faces_mtl_age_gender.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from scipy.io import loadmat\n",
    "from skimage import io\n",
    "from skimage.transform import resize\n",
    "from sklearn.model_selection import train_test_split\n",
    "import keras\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from os.path import join\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from IPython.display import Image\n",
    "from IPython.display import clear_output\n",
    "\n",
    "from datetime import datetime\n",
    "from dateutil.relativedelta import relativedelta\n",
    "from datetime import timedelta\n",
    "\n",
    "sample_size = 20000"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`PlotProgress` subclasses the Callback class to plot graph after each epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "class PlotProgress(keras.callbacks.Callback):\n",
    "    \n",
    "    def __init__(self, entity='loss'):\n",
    "        self.entity = entity\n",
    "        \n",
    "    def on_train_begin(self, logs={}):\n",
    "        self.i = 0\n",
    "        self.x = []\n",
    "        self.losses = []\n",
    "        self.val_losses = []\n",
    "        \n",
    "        self.fig = plt.figure()\n",
    "        \n",
    "        self.logs = []\n",
    "\n",
    "    def on_epoch_end(self, epoch, logs={}):\n",
    "        \n",
    "        self.logs.append(logs)\n",
    "        self.x.append(self.i)\n",
    "        self.losses.append(logs.get('{}'.format(self.entity)))\n",
    "        self.val_losses.append(logs.get('val_{}'.format(self.entity)))\n",
    "        self.i += 1\n",
    "        \n",
    "        clear_output(wait=True)\n",
    "        plt.plot(self.x, self.losses, label=\"{}\".format(self.entity))\n",
    "        plt.plot(self.x, self.val_losses, label=\"val_{}\".format(self.entity))\n",
    "        plt.legend()\n",
    "        plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Helper functions to extract data from the matlab metadata file and read image from path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "code_folding": [
     0,
     5,
     12
    ]
   },
   "outputs": [],
   "source": [
    "def get_date(matlab_datenum):\n",
    "    matlab_datenum = int(matlab_datenum)\n",
    "    python_datetime = datetime.fromordinal(int(matlab_datenum)) + timedelta(days=matlab_datenum%1) - timedelta(days = 366)\n",
    "    return python_datetime\n",
    "\n",
    "def get_age(row):\n",
    "    try:\n",
    "        age = relativedelta(row['photo_taken'], row['dob']).years\n",
    "    except:\n",
    "        age = None\n",
    "    return age\n",
    "\n",
    "def get_image(path, base='wiki_crop'):\n",
    "    return io.imread(join(base, path))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Read and Prepare Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_metadata = loadmat('./wiki_crop/wiki.mat').get('wiki')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "flattened_metadata = np.squeeze(np.array([row for data in raw_metadata for rows in data for row in rows]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Take only a sample for example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "metainfo_df = pd.DataFrame(\n",
    "    flattened_metadata.T[:sample_size, :4],\n",
    "    columns = ['dob', 'photo_taken', 'full_path', 'gender'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "metainfo_df['photo_taken'] = metainfo_df.photo_taken.apply(lambda x: datetime(x, 1, 1))\n",
    "metainfo_df['dob'] = metainfo_df.dob.apply(get_date)\n",
    "metainfo_df['age'] = metainfo_df.apply(get_age, axis=1)\n",
    "metainfo_df = metainfo_df[metainfo_df.age > 0]\n",
    "metainfo_df = metainfo_df[~metainfo_df.gender.isnull()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(                           full_path gender  age\n",
       " 0  [17/10000217_1981-05-05_2009.jpg]      1   27\n",
       " 1  [48/10000548_1925-04-04_1964.jpg]      1   38\n",
       " 2    [12/100012_1948-07-03_2008.jpg]      1   59\n",
       " 3  [65/10001965_1930-05-23_1961.jpg]      1   30\n",
       " 4  [16/10002116_1971-05-31_2012.jpg]      0   40, (18847, 3))"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metainfo_df.drop(columns=['dob', 'photo_taken'], inplace=True)\n",
    "metainfo_df.head(), metainfo_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 54.2 s, sys: 3.21 s, total: 57.4 s\n",
      "Wall time: 2min 13s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "metainfo_df['data'] = metainfo_df.full_path.apply(lambda path_array: get_image(path_array[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Filter images with 3 channels (RGB)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "metainfo_df = metainfo_df[metainfo_df.data.apply(lambda x: len(np.shape(x))) == 3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Resize images to 100 x 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/rahul/.environments/tensorflow-gpu/lib/python3.5/site-packages/skimage/transform/_warps.py:105: UserWarning: The default mode, 'constant', will be changed to 'reflect' in skimage 0.15.\n",
      "  warn(\"The default mode, 'constant', will be changed to 'reflect' in \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 4min 7s, sys: 3min 35s, total: 7min 42s\n",
      "Wall time: 2min 34s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "metainfo_df['data'] = metainfo_df.data.apply(lambda img: resize(img, (100, 100), anti_aliasing= True) )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Check for missing values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "full_path    0\n",
       "gender       0\n",
       "age          0\n",
       "data         0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metainfo_df.isnull().apply(sum)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "faces   = np.array([img for img in metainfo_df.data])\n",
    "labels  = metainfo_df[['age', 'gender']].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split data into train and test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, Y_train, Y_test = train_test_split(faces.astype('float32'), \n",
    "                                                      labels.astype('float32'),\n",
    "                                                      test_size=0.2,\n",
    "                                                      random_state=123)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_age    = Y_train[:, 0]\n",
    "y_test_age    = Y_test[:, 0]\n",
    "y_train_gender = Y_train[:, 1]\n",
    "y_test_gender = Y_test[:, 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.models import Model\n",
    "from keras.layers import Input, Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "\n",
    "from keras.utils import plot_model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keras Functional API\n",
    "- This is the main branch\n",
    "- These layers are common to both the tasks\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "filter_size = (5, 5)\n",
    "maxpool_size = (2, 2)\n",
    "dr = 0.3\n",
    "\n",
    "inputs = Input(shape=X_train[0].shape, name='main_input')\n",
    "\n",
    "main_branch = Conv2D(16, kernel_size=filter_size, padding=\"same\")(inputs)\n",
    "main_branch = Activation(\"relu\")(main_branch)\n",
    "main_branch = MaxPooling2D(pool_size=maxpool_size)(main_branch)\n",
    "main_branch = Dropout(dr)(main_branch)\n",
    "\n",
    "main_branch = Conv2D(8, kernel_size=filter_size, padding=\"same\")(main_branch)\n",
    "main_branch = Activation(\"relu\")(main_branch)\n",
    "main_branch = MaxPooling2D(pool_size=maxpool_size)(main_branch)\n",
    "main_branch = Dropout(dr)(main_branch)\n",
    "\n",
    "main_branch = Flatten()(main_branch)\n",
    "main_branch = Dense(32)(main_branch)\n",
    "main_branch = Activation('relu')(main_branch)\n",
    "main_branch = Dropout(dr)(main_branch)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- This is where the network branches for multiple outputs/tasks\n",
    "- gender is n x 2 output where as age is n x 1 output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "age_branch = Dense(1, activation='relu', name='age_output')(main_branch)\n",
    "gender_branch = Dense(2, activation='softmax', name='gender_output')(main_branch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "main_input (InputLayer)         (None, 100, 100, 3)  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_8 (Conv2D)               (None, 100, 100, 16) 1216        main_input[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "activation_11 (Activation)      (None, 100, 100, 16) 0           conv2d_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2D)  (None, 50, 50, 16)   0           activation_11[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dropout_9 (Dropout)             (None, 50, 50, 16)   0           max_pooling2d_8[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 50, 50, 8)    3208        dropout_9[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_12 (Activation)      (None, 50, 50, 8)    0           conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_9 (MaxPooling2D)  (None, 25, 25, 8)    0           activation_12[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dropout_10 (Dropout)            (None, 25, 25, 8)    0           max_pooling2d_9[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "flatten_4 (Flatten)             (None, 5000)         0           dropout_10[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 32)           160032      flatten_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_13 (Activation)      (None, 32)           0           dense_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dropout_11 (Dropout)            (None, 32)           0           activation_13[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "age_output (Dense)              (None, 1)            33          dropout_11[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "gender_output (Dense)           (None, 2)            66          dropout_11[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 164,555\n",
      "Trainable params: 164,555\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Model(inputs = inputs,\n",
    "     outputs = [age_branch, gender_branch])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 47,
     "metadata": {
      "image/png": {
       "height": 662,
       "width": 186
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "plot_model(model)\n",
    "Image(retina=True, filename='model.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multi loss optimizations\n",
    "\n",
    "- mse for age and cross-entropy for gender\n",
    "- we can weight these individual losses\n",
    "- Loss = weight1 * loss1 + weight2 * loss2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt = keras.optimizers.RMSprop(lr=1e-3)\n",
    "model.compile(optimizer='rmsprop',\n",
    "              loss={'age_output': 'mse', 'gender_output': 'sparse_categorical_crossentropy'},\n",
    "              loss_weights={'age_output': .001, 'gender_output': 1.})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xd4VFX6wPHvmWTSeyEBQgg9dNDQq7AsYAEbYlcW7GUtP1fX1V3XVdcVXcvaC6KIKIKrWNFFpEkLvfeWUJJQQhJInfP748yQSR+SCWEm7+d5fIaZuXPvmXHy3jPvOfc9SmuNEEII72Jp6AYIIYRwPwnuQgjhhSS4CyGEF5LgLoQQXkiCuxBCeCEJ7kII4YUkuAshhBeS4C6EEF5IgrsQQngh34Y6cExMjE5KSmqowwshhEdatWpVltY6tqbtGiy4JyUlkZqa2lCHF0IIj6SU2ufKdpKWEUIIL1RjcFdKTVFKZSilNlazzVCl1Fql1Cal1AL3NlEIIcTZcqXnPhUYVdWTSqkI4E1gjNa6MzDOPU0TQghRWzXm3LXWC5VSSdVscj3wpdZ6v337DPc0TQjhjYqKikhLSyM/P7+hm3JeCwgIICEhAavVWqvXu2NAtT1gVUr9CoQCr2qtP65sQ6XU7cDtAImJiW44tBDC06SlpREaGkpSUhJKqYZuznlJa83Ro0dJS0ujVatWtdqHOwZUfYELgUuAkcCTSqn2lW2otX5Xa52itU6Jja1xJo8Qwgvl5+cTHR0tgb0aSimio6Pr9OvGHT33NOCo1joPyFNKLQS6A9vdsG8hhBeSwF6zun5G7ui5fw0MVEr5KqWCgD7AFjfst1LbDufw4txtHM0tqK9DCCGEx3NlKuQMYCnQQSmVppSaqJS6Uyl1J4DWegvwI7AeWAG8r7WuctpkXe3OzOX1+TvJyJHgLoSonZCQkIZuQr1zZbbMdS5sMxmY7JYW1SDQzweAU4XF5+JwQgjhkTzuCtUgP3M+OlVY0sAtEUJ4Oq01jzzyCF26dKFr1658/vnnABw6dIjBgwfTo0cPunTpwqJFiygpKeHWW289s+3LL7/cwK2vXoPVlqmtoDM9dwnuQni6v3+zic0HT7p1n52ahfG3yzq7tO2XX37J2rVrWbduHVlZWfTq1YvBgwfz6aefMnLkSP7yl79QUlLCqVOnWLt2Lenp6WzcaLLOJ06ccGu73c3jeu6OtMxpCe5CiDpavHgx1113HT4+PsTFxTFkyBBWrlxJr169+PDDD3nqqafYsGEDoaGhtG7dmt27d3Pffffx448/EhYW1tDNr5b03IUQDcbVHva5NnjwYBYuXMh3333HrbfeykMPPcTNN9/MunXrmDt3Lm+//TYzZ85kypQpDd3UKnlczz3I6si5y4CqEKJuBg0axOeff05JSQmZmZksXLiQ3r17s2/fPuLi4rjtttuYNGkSq1evJisrC5vNxlVXXcUzzzzD6tWrG7r51fK4nrukZYQQ7nLFFVewdOlSunfvjlKKF154gfj4eD766CMmT56M1WolJCSEjz/+mPT0dCZMmIDNZgPgn//8ZwO3vnoeF9z9fC34WhSniiS4CyFqJzc3FzBXgU6ePJnJk8vO5L7lllu45ZZbKrzufO+tO/O4tAyY3rv03IUQomoeGdyD/Hwk5y6EENXw0ODuK7NlhBCiGh4Z3AOtkpYRQojqeGRwN2kZCe5CCFEVjwzugX4+MltGCCGq4ZHBPdjPl9MyoCqEEFXyyOAuaRkhxLlSXe33vXv30qVLl3PYGtd5ZHCXee5CCFE9j7tCFaTnLoTX+OExOLzBvfuM7wqjn6/y6ccee4wWLVpwzz33APDUU0/h6+vL/PnzOX78OEVFRTzzzDOMHTv2rA6bn5/PXXfdRWpqKr6+vvz73//moosuYtOmTUyYMIHCwkJsNhuzZ8+mWbNmXHPNNaSlpVFSUsKTTz7J+PHj6/S2y/PI4B7o58vpohJsNo3FIgvtCiFcN378eB544IEzwX3mzJnMnTuX+++/n7CwMLKysujbty9jxow5q0Wq33jjDZRSbNiwga1bt/L73/+e7du38/bbb/PHP/6RG264gcLCQkpKSvj+++9p1qwZ3333HQDZ2dluf58eGdwdZX/zi0vOrMwkhPBA1fSw60vPnj3JyMjg4MGDZGZmEhkZSXx8PA8++CALFy7EYrGQnp7OkSNHiI+Pd3m/ixcv5r777gMgOTmZli1bsn37dvr168ezzz5LWloaV155Je3ataNr1648/PDDPProo1x66aUMGjTI7e/TI3PujuCeVyCpGSHE2Rs3bhyzZs3i888/Z/z48UyfPp3MzExWrVrF2rVriYuLIz8/3y3Huv7665kzZw6BgYFcfPHF/PLLL7Rv357Vq1fTtWtXnnjiCZ5++mm3HMuZR3Z7A61S9lcIUXvjx4/ntttuIysriwULFjBz5kyaNGmC1Wpl/vz57Nu376z3OWjQIKZPn86wYcPYvn07+/fvp0OHDuzevZvWrVtz//33s3//ftavX09ycjJRUVHceOONRERE8P7777v9PXpkcD+zSHaRzHUXQpy9zp07k5OTQ/PmzWnatCk33HADl112GV27diUlJYXk5OSz3ufdd9/NXXfdRdeuXfH19WXq1Kn4+/szc+ZMpk2bhtVqJT4+nscff5yVK1fyyCOPYLFYsFqtvPXWW25/j0pr7faduiIlJUWnpqbW6rXzt2YwYepKvry7PxckRrq5ZUKI+rRlyxY6duzY0M3wCJV9VkqpVVrrlJpe65E5d1mNSQghquehaRlZJFsIce5s2LCBm266qcxj/v7+LF++vIFaVDMPD+6ScxfCE2mtz2oOeUPr2rUra9euPafHrGvK3EPTMuacJGkZITxPQEAAR48erXPw8mZaa44ePUpAQECt9+GZPXerpGWE8FQJCQmkpaWRmZnZ0E05rwUEBJCQkFDr13tkcD8zoCo13YXwOFarlVatWjV0M7yeR6Zl/H0tWJTk3IUQoioeGdyVUrJIthBCVMMjgztITXchhKiOxwb3YKnpLoQQVfLY4B4oaRkhhKhSjcFdKTVFKZWhlNpYw3a9lFLFSqmr3de8qgX5+XBaCocJIUSlXOm5TwVGVbeBUsoH+Bfwkxva5BJZak8IIapWY3DXWi8EjtWw2X3AbCDDHY1yRaBVBlSFEKIqdc65K6WaA1cA7i9IXA3puQshRNXcMaD6CvCo1tpW04ZKqduVUqlKqdS6XnosA6pCCFE1d5QfSAE+s1d4iwEuVkoVa62/Kr+h1vpd4F0wi3XU5aCm5y4DqkIIUZk6B3et9ZkiEUqpqcC3lQV2dzOzZUo8rnSoEEKcCzUGd6XUDGAoEKOUSgP+BlgBtNZv12vrqhHo54PWkF9kO1NITAghhFFjcNdaX+fqzrTWt9apNWehtOxvsQR3IYQox2OvUA2yL9ghg6pCCFGRxwZ3qekuhBBV89jgLotkCyFE1Tw2uAfKItlCCFEljw3uQbJIthBCVMmDg7ukZYQQoioeG9wD7VMhpecuhBAVeWxwD5KcuxBCVMmDg7t9nrtMhRRCiAo8NrgHWC0oJWkZIYSojMcGd6UUQVap6S6EEJXx2OAOUtNdCCGq4tHBPcjPh9MyoCqEEBV4fHCXnrsQQlTk0cE90L5ghxBCiLI8OrgH+fmQVyBpGSGEKM+jg3ugVQZUhRCiMh4d3IMkLSOEEJXy+OAuPXchhKjIo4N7oJ+PXKEqhBCV8OjgbnruxWitG7opQghxXvHw4O6LTUNBsa2hmyKEEOcVjw7uUtNdCCEq59HB/UxNd5kxI4QQZXh0cHcski31ZYQQoiyPDu5nFuyQtIwQQpTh4cFdFskWQojKeHRwL03LSHAXQghnHh3cpecuhBCV8+jgHnwm5y4DqkII4cyjg3ug9NyFEKJSHh3cIwKthPr7suXQyYZuihBCnFc8Orj7+lgY3D6WX7ZmYLNJfRkhhHDw6OAOMCy5CRk5BWw6KL13IYRwqDG4K6WmKKUylFIbq3j+BqXUeqXUBqXUb0qp7u5vZtWGdohFKZi39ci5PKwQQpzXXOm5TwVGVfP8HmCI1ror8A/gXTe0y2XRIf70bBHB/K0Z5/KwQghxXqsxuGutFwLHqnn+N631cfvdZUCCm9rmsuEd41iXlk1GTv65PrQQQpyX3J1znwj84OZ91mhYchMAft2aea4PLYQQ5yW3BXel1EWY4P5oNdvcrpRKVUqlZma6LxAnx4fSLDxA8u5CCGHnluCulOoGvA+M1VofrWo7rfW7WusUrXVKbGysOw7tOD4XJTdh0Y4sCorlgiYhhKhzcFdKJQJfAjdprbfXvUm1M7xjE04VlrB8d5XDA0II0Wj41rSBUmoGMBSIUUqlAX8DrABa67eBvwLRwJtKKYBirXVKfTW4Kv3bxBBgtfDz5iMMbu++XwVCCOGJagzuWuvranh+EjDJbS2qpQCrDxd3acpnK/dzbe8WdG4W3tBNEkKIBuPxV6g6e/LSTkQG+fHAZ2vJl3VVhRCNmFcF98hgPyaP686OjFz+9ePW0idsJbB6GpRIaWAhROPgVcEdYEj7WG7p15IPl+xl0Q77dMt9v8Gce2Hf4oZtnBBCnCM15tw90WOjO7J4Zxa3fZxKx6ZhXOG/lpuBolMnzUiwEEJ4Oa/ruYNZxGPKrb0Yn9ICf18Le9LSAZizssFmagohxDnllT13gJbRwfx9bBdz57dV8BOs3plO0IZDjO7atGEbJ4QQ9cwre+4VFJha720jFH+avZ79R081cIOEEKJ+NY7gnp8NwJWdI1DAvTNWS5kCIYRXa1TBPdyngMnjurM+LZvRry7i9V92cOCY9OKFEN7Ha3PuZeTbl+ArzGNk53j+c11Ppi3bx4s/befFn7bTPi6Edk1CadMkhK7Nwxme3ASLRTVsm4UQog4aSXA3PXcK8wC4rHszLuvejLTjp5iz7iCr9h5n48Fsfth4CJuGC1tG8uwVXUiOD2vARgshRO01ruBeVDYFkxAZxN1D25ZuVlTCnHUH+ef3W7jktcVMHNiKh0a0J8Dqcy5bK4QQddaocu4U5la7WYDVh2tSWvDLw0MZd2EC7y7czcMz12Gz6XPQSCGEcJ/G0XMvKJuWqUlksB/PX9WN1rHBPPf9VhKjg3h0VHI9NlAIIdzL+4O7zVZmQPVs3DaoNXuPnuKtX3fRKjqYcSkJLN11lOkr9mOzaf4+pjNNwgLqodFCCFE33h/cC3MAe1qlhrRMeUop/j6mMweOneLx/27grQW72JOVR0SQlYIiG2NeX8I7N11I9xYR7m+3EELUgffn3B29dh9/KDz7Oe1WHwtv3HABXRPCiQyy8tK47iz783C+vLs/vj6Kce8sZdaqNMnLCyHOK0rrhglKKSkpOjU1tf4PdHgjvD0AIpMg5wg8cdhtuz6WV8jd01exbPcxmoUHcEm3plzarRndEsKxLzkohBBupZRa5cpSpo2g524fTA1tBsWnzcIdbhIV7Me0iX14eXx3kpuGMfW3vYx9YwkTP0ol42S+244jhBBnq/EE97Bm5vYsB1VrYvWxcEXPBKbc2ovUv4zg8YuTWbIzixEvL2TOuoNuPZYQQrjK+4O7vSJkfQV3Z+FBVm4f3Ibv/ziIVjHB3D9jDY/NXi/5eCHEOef9wb18z72oHguFFeXDu0Npk72cWXf2466hbfhs5QGe/HojDTW2IYRonLx/KuSZnLt9gY6znA55Vo5shINrIH01vm2H86eRHbBpzTsLdhPk58PjF3eUgVYhxDnROIK7NRgCws39ekzLcHCNubVfEauU4rFRyZwuLOG9RXsI9PPloRHt6+/4Qghh1ziCe0AY+IWY+/UZ3A+ttR/z5JmHlFI8dVlnTheW8Nq8HVgUPPA7CfBCiPrVSIJ7OPgFm/v1mZY5uM7cFpws87DFonj+qm7YNLzyvx1oDQ+OaI/Wmv9tyeCDxbu5rnciY3s0r7+2CSEalUYY3OtpQLUoHzK32I95ssLTPhbFC1d3Qyl4dd4OMnIK2Hwwm3Vp2VgU7M40C4lIeWEhhDt4/2yZgpP24F7PaZmMTWArBhQU5FS6iY9F8cJV3Rh3YQIzVuwnK7eQF67qxsd/6ENGTgEzVuyvn7YJIRodz+u5718GC16Aaz4C/9Cat8/Phqg24Bdk7tdXWuagPd/etHuFtIwzi0Xxr6u6cU2vFnRPiMDP15xf+7SK4q1fd3Fd70TpvQsh6szzeu5aw655sH2ua9s70jK+AaAs9ddzP7QWAiMhrkulaRlnFouiV1LUmcAOZpA1I6eAT5dL710IUXeeF9xb9DFz1jf9t+ZttTaBNiAclDKpmXoL7uugaQ8zM6eanntV+rWJpm/rKN5asIv8IvfVvxFCNE6eF9wtFuh0Oez4ucYeMkWnwVZkAi6YQdX6SMsUF8CRzSYl4x9qjlGLAmV/HN6ezJwCpkvvXQhRR54X3AE6Xw4lBTWnZhxXpzouYPILrp/yAxmbzUmkWQ/wt59Iatl7798mmpd+2sbaAyfc3EghRGPimcE9obcp4VtTaqZ8cLcG1U9a5sxgao/SXwlVzJipySvjexAT4s+tH65g2+HSfWitST9xWmrUCCFc4pnB3WIxvfedNaRmHL3nMz33esq5H1pnjhGZVNpzryllVIUmYQFMn9QHf18LN32wnG2Hc/h85X4ue30xA57/hb9/s1kCvBCiRjUGd6XUFKVUhlJqYxXPK6XUa0qpnUqp9UqpC9zfzEp0vgJKCmHbD1Vv4+i5+zulZeoj535orcm3K+XUc69dcAdoERXEJxP7UFRiY+QrC3l09gaKijWjOscz9be9vP7LTjc1XAjhrVzpuU8FRlXz/Gignf2/24G36t4sFzRPgbCE6lMzleXc3d1zLy6EI5tMSgZKTyTV9dy1Niclm63KTdrFhTJtYh9u6tuSmXf048cHBvHmDRdw5QXNeenn7Uxbts+Nb0II4W1qvIhJa71QKZVUzSZjgY+1yRUsU0pFKKWaaq0PuamNlXOkZla8C6dPQGBExW3y7YOSZdIybh5QzdxifkE0swd3V3ru+5fCjGvhxi+h7fAqN+vSPJwuzcPP3FcK/nVVN06eLuKvX29k55EcOjcPp0NcKB3iQ+XiJyHEGe7IuTcHDjjdT7M/VoFS6nalVKpSKjUzM7PuR+50uQms23+s/HlH7/nMVMgg96dlnAdTofSqWcevhsqctC+/d3zPWR/O6mPh9esvYGSneD5beYA/zVrP2DeW0PvZ//HG/J2cKiw+630KIbzPOS0/oLV+F3gXICUlpe6jgs0vhKBo2LMQul9b8fn8bPDxM1enQv2kZdJXQUAERLU2912ZCnnqqLk9caDqbaoRYPXh7ZsupMSm2X/sFNsO5zBr1QEmz93Gh0v2cO9Fbbm5XxIWiywMIkRj5Y7gng60cLqfYH+s/lkskNgP9i2p/HlH6QHH6kd+wWY+enEh+Pq5pw3pq6H5BaXHsAaCxbf6qZB5Web2RN0uVvKxKFrFBNMqJphRXeJZte84k+du5alvNrNy33H+fU13/H0lVSNEY+SOtMwc4Gb7rJm+QHa959udtRwAx/dCdiXnk4KTpT1pcKoM6abUTOEpcwFT8wtLH1PKHLO6AdU8e0qqjsG9vAtbRjLjtr78eXQy360/xM0frCD7dJFbjyGE8AyuTIWcASwFOiil0pRSE5VSdyql7rRv8j2wG9gJvAfcXW+trUzSAHO777eKzzl67g6Omu6uXqVaUwmBw+tBl5QN7lBzfZlT9p57du3SMtVRSnHHkDa8em0PVu8/zri3f2PxjiwKiqVejRCNiSuzZa6r4XkN3OO2Fp2tuC6mp7xvMXQbV/a58sHd6ij760LePTcDXukGN3wBrQZVvk36KnPbrNzU/hp77vace+4RU//GGlhze87S2B7NiQnx565PVnHjB8sJtPrQv000v+8cxyXdmhHi73nVnoUQrvP8v3CLDyT2rbrnHuY0ceds0jJZO6D4tLlAqbrgHpYAoXFlHw8Ir7nnriygbZCdBjHtam5PLQxoG8Oyx4ezbPdRft2WyfxtGczbmsFTczZzcdemjOnRjM7NwogJ8a+X4wshGo7nB3cwefcdP5nedkiT0sfzT5ZOgwSnpfZc6bkfNrfZaVVvk77KDKaW5x8Kx6u5yCgvC2KTTb7+xP56C+4AQX6+DEuOY1hyHFpr1hw4wRepB/hm3SFmrzbvLSbEj45Nw7ilXxLDOzZBKZllI4Sn857gDqb33vny0seryrm7EtxzagjueUfNQO6FEyo+5x9W9WyZkmI4fQySLy4N7ueIUooLEiO5IDGSv17amVX7jrPtSA7bD+fw2+4sJn2cSkrLSB4dnUyvpKhz1i4hhPt5R3Bv1sPk052De3GhSauUCe6VrKN6eIO5n9i37D5z7BN+TlYxq/PgGnNbfjAV7AOqVVzEdPqYuY3raqZMnsPg7izQz4eB7WIY2C4GgKISGzNTD/Dq/3Yw7u2l3NS3JU+N6YyPzJUXwiN5ZlXI8nys0KJ32fnujpy3v3Nwr2RA9acnYM79Ffd5pudeRXBPXwWo0rIDzhw998qqNzrmuIc0MeMB9TBjpjasPhZu6NOSBY9cxKSBrZi2bB/3z1gjs2yE8FDe0XMHk5qZ/xycOgZBURWLhkHlaZlje8y8c61LL0SC0uCel2FWWvItN+iYvgpiO1S+SHdAmBksLcyt+LxjGmRwDEQkNljPvSqBfj48cWkn4sICePb7LZzML+Kla7qTcbKAPVl5nC4sYWTneMKDrA3dVCFENbwruKNh/zKTzy5fNAzAWi64lxSbnLougdPHzUnBIecwoMw+T6aXlhcAcyJIXwXtR1beFuea7uWDu6PnHmQP7rt+qcWbrX+3DW5NeJCVx2avp/ez88o897c5m7i8Z3Nu7Z9Eh/hKTm5CiAbnPcG9+YWmjszexfbgXknP3dfPbOOYCnnSHtjBpEfKB/cmHc2gZ3a54H5iv+mBVzZTBspVhixXQ81RVyY41gT3nMOV/zI4D1yT0oLEqCBW7TtO65hgWsUGU1Ss+WTZPr5cncaMFfu5Y3Br/jQqWXLzQpxnvCe4WwOgzTBY+T60HQYF9gDuPBUSyhYPO7639PHsdLPgBpjXFuaYE0bG5oqDqo6LlyobTAWn4mGVzJjJywSUOZFEJALa/HqIbuPiGz23+raOpm/r6DKP/evqbjw2OpnJP23jnYW72ZWZyyvX9pQLo4Q4j3jHgKrD5W9BbHuYcT1s+cY85txzB5OacZQfcJ6L7jzlMfeIuXUE7/KDngdXg48/NOlceTuqW2ovLwsCI83FVxGJle/fA0QG+/HcFV15emxn5m/L5Oq3fmP/0XpYfFwIUSveFdyDouDmOeaioI2zzGPlg7vzUnvH95rpiD7+ZQOsYxpkVCtTUrj8jJnDG0zKpqrKkmfSMpVMhzyVZQZTAcLtxTTPs0HVs3FzvySmTuhF+onT/P6VBbwxfyeFxVWvMCWEODe8K7hDaYBv0hl8A0vntjuUT8uEt4Dw5mV77o6ZMqFNzXTF8mmZI5tNTZuqVNtzP2oGU8HsW/l4dHAHGNQulh8fGMyQ9rFMnruNUa8uZN6WI5TYZCFvIRqKdyZJg6PhDz+YtEv5S+mdg/uJfRCZZAZVywR3e889JM4Ef+fcfG6mmR4Z16nq41e31N6prNJyAz6+JsDXctGO80nziEDeuSmF+dsyeGrOJiZ+lEqz8ACuvCCBYR2bsCczjzUHjrPlUA7dEyK4vk8ibZuE1LxjIUSteGdwB5OOadqt4uN+IWaWDJig3fEyKCmCXfNLt8k5bHr9AeGmV793celzGZvMbVwV+XbHMZSlip57JrTsX3r/PJzrXhcXdWhC/wej+XnzEb5ITePNX3fy+vydAIT4+9IuLoRpy/YyZcke+rSK4t5hbRnULrbMPnILivl+wyEu7tpUBmmFqKXG95fjF2QW2SjIMdMSI5NM2d2cQybI+1hNcA+NN73+sOYmd16QY+asH7EH96oGU8G+YEdoxZ67rcR+kVVM6WMRLWDPIre/zYbk7+vDpd2acWm3ZhzOzid13zHax4XSJjYEH4siK7eAL1LT+HTFPm6esoKHR7TnnovaopRiT1Yet3+cyo6MXKYs3sN7N6fQIiqood+SEB7H+3LuNXGkZRwzZSJaQngC5mIl+8LVOYdNvh3sz1E6qHpkMwQ3gZCyvc0KKisedvq4OU6wc3BPhJyD5sTiheLDA7i0WzPax4WemQsfE+LPXUPb8NMDQxjbvRkv/rSdu6ev5rv1hxjz+mKycgt4/OJkDp44zdg3lrB899EGfhdCeJ5GGNxD7MF9r7kfmeQUwO3pmpxDpTXaHfXgHamcIxurz7c7VLZgx5mrU53mjUckmlIFVRUo82KBfj68PL4HT1zSkbmbDnPPp6tJjArim/sGcvvgNnx1zwAigqzc+MFy3l+0WwZohTgLjTC426dClgnu9imJjuCee6SSnnuaSatkbq1+poxDZUvtOdeVcfCC6ZB1oZRi0qDWfDKpD/cPb8fsu/qTEGnSMK1jQ/jv3QMY0j6WZ77bwri3f2NnRi4lNs28LUe46YPlDHj+F7JyCxr4XQhx/mmEOfdgQJsg7R9mLijyDTDPZR8wqZTCXJNzBxPklcWkZY7tgeJ8aOJizz2n3DrhjoWxg51SOo4LmQ6sMGmbQ+shvgt0vqJOb9PT9G8TQ/82MRUeDw+08t7NKXy1Nt2sIPXaIpqE+pN2/DRNQv3Jyi3gnQW7+MslLvw/EaIRaXzB3VE8LGMzRLY0g59+QfaLldJK57iH2IO7j6/598l0k5KB6mfKOASEQda2so85Fw1zCE8wc91/+UfpY0HR0HEsWBrfD6vKKKW4omcCA9rG8M/vt5KRk89jo5MZ2TmeR2etZ9qyfdw2uDVNQgMauqlCnDcaX3B3lP3N2GJq0TiEJ9iDu7237ei5g/0ipwPmhKAsptRvTSrLuTuKhjkXKPOxwlXvmW2bdoPDG+Gb++Hw+sprxTdiTUIDeHl82c/kvuHt+HrdQd5ZsJsnL5XeuxAOja9r6AjuRadMvt0hvIU9uNvryjhy7mAP/OlmGmR0W7AG1nwhNZo4AAAgAElEQVQc/9CKC3bkZUFAhAnozrpcBSkTTC2b9qPMY+4oBVxcAB+MhN2/1n1f56lWMcFc3qM5nyzbR8bJ/IZujhDnjUYY3J2uiiwT3KvpuTtKEBzZ6Fq+HUxaxlZkcvQOznVlqhIaZwZsd8+vfjtXHN0FB5bBzv/VfV/nsfuHt6XYpnlrwa6GbooQ543Gm5aBisG9MMfkya1BZRfZCE8wQfr4Xuhxo2vHca4v4+jp52WVzbdXpfVQWPGuudjKrw4X8BzdYW6P7an9PjxAy+hgruzZnOnL9xMT4k9YgC8hAb70ax1DfLjk4UXj1AiDu1OwjGhZ+m/HlMe01NKrUx3CnBbccGWOO5RWoyw4WTpnPi/LtbrtbS6Cpa+bBb/b/c6141XmqLns39uDO8D9w9uxaEcWk+eWDmKHBvjy/JXduKRb00pfk1tQTNrxU7SOCcHPt/H9iBXerREGd6e0jGMaIpTON8/cVrb2C5QGfnBtpgxUXhnyVBYk9qn5tYn9zYpRu+fXLbhnOYL77oprxHqZFlFBLP3zMAqKbeTkF3PkZD5PfLWRez5dzaIdLXhwRHu2Hs5hzf7jrE/LZtvhHNJPnAagc7MwXruuJ21ipZCZ8B6NMLjb0zKhTc3qTQ5nArgum293fs4vBMITcUn5mu42W8W6MlW2MQgS+9V9UNXRcy8+baZ4hlXegy3DZoPXL4Q+d0Gf2+t2/HNMKUWA1YcAqw+xof58cWc/Xv55O28t2MVnKw/Yt4F2TUK4sGUk1/VuQViglX//vJ1LX1vMU2M6cU1KC5QXnwRF49F4g7tzvh1MvRiL1QyChpQL7kExpifdpJPrc8/LL7WXf8KUFq5pQNWhzUXwv6dKi5jVxtGdpVUnj+12LbhnbjXb7v/N44J7eVYfC38alczQDk1Yte843RLC6ZYQTmhA2dlKv+8Uz0Mz1/Lo7A0s2XmUF67uRoDVp4FaLYR7NL5Eo9Wec3fOt4MJ2uH23Hr5YGqxQIs+JuC6yjEg60jLVHYBU3Va249V22mMp47B6WPQbqS5f2y3a69LW2FuHb1+L9C7VRR3DW3DgLYxFQI7mOJmn0zswyMjO/DN+oNc885SDmfLtErh2RpfcLf4QJvh0LaSXLYj7x5aSQ/31m/hosddP075BTvOlB5wMbjHdzNXqtY2NeMIzm0uMksJHndxUPXASvvrd5Wdo+/lLBbFPRe15d2bUtiVkcuY1xez9sCJhm6WELXW+II7wE1fQrdxFR935NZrmwZxVn5AtbKiYdWxWMyUyB0/w+KXYcV7sGGWuTDJFVn2aZCxySY1c7Y996JTFWvjNAIjOsUx++7++PlauPLNJdwzfbUEeeGRGl/OvTruDO4WHzMAW1DLtAxAl6thyzcm9+5w2atw4a01v/boTtNjj0iEqNauBfdTxyBrOyQNgr2LzD7CmrneXi+RHB/GN/cO5J2Fu5m+fB/fbThEz8QIEiKDsPoo/HwsBPn5Ehpg/osO8aN5RBDNIwOJDws4U7deiIYkwd1Z0iDTU3akZ+rKUV9m31LT64aytdxrknwxPJFhVooqzIO3B5pVm1wK7jsgspUpdRDV2lSdrGk6ZPoqc9vj+tLg3mqw6+31IpHBfjw2Opl7h7Xl85UH+GpNOpvSsykotlFYYuNUQTF5hSUVXtck1J/Zd/WX1aNEg5Pg7qz1ELhjgfv2FxAG62bA2k/APxyGPAq+fme3D0fVSr8gSBoI+5a4Nmf96C5TBwdMcC84aQqXVZcWOrDCFEbreBl8+6DZRyMX4u/LxIGtmDiwVYXnSmya3IJisnILSD9+mv3HTvH8D1v5vy/WMeO2vlikBy8akEvBXSk1CngV8AHe11o/X+75ROAjIMK+zWNa6+/d3FbPkzTIlB64cAJ0vbps6YNa7W8AbJxlUizVXelqKzGBue1wcz+qtbk9tqf64J62wlyk5R8KUW28asZMffCxKMIDrYQHWs9cAOXna+FPs9YzZckeJg1qfWbbwmIbSpnpmUKcCzUGd6WUD/AGMAJIA1YqpeZorTc7bfYEMFNr/ZZSqhPwPZBUD+31LJe86N79JQ0yt3sXVR/cs9OgpKC05x5p73Ue2w0telX+GlsJpK2CbteY+9FtTIljcVbGXZjAT5uO8MLcbQxuH0tSdDAf/baX1+btwGJRjOgUxyXdmtK/TTT+vmYuvdaag9n5bEjLZu/RPC7p2lTSOqLOXOm59wZ2aq13AyilPgPGAs5/+RqwTw8hHDjozkYKu+i2EBIHexdXn3d3FAw7E9xbAqr6QdXMbaZwWkKv0tdu+94s3F2+RLGoklKK56/qysiXF3L39NUUl9jYe/QUQzvEEhXkx9yNh5m1yiznaPVRBPn5orXmZH7xmX18/NtePr+jnwR4USeuBPfmwAGn+2lA+QIpTwE/KaXuA4KBOhREEVVSCloOgL015N0dufLodubW198MElcX3B1TIFv0tr+2LdiKzdWtrhQ7E2fEhPjz7BVdufOTVbRtEsLUCb0Y2qEJAAXFJSzekcXmgyc5VVTC6cISSmya9vGhdGlm+ke3friS695bxud39KN5hAtrBwhRCXcNqF4HTNVav6SU6gdMU0p10VrbnDdSSt0O3A6QmOhijRZRVtJA2PRl2bx7fnbZipNHd4JfKIQ0KX1dVKvqg/uBlWYmjyM/H9OudF8S3M/aqC7xzHt4CIlRQWXy7P6+PgzvGMfwjnFVvvaTiX24/v1lXP/eMqbc2otW0cEyOCvOmiujO+mA89zABPtjziYCMwG01kuBAKDCyJ3W+l2tdYrWOiU2Nrb808IVZ/Lui82trQSmXQlv9TeLa4O5gCmmbdmefVTr6q9STVthUjKO1zhSOjKoWmttYkNqNYDaNSGcj/7Qm6ycAoa/tIDkJ3/kohd/5YHP1pCZ4+JFbKLRc+WbtxJop5RqpZTyA64F5pTbZj8wHEAp1RET3DPd2VBhF9POFDnbt8TcX/EepKeaC5Y+vxFOHy87DdIhqrWZCnm6kqst846ai5cSnAZbg6IgMFKCewO5IDGSb+8fxDOXd2HCwCSS40P5YeNhRr+6iCU7s85sZ7NpdhzJIb+o4px70bjVmJbRWhcrpe4F5mKmOU7RWm9SSj0NpGqt5wAPA+8ppR7EDK7eqnUjKkxyLillpkTuXWzy4fOehrYjzBz6D0fDrD+Yxbyjbyj7uij7jJnjeyCwp/m3rQTWfw6/PGvuly+MFt1WgnsDahUTTKuY0umzWw+f5J7pq7nxg+Xc3Lcl2aeLWLgji2N5hXRuFsbHf+hNdIh/A7ZYnE9c+s2otf5ea91ea91Ga/2s/bG/2gM7WuvNWusBWuvuWuseWuuf6rPRjV7SQLOm6+f2Jf8u/beZ4jj6eXuhMV15zx1K8+57FsI7g+Gru0xu/pZvzQLdzqLbyoVM55Hk+DC+uW8gV1+QwEdL97FwRxaD28Xw6KhkdmbkMv7dZdVWs9x3NI+c/KJz2GLRkOQKVU/UcqC5PbQORv2rdEWplIlmYHT9Z6ZgmDNH/fq0VNjxP1j3qSl7fPWH0PmKymfeRLcxV9gW5tV8AVbmdpPGCZGxlPoU5OfL5HHdeWRUB2KC/c8MtPZMjGDSR6mMe+c3pv2hD0lOPf7D2fk8/e0mvt9wGIDWscF0ax7OjX1bkpIU1SDvw+tt/R7WTofxnzTYCmgS3D1RbAdTljisOfS+rfRxpWDMa9D9WojvUvY1fsHmNcveNPn5Qf8Hg/+vdPHuyjh6/8d2Q3xXE+SztkOznmW3K8iFD34HrYbA+GnueY+iWk1Cyy783bd1NNMn9eGWD1cw7KVf6ZkYybDkJvhaFK/N20GxTXP/sLb4+VpYl5bNwh1Z/LjpMFMn9KZv67OodyRcs24GbP3W/L3EdmiQJkhw90RKwYQfIDDCVJ905utf9aIiyZeYL9voF6BJx5qP4zxjJqIlfHIlpK2ESb9AglMKZ/1nZjrmznmmJLGv5H0bQvcWEXxz70BmrUrjl60ZZxYLH9I+ln+M7UJidOlFUUdzCxj/7jImTl3JxxP7cGHLyIZqtndKSzW3+5Y0WHBXDTXumZKSolNTUxvk2MJFhXnwXDMY8ADsX2qqRvoGmgHd6z8322gNb/Y1JQ8Kc+HGL0tr2jQGaz4xM5T639fQLakg42Q+h0/m07V5eKXrwmaczOead5ZyNK+Q6ZP60C0hoszzh7Pz+WptOiM6xcni4WcjOx1e7mT+3XUcXPW+W3evlFqltU6paTupYiSq5hdsUj9LXjU9kas+gAF/hO0/mnw/mIHZzK0w4mmzhOH2H13f/49/NrN9PJXWMP+fsHCyWVj8PNMkLIBuCRFVLvjdJCyA6bf1JSzAyuVvLOHad5cybeleUvce46GZaxn0wi88/8NWrnhjCb85Tb8EKC6xyeBsVdLsq5lFtS69mrwBSHAX1Yu2Xwx11XvQ+XKzaLZ/uAloACveNVe29rjBrBy1/UfXvswFubDyfVj8iufOyDmyEU6mmZRU5taGbk2tNI8I5Mu7+3PvsHZk5hTw5NebuPrtpfy48TA39GnJzDv6ER8ewM1TVjBz5QGO5RXyxvydDH5hPn2em8fiHVk1H6QuCk/Bmunn5uSZtdM9gThtJfj4Q6/bIOcgnNhX933WguTcRfVGPmfSDq3sV8YGhEOfO2DhC7B9rikuNuABsAZA+5HmfsYWiOtU/X73LICSQvPvRS/B5W+WPldcADt+gvajwafcVzQ3wzwf4aYFVerC+VfKgeU1v+fzVFxYAA+NaM+Dv2vH9iO5bDqYzbDkJkQEmbUHZt3Vn3umr+ZPs9fj95WFwhIb/dtEczS3kD9MXclr1/VkVJcqVi/77XWzXOPIZ2vXuHUz4LuHzIwwx3ewPmRsNenFq9435bnrIi0VmvUwnR0wvXfHbLVzSHruonrxXSr+UfW9yywhOPMWQEGviebxdiPNrSupmW0/mJWqet8O6z4r7b1rDd8+ZObwr/m44us+vxGmjHR9Ldn6tH2umTkUHGuCu4dTStEhPpQrL0g4E9gBwgKsTLm1F3cOacO1vVvw04OD+fS2vsy8ox9dmodx9/RVzEw9gPP4ndaaHYdPkvfrK7D0dbL3rK5do/YvM7eOchv1ZfevgDbltOuiuBAOrTVXe8cmm+nB+35zRwvPmgR3cfaCoqDXJCg+bWbgONaeDWsKTXuYoFcdm830zNsON1Myfaym9w6Q+oFZucrHD1Z9VPZ1h9abIHoy3QxkNqTcTNNDaz8aWvTxiuBeHauPhcdGJ/P02C60jwsFIDzIyieT+jCgbQx/mrWeLn+byyWvLeLOaasYMvlX7nt1OsGFpgrJ0g8f49FZ69l88OTZHfjAOQrujqB+YEXd9nNkIxTnQ0KKWeS+5YDSUiHnmAR3UTv97zNFzAb/X9nH248yRcjyjpr7p0/Ath/L5kwPrYXcI2bb0Dhz8dW6z2Dtp/DDo+YXwIh/mO0Oril93aoPwTcA4rvB4pdNL8kV6ath01d1e7/l7fgJ0NBhlAnux3ablFEjE+Tny/u3pPDPK7syLqUFsaH+bD+SQ5vYYJ7rai6ayu4wnt9bVrBp3TIufm0RV765hC9SD3C6sIRThcVsTM/mqzXpzNtyhIwcpytsTx4yJTYCwk0eu6jqq2/rxGYzAdjia1KK5esv/fYf+PBi147vmALpqNPUsr8p+XHy3C9xITl3UTvBMXDrtxUfbz8SFjxvgp+P1cyIycuAsW9AT3u5hO1zAWVq4oCZgZP6gSmFENUarnzXPP6/v5nee7OeUJAD62dC5yuh61XwyVXmKtuaFgs/ugs+vhyK8kzZhuqWGaxK3lH49Tnof7994RNM6im0qTnROFJEB1ZAx0vPfv8ezt/Xh+t6V1LCe+rTENeF8DH/hFe+5cuOS/m42ZN8umI/H86eQ+ycL3ihaBybdVKZl8WHBdAqJpgBhYu4F/g59HJGZH5kCuQlDXT/G8jYZMaVul9vvlNpqdDOaUmKVVPNtR7zn4Xf/6P6faWtLL3AEExwB5OaqWsu/yxJz124V9MeEBIP3z0MsydCeHPz2M9/K+0Rbf/RLAoSbL8yMjTO5PH9w2D8dHNxVmCECeQbvjAzazZ8YebRp/wB2gw3dXAWvWRWiqpKQS58dgNom1l4ZMMXtXtPy940M3tmXGtOMsWFpoZP+5FmJlHT7mZ2hCOF4KwwDzbMgpk3m9vGoiDH5MvbDjf/n3tNxG/rV0zqaGPeyON8E/wPhlrW8EHMDN66vic/PTiYL+7sx5OXdqJP6yiKSmwk5m4gH3/+kTmIEq2Y/eVnLN99tOpjuvpLrjxHymfgA2aBeOcUW9ZOE9jDmpsefE3587SVJiXjmH4a382srdAAqRkJ7sK9LBbocpX5Ixk9GSbNgzH/gdPH4Nd/mp/ah9aalIyz4X+Dh8rNsrnwVhPQN86ClVMgrmvpH86QR81P9vWfV94OreHruyFrmymJ0LS7SfucraJ8kw6K7WiWIpw9CfYuNO1qP9ps4+tvfl3sdwoKhafg63thcjtzkts8B+b9/bycD18v9iwCWxG0tfeA+99nToCfjkPNuhWf+K4w/G80zdnAaMtS2seF0ispiokDW/HqtT2ZdVd/xkTtJ6BlL3748xUcC0umZc5qxr+7jGEv/coLP25l3YETZhDXZoP/3gkvtjMX2p2tvYvNbJbYDhDXpexJeod9/OiGWWab/95pTlyVyc00KRjn0tkWH0js2yCDqhLchfuNeBr+tMvMibf4QNNucOEEMyd+8ctmm/LBXSnwL3cVZIveJqjOfw6ObICUCaU9ona/N78IFk4uze87aG0e3/w1/O7vphxD9+vh8Ho4suns3suGL0wd/ItfgNH/Mr86/nuXyf23Gly2rYfWluZlF/wL1kyDLleaiptXvG1ORg00uFZrte0N7/yfmVHVoq+5H9LE/P87tht63mRSegP+aE7Y/3uqYj67MM8MoCf2Idjfl9guw7nQspN/jmlH0/AA3lm4m7FvLKHn0z/x879vhXUzKCqxUTx1LPs3LOJ4novtttlMcHcsgtOij1kovsS+pu32uWbWS1wn8/8w+wDM/Uvl+0ovl293aNnfXAeRc9i1NrmJBHfhfj6+FevLDHsCAiJgxTsQnuhabRulTO8994gJFN2uKfvciKfNL4F3Bps/SDC50y9uNfnRruNKywJ0vdoMmK2b4fr70BqWv216c0mDTJG2XpPMGEKrIeDntIB1Yl8zb//QWji8wfyE73kjjH3dTCXtOMaknWrz66E6xYXm/X94iUn71HWKaOZ2eH8EvNQRnomDZ2Lh7UGw7C2zlKMrtIadP5vPyLd0SiW/ewr+8JP5Jefrb078I58xJ73lb5fdR/oq0CWlJ4ekQaiSAq5rlsH0SX1J/cvveGlcd16M/YERuV/zfsklDMl5hoMFAUTMuoabn3mHQS/8wkOfr2XGiv3szcqjfKmVU4XFZO1aDfknygb3ojyTh88/aU7G7e1TfBP7mhPS6o9gfSUpvrSVoHxMp8NZu9+bX7JvDzRjSLZzs7CKDKiKcyMoCob/Fb59oDRX7Yru402Jgm7jwT+07HOth8DEuSafPWWkCeTrPzcng989ZQZAHccJjjF/ZOtnwvCnKl4cVZm9i8zUtjH/Kd3PqOfNiabjmLLbJtgXFt/3G2z5xrzfEU6Db35BprTyhi/Mr4Dy76W2Nn1pSkGExJn0T1CM+RwG/LH6z3j/MjPw5xggdpj3d9PL7DgGgiJNLaEdc+HHx+CnJ8x+h/+1+jYd3WUC9oAHyj7u6w+Jfco+1nqo+RW36CVzMnQMeDtSXC3sveDEvoCy97IHEhnsx1VF30DGh9DjRm4Y/Sr9j57iSGZPoudexyzr87wW8RyfbS/hyzVmVdCEyEAGto0h2N+X1L3H2HjwJLeo7/mrFbJiUsy6oI727V8Ox/easRrnX5lDHzcD51/fYwb/HQX08rJgy7cQ17nsSR/MtSKT5sHcx+Gb+83qaaP+Wb8XZSE9d3EuXXAzDHsS+t3t+msCI+HupVVf4disJ9y+wKReFv/bpEsm/gwDH6xYMbP7dSbw755f+b7SV5mLWRx58WVvQ2CU+QXg4GOFEX8vWxUTTB37qDaw6N9wcLU5CQSVq5Xe4wYoOmXSRQ62ErPmbW1oDUtfN2mDh7aYom3NLzCzjP57Z9UpleP74KPLTJVP520ytpoytX3vgsvfgN8/A8P+AncshLuWQqexJghv+m/17dr5s7l1tYDciH+YNMz3j5Re/n9gmUnJBdqrVQZGmPSeY/Bz6RvmhJN8KVz2KoH+vnRqFkav7t0Ivv0H/MPjeeTII6Rekcu8h4fwj7Gd6dQ0jO82HGLasn34W324c0hrxsfuY5+OY+i7O3lnwS52FkSiQ5qaQdXtc82vTceJG8wvkWumoUPjyf9kPH+bNpdNqxfDuxeZMgND/1z5e2x+gankOm6qKVeRVsf59C6QqpDCO9hssPsX8zO+fO7eobgAXuoAbYbB1VPKPrfpKzNYaiuC8BYmkC19w5wkfvc319rw1d1mgYa2I+CGLyr2nLWG11NML3vC92amz+yJJthf+6m5IMz5/cy41pwMLn+r8nILexaaID3mP+bE6TjGwhdh/jOl9fUDwsu+7svbYeNs0yv93d/NLBGAL+8wvzoe3FjxxASmvVNGmbLRdy4qvaR+01fw6/MQ1sycbHfNMzOV7juLv+9FL5lfaIMehouegH8lQZcr4LJXS7eZ+xfT6x30sJma2nGMKWbnnPpxOHUMZlxnThIjnj7zK67EprFpbRYut5XAC63IaX0x9+f9gfnbzAVXb1pf4QKf3firItb4dOe5oP8jwGohPiyAuLAANLBt/Qo+sj3OER1JM8sxfIMjsV4/wwTxmhTlm+9GLUtjS1VI0bhYLGZmRlWBHcwfU5erYet3JiA5Bs3WfgqzJpjplVe+BzHtTWC3+Jocu6s6jDaB+5KXKk+JKAU9rjd53MxtpnzD5q9N4bUfHjMzbBxWf2TSIfuXwTuDTLmG8pa+YdIwXcuNRQx5xJwQ9i0xF98458oPrTepqf73QYeLYcEL5gKb43tNyihlQuWBHcyvlqunAMqs1VuYZ9r9xS3muLkZZsD84BpzcdfZGPiQGV9Z9JLpkRdkl+bbHZIGQkmBCexdrjariFUW2MG8h5u/hk6Xw89/hTn3QtFpfCzKBHYwKbf8bEKTh/HhhN58e99AXh7fneC2A4kni0idzb7ogXSICyU2xJ/0E/l8v+EQs1el0bRdT/YMeZXWlsPsUC25sugZjoS6WFvIGnBO1jyQnrtoXI7ugmlXmJ/QYQnQZqgpZdB6qOk9O5YTzE4z8/LLr2hVE62rz3Vnp8MrXUxv+vRxuPhFM7g89RIY/IgZeM45DK/3NmmIy141A8SH10Ofu8zz/iEmlfN6ikkDDH2s8mPtnGfm+Ue3gVvs4wDTrjDB9/615vhv9IGOl5n2rJkGf1xneuDV2fy1GecIijYzifrcZXrHvn5QdNqcuGLaV8w916SkGD67vnT64f1rStf+BfP/4z8XmCmoY16rmHarjM1mBtcXvWhm5lzzEUS2MmMVv/wDThwwv1Sc33PaKnh/mBkEfWRXhZOd1rq0jPLxvWzKCeaa91eRGB3MzDv6EhpgrbZJR3MLsGmIDa3fnrsEd9H42EpMT3j522bQtMMlpkdqDaj5te4w7UpzEdRlr5ReYTv7Ntj8Fdy9zAxqbvvRjDVEtzHppJ+eMFNJQ5uZqyT3LTGlcB/cVP26tbt+gU+vNXO4Bz1kThQjn4N+95jn5z9npm1afM2ApnMapDo/PGp+8Yz5jykF7S6FeTD1UsjLhAc2VDxRFhdW3Vuvzva5Jh2lbWYQ+fAGaNLZzNZpM6ziMZ5vYWa9TKyhTpLdgu2Z/GHqSsIDrfRoEUGX5uG0iQ3GYm//6cIS1hw4zoo9x9iVmcc9F7XhkZHJNey1chLchXBFdrqZNWI5hxnKk4fML4MWTvOhcw7Df1JMoD6228xIGfRw2dftXw7f/5/pxYOZLz729ZqPt+Nn0yMuKTTTUO9LLU0LFJ4yvfeTaXDfqrI95epobXLwtQm0NSk6baYhhsa5d78n9sMXE0z6aNhfzEB5Vb3/tZ+aHn7Lfi7vfsH2TOasPciG9BPszMjFVi60hgX4kpIURa+kKIZ2iKVj07BavQ0J7kJ4mqVvwtw/Q5NOZoaKTyU/720lJh+/ZrqpwRPdxrV9b/3e9FzHvl6xp31kMxzbZdIz3s4R71ydiltLpwqLOXji9Jn7vhYLiVFBWCx1P64EdyE8TUmxSZF0uQqa1O4ne7VsJa7lqcV5zdXgLhcxCXG+8PE16YL6IoG9UZGpkEII4YUkuAshhBeS4C6EEF5IgrsQQnghCe5CCOGFJLgLIYQXkuAuhBBeSIK7EEJ4oQa7QlUplQnsq+XLYwAX1/zyavI5yGfgIJ9D4/kMWmqtq6kWZzRYcK8LpVSqK5ffejv5HOQzcJDPQT6D8iQtI4QQXkiCuxBCeCFPDe7vNnQDzhPyOchn4CCfg3wGZXhkzl0IIUT1PLXnLoQQohoeF9yVUqOUUtuUUjuVUlWsDOxdlFItlFLzlVKblVKblFJ/tD8epZT6WSm1w34b2dBtPReUUj5KqTVKqW/t91sppZbbvxOfK6XqYe2384dSKkIpNUsptVUptUUp1a8xfheUUg/a/x42KqVmKKUCGtt3oToeFdyVUj7AG8BooBNwnVKqU8O26pwoBh7WWncC+gL32N/3Y8A8rXU7YJ79fmPwR2CL0/1/AS9rrdsCx4GJDdKqc+dV4EetdTLQHfNZNKrvglKqOXA/kKK17gL4ANfS+L4LVfKo4A70BnZqrXdrrQuBz4CxDdymeqe1PqS1Xm3/dw7mj7k55r1/ZN/sI8CNy9Cfn5RSCcAlwPv2+woYBsyyb+LVn4NSKhwYDHwAoHvFvDAAAAIlSURBVLUu1FqfoBF+FzAryQUqpXyBIOAQjei7UBNPC+7NgQNO99PsjzUaSqkkoCewHIjTWh+yP3UYcPNy8eelV4A/ATb7/WjghNa62H7f278TrYBM4EN7aup9pVQwjey7oLVOB14E9mOCejawisb1XaiWpwX3Rk0pFQLMBh7QWp90fk6baU9ePfVJKXUpkKG1XtXQbWlAvsAFwFta655AHuVSMI3kuxCJ+bXSCmgGBAOjGrRR5xlPC+7pQAun+wn2x7yeUsqKCezTtdZf2h8+opRqan++KZDRUO07RwYAY5RSezEpuWGY/HOE/ac5eP93Ig1I01ovt9+fhQn2je278Dtgj9Y6U2tdBHyJ+X40pu9CtTwtuK8E2tlHxP0wAyhzGrhN9c6eV/4A2KK1/rfTU3OAW+z/vgX4+ly37VzSWv9Za52gtU7C/L//RWt9AzAfuNq+mVd/Dlrrw8ABpVQH+0PDgc00su8CJh3TVykVZP/7cHwOjea7UBOPu4hJKXUxJu/qA0zRWj/bwE2qd0qpgcAiYAOluebHMXn3mUAipsLmNVrrYw3SyHNMKTUU+D+t9aVKqdaYnnwUsAa4UWtd0JDtq09KqR6YAWU/YDcwAdNRa1TfBaXU34HxmNlka4BJmBx7o/kuVMfjgrsQQoiaeVpaRgghhAskuAshhBeS4C6EEF5IgrsQQnghCe5CCOGFJLgLIYQXkuAuhBBeSIK7EEJ4of8Hkl6rEcr6CGoAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 95/100\n",
      "4480/9719 [============>.................] - ETA: 2s - loss: 0.8105 - age_output_loss: 538.3462 - gender_output_loss: 0.2721"
     ]
    }
   ],
   "source": [
    "plot_progress = PlotProgress(entity='loss')\n",
    "\n",
    "try:\n",
    "    model.fit({'main_input': X_train},\n",
    "              {'age_output': y_train_age, 'gender_output': y_train_gender},\n",
    "              epochs=100, batch_size=128,\n",
    "              verbose=1,\n",
    "              callbacks=[plot_progress],\n",
    "              validation_split=0.2,\n",
    "             )\n",
    "except KeyboardInterrupt:\n",
    "    pass"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}