DOZ8vguzxv8jDfwxy3q4D+C/QJm9baK0fB/CHoF3+LMiB/LBzyn5c+Tv/CMBhkGb0BQA/r7Ve1lqXAfwmgK+D5OCnQc54w2HQ96+A/n//Rmv9xOvmtgZytP+MUupPnf+DIwDO8zO/DJJZAPAN/lxWSr14pe9tPN8Co5T6JZBa+wfbPP/PAExorT+z5cmCIOwoeDP5Da31e7s9lxuNLA5XCZuSMiDN4j6QevjrWut/7+rEBEEQbiCSYXj19IFMSXtAuRpfBIWbCYIg7BiuS3PgjLsvgeL/v6y1fnSLSwRBuIURmdE7XPPioJTyAZwCOUOmATwH4FPsoBIEQdiEyIze4nrMSvcDOKO1PgcASqmvAfg4gJ/4i85k0zqfz8L3bMivjmlxMlHAvhNAVUhnAQDFbB4AUK7XkrFyo07Xpeh8d5Ezd1D8nDh5mJ2LMj/75pz2RTLFd0ppO9++EkXbrlbK9J0KSfkVUCSefU7Usrk4UUwRaCvLG0taazepRxBuFa5KZuTzBd3fNwA/5duD/J57Hh3TcfJ2I2w1AADNRhMAEMc26tOkGXj86cqLmO+h+cWNeWyTROB/KBh5ZWVC8rNqT5sIw9amc5SbWqHMOSEAIIrsdzG3irXumry4nsVhLzZnB0+DUrc3oZR6BMAjAJDLZ/DgQ3ejmMsm462AfqFp/mX3w469Z+IOAMB9h+8GADz50gvJ2PcvHAMAeCOUHxM168lY3qd7eTmqs1fnX74f2YUnTX8/0AN0rIq2pEqMalqUhlt2Th94/0MAgG8+9TgAYPKetyVjQUiLl9/ihWDRRrBWqjT2T1/99uszqAXhVmFLmeHKi75SPz71yV/F0NBQMh6xIO0r0nvfaFSTsZk5erUuXTgPAKjX7Vg2ld70GUdhMlYubwAAQpYBDR4LndWhxZvYHKcZ+GkrOtO8ifVZ7mhlF7O5OUoE9700n5O2N+XN6/w8nbOxsWHv6dP9y4161+TF9SwOb5Rd2Lb91lo/BuAxAOgfKuq610J/2iY4F3hXENZpkdCevcVKg3bnR8+9BgA4cdb29Mj10459PaA/gIG+gh3je7ZYK1C8gahU7AIymM9vmmcuYxcAzX8cLZ5TKmPPPXWG5hC26F7ljZVkLOK5e7xzCQKr6TTDBgThFmdLmeHKi/Gx3RoqRj5rBWrMQpPC+YG0b2952z4qVjBQIvlSrdrFoVquAABSfH4hZws0r62tAgAiFvyNkN5f5TubSbZepP12C0OtRmlXtYA+lbM47NlDFYEKvJhpxwqRydAclpaWeL5WPk1PU6HZoye6Z3G7niS4aWwuHbEPtnSFIAjC6xGZ0UNcz+LwHIDDSqmDivokPIzN2X2CIAguIjN6iGs2K2mtW0qpzwL4Digs7Sta6+NXvEgB8D00GlZ98j2aQjMmVS6VslNaZ7OSz36aZsr6BdIlMiNl2VehXBsgO6nrNVIr00UyGUV+MzknypGaGDfIhOQ76qJu8HNadGzvpC1oeO48mQA12ycvXjrvfEGaaIp9HHFoHUzal5QS4dbmamVGs9nA1NRZvHr0leRYlt8jjy1UBScgZHSMipmul8nU6/oqjDVnaZXKs42ODCdjis3QdfY9NBpsTnZkUVAhE3HIkSyDg0mfIAwO070mcjQXz3GgG2f30NAuAEDkOMIrFZJP8TKdU6lXkrHVjVV0m+uSWFrrb4MyhAVBELZEZEbv0NntbKThVxpQBWvNCmLaudfZa6ydqKEmr9L9AxQ+mnYcUxt8XWaQNAjfWZGjOmkIRV75swV2gLdsaJvHfrBqSMfcXUKTdwljJdp57NpldyAvH6N6fLlhuqdWTpFEZe7Nn1nHoZWxOxxBELaDhtIttEJbZ9Nr0Xsa1Mn6kPGsBhCwpWBpYREAMDdn2zNE/HJ6/Eq25u09Q4509FimGFmgQ2tpMCGzQUz3WVxcTsbMc0w4/OTkZDJmNIc8O8nrdWs1ueMOisacmroAADh79nQyVq93P4Dlza7KKgiCIPQgHdUcfCiUVBZZz4aRldnXEKbomHYSx2otWt37D/bTgdhqBzUOUysO0I684ISm+i32OTT4+iLZB9//wEPJOQtTFCTxvWd/RPdbt2GnqNEOYu8hCkO7fNn29Firkh+kbxc9L1+wIbBra2TrTHMomxsL7dohBUHYmihqYWN9FRlHqzfx/wNFygvLZKw14dIU+QNbphGck5QWtuidzrFfoF617/vgAIWZZtN0ftikXfvGejk5p9kkObVQpuvcnASfw2NNCKzJW6Axeu9X1snX4YbXzs7OAgAuXrzI07XzLRZJvqyu2ed0GtEcBEEQhDZkcRAEQRDa6KhZSXke0oUcYs+aWIKAVEBT6iJynMZNdhD1lcisNNw/kIzVQlL54oDOabScGiscilocIHPSxz5BzaDuHDuQnBMdouuPXSLz0tSpU8nYaI4c4HtGKYT1iaeeTMbMPD3OqK5UrHpqMiMTE5LzPaVvhiBcHZlMBvv2HUAubc3QPreH7stRIEoU2zIYuTyHxUdk3jGmIMA6mU0IalCzYaP1GsmCtZXlTee6pS5W15f4GL3To06QynqZrq+vkSxIOe99is1ep1+jygojI2PJ2LGXuQQQy439e/YmY8vL3Q9lFc1BEARBaKOjmkOkNKoqQsFJCEtxIayQQ0pznl2tc2naHZiVuJSzTucBTogLcym+t31Ok8PAJljTKC/Ryv4nj/5Rck5pg54X3kH39LRdJ0eHKWFlih1c04vWwbT7LQcBAAE7uFzH1K5h2pWYcLlaYMPRqk5dJ0EQtoEG0ALqTRt2WiqSVp/KkjaxMG3DVRO/dUyyYYhD4AHrgG7w+6qU1eSznIg7MkRJdAE7lst1+9wGV+HLpUkWmbpIADDYbwJmyHpRC6zGks2yFqJIrs1M22ohTZZ55l6plL1nM5BQVkEQBOEmpKOag+d5yBSz0JFdtX2ufBhE7GtIWxWgySFl1WUKEc071wWVdbqe/QIr69ZGV+BwtcocXfcvf/m3AIDiqpMEx2W4X9s4AwC4jbUFALhtbDcA4Ic/fBoAkBuwFV/9El23wOnt6bTVdExKv4lIi5yywEHD7kIEQdgarTXCMETasf0XCvQu7t9P9fumpmz5mgr3WKlUSTZEofUL7BqkZLnFhQUAwNjISDJWi0jTmJ8lC8H8CvkX8o6Ps8nyKc8vd9qxfizNk6+izrv9ifEJex37Pz1T9sNJmlWazl9f3dj0CQDFoq1c3S1EcxAEQRDakMVBEARBaKOjZqU4itCsVDa13WyajOgsOXoCZU0x4AzjhRlqHjVatOadzAaZmMpcydB3MpX9Jt0rxZnOHz5yDwCgv2KfGy2SCvriMcpOvH2iPxmbLJI6+jybhQqlkp1Ths1gHCaXcbraeeyJbrIJqa/POsRiJ0RXEIStyWSyODg5iWbDyoS93DxncpICQ9ZWbLfFhQVy9jZDyp7WTttN825GbB4y7TsBIJcluVLmkNT9+24DAMwv2/pJQ4NkdjZhr8WCDVc1pq46O7AzaVtHrczmZ9Px0ulqiiw3EOrvH+Q52e8Zuyd2CdEcBEEQhDY63GRAI4pC6IZdFfPs0PXYMd2AHasr2p1XQZ/jJbu7/+A9DwAALoNW5FcuWseU2aWns7Srr8fcvm/D1kqZO34CAHAkpN39oYrdCZQWSRsZ0FwB0u15nfSQpfm6yW1Rc3N4a0lZjaPU130HkyD0Er7vodhXQCZjd/nvete7ANgabEeOHEnGTp6k9/TUKWor7DmO7I0KvdO5PL2Hpi8LAIyOkVZQ5DDZElsKTNtPup5CYT0Oqy+XbfJri/u+HDxEVVaNRgAAc9xH3mguDScwxSTbmWQ9t7bSwAA7w92O2x1GNAdBEAShjc5qDp4Pr1RECjbBIwdaiU0iSuDMqMl2wXMVCj+rrNlw1Q+/9wMAgPeM7wMAFC7bFfnozDkAwEaajs1xAspb0nZFz3GY7AdjKqmRWbLrZHSa6sEPNelYNWOvW+AyHRnuz5CG1TiMPdPsBAKnJrubNCMIwtZoHSMMG9i/35a9GRoi+/yPn38BAFCrOtaAGapyWq1Swmna0fgXVtj2z6Gl1YZNVBvmDXupnzSHCmv+vrOTbwV0T8UVpUPHh5hK03OMTJhdsEmzpqtcizUd8wwA2HeAymUYubHmhLJ6fverOIvmIAiCILQhi4MgCILQRhcc0hopJ7uw0SSVqmimElt1rd4kVWyDnUChM/bMS9R0/ONvpTCy+5u2XeCukJxFiz5d31hg1a5iG22Ma1IFf9qj0LgZzpwEgFcXuPlGgcxSummd5OWY7pFiU1PGqQWlTXgrq5faDdkNnBBdQRC2JJfP466334W8Y9at1+n901w/6dIl67E17Tbffc97AACX523dpWqN3uVlDk8dG7UVESJN73edTUCNupETdu+s+JyxMcp+rlRsVdf1dcrI3qjSMbehj3FEF0r0HQ4dmkzGTOiqyYZ267S5P3cL0RwEQRCENjqrOcQaXhCi7rQCNYGgGU548+wQ0rxjV0XaiasxGxp6YoZ2BZPHjwMAjixZ59NdZ2iXcNcArcgNbu4d12xl1Mnd7wQAZC+Tg2tXfS0ZC6q0aqf3cqNxRwMwP5u67qFTgVGzA91UV9TofiKLIPQqrVaIldUl3Pfu+5JjF8+TVv/yUbIcNAP7ThuNYWCYwkC/9T//nYxNs7PaJKbu5mQ6AChwAq5xZYBURQAACqNJREFURBf6SM6sLVuZYMJOl5aWNv2b5knvvQlrzzsti2PWcA4fPrzpHABYXFzk88nS4SbBra5KPwdBEAThJqSzneA0kI11EqIKAHGObPYBp4uXnNDQYopW4Bb7HtYzNvEkO0zaxGumtMZF+1XeuU5axIgmjWFtiWyJYcZqHqPeIE+A1kflJMwEnOhm5lZ31Bkvw93eWOdpVK3t0fSbMF3i6k0byhqF4nMQhKuhVOrDgw8+iGeefiY59vRTPwAAVNnm/74HfioZm9hL/oBnn30WAHB5xia6DXFV1jx3kGuF9t1cZ+0/1twjpkUyIfasxWCVS2tU6vQe53K2RMbgMJXbyXBvmlWnD0SLQ989lmtnz51NxoyvwWgJ5noACOrWb9EtRHMQBEEQ2pDFQRAEQWijs1VZEaOi61CedcrkOaRUceZi03H+ZrDZ+avhtPYrkAoWl9jk5FnHcN2Elyq692CazUn9NtwVG+zISnN114y99wpIdazm6T4bsVUTNZuMFJuenN5EKLIZCik6GDutR7UnGdKCcDU0GgFOnzmDJ5/+fnJsbpYCUUoFMskUB229teVVClM9c/4UACDnZEinuXZbGJA5auaiNTmlORgmirk2Wo1MOqU+p9kPB5cY01M9sGafiewYACCTofffDWUdG6Ox+dc5xAFbP8lkSJtPmjuZrcqN7rUXFs1BEARBaKOjmoP2FcJiCvnA7tJzHK7al6OddVPZ8M9qRKtmgZ3AqdjROFiL8HiXnhm3O4hKwDv3nHEecx+I2F0LOdwsRTuJSsppQcoVV7Nj5GhKDVj1IFviZuABrfIqYx3oJlGmxrVdIieU1dSDFwRhe9SDACdOnEwS1wCbCPuhj3wQAPDiSy8kY0abv3CBaquNjY4mY62kXSe9r4tOgtyePbatJwB4XOetqe3722IrRo61g4MHDyZjBw5Q7acUWxWmLtjEvEOHDgEAXn75KADg3nvvTcZMWGw2RdfNLs4687WWkG4hmoMgCILQRkc1Bz8C+jYU4rTTA6FAK3K5SSnoWaeLUsh9HEKPVtZdBWsDzLE9/3REO4CZYWvnu2+Qmo+XVui6/ats56tZ30Hk0/l+jSuv+naXMDPI9skB2mUMZmyI2UiJtInp+QsAgAHH5mmSYZoB3cvYOQEg4/gmBEHYmqgVY3mpgrfcfldybGKUQtAvnT0JADj6/HPJ2G7u4JbP0TkqZf18u4boPeVqFlhvrCRjlYjkgp8i2bNeobDV1rrVWDxNWsmd7yCNYShr+7MUInq5T50kjaVveDAZm1smbSAMKVz19HE73/U6WRg0P3dhzT7v4DhpI6tB90JaRXMQBEEQ2pDFQRAEQWijw1VZFXzlwemhkbTGMw213dojBjPm1h7Js/nJ4zA0HTkVUE2NI2MOMkNOqJhiRzZadE+Vtdf7fF3Mny1nvnVWBTMcJufWWDEVGPs487ER2cxqt4qjIAhbU6/Vcezoq7j/nncmx956550AgO9eoEzjdNqafE3VgkFuqDM9czkZi0cpjH2EzVJ9RRtSqmMyH4+OkAO7WqZ3dXjIOqr37t0NADh4gD6fftyG18a7SEAsr5KpqhZZZ/LKApnLM3WusTQ5Zq/jatN+P80pr6wZLO1Ue+4WojkIgiAIbXS2thIooc1PO+GfrDlE7PBxW/Olue+D2ZG7SSLghTVXoJDUdNNeV1un3X1SG8U4hh0HsanhbnYbrr5iwt1M8p1ywmsDrvluklTcpuAmwSVxTDvJMFG7QiQIwhVoRRHWVtbdFi8oc42jidERAMDcgA0IWZif5evoZZsYGUnG1nhXX8zS7ry+YYNTPEVy5p67jwAA9ozQ7n5wyNZiG+ijcPglDoHNp2zgzI9feRkAEOfp/Fpk791cp/keGSON4/a8TcTNFun8+RZpGuMl+13y3A71lalT6BaiOQiCIAhtbKk5KKX2A/gHABMAYgCPaa2/pJQaBvCvACYBXADwSa31lkXIPQBK2TUpNjt3btitnB12hjUH6zKwWwjTPNxv0U5eR3Z3Xw9pJdZJ9jzf1H+DtZC1i8jRHTJ52hVEfL7blyHge/ezz0PDag5mfrUaVY8NaraKrJeS8hnCrcGNkhmpVBoju8Yxussms50+fRoAUFknTaBa23DOp/e1v5925BPjtmdDJkX9FE6fpJ14FFh/4PiecTrGiWdz0+SrGO67PTlHs2+ywT1h3n73O5Kx1RrJoCWuHl1Ztu/97btpDrcN0TNSFatV7BvfCwC4cOrEpu8EADrX/X37dmbQAvA7Wuu3AngAwG8opd4G4PcBPK61Pgzgcf63IAiCyIwdwJaLg9Z6Vmv9Iv9cBnACwF4AHwfwVT7tqwA+8WZNUhCE3kFkxs7gqhzSSqlJAO8G8CMA41rrWYD+GJRSY1e4FHwewjCEl7IOaROm2mxyw42s430y1VvZweQ6dU0IaZVNN3HdXjcQkre6prihh4kibdhzWhn6OcWNOUwdJwBIDdH1TQ5FLdesY7nFbQn9Ik2mkLaVH03bvwbf03w3AAhr3auuKAjd4npkRtgMcenyDF47dTo5dun8MQBAo0bWqIUFW48o4KY9hRJVUjCBLACQ8+k9nZnm853Q90KaglounqeaSMd+zA7m0IafxxHJkgxXg20EzrvNcimI2PQUOe0+16jV6DLXdQscs1KLTdlzs/Tc5bK1sJV195uDbXtxUEqVAHwTwG9rrTfcKJ0trnsEwCMAkM91P3ZXEITOcC0yw5UXafHTdRX1RklnbScplQbwLQDf0Vr/BR87CeAh3gHsBvCE1vrOLe6zCKAKYOm6Z955RnBj5n2b1np069MEoXe5ETKD5cUUbty712luxLy7Ji+2E62kAPw9gBPml8z8J4DPAHiUP/9jq3tprUeVUs9rre/d6tybjV6dtyB0mhslM4xQ7NV3r1fnbdiOWel9AH4ZwFGl1Et87POgX/DXlVK/BuAigF94c6YoCEKPITJjB7Dl4qC1fhrATzIWfujGTkcQhF5HZMbOoBuZFo914Zk3gl6dtyD0Or367vXqvAFs0yEtCIIg3Fp0P0dbEARBuOmQxUEQBEFoo2OLg1Lqo0qpk0qpM0qpm7qmilJqv1Lq/5RSJ5RSx5VSv8XHh5VS/6uUOs2fQ92eqyDsVHpFZuxUedERn4NSygdwCsBHAEwDeA7Ap7TWr77pD78GOEFnt9b6RaVUH4AXQHVgfgXAitb6Uf5jHdJaf66LUxWEHUkvyYydKi86pTncD+CM1vqc1roJ4GugIlw3JVI4TBC6Ts/IjJ0qLzq1OOwFcMn59zQfu+m5UuEwAFsWGxQE4ZroSZmxk+RFpxaHN0qIueljaF9fOKzb8xGEW4iekxk7TV50anGYBrDf+fc+ADMdevY1wYXDvgngn7XW/8aH59m+aOyMC92anyDscHpKZuxEedGpxeE5AIeVUgeVUhkAD4OKcN2UbKNwGLDNYoOCIFwTPSMzdqq86FiGtFLqZwH8FQAfwFe01l/oyIOvAaXUgwC+D+AokDSQ/jzIjvh1AAfAhcO01itveBNBEK6LXpEZO1VeSPkMQRAEoQ3JkBYEQRDakMVBEARBaEMWB0EQBKENWRwEQRCENmRxEARBENqQxUEQBEFoQxYHQRAEoY3/B51tCjZW6VbpAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "incorrect = np.where(predicted_classes!=test_Y)[0]\n",
    "print(\"Found %d incorrect labels\" % len(incorrect))\n",
    "for i, incorrect in enumerate(incorrect[0:9]):\n",
    "    plt.subplot(3,3,i+1)\n",
    "    plt.imshow(test_X[incorrect].reshape(21,28,3), cmap='gray', interpolation='none')\n",
    "    plt.title(\"{}, {}\".format(deportes[predicted_classes[incorrect]],\n",
    "                                                    deportes[test_Y[incorrect]]))\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-08T00:21:00.968727Z",
     "start_time": "2018-11-08T00:21:00.947262Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "    Class 0       0.71      0.93      0.80      1985\n",
      "    Class 1       0.81      0.90      0.85      1832\n",
      "    Class 2       0.97      1.00      0.99      1783\n",
      "    Class 3       0.94      0.94      0.94      1018\n",
      "    Class 4       0.96      0.99      0.97      1513\n",
      "    Class 5       0.84      0.77      0.80      1526\n",
      "    Class 6       0.95      0.78      0.86      1490\n",
      "    Class 7       0.69      0.65      0.67      1878\n",
      "    Class 8       0.74      0.37      0.50      1002\n",
      "    Class 9       0.88      0.93      0.91      1399\n",
      "\n",
      "avg / total       0.84      0.84      0.83     15426\n",
      "\n"
     ]
    }
   ],
   "source": [
    "target_names = [\"Class {}\".format(i) for i in range(nClasses)]\n",
    "print(classification_report(test_Y, predicted_classes, target_names=target_names))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prediccion de una nueva imagen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from skimage.transform import resize\n",
    "\n",
    "images=[]\n",
    "# AQUI ESPECIFICAMOS UNAS IMAGENES\n",
    "filenames = ['test/golf_0113.jpg']\n",
    "\n",
    "for filepath in filenames:\n",
    "    image = plt.imread(filepath,0)\n",
    "    image_resized = resize(image, (21, 28),anti_aliasing=True,clip=False,preserve_range=True)\n",
    "    images.append(image_resized)\n",
    "\n",
    "X = np.array(images, dtype=np.uint8) #convierto de lista a numpy\n",
    "test_X = X.astype('float32')\n",
    "test_X = test_X / 255.\n",
    "\n",
    "predicted_classes = sport_model.predict(test_X)\n",
    "\n",
    "for i, img_tagged in enumerate(predicted_classes):\n",
    "    print(filenames[i], deportes[img_tagged.tolist().index(max(img_tagged))])"
   ]
  }
 ],
 "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}