2D-CNN-classifier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import division, print_function, absolute_import\n",
    "\n",
    "from keras.models import Sequential, model_from_json\n",
    "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D, BatchNormalization\n",
    "from keras.optimizers import RMSprop\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from keras.callbacks import ReduceLROnPlateau, TensorBoard\n",
    "\n",
    "import h5py\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "sns.set_style('white')\n",
    "\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Hyper Parameter\n",
    "batch_size = 86\n",
    "epochs = 30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up TensorBoard\n",
    "tensorboard = TensorBoard(batch_size=batch_size)\n",
    "\n",
    "with h5py.File(\"./3d-mnist-kaggle/full_dataset_vectors.h5\", 'r') as h5:\n",
    "    X_train, y_train = h5[\"X_train\"][:], h5[\"y_train\"][:]\n",
    "    X_test, y_test = h5[\"X_test\"][:], h5[\"y_test\"][:]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((10000, 4096), (10000,), (2000, 4096), (2000,))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, y_train.shape, X_test.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = to_categorical(y_train, num_classes=10)\n",
    "# y_test = to_categorical(y_test, num_classes=10)\n",
    "\n",
    "X_train = X_train.reshape(-1, 16, 16, 16)\n",
    "X_test  = X_test.reshape(-1, 16, 16, 16)\n",
    "\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.15, random_state=42)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((8500, 16, 16, 16),\n",
       " (8500, 10),\n",
       " (1500, 16, 16, 16),\n",
       " (1500, 10),\n",
       " (2000, 16, 16, 16),\n",
       " (2000,))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, y_train.shape, X_val.shape, y_val.shape, X_test.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Conv2D layer\n",
    "def Conv(filters=16, kernel_size=(3,3), activation='relu', input_shape=None):\n",
    "    if input_shape:\n",
    "        return Conv2D(filters=filters, kernel_size=kernel_size, padding='Same', activation=activation, input_shape=input_shape)\n",
    "    else:\n",
    "        return Conv2D(filters=filters, kernel_size=kernel_size, padding='Same', activation=activation)\n",
    "\n",
    "\n",
    "# Define Model\n",
    "def CNN(input_dim, num_classes):\n",
    "    model = Sequential()\n",
    "\n",
    "    model.add(Conv(8, (3,3), input_shape=input_dim))\n",
    "    model.add(Conv(16, (3,3)))\n",
    "    # model.add(BatchNormalization())\n",
    "    model.add(MaxPool2D(pool_size=(2,2)))\n",
    "    model.add(Dropout(0.25))\n",
    "\n",
    "    model.add(Conv(32, (3,3)))\n",
    "    model.add(Conv(64, (3,3)))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(MaxPool2D())\n",
    "    model.add(Dropout(0.25))\n",
    "\n",
    "    model.add(Flatten())\n",
    "\n",
    "    model.add(Dense(4096, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "\n",
    "    model.add(Dense(1024, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "\n",
    "    return model\n",
    "\n",
    "# Train Model\n",
    "def train(optimizer, scheduler, gen):\n",
    "    global model\n",
    "\n",
    "    print(\"Training...\")\n",
    "    model.compile(optimizer = 'adam' , loss = \"categorical_crossentropy\", metrics=[\"accuracy\"])\n",
    "\n",
    "    model.fit_generator(gen.flow(X_train, y_train, batch_size=batch_size),\n",
    "                    epochs=epochs, validation_data=(X_val, y_val),\n",
    "                    verbose=2, steps_per_epoch=X_train.shape[0]//batch_size,\n",
    "                    callbacks=[scheduler, tensorboard])\n",
    "def evaluate():\n",
    "    global model\n",
    "\n",
    "    pred = model.predict(X_test)\n",
    "    pred = np.argmax(pred, axis=1)\n",
    "\n",
    "    print(accuracy_score(pred,y_test))\n",
    "    # Heat Map\n",
    "    array = confusion_matrix(y_test, pred)\n",
    "    cm = pd.DataFrame(array, index = range(10), columns = range(10))\n",
    "    plt.figure(figsize=(20,20))\n",
    "    sns.heatmap(cm, annot=True)\n",
    "    plt.show()\n",
    "\n",
    "def save_model():\n",
    "    global model\n",
    "\n",
    "    model_json = model.to_json()\n",
    "    with open('model/model_2D.json', 'w') as f:\n",
    "        f.write(model_json)\n",
    "\n",
    "    model.save_weights('model/model_2D.h5')\n",
    "\n",
    "    print('Model Saved.')\n",
    "\n",
    "def load_model():\n",
    "    f = open('model/model_2D.json', 'r')\n",
    "    model_json = f.read()\n",
    "    f.close()\n",
    "\n",
    "    loaded_model = model_from_json(model_json)\n",
    "    loaded_model.load_weights('model/model_2D.h5')\n",
    "\n",
    "    print(\"Model Loaded.\")\n",
    "    return loaded_model\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Isaac\\Documents\\miniconda3\\envs\\tensorflow\\lib\\site-packages\\keras_preprocessing\\image.py:1183: UserWarning: Expected input to be images (as Numpy array) following the data format convention \"channels_last\" (channels on axis 3), i.e. expected either 1, 3 or 4 channels on axis 3. However, it was passed an array with shape (8500, 16, 16, 16) (16 channels).\n",
      "  ' channels).')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Isaac\\Documents\\miniconda3\\envs\\tensorflow\\lib\\site-packages\\keras_preprocessing\\image.py:1404: UserWarning: NumpyArrayIterator is set to use the data format convention \"channels_last\" (channels on axis 3), i.e. expected either 1, 3 or 4 channels on axis 3. However, it was passed an array with shape (8500, 16, 16, 16) (16 channels).\n",
      "  str(self.x.shape[channels_axis]) + ' channels).')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      " - 10s - loss: 2.0765 - acc: 0.2853 - val_loss: 1.3734 - val_acc: 0.4960\n",
      "Epoch 2/30\n",
      " - 9s - loss: 1.4759 - acc: 0.4768 - val_loss: 1.1494 - val_acc: 0.6093\n",
      "Epoch 3/30\n",
      " - 9s - loss: 1.2973 - acc: 0.5379 - val_loss: 1.1178 - val_acc: 0.5967\n",
      "Epoch 4/30\n",
      " - 10s - loss: 1.2355 - acc: 0.5636 - val_loss: 1.0950 - val_acc: 0.6087\n",
      "Epoch 5/30\n",
      " - 10s - loss: 1.1721 - acc: 0.5836 - val_loss: 1.0397 - val_acc: 0.6113\n",
      "Epoch 6/30\n",
      " - 9s - loss: 1.1574 - acc: 0.5875 - val_loss: 1.0263 - val_acc: 0.6120\n",
      "Epoch 7/30\n",
      " - 9s - loss: 1.1334 - acc: 0.5975 - val_loss: 1.0322 - val_acc: 0.6147\n",
      "Epoch 8/30\n",
      " - 10s - loss: 1.1138 - acc: 0.6024 - val_loss: 0.9909 - val_acc: 0.6427\n",
      "Epoch 9/30\n",
      " - 10s - loss: 1.0947 - acc: 0.6126 - val_loss: 1.1797 - val_acc: 0.6020\n",
      "Epoch 10/30\n",
      " - 10s - loss: 1.0744 - acc: 0.6145 - val_loss: 0.9982 - val_acc: 0.6467\n",
      "Epoch 11/30\n",
      " - 10s - loss: 1.0447 - acc: 0.6307 - val_loss: 0.9332 - val_acc: 0.6647\n",
      "Epoch 12/30\n",
      " - 10s - loss: 1.0610 - acc: 0.6224 - val_loss: 0.9514 - val_acc: 0.6540\n",
      "Epoch 13/30\n",
      " - 10s - loss: 1.0345 - acc: 0.6285 - val_loss: 1.0039 - val_acc: 0.6500\n",
      "Epoch 14/30\n",
      " - 10s - loss: 1.0179 - acc: 0.6367 - val_loss: 0.9753 - val_acc: 0.6567\n",
      "\n",
      "Epoch 00014: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
      "Epoch 15/30\n",
      " - 10s - loss: 0.9954 - acc: 0.6428 - val_loss: 0.9741 - val_acc: 0.6653\n",
      "Epoch 16/30\n",
      " - 10s - loss: 0.9425 - acc: 0.6629 - val_loss: 0.9105 - val_acc: 0.6820\n",
      "Epoch 17/30\n",
      " - 10s - loss: 0.9697 - acc: 0.6554 - val_loss: 0.9463 - val_acc: 0.6733\n",
      "Epoch 18/30\n",
      " - 10s - loss: 0.9201 - acc: 0.6743 - val_loss: 0.9242 - val_acc: 0.6800\n",
      "Epoch 19/30\n",
      " - 10s - loss: 0.9387 - acc: 0.6638 - val_loss: 0.9466 - val_acc: 0.6753\n",
      "\n",
      "Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
      "Epoch 20/30\n",
      " - 10s - loss: 0.8952 - acc: 0.6798 - val_loss: 0.9173 - val_acc: 0.6847\n",
      "Epoch 21/30\n",
      " - 10s - loss: 0.8946 - acc: 0.6794 - val_loss: 0.9231 - val_acc: 0.6793\n",
      "Epoch 22/30\n",
      " - 10s - loss: 0.8704 - acc: 0.6889 - val_loss: 0.9194 - val_acc: 0.6787\n",
      "Epoch 23/30\n",
      " - 10s - loss: 0.8719 - acc: 0.6856 - val_loss: 0.9089 - val_acc: 0.6907\n",
      "Epoch 24/30\n",
      " - 10s - loss: 0.8768 - acc: 0.6892 - val_loss: 0.9130 - val_acc: 0.6860\n",
      "Epoch 25/30\n",
      " - 10s - loss: 0.8646 - acc: 0.6866 - val_loss: 0.8703 - val_acc: 0.7000\n",
      "Epoch 26/30\n",
      " - 10s - loss: 0.8545 - acc: 0.6949 - val_loss: 0.8877 - val_acc: 0.6913\n",
      "Epoch 27/30\n",
      " - 9s - loss: 0.8512 - acc: 0.6924 - val_loss: 0.8861 - val_acc: 0.6960\n",
      "Epoch 28/30\n",
      " - 10s - loss: 0.8459 - acc: 0.6955 - val_loss: 0.8970 - val_acc: 0.6907\n",
      "\n",
      "Epoch 00028: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n",
      "Epoch 29/30\n",
      " - 10s - loss: 0.8292 - acc: 0.7060 - val_loss: 0.8872 - val_acc: 0.6867\n",
      "Epoch 30/30\n",
      " - 10s - loss: 0.8216 - acc: 0.7093 - val_loss: 0.9223 - val_acc: 0.6900\n",
      "0.6875\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABBkAAARdCAYAAAAUtpqIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzs3XmclQW9P/DPLOwzikgmbuSCa5KRqRVZpKSZ5s+lNAu9buUSLrmkqKC5YC6pV9y1tDI1JBUV1ETNJdOudS31XrewTLmWgsIM+5zz+4N7MUspm+fM4xze79eL14s5Z+aZz3l8OM585nu+01CtVqsBAAAA6KTGsgMAAAAA9UHJAAAAABRCyQAAAAAUQskAAAAAFELJAAAAABRCyQAAAAAUornsAAAAAEDXWbRoUcaMGZOXXnopCxcuzMEHH5xVV101Bx10UD7wgQ8kSb785S9nhx12yIQJE3Lfffelubk5Y8aMydChQ5d5bCUDAAAALEcmT56c/v375+yzz86sWbOyyy675NBDD82+++6b/fbbb+n7Pfnkk3n00UczceLEzJgxI6NHj86kSZOWeWwlAwAAACxHtt9++2y33XZL325qasoTTzyR6dOnZ9q0aRk8eHDGjBmTxx57LMOHD09DQ0NWW221dHR0ZObMmRkwYMA7HrvwkuH1L48o+pD8lYGTnik7AhSisaGh7Ah1r8E57hKVSqXsCHWvWnaA5YBni9pzHVNPFi98qewINbPo1d+XHaFmegxcZ+nf+/XrlyRpa2vLYYcdliOOOCILFy7MF7/4xXzwgx/MJZdckosuuiitra3p37//Wz5uzpw5yywZLH4EAACA5cyMGTOy9957Z+edd85OO+2UkSNH5oMf/GCSZOTIkXnqqafS0tKS9vb2pR/T3t6e1tbWZR5XyQAAAADLkVdffTX77bdfjjnmmOy+++5Jkv333z+//e1vkyQPP/xwNtlkkwwbNiwPPvhgKpVKXn755VQqlWVOMSR2MgAAAMBy5dJLL83s2bNz8cUX5+KLL06SHHfccTnjjDPSo0ePDBw4MKeeempaWlqy+eabZ4899kilUsnYsWP/4bEbqtVqoS8Rs5OhtuxkoF7YyVB7djJ0DTsZas9r2WvPs0XtuY6pJ3YydE9/vZOhlkwyAAAAQJJUOspO0O3ZyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIkqqFzp1lkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQJBU7GTrLJAMAAABQCCUDAAAAUAglAwAAAFAIJQMAAABQCIsfAQAAIEm1avFjZ5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACRJxU6GzjLJAAAAABRCyQAAAAAUQskAAAAAFMJOBgAAAEiSqp0MnWWSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJAklY6yE3R7JhkAAACAQigZAAAAgEIoGQAAAIBCKBkAAACAQlj8CAAAAElSrZSdoNszyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIkoqdDJ1lkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQpFq1k6GzTDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAkqRiJ0NnmWQAAAAACqFkAAAAAAqhZAAAAAAK0e1LhqZ1N0rLSef9/e3rbJCWcRekZdy/p+8RJyc9erzrYzcP+1haTrskLadMSM/PfH7JjX36pd/Rp6dl7PlpOWVCmoZs3MlHUL8aGhpy0YQz8+D9kzPtZxOz7rofKDtSXXKea6+5uTnf/94FuWfapDz04G3ZcceRZUeqO42NjbnssnNy770/zd1335h11hlcdqS6tcVHP5y7fzax7Bh1y3Ny13Et147ruPacY+rZP734sVKppLHxvdVJ9Nppz/QcPjLVBfP/7r6+Bx6d9vPHpfLKy+k5Yoc0Dlw1lRkv/vMHb2pKn1GHpu3Eg1KdPz8tp1yYRY/9Ir1G7pzFT/46C6ZOSuOgNdN39IlpG/P1Ah9V/dh55+3Tu3evDN/6C9lyi2E5+6yx2XW3/cqOVXec59rba69d89rMWdl3v8MzYED/PPrInbnttp+VHauu7Pj5JcXNiBG7Zuutt8pZ3xmb3b+4f8mp6s9RRx2cr35lt7S3zys7St3ynNw1XMu15TquPef4Paxq8WNnLbM1ePHFF3PIIYdk6623zrbbbptPf/rT+drXvpbp06d3Vb5lqrzyctrPG/t3tzcOWjOVttnp9bnd0zL2/DS0rLCkYGhqSp+vHbNkCmHcv6d5ow+95eNWuGTSm8dYfXAqr7yUantb0rE4i5/+XZo3HJoFUyZmwd23LnmnpqZk0cKaPsbubPjHt8idd92bJHnk0V/nI8OGlpyoPjnPtTdp0m05+eSzl769uGNxiWnq0+Rb78whh3wrSbLWWmvkz3/+S8mJ6tPvf/+HfPFLB5Ydo655Tu4aruXach3XnnNMPVvmJMMJJ5yQo446Kh/60JvfjP/nf/5njj/++Fx//fU1D/ePLHr0/jQOfP/f3d7QumKa198kc67+91T+50/pd8z4dPz+mTQOWiPVOW+k7fKz09CyQlrGXZA5x+ybft86Mw09e6WhpTUtJ52XysxXs+DuW1Kd2/7mQefNS0Pffktva1hxpfQ9dEzm/eCirnq43U7rCi2Z/cacpW93dFTS1NSUjo6OElPVH+e59trb5yZJWlr65frrLs/J487+Bx/Bv6KjoyNXXvnd7PyF7fPlvQ4qO05duummKRk8eI2yY9Q1z8ldw7VcW67j2nOOqWfLLBkWLlz4loIhSTbbbLOaBipCte2NVP7npVRe+kOSZPHjj6ZpnfXT+L5V07zBpmleb6Ml79jYmIaWFdL+neOSLJlkaDv1yCV3rbVOGnr3efOgffosmWpI0rjm2ul32NjM+9El6fivx7vugXUzc2a3paW1ZenbjY2NnjhrwHnuGmusMSgTf3JlLr3sB7n+hpvLjlO3Djjgmznh/ePzwAOTs9lmn8ncuUah6V48J1MPXMe15xxTz5b5cokNNtggxx9/fKZMmZIHHnggd9xxR44//vhssMEGXZXvX1J5ZUbSu08a379akqRpw03T8acX0vHSH7PwF/ek7dQj03bmt7LokZ+n2j7n7Y/x0h/SuOoaaejXmjQ1p3nDD2Xxs0+lcfXB6Xf4yWm/8LQsfvzRrnxY3c5DD/8qn9v+M0mSLbcYliee+K+SE9Un57n2VlllYKbc/uOMOeGMXHPNDWXHqUt77bVrjjnm0CTJ3LnzUq1U0tHhNZF0P56TqQeu49pzjt/DKh31+6eLLHOS4eSTT87dd9+dxx57LG1tbWlpacmIESMycuR7c7N6j49vk4befbLwntsy7/Kz0/cbJyYNDel45sks/s0vk+Ye6Xvg0WkZe37Sp28W/uyWpFpd+vGzD97tzYN1dGTejy5Ov+PPShoas/C+qanOejV99zsiDT17pu8+30iSVOe2p/3cE7v6oXYLN988Ndtus3Ue+PktaWhoyP4HHll2pLrkPNfet741Ov37r5gxxx+RMccfkSTZ6QujMn/+3y+d5V9z881Tc8Xl5+buu29Mj+bmHH30KVmwYEHZseBd85xMPXAd155zTD1rqFb/6rvsArz+5RFFHo6/MXDSM2VHgEI0NjSUHaHuNTjHXaJSMXFRa4V+ocLb8mxRe65j6snihS+VHaFmFvz3z8uOUDO9NvxUl3ye99bvpAQAAAC6rWW+XAIAAACWG1UTip1lkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQJH5rVKeZZAAAAAAKoWQAAAAACqFkAAAAAAqhZAAAAAAKYfEjAAAAJEnV4sfOMskAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASJKKnQydZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKRa7Sg7QrdnkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQJNVK2Qm6PZMMAAAAQCGUDAAAAEAhlAwAAABAIZQMAAAAQCEsfgQAAIAkqVj82FkmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJUrWTobNMMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAACSpNJRdoJuzyQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEmqlbITdHsmGQAAAIBCKBkAAACAQigZAAAAgEIoGQAAAIBCWPwIAAAASVKx+LGzTDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAkqRqJ0NnFV4yDJz0TNGH5G/Me/mBsiPUvT6rfbLsCHWvUq2WHaH+OcfAP6mhoaHsCHWv6jkZWE54uUQ3o2AAAADgvUrJAAAAABTCTgYAAABIkoqdDJ1lkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQ2MlQAJMMAAAAQCGUDAAAAEAhlAwAAABAIZQMAAAAQCEsfgQAAIAk1WpH2RG6PZMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgCSpVMpO0O2ZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkSdVOhs4yyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIkoqdDJ1lkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQJFU7GTrLJAMAAABQCCUDAAAAUAglAwAAAFAIJQMAAABQCIsfAQAAIEkqFj92lkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJU7WToLJMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgCSp2MnQWSYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAImdDAUwyQAAAAAUQskAAAAAFELJAAAAABRCyQAAAAAUwuJHAAAASJKqxY+dtdyWDA0NDZlw4fh8aOjGWbBgQb520DF5/vkXyo71nrNo8eKcdMZ5eXnGK1m4aFG+vs+XM+KTWy29/5rrf5qf3npnVlppxSTJuGMOy9qD13hXn+PGyVPzk5unprmpMV/7ty/n05/YMjP+5885afx5WdzRkVSrGXfs4e/6uMsL13LtOce15xzXnnPcNZzn2mtubs4Vl5+bwYPXSK9evTL+zAty220/KztWXXEd155zTD1bbkuGnXfePr1798rwrb+QLbcYlrPPGptdd9uv7FjvObfdeU/6r9CaM8cek9ffmJ3d9/3GW0qGp55+LmecdHQ22XDIv3T8V1+bmWsnTs4NV12QBQsXZe+Dj87HP/rhXHjFD/Ll3XbKNlt/PA898ljOv/T7uWD8SUU9rLriWq4957j2nOPac467hvNce3vttWtemzkr++53eAYM6J9HH7lTyVAw13HtOcfUs+W2ZBj+8S1y5133JkkeefTX+ciwoSUnem/absQn89lPD1/6dnNT01vuf+rp53LlD2/IqzNnZeuPbZED994jc9raM3b8eXn9jTlJkuOPPCjrr7t2kuSlGa/kmLHj8+Mrzk+S/O6pZ7LZphunZ8+e6dmzZ9ZcY1Cefn56jhl9YFpa+iVJFnd0pFevnl3xcLsl13LtOce15xzXnnPcNZzn2ps06bb89Ke3L317ccfiEtPUJ9dx7TnH1LPltmRoXaEls//3m+Ak6eiopKmpKR0dHSWmeu/p27dPkqS9fW6OPOH0jD5w77fc/7ltP5Uv77pjWvr1zWHHn5r7Hnokv378iWy5+WbZc5cd84cXX8qJZ3w3F5/97Yw+7pQsXLAwz7/wx/zbN47NJhsMyYbrr5vWlr5Lj9evb9+0tc3NSv2XvPxi+h/+lHMmXJl/Hz+26x50N+Narj3nuPac49pzjruG81x77e1zkyQtLf1y/XWX5+RxZ5ecqP64jmvPOX4Pq9jJ0FnLbckwZ3ZbWlpblr7d2NjoH/U7mPHKX3L48admz10/n89/dsTS26vVakZ96f+l9X8nDj718S3y3888n2effyGPPPZ47ph2f5Jk9uy2tLb0y9UTzlo6yXD1hLOSJPc+8Mu0z5239Jjtc+cuPd6jjz2eU8+9KONPOto+hmVwLdeec1x7znHtOcddw3nuGmusMSgTf3JlLr3sB7n+hpvLjlN3XMe15xxTz5b5KyxHjRqVPffc8y1/9thjj+y5555dla9mHnr4V/nc9p9Jkmy5xbA88cR/lZzovenVmbPytSNPyDcP2Te77rjdW+5ra5+b/zfqoMydOy/VajWPPPaf2XiD9bL24DWz9x675OoJZ+XcU8e8pZj4W5tuvH5+/fiTWbBgYea0tWf6Cy9myDofyKOPPZ7xF1yay757aj640fq1fpjdmmu59pzj2nOOa8857hrOc+2tssrATLn9xxlzwhm55pobyo5Tl1zHteccU88aqtVq9Z3ufPzxx3PiiSfmoosuStPfvBZ/9dVXf9uPae759re/1/zfRtehm26UhoaG7H/gkXn66efLjvUPzXv5gS79fOPPvzR3TLv/LZMEu++0febNn58v7rxDJt8xLddOvCU9e/bIlh/ZLN84YFRef2N2xo4/P7Pb2tLePjeH7PfVtyyL/Fs3Tp6aibdMTbVazYGj9sjIEcOz6z6HZNHCRVl55ZWSJGuvtUbGHXtYzR/v/+mz2ie77HN1Vne9lrsT57j2nOPac467Rnc9z40NDWVH+Kede+4p+eLuO73lvO70hVGZP39+ian+sco7f8n9ntNdr+PupLuf48ULXyo7Qs3Mm3xO2RFqps8Xju6Sz7PMkiFJrrzyygwePDgjR478pw7YXUqG7qqrS4blVXcqGQCgs7pTydBddaeSAf6Rui4Zbjmr7Ag102fnY7vk8/zDnQwHHHBAV+QAAAAAurll7mQAAAAA+GcpGQAAAIBCLLe/whIAAADeolIpO0G3Z5IBAAAAKISSAQAAACiEkgEAAAAohJIBAAAAKITFjwAAAJAkVYsfO8skAwAAAFAIJQMAAABQCCUDAAAAUAg7GQAAACBJKnYydJZJBgAAAKAQSgYAAACgEEoGAAAAoBB2MgAAAEBiJ0MBTDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAkqRaLTtBt2eSAQAAACiEkgEAAAAohJIBAAAAKISSAQAAACiExY8AAACQJJVK2Qm6PZMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgMROhgKYZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkSdVOhs4yyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIkoqdDJ1lkgEAAAAohJIBAAAAKISSAQAAACiEkgEAAAAohMWPAAAAkCTVatkJuj2TDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAkqVTKTtDtmWQAAAAACmGSAQAAAJYjixYtypgxY/LSSy9l4cKFOfjgg7PeeuvluOOOS0NDQ4YMGZJx48alsbExEyZMyH333Zfm5uaMGTMmQ4cOXeaxlQwAAACwHJk8eXL69++fs88+O7Nmzcouu+ySDTfcMEcccUS23HLLjB07NtOmTctqq62WRx99NBMnTsyMGTMyevToTJo0aZnHLrxkWKFX36IPyV9Zaa1tyo6wXHh2o43LjlD3PvaHP5Udoe69sWBu2RGWCz0am8qOUPcWdiwuO0Ld22zAOmVHqHuPz5pedoTlwiLPF3TWcrKTYfvtt89222239O2mpqY8+eST2WKLLZIkW2+9dR566KGsvfbaGT58eBoaGrLaaqulo6MjM2fOzIABA97x2HYyAAAAwHKkX79+aWlpSVtbWw477LAcccQRqVaraWhoWHr/nDlz0tbWlpaWlrd83Jw5c5Z5bCUDAAAALGdmzJiRvffeOzvvvHN22mmnNDa+WQ+0t7dnhRVWSEtLS9rb299ye2tr6zKPq2QAAACA5cirr76a/fbbL8ccc0x23333JMnGG2+cRx55JEly//33Z/PNN8+wYcPy4IMPplKp5OWXX06lUlnmSyUSix8BAABgierysZPh0ksvzezZs3PxxRfn4osvTpKccMIJOe200/Ld734366yzTrbbbrs0NTVl8803zx577JFKpZKxY8f+w2M3VKvVapFhB7QOKfJw/A3Lr7rG74asV3aEumfxY+1Z/Ng1LH6sPf/vqz2LH2vP4seuYfFj11i88KWyI9TMvCu/WXaEmulzwHe75PN4uQQAAABQCCUDAAAAUAglAwAAAFAIix8BAAAgSbVS6MrC5ZJJBgAAAKAQSgYAAACgEEoGAAAAoBB2MgAAAECSVCplJ+j2TDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAkqRqJ0NnmWQAAAAACqFkAAAAAAqhZAAAAAAKYScDAAAAJEmlWnaCbs8kAwAAAFAIJQMAAABQCCUDAAAAUAglAwAAAFAIix8BAAAgSSqVshN0eyYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAImdDAUwyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIkmq17ATdnkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJUKmUn6PZMMgAAAACFUDIAAAAAhVAyAAAAAIVQMgAAAACFsPgRAAAAkqRSLTtBt2eSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJAk1UrZCbo9kwwAAABAIZQMAAAAQCGUDAAAAEAhlsudDB/Z/EMZ9+1j8oUdvrr0ttPHj8mzz07P1d+7rsRk9eUrX90tX/3q7kmSXr17ZejQjbPu2h/NG2/MKTnZe1OvTTfMSocfkP854Oi3vX/lk45IZfaczLrgqnd97D6f2iorfe2rqXZ0pO3mOzLnp1PT0NI3q5xxXBr69UtDj+bMPOfSLPjtf3X2YdSdD39kaE465ajsuuM+2fRDG+cH112c6b//Q5Lkmquuzy03TS05Yf14+OEpmT17dpLkhRdezNe/fkzJiepLc3NzLrvinKy11urp6Khk9DeOz7PP/L7sWHXpfe9bOQ8/fHt22OEreeaZ58uOUxcaGxtz/DlHZ/C6a6ajo5LTjvxODjrugKz8vgFJkkFrrponfv1UTjr42yUnrR+ek2uroaEhEy4cnw8N3TgLFizI1w46Js8//0LZsUiSSrXsBN3eclcyjD7iwOyx586ZO3dekmTlgQNyyWVnZd311s6zF1xZcrr6cu2PJuXaH01Kknz3vG/nhz+YqGB4Byv+25fSsuO2qcyb/7b3t+7++fQcsnbmP/bbd3/w5qasfPRBeXmvb6Qyb35Wu+b8zP35L9P6pZ0y75HfZPa1N6XH4DXyvu+Myct7HtLJR1JfDj1s/+y+xxeWPl9sOnTjXHbx1bl0wtXlBqtDvXr1SpJst92eJSepX5/d7tNpamrKyG2+mBGfGZ6x447OqK/4N1+05ubmTJgwPvPe4fmcf83wz348SfK1nUdn2Mc2y+EnH5Jj9z0xSdK6YksuuvH8nD9uQpkR64rn5Nrbeeft07t3rwzf+gvZcothOfussdl1t/3KjgWFeNcvl1i4cGEtcnSZF37/x+z9lUOXvt2vX998Z/yF+cn1N5eYqr59eNim2WijIfm+KZF3tOjFl/PKN0952/t6Dd0ovYZulDk33v7mjc1NGXjyNzPoe+dm0NXnpffmQ9/yMWtOu2Hp33usvVYWvfhyKnPaksWLM/83T6TXsE0z+0eT3jxmc1OqC7r3v+1aeOGFP2a/UYctfftDm22SbT/7qdw05Yf57oWnpV9L3xLT1ZehQzdK3769c+utP8zUqddliy0+XHakuvPcc9PT3NychoaGtLa2ZPGiRWVHqktnnnlirrjiR5kx45Wyo9SV++94MGcec26SZNU13p+Zf5m19L4Dj943E7/307z255llxas7npNrb/jHt8idd92bJHnk0V/nI8OG/oOPgO7jHUuGe+65JyNGjMjIkSMzZcqUpbcfcMABXRKsVm6dfGcWLVq89O0//uFPeew/Hi8xUf075phDMv6MC8qO8Z42d9qDyeLFf3d708AB6X/w3nntjAvfcnvrLp9Lx+uzM2O/o/LK4WOz8pjRSZL3X3R6Vr3ynDSt2JpVrzwn7zvjuDS29E2lrX3px1bmzktjS79U5rSnumBhmlZeKe8747jM+vfv1fZBdkO3T/5ZFi9+8xux3/z6t/n2SWdnlx1G5Q8vvJijv3XoMj6ad2Pu3Hk5//zLs9NOozJ69Jh8//sXpKmpqexYdaW9bW7WGrx6HvvN3blwwhm55JJryo5Ud0aN2j2vvvpa7r77/rKj1KWOjo6cdP5xOeq0w3LP7T9Pkqy0cv9sPnxYbr/hjpLT1RfPybXXukJLZv/VhG9HR8U5pm6848slLr300tx0002pVqs5/PDDs2DBguyyyy6pVr1GhX/eiiu2Zv3118399/+y7CjdUr/Pbp2m/ivk/RNOT9PAldLYu3cWTX8xPYesnd7DNk2vD2645B2bmtK4YmteOfSEJEsmGf5vt0OPIWunse+bP3Fv7NtnyVRDkh7rfSCrnHVCZp57+b/2UozlzJTb7l76BcHU2+7O6WedUHKi+vHss9OXvhb1ueemZ+bMWRk0aJX86U8zyg1WRw4dvV+m3f1AThl3dlZffVBum3Jtttpi+ywwxVSYffbZI9VqNZ/5zPAMHbpxrrrqvOy++/555ZW/lB2tbpx6xJm56PTLc9XtF+fLn/q3fGbHT+Wum6alUvF77YvkObn25sxuS0try9K3Gxsb09HRUWIi/k/V80mnvWPJ0KNHj/Tv3z9JcvHFF2efffbJoEGD0tDQ0GXh6P4+MXzL3HvvQ2XH6LZm//jmzP7xkpfytHzhs+mx9pppm3xXGlv6ZvErr+aNq65LQ6+e6X/AXqnMbnvbYyya/sf0WGv1NK7Qmsrceen9kU3zxg8mpsc6a2WVc07KX449PQstf/unXD/pipxw7On5za9/l09+aqv89j+fKjtS3dhnny9lk002zBFHnJhBg1ZJa2tLZsz4c9mx6srrs97Iov+dmJo16/U092j2U7OCbbvtF5f+/a67bsg3vjFGwVCQ7XcbmVUGvS8/mPDjzJ83P9VKNZVKJR/95Efy/Qt+WHa8uuM5ufYeevhX2fHzI3Pjjbdmyy2G5YknLN+mfrxLpY6TAAAgAElEQVRjybD66qtn/PjxOfzww9PS0pIJEyZk//33X7plFv4ZQ4ask+nT/1h2jG6n3+dGpLFvn8yZNOVt75898fYMHHdkVr3q3DS29M2cGyYnfzVl9OI2e7z5zos78tq5l2bVS8YnjQ2Zc/Od6fjza1l5zGFp6NkzA45dsvit0taePx8xrqaPq7v71lHfzhlnnZhFixblz6+8mqOPGFt2pLpx9dU35Iorzs20aTemWq3moIOO9ROdgl004Xu56JLv5I67bkjPnj3y7ZPPWbrUFN7r7pvyQE4871u55KcXpLm5OeeNm5CFCxZmrXXXzEt/8NP1onlOrr2bb56abbfZOg/8/JY0NDRk/wOPLDsSFKah+g6vf1i8eHEmT56cz33uc+nTp0+S5NVXX81ll12WE0545xHhAa1DapOUJMnCjr9/3T7F+92Q9cqOUPc+9oc/lR2h7r2xYG7ZEZYLPRpNA9Sa//fV3mYD1ik7Qt17fNb0siMsFxZ5vugSixe+VHaEmmkfv0/ZEWqm3/Fds4/pHScZmpubs+uuu77ltoEDBy6zYAAAAACWX+9YMgAAAMBypeIXHXTWO/4KSwAAAIB3Q8kAAAAAFELJAAAAABTCTgYAAABIkmql7ATdnkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJUqmUn6PZMMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAACSpFIpO0G3Z5IBAAAAKISSAQAAACiEkgEAAAAohJIBAAAAKITFjwAAAJAklWrZCbo9kwwAAABAIZQMAAAAQCGUDAAAAEAh7GQAAACAJKlWyk7Q7ZlkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACRJpVp2gm7PJAMAAABQCCUDAAAAUAglAwAAAFAIOxkAAAAgSbVSKTtCt2eSAQAAACiEkgEAAAAohJIBAAAAKISSAQAAACiExY8AAACQJJVq2Qm6PZMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgMROhgKYZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkSbVSdoJuzyQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEkq1bITdHsmGQAAAIBCKBkAAACAQhT+col5ixcWfUj+SlODXqgrbPXCi2VHqHt39l+/7Ah1b8RrT5UdYbnQ4Vdd1VzFOa65GQtmlh2h7q3Qs0/ZEZYLbYvmlx0Blnu+YwUAAAAKYfEjAAAAJKla/NhpJhkAAACAQigZAAAAgEIoGQAAAIBC2MkAAAAASWInQ6eZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkSaVSdoJuzyQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEkq1bITdHsmGQAAAIBCKBkAAACAQigZAAAAgEIoGQAAAIBCWPwIAAAAicWPBTDJAAAAABRCyQAAAAAUQskAAAAAFMJOBgAAAEhSrdrJ0FkmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJUrGTobNMMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAAASOxkKYJIBAAAAKISSAQAAACiEkgEAAAAohJIBAAAAKITFjwAAAJCkavFjp5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACSJnQydZpIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkCSVsgN0fyYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAEmqlWrZEbo9kwwAAABAIZQMAAAAQCGW65Lh4Yen5M47r8+dd16fyy47u+w4decrX90tU++4LlPvuC733PfTvDrzv7Piiq1lx6obwz4yND+97QdvuW3X3XfM7T+7vqRE7319N1s/611/+t/dvuLnPpb1J5+bIbeckwF7jvyXjr3CNh9dcoybzsqAPT+bJGls7Zu1rzox691wRobcdFb6DtugU/nr1Uc2/1BunXptkmTtdQZn6l3XZ8pd1+Xc809JQ0NDyenqR3Nzc676/vn52bSJueOuGzJk/XXKjlSXmpub8/3vXZB7pk3KQw/elh13/NeeU3ir5ubmnHfJ6Zl429W55WfXZtvtP53Ba6+ZG2+/OhNvuzqnnXOi54uC+Pqi6xx99CG5596f5sGHbs3e+3yp7DhQmHe1k2H+/PlpbGxMz549a5Wny/Tq1StJst12e5acpH5d+6NJufZHk5Ik3z3v2/nhDybmjTfmlJyqPhx6+P754h47Z+7cuUtv22TTDbPX3rvF11hvb5Wv75qVdv10KnMXvPWOxsas9q198vRO30ylfX42vHtC3rjzl+mY9S6u1eamrD72gDyz0zdTmbcgQyZ9J7OnPZqBX/1c2h76bf7yvcnptc7qGXzh0Xnm80cW+8C6ucOOODB7fPn/pb19XpLk9PFjctqp5+WhBx7Jdy/4dnbYcdvcfuvPSk5ZHz673afT1NSUkdt8MSM+Mzxjxx2dUV85pOxYdWevvXbNazNnZd/9Ds+AAf3z6CN35rbbXMOdtcuXPp9ZM9/IkQefkP4rrZgp9/0kT/3u6ZxzxoT88qH/yOnnnJjP7jAid95+T9lRuzVfX3SdT35yq2y51bBs85nd0rdvnxx+xNfKjgSFWeYkw4svvphDDjkkY8eOzS9+8YvssMMO2WGHHXLvvfd2Vb6aGTp0o/Tt2zu33vrDTJ16XbbY4sNlR6pbHx62aTbaaEi+/73ryo5SN16Y/mL2/eropW+vtFL/nHjyUTnpuPElpnpvW/DHGZn+9bc5P5VK/mubQ1KZMzfNK7UmDQ2pzJ2fNDdlzbNGZ72fjM96N56Zlq0++JYP2+RX1yz9e+/11syCF2akY3Z7qosWp+1XT6XfRzfJX66anFevvSNJ0tDUmOqChTV9jN3R9Ol/zKi93vxG90Mf3iQPPfBIkuRnd92fT4/4RFnR6s5zz01Pc3NzGhoa0traksWLFpUdqS5NmnRbTj75zenIxR2LS0xTP26/5a6cO37C0rc7Fndk0802yi8f+o8kyX3THszwT21VVry64euLrrPttlvnySefzvU3XJ6JN16VqVOnlR2J/1Op1u+fLrLMSYYxY8Zk9OjReemll3LYYYflzjvvTK9evXLAAQdkxIgRXZWxJubOnZfzz7883//+9VlvvbVzyy3XZOjQEeno6Cg7Wt055phDMv6MC8qOUVdun3xX1lxr9SRJY2NjzptwWsYePz7z5y/4Bx+5/Hpj6sPpucYqb39nRyUrbv+xrHHq1zP7nv9IdVFHVt5ruyyeOTsvHnthmvq3Zr2J4/P0yG9knWvGpbFXzzT1b8l615+eha+8ltd+ODUdc9qXHq7SPi9NK/RNx+wltzW/r3/WuuCovHTKFV3xULuVW2+5c+m1nOQt485tbW1ZYQUvsSpKe9vcrDV49Tz2m7uz8sor5Yu7H1B2pLrU3r7kJ8AtLf1y/XWX5+RxXo5ZhLn/O+3Ur6VvLv3+uTnnjAk54dvfXHp/e9vctK7QUla8uuHri66z8sCVstaaq2e33fbPBz6wZn4y8Yp8eLNtyo4FhVhmybB48eJsscUWSZJHHnkkK6+88pIPau7+v/ny2Wen5/nnX0iy5Kc7M2fOyqBBq+RPf5pRbrA6s+KKrVl//XVz//2/LDtK3frQZptk7XUH5zvfPTm9e/fM+husl1PHH5+TjvdTh3fjjTsezht3/jJrnXt4Buw2In02HJx+H904/TZbP8mSSYSm/q35/T6nJFkyyfDcnickSXpv+IE0tfRZeqzGfn2WFgy9NxicwROOycunfy/tjzzZxY+q+6n8Vcve0tKSN96YXWKa+nLo6P0y7e4Hcsq4s7P66oNy25Rrs9UW22eBCZvCrbHGoEz8yZW59LIf5Pobbi47Tt0YtNr7c/kPz88Pr7oht0yakuNPfvPlZ/1a+ma2l2QWytcXtTXztdfzzNPPZ9GiRXn22d9nwfyFed/7Vs5f/vJa2dGg05b5com11147J5xwQiqVSs4888wkyeWXX56BAwd2Sbha2mefL+XMM09KkgwatEpaW1syY8afS05Vfz4xfMvce+9DZceoa7/59e/yqa12yq477p2v73dUnnn6OV8AvAuNLX2y3g1npKFnc1KtpjJ3QaqVauY/96fMmvxAntvzhDy/zyl5fcpD6Xij7W2PMf+5F9PrA6ulacWWNPRoTsuWm2TuY/+dXkPWzAcu/lb+cNg5mXPfr7v4kXVPv3v8qXzik1smSUZ+dus8/ItflZyofrw+643Mnr3km7BZs15Pc4/mNDU1lZyq/qyyysBMuf3HGXPCGbnmmhvKjlM3Br5vQH406bKcecr5+cmPlxQ3T/72v7PVJzZPknx6m+F59GHPs0Xy9UVt/eLhX2XkyE8lSVYdtEr69uuT116bVXIqKMYyRxJOO+203HPPPWlsfLOLeP/7359Ro0bVPFitXX31DbniinMzbdqNqVarOeigY71UogaGDFkn06f/sewY8Hf677x1mvr2yWvX3ZlZN/886/1kfKqLOzL/v1/IrJvuS0NzY9Y88xtZ74Yz0tjaN6/+YEpSffOn7E9+dJ83D7a4Iy+delXW/eEpSWNDZv7k7ix6ZWbWPu3gNPbqkTXGHZgk6ZgzN9MP/PvfbsGbThxzRi6YcEZ69OiRZ55+LrfcdEfZkerGRRO+l4su+U7uuOuG9OzZI98++ZzMnTuv7Fh151vfGp3+/VfMmOOPyJjjj0iS7PSFUZk/f37Jybq3Q488MCusuEJGH/W1jD5qyYK8U8Z8JyePPy49e/bIc8/8PlMmW7BJ93HH1Hsy/BNb5P4HbkljY2O+eeTYVCqVsmORJP4zdFpDtVotdANEnz6Dizwcf6OpYbn+raNdpl+PXmVHqHt39l+/7Ah1b8RrT5UdYbnQUfXVSK3NX+wlHbU2qGVA2RHqnuu4a7QtUuh1hfa5L5QdoWZe36N77x5clv43dM0vcPAdKwAAAFAIJQMAAABQiO7/ayIAAACgANVKodsElksmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJ4jdTd5pJBgAAAKAQSgYAAACgEEoGAAAAoBBKBgAAAKAQFj8CAABAkmqlWnaEbs8kAwAAAFAIJQMAAABQCCUDAAAAUAg7GQAAACBJKmUH6P5MMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAACSVO1k6DSTDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAksZOh00wyAAAAAIVQMgAAAACFUDIAAAAAhVAyAAAAAIWw+BEAAACSVC1+7DSTDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAksZOh00wyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJJU7WToNJMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgNjJUASTDAAAAEAhlAwAAABAIZQMAAAAQCGUDAAAAEAhLH4EAACAWPxYBJMMAAAAQCGUDAAAALAcevzxxzNq1KgkyZNPPplPfvKTGTVqVEaNGpUpU6YkSSZMmJDdd989e+65Z37729/+w2N6uQQAAAAsZ6644opMnjw5ffr0SZI89dRT2XfffbPffvstfZ8nn3wyjz76aCZOnJgZM2Zk9OjRmTRp0jKPW3jJUK1Wiz4kf2VxtSN9e/QqO0bdW1TpKDtC3dt8xmNlR1guPLfxxmVHqHubPPNs2RHqXlNjU9kR6l7fpt5lR6h7fZt658W2v5Qdo+41NRjUppOqDWUn6DJrrbVWLrzwwhx77LFJkieeeCLTp0/PtGnTMnjw4IwZMyaPPfZYhg8fnoaGhqy22mrp6OjIzJkzM2DAgHc8rn+F3YyCAfhnKRgA3jsUDMB7zXbbbZfm5jfnDoYOHZpjjz021157bdZcc81cdNFFaWtrS0tLy9L36devX+bMmbPM4yoZAAAAYDk3cuTIfPCDH1z696eeeiotLS1pb29f+j7t7e1pbW1d5nGUDAAAALCc23///Zcudnz44YezySabZNiwYXnwwQdTqVTy8ssvp1KpLPOlEonFjwAAAJAkqVbKTlCek08+Oaeeemp69OiRgQMH5tRTT01LS0s233zz7LHHHqlUKhk7duw/PE5DteBNjb17r1Xk4fgbdjJQL2YvmFt2hLpnJ0PXsPix9iqWStfcB1rfX3aEumcnQ9dobFh+lvaVaXb778uOUDP/s/Wny45QM6vef1+XfB4vlwAAAAAKoWQAAAAACmEnAwAAACSpVrzkprNMMgAAAACFUDIAAAAAhVAyAAAAAIVQMgAAAACFsPgRAAAAklQrZSfo/kwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJJUqw1lR+j2TDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAklQrZSfo/kwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJJUKw1lR+j2TDIAAAAAhVAyAAAAAIVQMgAAAACFUDIAAAAAhbD4EQAAAJJUq2Un6P5MMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAACSVCsNZUfo9kwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAGInQxFMMgAAAACFUDIAAAAAhVAyAAAAAIWwkwEAAACSVKtlJ+j+TDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAklQrDWVH6PZMMgAAAACFUDIAAAAAhVAyAAAAAIVQMvD/2bvvaLkKcn38zykpJOeE0AklEJoCEiBgQAgoHaneANJBqhCqAiJIqELoRVCk9xoSSoCQEOBKkeK1AlcFNDRFQBLSTuqZ+f2R+w3yU0HNPtlk5/NZK2tl5pyz88ws2Jk88+53AAAAoBAL/OLHJZZYLM8++1C2227vvPLKH8qOUxnrrb92TjvzhOy03T5z7jt7yMl59dWxufH6O0pMVh2e43mnoaEhV1w+JGv3XSPTp0/PoYedkD/84fWyY30mdf7C57PosQfnLwcf/w+/vtjgY1ObMCnjf3Ddv33shTbdMD2/uU/qs9oz+f5HMnn4yDS0dMsSZ383jd27J52aM/6iH2f6b347tw+jso4/flC2237LdO7cKVdffUtuvunusiNV0rPPPpyJEycmSV5//a1885snlJxo/tfc3JSzLxucZZZfJp07d8qPL7k+OwzcJosvuViSZNnle+XXP38px33zlJKTVsPe++ySffbZNUnSpWuX9O27Rlbu88VMmDCp5GTV0dzcnKuuuTC9ey+b9vZajjrypLz6yh/LjkWSet3ix7m1QJcMzc3NueKKIZk6dVrZUSrlqGMPye577Jy2tqlJksUWXzRXXnV+Vl6lT1697NqS01WD53je2nnnbdO1a5cM2HSnbNC/Xy44/9QM3OXAsmN95vT4xtfTsv2Wqf+Tc2rLLtun0yp9Mv3nv/n3D97clEWPPyzv7H1kalOnpddNl2bqT55L69d3zLQXfpmJt92b5hWWyxLnnpx39hw0l4+kmjbZZMNssGG/bLH5LunWbaEcc+yhZUeqpC5duiRJttlmj5KTVMuOu341H46bkBOPOD09F1k4wx67JVv02ylJ0mPh1tx475U5d/AlJaesjttuHZbbbh2WJLn4kjNzy81DFQwF23qbr6SpqSlbbbFbNtt8QE497fjsu7e/v6iGf/lyiQ8++KAjc5Ti3HNPyTXX3Jp33nm37CiV8vof38x+ex8x53b37t1y3pDLc/ed95WYqlo8x/PWgI36Z9ToJ5Ikz7/wi6zXr2/JiT6bZr3157x33Bn/8Gtd+q6eLn1Xz+RhD310Z3NTFjvt21n6uouy9A2XpOv6H39elx9z15zfd+rTO7Pe+nNqkyYns2Zl2i9fSpd+a2XircMy6Z7Zx2xobkp9xoziH1hFbLnlpnn55d/nzruuztB7rsvIkY+VHamS+vZdPd26dc2IEbdk5Mg70r//umVHqoRRIx7LZedeNed2e3v7nN8f+Z1Dc9u1d+f996r3WrVs6/ZbK6uvvmpuMCFZuNdeG5vm5uY0NDSktbUls2bOLDsSFOaflgxjx4792K/DDz98zu+rYN99d81f//pBxox5suwolTPigVGZOXPWnNtvvvF2fv4/vy4xUfV4juet1h4tmfg37+C0t9fS1NRUYqLPprbHnk5mzfq7+5sWXzQ9D9sv44Zc/rH7W//rq6l9ODF/Oei4vHfsqVn0pKOSJEtecXaWvvbCNC7cmqWvvTCLn/PdNHbvltrkKXN+tj5lahpbuqc2aUrq02ekabFFsvjZ3834H1zfsQ9yPrbY4ouk37prZZ+9B+WYo7+X66+/tOxIldTWNjWXXnp1dtxx3xx11Mm54YbLnC8K0DZlatqmtKVb92659LohuWzIj5Mkiy6+SL60yRdz750Plpywmk44YVCGnHNZ2TEqacrktvReYdn8/JdjcvkV5+TKK28qOxIU5p9eLnHAAQeka9euWXLJJVOv1zN27NiceuqpaWhoyM033zwvM3aI/fffPfV6PZtvPiB9+66R6667JLvuelDefff9sqMBnzGTJk5OS2vLnNuNjY0fexeNT9Ztq03T2LNHlrri7DQttkgaunbNzNffSqdV+6Trumuly1qfT5I0NDWlceHWvHfk95LMnmT4f7sdOq3aJw3du805ZkP3hWZPNSTptMqKWeK872X8xVf/Z5diLCDGffBhXvn9HzJz5sy8+uofM33ajCyxxGJ5/33v/hbp1VfHztnZ8tprYzNu3Pj06rVk3n77nXKDVcDSyyyZy2+8IHfccE8eGj4qSbLNjpvnweGjUqvVSk5XPQsv3JrVVls5Tz75XNlRKumIow7MY2OeyhmnXZBll+2VBx++LRv23zbTp5vIK1vd6WSu/dOSYdiwYTnttNOy5557ZuONN86+++6bW265ZV5m61BbbrnbnN+PHn1XjjzyZAUD8A898+zPssP2W+Wee0Zkg/798tJLFgv+OybdcV8m3TH7Up6WnbZOpxWXz+QHRqe1e7e0v/vXTLjujjR06ZyFD94rtYmT/+ExZo59M516L5vGHq2ptU1N135rZeLNQ9Nppd5Z8oLBee/EszPTwqxP9NNnf5YjBh2QH/zg2izda8l0675QPvhgfNmxKmf//b+eNdf8fI499pT06rVkWltb8s4775Uda7632BKL5tq7L8/3T7owzz31szn3f2nT/vnxxSaYOsLGAzbIE088U3aMyvpw/ITM/L/pv/HjP0xzp2ZTT1TGPy0ZFltssVx66aU577zz8uKLL87LTACfKffdNzJbbrFpnvrJ/WloaMhBh3yr7Ejzhe5f3SwN3RbK5GEP/8OvT7rnoSx+6rey9LUXpaGlWybd/UBSr8/5+ltb7v7RN89qz7gLf5ylrhySNDRk8v2j0v7eB1nspKPT0KVzFvvO7GVZtUlT8t63TuvQxzW/emTk4xmwcf88+dT9aWxszLe/dap3fzvAjTfelWuuuSiPPXZP6vV6DjvsOyafCnDoMd9Ij549cvi3D8zh3569ePfQPY9Nn5VXyFtv/KnkdNW06qorZezYN8uOUVk/vOL6/PDK8/LI6LvSuXOnnHn6hXMWesP8rqFe/5tXdP/E8OHDM3z48Nx6662fesCuXXsXEox/rFunLmVHgEJMnN5WdoTKe22NNcqOsEBY85VXy45QebVPf6nCXFqxdamyI1TeW5NNzM4LjQ0+fnBemDilutODr62xTdkROswq/ztqnvw5/9JHWA4cODADBw7s6CwAAABQmlpdUTW3/uWPsAQAAAD4JEoGAAAAoBBKBgAAAKAQ/9JOBgAAAKi6up0Mc80kAwAAAFAIJQMAAABQCCUDAAAAUAglAwAAAFAIix8BAAAgSb1m8ePcMskAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASFKvl51g/meSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJCkXmsoO8J8zyQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEmtbifD3DLJAAAAABRCyQAAAAAUQskAAAAAFELJAAAAABTC4kcAAABIUrf4ca6ZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkqdfLTjD/M8kAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASFKrN5QdYb5nkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQpG4nw1wzyQAAAAAUQskAAAAAFELJAAAAABRCyQAAAAAUwuJHAAAASFKvl51g/meSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJCkVm8oO8J8zyQDAAAAUAglAwAAAFCIwi+X6NzkCoyONL19ZtkRFggLNXcuO0LlLbZQa9kRKq/vq38oO8IC4YpFBpQdofIOff+JsiNU3sSZU8qOAIXwWhnKpxEAAACAJHU7GeaayyUAAACAQigZAAAAgEIoGQAAAIBC2MkAAAAASWp2Msw1kwwAAABAIZQMAAAAQCGUDAAAAEAhlAwAAABAISx+BAAAgCT1sgNUgEkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJavaHsCPM9kwwAAABAIZQMAAAAQCGUDAAAAEAh7GQAAACAJHU7GeaaSQYAAACgEEoGAAAAoBBKBgAAAKAQdjIAAABAklrZASrAJAMAAABQCCUDAAAAUAglAwAAAFAIJQMAAABQCIsfAQAAIEk9DWVHmO+ZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkqdXLTjD/M8kAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASFJLQ9kR5nsmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJ6nYyzDWTDAAAAEAhlAwAAABAIZQMAAAAQCGUDAAAAEAhLH4EAACAJLWyA1SASQYAAACgEEoGAAAAoBBKBgAAAKAQdjIAAABAknoayo4w3zPJAAAAABRCyQAAAAAUYu+zhyoAACAASURBVIEtGZqbm3PdDZfm0ceG5pHRd2XV1VYqO1IlHX/8oDz+xPA8/cyI7Lf/18uOUyn91u+b+x+6JUnyhbVWz4hHbs/9D92Su++9LksssVjJ6aqj33p9M/zBmz9238Bdd8hDj95ZUqJqck7uOI2dm7PpFYOy/YjTs/XtJ6ZHn6XSa5M1s+PDZ2b7Eaen33d2LTti5fT/4roZ8+jQsmNU0rrrrZWhI25IkvzougsydMQNGTrihjz369H50XUXlJyuOvbeZ5eMfOSOjHzkjjz+38Pz13G/y8ILt5Ydq1IaGxtz1VUX5oknhmfMmHuy0korlB0JCrPA7mTYepuvpKmpKVttsVs223xATj3t+Oy796CyY1XKJptsmA027JctNt8l3botlGOOPbTsSJVx1DEHZ7c9dk5b29QkyTnnfS8nnXBWXnrxt9n/gN1z9LcOzeCTh5Sccv53xDEHZbfdd05bW9uc+9Zc6/PZa79d0uByvUI5J3ecz+21WWZNmZaHdjw9PVbulQ2/v3+6Lt4jPznyR5nw6p+z3b2Ds8jnl8v4371ddtRKOO64w7PP3rtkypSpZUepnMOPPjC77L5jpv7fczvooBOSJAsv3CNDR9yQ008+r8x4lXLbrcNy263DkiQXX3Jmbrl5aCZMmFRyqmrZYfutkiSbbTYwm266Yc4/79TsuttBJaciSWplB6iAf3mSoVar5d13302tVo2n/bXXxqa5uTkNDQ1pbW3JrJkzy45UOVtuuWlefvn3ufOuqzP0nusycuRjZUeqjLFj38w39jlyzu1DDvhWXnrxt0mSpubmTJs2vaxolfL62LdywD5Hzbm9yCI9c8rpx2XwdxU4RXNO7jgLr7Zs3n7iN0mSiX94Jwuvukw+eOmNdOnZksZOTWnq0im19mr83f5Z8Mc/vpHdvn5I2TEq6Y2xb+WQfY/5u/uPO+mIXH/1bXnv3b+WkKra1u23VlZffdXccP0dZUepnAdGjMqgQScmSXr3Xi7vvfd+yYmgOJ84yXDyySfnnHPOya9//escf/zx6dmzZ6ZMmZJzzjkn66yzzrzK2CGmTG5L7xWWzc9/OSaLLbZIdtv14LIjVc5iiy+S3ssvm112OSgrrrh87h56TdZdZ4uyY1XCgw+MzvK9l51z+913Z//F9MX+6+bgQ/fJjtvuVVa0Snnob57nxsbGXHLF93PqSUOUOB3AObnjjHv5jSy/5Tp585H/yRL9Vk63pRfNh6+8nS1vOi7Tx0/O+N++lQmvvVN2zMq4996Hs8IKy5Udo5IeHvFollt+mY/dt9jii2bAphuaYuggJ5wwKEPOuazsGJXV3t6ea6+9ODvvtG323OuwsuNAYT5xkuHtt2ePTl5yySW55pprMnTo0Nxwww258MIL50m4jnTEUQfmsTFPpd86W2SjDbfPVVdfmC5dOpcdq1LGffBhxox5MjNnzsyrr/4x06fNsCugA31t4Ha58NIzsuduh+aDD8aXHady1l5nzfRZeYWcd/Hpuer6i7La51bJWUNOKjtWZTgnd5xX7/xJZk6emm2Hfi/Lb9Uvk954N2sN2iH3bf7dDNv4uEwc+5d84ZvblR0T/iM77Lx17hv2UGUmbT9LFl64NauttnKefPK5sqNU2sEHfztfWOvL+dGPzku3bguVHQcK8S9dLtHU1JQVV1wxSbLUUktV4kT+4fgJmThx9rVl48d/mOZOzWlqaio5VbX89NmfZautvpwkWbrXkunWfSH/+O0gu+2+Uw46dJ/svP2+eeP1t8qOU0m//MWL+fKGO2bgDvvlmwcel1d+/1oGn+SyiaI4J3ecxddZKe++8Eoe2e3svDHyf/Lh7/+UWVOmZ+aUaUmStvc+TOee3UtOCf+ZAV/eMI8/+lTZMSpp4wEb5Iknnik7RmXttdfAnHDCEUmStrapqddqaXfp2mdCrcK/5pVPvFxi0qRJGThwYNra2jJ06NDstNNOOffcc7PMMst80o/NF354xfX54ZXn5ZHRd6Vz50458/QL5yzRoxiPjHw8Azbunyefuj+NjY359rdOrURB9VnT2NiYc84/JW+/9U5uuvWKJMlPn/lZzjvnByUng3+dc3LHmfjHv6TfCbvmC4dtlxkT2vL08ddkiXVXyTa3n5j26TMzY2JbnvrWVWXHhP/Iyqv0yZuvW1raEVZddaWMHftm2TEq6777Ruaaqy/KmDH3pFNzc44//oxMn+5yTKqhoV6v1z/pG2bMmJHf/e536dq1a1ZcccUMGzYsu+66azp16vQPv79Hdx871pHa6/6RPi8s1GxMu6M1NSywn6A7z0ydNaPsCAuEy3puVHaEyjv0/SfKjlB5S3bvWXaEypswve3Tv4m5NrM2q+wIC4Tp06o7OfvwUnuUHaHDbPfuvPkI9k/9CMvOnTunb9++c27vueeeHRoIAAAAmD95KxEAAAAoxKdOMgAAAMCCoJ6GsiPM90wyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUrGSYayYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAElqsZRhbplkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACSplx2gAkwyAAAAAIVQMgAAAACFUDIAAAAAhVAyAAAAAIWw+BEAAACS1MoOUAEmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJag0NZUeY75lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACSplx2gAkwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUyg5QASYZAAAAgEIoGQAAAIBCKBkAAACAQigZAAAAgEJY/AgAAABJag1lJ5j/mWQAAAAACqFkAAAAAAqhZAAAAAAKYScDAAAAJKnFUoa5ZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKRedoAKMMkAAAAAFELJAAAAABRCyQAAAAAUQskAAAAASWoN1f31j/z617/OvvvumyR54403sueee2avvfbKaaedllqtliS54oorsuuuu2aPPfbIb37zm099DpUMAAAAsIC55pprcsopp2T69OlJkiFDhuTYY4/N7bffnnq9nsceeywvv/xyXnjhhQwdOjQXX3xxzjjjjE89buGfLtG9U9eiD8nfmDijrewIC4Rps2aWHaHyunfqUnaEymtq0CPPC4PGPVV2hMo7d+nNyo5QeZdM/lXZESqvvV4rO8ICodHfffAv6927dy6//PJ85zvfSZK8/PLL6d+/f5Jk0003zTPPPJM+ffpkwIABaWhoyDLLLJP29vaMGzcuiy666D89rv8LAQAAYAGzzTbbpLn5o7mDer2ehobZ11V07949kyZNyuTJk9PS0jLne/7f/Z9EyQAAAAALuMbGj+qBKVOmpEePHmlpacmUKVM+dn9ra+snH6fDEgIAAMB8pFbhX59mjTXWyPPPP58kefLJJ7P++uunX79+efrpp1Or1fLnP/85tVrtEy+VSDpgJwMAAAAwfznxxBMzePDgXHzxxVlppZWyzTbbpKmpKeuvv35233331Gq1nHrqqZ96nIZ6vV4vMlivnmsUeTj+fyx+nDca8k8+44XCWPzY8Wa0zyo7wgJhWrtFsR3t+0tsUnaEyrP4seN9MPWTr2GG+cnUqW+UHaHD3LjsPmVH6DDf+NOt8+TPcbkEAAAAUAiXSwAAAECSQsf8F1AmGQAAAIBCKBkAAACAQigZAAAAgELYyQAAAABJaj5kbq6ZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkqZUdoAJMMgAAAACFUDIAAAAAhVAyAAAAAIVQMgAAAACFsPgRAAAAYvFjEUwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUG8pOMP8zyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIUis7QAWYZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAADEToYimGQAAAAACqFkAAAAAAqhZAAAAAAKoWQAAAAACmHxIwAAACSplx2gAkwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUGspOMP8zyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIUis7QAWYZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAADEToYimGQAAAAACrFAlgzrrtc3wx68MUmy2OKL5obbr8i9D9+c+x+5NSusuHy54Srm+OMH5fEnhufpZ0Zkv/2/XnacStp7n10y8pE7MvKRO/L4fw/PX8f9Lgsv3Fp2rMrot17fDH/w5o/dN3DXHfLQo3eWlKh61lt/7YwYeVuSpM9KK2Tk6Dvz8Og7ctGlZ6ShoaHkdNXy7LMPZ9SoOzNq1J256qoLyo5TGY3NTdnuB4Oy5/BTs/s9g7Poyr2yxBq9s+d9p2ePYYOzzQWHJP5bLsS6662VoQ/ckCRZ4wufywOjb8vwh2/OhZef5XxRMOeLecPzTBUtcJdLDDr6wOy6+05pa5uaJBl85nEZfveDGXHfI9lok/5ZZbU+eeP1t0pOWQ2bbLJhNtiwX7bYfJd067ZQjjn20LIjVdJttw7LbbcOS5JcfMmZueXmoZkwYVLJqarhiGMOym6775y2trY596251uez1367+PdCQY4+9pDsvufXMmXK7HPy2UNOzvfPuiTPPPV8Lr7szGy3w5Z5aMSjJaeshi5duiRJttlmj5KTVE+fzdZOY3NT7hh4ZlbY5AvZ+ITd0tDYmOcuuzdjn/h1trvs8Ky0xTr545hflh11vnb4UQdkl913nPMa7tvfGZRLz/9xHh/zVC6/6txssfWmGTPqJyWnrAbni3nD80xV/VuTDOPGjUu9Xu+oLPPEG6+/lYP2PWbO7S9u0C/LLLtU7rrvugzcbYf89OmflZiuWrbcctO8/PLvc+ddV2foPddl5MjHyo5Uaev2Wyurr75qbrj+jrKjVMbrY9/KAfscNef2Iov0zCmnH5fB3x1SYqpqGTv2zey716A5t9ded80889TzSZJHRz+Zr2y2cVnRKqdv39XTrVvXjBhxS0aOvCP9+69bdqTKGD/2L2lsakwaGtK5ZaHUZrXnvZdfT9eeLUky+76Z7SWnnP+98fpbOWS/Y+fcfunF36bnIgsnSbq3dM+sWbPKilY5zhfzhueZqvrEkmHYsGG54oor8vLLL2fbbbfNAQcckG233TY//elP51W+wj30wKOZOWvmnNvL914mH344Mbt/7aD86e13cuSxB5WYrloWW3yR9Ft3reyz96Acc/T3cv31l5YdqdJOOGFQhpxzWdkxKuWhB0bPedHa2NiYS674fk49aUgmT55ScrLqGHH/qMyc+dE/DP523Hny5Mnp0cOlP0Vpa5uaSy+9OjvuuG+OOurk3HDDZWlqaio7ViXMnDItPZZbIgc8cX62Ou+g/PKG0flw7LvZ7Ix9c8Dj56fb4j3y1nO/LTvmfO/hEWM+dr4Y+4c3cua5J+W/n3sgSyy5WJ71RlFhnC/mDc/zZ1O9wr/mlU+8XOL222/PLbfcksMPPzxXXnll+vTpk3fffTeDBg3KRhttNK8ydqjx4yZk9MOPJ0kefeS/891TjvmUn+BfNe6DD/PK7/+QmTNn5tVX/5jp02ZkiSUWy/vvf1B2tMpZeOHWrLbaynnyyefKjlJZa6+zZvqsvELOu/j0dO3aOat9bpWcNeSkDD7JVEORarWP/gpsaWnJhAkTS0xTLa++OjZ/+MPrSZLXXhubcePGp1evJfP22++UG6wC+h381bz+5G/y9Hl3p7XXotntzpPTpXWh3LXrWfnglT9lnf22zFdO2SuPDb6p7KiVcsaQ72bg9vvlld/9IfsftEcGn3VCTvnO2WXHqgTni3nD80xVfeIkQ6dOndKtW7d07949yy8/eyHiUkstVanFOi889/NssfWmSZINN1ovv//dayUnqo6fPvuzbLXVl5MkS/daMt26L5QPPhhfcqpq2njABnniiWfKjlFpv/zFi/nyhjtm4A775ZsHHpdXfv+agqEDvPjr/83Gm2yQJNlq603z7E+9M1mU/ff/es49d3CSpFevJdPa2pJ33nmv5FTVMH3ClMyYNHtPwNQPp6SxuSnTJ02dc9/kdz9Ml4W7lxmxkj4cPyGTJs2eLHv3L++nZ88eJSeqDueLecPzTFV94iTD5ptvnsMPPzyrrbZavvnNb2aTTTbJU089lQ033HBe5etwZ5xyQS78wZnZ78A9Mmni5Aw6+ISyI1XGIyMfz4CN++fJp+5PY2Njvv2tU1Or+eTZjrDqqitl7Ng3y44Bc+2Uk8/JZVeck06dOuWV37+W++99pOxIlXHjjXflmmsuymOP3ZN6vZ7DDvtO2tvtCSjCz68dmW0uPDS73zM4TZ2a8vT5d2fSnz/I9j88MrVZ7anNnJXRJ15XdszKOeGY0/Kjay9I+6xZmTFzVr5zzGllR6oM54t5w/NMVTXUP2WT4wsvvJCnn34648ePT8+ePbPeeuvlK1/5yj/9/l491yg6I39j4oy2T/8m5lpDqjOt81nVvVOXsiNU3ox2S9DmhWntMz/9m5gr319ik7IjVN4lk39VdoTK+2CqT36iOqZOfaPsCB3m/BX2KTtCh/nOG7fOkz/nUz/Csn///unfv/+8yAIAAADMx/6tj7AEAAAA+GeUDAAAAEAhPvVyCQAAAFgQWFM/90wyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUyw5QASYZAAAAgEIoGQAAAIBCKBkAAACAQigZAAAAgEJY/AgAAABJalY/zjWTDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAktbIDVIBJBgAAAKAQSgYAAACgEEoGAAAAoBB2MgAAAECSetkBKsAkAwAAAFAIJQMAAABQCCUDAAAAUAg7GQAAACBJrewAFWCSAQAAACiEkgEAAAAohJIBAAAAKISSAQAAACiExY8AAACQpNZQdoL5n0kGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJa6mVHmO+ZZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkNjIUwCQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEmt7AAVYJIBAAAAKISSAQAAACiEkgEAAAAohJIBAAAAKITFjwAAAJCklnrZEeZ7JhkAAACAQigZAAAAgEIoGQAAAIBCFL6T4a9tE4o+JH+jpfNCZUdYIExrn1l2hMobP21y2REqr0eXbmVHWCDUZtXKjlB5J/7libIjVN77O69adoTKW/L+8WVHWCA0NnoPlbljI8Pc838hAAAAUAglAwAAAFAIJQMAAABQiMJ3MgAAAMD8yKaluWeSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJCklnrZEeZ7JhkAAACAQigZAAAAgEIoGQAAAIBCKBkAAACAQlj8CAAAAIm1jwUwyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABIUis7QAWYZAAAAAAKoWQAAAAACqFkAAAAAAphJwMAAAAkqadedoT5nkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJJa2QEqwCQDAAAAUAglAwAAAFAIJQMAAABQCCUDAAAAUAiLHwEAACBJLfWyI8z3TDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAEhsZCmCSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJCkZivDXDPJAAAAABRCyQAAAAAUQskAAAAAFMJOBgAAAEhSKztABZhkAAAAAAqhZAAAAAAKoWQAAAAACqFkAAAAAAph8SMAAAAkqadedoT5nkkGAAAAoBBKBgAAAKAQSgYAAACgEAt8ydD/i+tmzKNDy45ROeutv3ZGjLwtSdJnpRUycvSdeXj0Hbno0jPS0NBQcrrqefbZhzNq1J0ZNerOXHXVBWXHqZzm5ubccP1lefyxYXnm6Qezww5blR2pUvqt3zf3P3RLkuQLa62eEY/cnvsfuiV333tdllhisZLTVUdjY2OuuurCPPHE8IwZc09WWmmFsiNVUkNDQ354xbl5+skH8tijQ7PyyiuWHekzq2mV1dNy+qV/f//Kn0vLmT9I61mXp/txZySdOv/bx+603pfSOuTHaT37h+m8xfaz7+zWPd1PPCctZ1ya1rN/mKbV1pjbh1B5Xid3HOfkz65ahX/NK5+4+HHy5MlpaWmZV1nmueOOOzz77L1LpkyZWnaUSjn62EOy+55fm/O8nj3k5Hz/rEvyzFPP5+LLzsx2O2yZh0Y8WnLK6ujSpUuSZJtt9ig5SXXttdfAfDBufA448JgsumjPvPD8qDz4oP+Gi3DUMQdntz12Tlvb7PPFOed9LyedcFZeevG32f+A3XP0tw7N4JOHlJyyGnbYfnY5ttlmA7Ppphvm/PNOza67HVRyqurZeedt07VrlwzYdKds0L9fLjj/1Azc5cCyY33mdNlpj3T58tapT5v2d1/rdtgJmXLRaan95U/pvPn2aVxiqdT+/Na/fvCmpiz0jSMz6bvfTH36tLSedUVm/vzZdNl6p8x68ReZ/vA9aVxm+XQ/ZnAmnXhogY+qWrxO7ljOyVTZJ04ybLzxxhk6tLrt5R//+EZ2+/ohZceonLFj38y+ew2ac3vtddfMM089nyR5dPST+cpmG5cVrZL69l093bp1zYgRt2TkyDvSv/+6ZUeqnGHDHszpp380ITKrfVaJaapl7Ng38419jpxz+5ADvpWXXvxtkqSpuTnTpk0vK1rlPDBiVAYNOjFJ0rv3cnnvvfdLTlRNAzbqn1Gjn0iSPP/CL7Jev74lJ/psqr3750y+YPDf3d+4zPKpT5qQLtvvmpYzLk1DS+vsgqGpKd0OPyEtZ1yW1rMuT/Ma63zs5xa+Zvic3zctu0Jqf/lT6lMmJ7NmZdbvXkzz59fK9AfvyfQxD/zfH9SUzJzRoY9xfud1csdyTqbKPnGS4fOf/3x++9vfZr/99suRRx6Z/v37z6tc88S99z6cFVZYruwYlTPi/lFZvveyc27/7eURkydPTo8erWXEqqy2tqm59NKrc8MNd2aVVfrk/vtvSt++m6W9vb3saJUxZUpbkqSlpXvuvOPqnH6aS1KK8uADoz92vnj33dkvsr7Yf90cfOg+2XHbvcqKVknt7e259tqLs/NO22bPvQ4rO04ltfZoycQJk+bcbm+vpampyTn5/2fm80+mcYml/+7+htaF0/y5L6Tt+h+k9s7bafnukLT/8ZU0LrNc6hMnZPKVF6ShpUdaz7wsE799QFpOPi/p3CUNLa1pOf3S1Ma9n+mjH0i9bfKcY9antaWhW8uc+xp6LpruR38vU2+8Yp493vmR18kdzzmZqvrEkqFLly459dRT8+KLL+bqq6/OmWeemS996UtZfvnls99++82rjMznarWPPmu2paUlEyZMLDFN9bz66tj84Q+vJ0lee21sxo0bn169lszbb79TbrCKWW65Xhl697X58VU358677is7TqV9beB2+dbxh2XP3Q7NBx+MLztO5Rx88LfzvaWG5KmnHsg662w+51IVijFp4uS0tH50qWljY6OC4d9QnzQxtb/8KbW330iSzPzVC2laabU0Lrl0Oq3eN02r/t8ehaamNLT0yORzZr8TvPA1wzP59GNnf6n3Smno2m3OMRu6dptTMDT27pOWY09N281XZtb//noePjL4x5yTP3vqqX/6N/GJPvFyiXp99hO81lpr5fLLL88dd9yRL33pS5k5c+Y8CUc1vPjr/83Gm2yQJNlq603z7E9/VnKiatl//6/n3HNnj5z26rVkWltb8s4775WcqlqWXHLxPPzQ7Tn5e+fkppvuKjtOpe22+0456NB9svP2++aN1/+Na7D5VHvtNTAnnHBEktkTUPVaLe3t83IN1ILhmWd/lq9uu3mSZIP+/fLSS78tOdH8pfben5OuC6Vx6dkTTs2r903726+n9qc3M+PpxzL59GMz+ZzvZMazP0l9yqR/eIz2P72Rxl7LpaGlNWluTvMafTPrlZfTuNwKafn2GZly2fcz61cvzMuHBX/HOZkq+8RJhoEDB37sdmtrazbffPMODUT1nHLyObnsinPSqVOnvPL713L/vY+UHalSbrzxrlxzzUV57LF7Uq/Xc9hh3/GuWcFOPPGo9Oy5cE4+6dicfNLsd8p23GnfTPsHC8v4zzU2Nuac80/J22+9k5tunT3G/NNnfpbzzvlBycmq4b77Ruaaqy/KmDH3pFNzc44//oxMn27nRdHuu29kttxi0zz1k/vT0NCQgw75VtmR5gudBmyRhq4LZcaYB9N25fnpfswpSRoy65WXMusXz2VWc6d0O+z42XsaFuqe6aPuS+ofvds44ZC/ec3a3p6pN/0wLd+7IGlsyIzHR6Y+7q/pdvCxSafOWeiAo5Ik9bbJmXL+KfP4kcJszslUWUO9Xi90HqRT52U//Zv4j7V0XqjsCAuEae2mdTpae00R0tF6dOn26d/EXJs0w2hrR2uveXevo72/86plR6i8Je9/tewIC4TGxk8c1KYg06dVd9rwgBV3KTtCh7nh9WHz5M/5xEkGAAAAWFCoteeeqg8AAAAohJIBAAAAKISSAQAAACiEnQwAAACQpFbs5yIskEwyAAAAAIVQMgAAAACFUDIAAAAAhVAyAAAAAIWw+BEAAACSWPs490wyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJLUbGWYayYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAEnqdjLMNZMMAAAAQCGUDAAAAEAhlAwAAABAIexkAAAAgCS1sgNUgEkGAAAAoBBKBgAAAKAQSgYAAACgEEoGAAAAoBAWPwIAAECSWuplR5jvmWQAAAAACqFkAAAAAAqhZAAAAAAKYScDAAAAJKnbyTDXTDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAktTKDlABJhkAUuxmXAAAIABJREFUAACAQigZAAAAgEIoGQAAAIBC2MkAAAAASer1etkR5nsmGQAAAIBCKBkAAACAQigZAAAAgEIoGQAAAIBCWPwIAAAASWqx+HFumWQAAAAACqFkAAAAAAqhZAAAAAAKYScDAAAAJKmVHWAe+9rXvpbW1tYkyXLLLZfdd989Z599dpqamjJgwIAceeSR//YxlQwAAACwgJk+fXqS5JZbbplz384775zLL788yy+/fA499NC8/PLLWXPNNf+t4xZeMnTr3LXoQ/I3aqln2qwZZceovEW7tpYdofLqNvd2uLaZ08uOAIVobGgoO0LlrTTy7bIjVN5di3657AgLhL0nPFN2BJhv/O53v8vUqVNz4IEHZtasWTnqqKMyY8aM9O7dO0kyYMCAPPvss+WXDHQsBQMAAABzq2vXrjnooIOy22675fXXX88hhxySHj16zPl69+7d89Zbb/3bx1UyAAAAQBasads+ffpkhRVWSENDQ/r06ZPW1tZ8+OGHc74+ZcqUj5UO/yqfLgEAAAALmHvuuSfnnntukuTdd9/N1KlT061bt7z55pup1+t5+umns/766//bxzXJAAAAAAuYXXfdNSeddFL23HPPNDQ05JxzzkljY2OOP/74tLe3Z8CAAVl77bX/7eMqGQAAAGAB07lz51x00UV/d//dd989V8dVMgAAAEBmf5ofc8dOBgAAAKAQSgYAAACgEEoGAAAAoBBKBgAAAKAQFj8CAABAknrd4se5ZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKRWdoAKMMkAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASFJPvewI8z2TDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAkNTsZ5ppJBgAAAKAQSgYAAACgEEoGAAAAoBBKBgAAAKAQFj8CAABAknrd4se5ZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKQWOxnmlkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJK6nQxzzSQDAAAAUAglAwAAAFAIJQMAAABQCDsZAAAAIEmtbifD3DLJAAAAABRCyQAAAAAUQskAAAAAFELJAAAAABTC4kcAAABIYu3j3DPJAAAAABRCyQAAAAAUQskAAAAAFMJOBgAAAEhSs5VhrplkAAAAAAqxwJYMnTt3zrXXX5Ixj9+Te++/MSutvGLZkSqnsbExV111YZ54YnjGjLknK620QtmRKmXd9fpm2IM3JkkWW3zR3HD7Fbn34Ztz/yO3ZoUVly83XIWsu17fDH/wpiTJWmuvkV/+739n+IM3ZfiDN2Xn//pqyemqo7m5OdfdcGkefWxoHhl9V1ZdbaWyI1WOc/K80dzcnBuuvyyPPzYszzz9YHbYYauyI1XKeuuvnREjb0uS9FlphYwcfWceHn1HLrr0jDQ0NJSc7rNnkXVXzibDT/mHX2taqHO+/MBpaVllmf/o2Etv1S+bPXJWvvzgGVlx782SJM2tC+VLNx+fTe4dnC8/eEYWXW/V/zj7guLZZx/OqFF3ZtSoO3PVVReUHQcKscBeLrH/AbtnypS2bLn5rlll1T658KLTMvBrB5Qdq1J22H72C6vNNhuYTTfdMOefd2p23e2gklNVw6CjD8yuu++UtrapSZLBZx6X4Xc/mBH3PZKNNumfVVbrkzdef6vklPO/I44+6GPP81p918hVP7oxP77ixnKDVdDW23wlTU1N2WqL3bLZ5gNy6mnHZ9+9B5Udq1Kck+eNvfYamA/Gjc8BBx6TRRftmReeH5UHH3y07FiVcPSxh2T3Pb+WKVNmn5PPHnJyvn/WJXnmqedz8WVnZrsdtsxDIzzX/8+qR+yQ3rsOSHvb9L/7Ws+1+2Td8w/KQr0W/Y+O3dDclL5n7pMnth2cWW3T8pURp+ed0b/ISt/YMu899VL+cM0jaVm5V/pfeWQe3/p7c/tQKqtLly5Jkm222aPkJFCsf2uSYcaMGZk2bVpHZZmnPv/5VfLo6J8kSV57dWxW+9wqJSeqngdGjMqgQScmSXr3Xi7vvfd+yYmq443X38pB+x4z5/YXN+iXZZZdKnfdd10G7rZDfvr0z0pMVx2vv/5mDtz36Dm3115nzWy59Zdz78O35OLLv5/uLd1KTFctr702Ns3NzWloaEhra0tmzZxZdqTKcU6eN4YNezCnn/7Ru5Gz2meVmKZaxo59M/vu9VH5uPa6a+aZp55Pkjw6+sl8ZbONy4r2mTTl9Xfz3IGX/sOvNXbulOcOuDiTXvvznPsampvS7+JDsum9g7Pp/adl8Y1W/9jPbPebH835feuqy2TK2Hczc8KU1Ge256/P/z6Lb/j5vHbVyIy95bH/O15j2qc7l3+Svn1XT7duXTNixC0ZOfKO9O+/btmRyOydDFX9Na98YskwduzYHH300TnuuOPyq1/9KjvuuGO23377PPzww/MqX4d58Te/zbbbzh7tWv+L62SZZZZKY+MCe/VIh2lvb8+1116cSy4+M8Pvnf//u/mseOiBRzNz1kd/cS/fe5l8+OHE7P61g/Knt9/Jkcd6d7IIDz3waGb9zfP8y1/8JmcOviD/td2+eeP1t3L8iUeUmK5apkxuS+8Vls3Pfzkml19xTq688qayI1WSc3LHmzKlLZMnT0lLS/fcecfVOf00489FGXH/qMyc+VFp87eXR0yePDk9erSWEesz688P/Sz1Wf+45Br3s1cy9c/jPnbfintvlunjJuXJ/zorz33joqwzZPaE70a3fyebDD8lnXu2ZJPhp2T9Hx6RTq3dMnNS25yfnTVlWjq1LpSZE9tSmzYzXZZYOF+84oi8fM6dHfcAK6CtbWouvfTq7LjjvjnqqJNzww2XpampqexYMNc+8V/VgwcPzh577JGtt9463/zmN3PzzTdnxIgRuemm+f/F3y03D83ESZPz4Mjb8tXttsivfvlSarVa2bEq6eCDv50vrPXl/OhH56Vbt4XKjlNJ48dNyOiHH0+SPPrIf2ftdb5QcqJqevjBMfnNr/83STLywTH5Qt/VP+Un+FcdcdSBeWzMU+m3zhbZaMPtc9XVF6ZLl85lx6ok5+SOt9xyvfLo6Ltz2+3Dcudd95Udp7JqtY/elWtpacmECRNLTDP/W3j15bP0Futkk+GnZIPrjk1DU2M6L9KSn+51fp4a+P3M+HBynhr4/fzPET/MzEltaW756PzR3L1rZkycXTr0+Pzy2eSe7+XlIXflr8/+rqyHM1949dWxueOOe5PMnugbN258evVasuRUMPc+sWSYNWtWNtpoo2y99dbp2bNnllpqqXTr1i3NzfP/Kod+6/XNc8/+T3b46t558IHRed3164Xba6+BOeGE2e/0trVNTb1WS3u7IqcjvPDcz7PF1psmSTbcaL38/nevlZyomu4cdk3W7bdWkmSTL2+Y3/x/7d17mFYFvS/wLzAgl0HxAm4URdRNcewY6anMhDTAS2UWYN62qFjneNneUuMiIqRy2WleMhVJotQE79dERPGGl215ozIVwUwkVECR+zDznj98jm2f3XF3zqyZFe98Ps/DH+8wvHxnnvW878x3/dZvPf+HkhNVj/dWvJ+VKz9IkqxY8V5q2tY4m1Mwr8nNo1u3bfLre3+V0edMyC9+MbPsOFVt/gt/yJf7fTFJMmj//nnyCZcKNsYHr76VN29/Mo8NviDzjpycxXc/nQ3vrf6/fm5tr39K2y6d0qptm2yzV58s/82r6dx7+3xx6ml55sQrsvShF5r5K9j0HHPMdzJp0rlJku7du6Vz59osWfJ2yamg8T6xLdh+++1zxhlnpL6+Pp06dcoll1yS2tradO3atbnyNZnXXns9Y849I6ec+t28//4H+deTRpYdqercccd9mXrNxZkz55a0ranJWWeNz/r1/3n5EI03fsyPctHlP8yw4Yfng5WrctJ3zy47UlUaceYPM+HfxqSuri5vL303Z50+tuxIVeOnV0zLT6+anFmzZ6Zdu7b54biLPlq4STG8JjePESNOSZcuW2T0qNMzetTpSZKDv3l01ey0+kcyZvSEXHbFhLRt2zavvLwgd94+q+xI/9B6fHvv1HRqn9evf+hv/v2i6x7M5y7+Xvrdfm7a1nbIwukPJJW/Tov8eve/7sOobKzPi+ddn31mjExatc6fZjycdX9Zkb6Tjkvr9m2z+wXDkiR1K9fkqWN/3LRf2CZs+vSZmTr14jz44C2pVCo54YQfpL6+vuxYLV6l0ny7C6pVq8onfBc3btyYRx55JDvttFM6deqU6dOnZ4sttsgxxxyTjh3/9sKzLWp3abKwJOs2big7QouwVXvXdTa1SjMun2mp1tT5BbI5rK+32Kyp+YGv6XVq277sCFXv2tovlB2hRTjq/XllR2gR1q79U9kRmsxe2+1bdoQm89RbDzfL//OJkww1NTUZMGDAR49HjnS2HwAAAPjb3E4BAAAAKISSAQAAACjEpn+bCAAAAChAg71hjWaSAQAAACiEkgEAAAAohJIBAAAAKISdDAAAAJCkYidDo5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACSpVOxkaCyTDAAAAEAhlAwAAABAIZQMAAAAQCHsZAAAAIAkDbGTobFMMgAAAACFUDIAAAAAhVAyAAAAAIVQMgAAAACFsPgRAAAAklQqFj82lkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJKG2MnQWCYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAEkqdjI0mkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJKGip0MjWWSAQAAACiEkgEAAAAohJIBAAAAKISSAQAAACiExY8AAACQpBKLHxvLJAMAAABQCCUDAAAAUAglAwAAAFAIOxkAAAAgSUPFTobGMskAAAAAFELJAAAAABRCyQAAAAAUwk4GAAAASFKJnQyNZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKShYidDYxVeMmyo31j0U/IftK9pV3aEFmHZ2pVlR6h6juWmt+vm25UdoUVYsPKtsiNUvfpKQ9kRql67Ns47NbUj33u87AgtwtVb9y87ArR4LpcAAAAACqFkAAAAAAqhZAAAAAAK4QI8AAAASFKJxY+NZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAkKShYidDY5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACSpxE6GxjLJAAAAABRCyQAAAAAUQskAAAAAFMJOBgAAAEhSqTSUHWGTZ5IBAAAAKISSAQAAACiEkgEAAAAohJIBAAAAKITFjwAAAJCkIZWyI2zyTDIAAAAAhVAyAAAAAIVQMgAAAACFsJMBAAAAklQqdjI0lkkGAAAAoBBKBgAAAKAQSgYAAACgEHYyAAAAQJKG2MnQWCYZAAAAgEIoGQAAAIBCKBkAAACAQtjJAAAAAEkqFTsZGsskAwAAAFAIJQMAAABQCCUDAAAAUAglAwAAAFAIix8BAAAgSYPFj41mkgEAAAAohJIBAAAAKISSAQAAACiEnQwAAACQpBI7GRrLJAMAAABQCCUDAAAAUAglAwAAAFAIOxkAAAAgSaViJ0NjmWQAAAAACqFkAAAAAAqhZAAAAAAKYScDAAAAJGmInQyN1aInGZ588te5//4Zuf/+GZky5Udlx6k67dq1y8+mXZI5D92S2++cnp132ansSFWnpqYmP592WR568NbMe/yefOMbg8qOVJVqampy7c8vzQMP3pxZs2fmn3vvXHakqtC6deuc9+NRmXbnlfnZ7VekR8/t0qv3Trn2zisz7c4rM3LimWndukW/TRXKcdx8zjrrpDw097Y8Pu/uDDvmO2XHqSp77Ll7brvnlx/72OCh38i9D8woKVF169p16yxY8FR6996l7ChVo3W7mvS/4qR8/e5xGfSrEenca9ts1/8zOfj+C3LQ7edm99MOKTsiNFqLnWTYbLPNkiQHHHB4yUmq1zHHHZbVq9dk4FeHZtd/7pWLLj4vg791XNmxqsqRRw7OsuUrctzw07LVVl3y70/fn3vueaDsWFVn/wP2TZs2bTJowKHZ76v7ZOx5Z+Xoo04qO9Ymr//+X06SDD/kpOz5pc/l++NOSSWV/HTilDz71AsZd+nofOWAfTL3vkdLTlodHMfNo1+/vfLFvfbIgK8OSceOHXLa6f+z7EhV4+TTjs+hhx2SNWvWfPSx3f77p3PksCFp1arEYFWqpqYmV1wxMWvXris7SlXpfeR+qVu9LvcePC6b79I9X5pwbDbfpXtmDb0wq954J/0uPzHdPt87bz/zStlR4f/b332KqNpu5bH77n3SsWP73H33dbnvvhvzhS98ruxIVefTn941D8x+JEmy4NVF6f2pXUtOVH1uvfWejBv31ymcjfUbS0xTvRYsWJSampq0atUqnTvXZmNdXdmRqsLDsx7LBWf/W5Kke49ts/zdFTn7+DF59qkXUtO2Jtt02zrL3llecsrq4ThuHgMH9s/vf/9yZsy8Jjffcm3uu+/BsiNVjdcX/TnH/cspHz3ecssuGTPuzJw7cmKJqarXpEljMnXq9VmyZGnZUapKl97bZ/HcF5MkK19bkm6f750N76/OqjfeSZK8/ZtXsu0XPlVmRGi0T5xkeOONNzJ+/PgsXLgwb7/9dnbbbbfssMMOGTlyZLp27dpcGZvEmjVrc+ml1+TnP5+RXXftlTvv/EV2332/1NfXlx2tasx/8aUceOB+uefu2fkfn++b7bbbNq1bt05DQ0PZ0arG6tUfns2pre2UGTdek3HnueynKaxetSY79tw+v31uTrbeesscOvS7ZUeqGvX19Rl/2TnZ76D++cH3xqShoSHde2ybq2ZemlUfrM6fXnuj7IhVw3HcPLbeZsvsuMP2GTLk+Oy00w656eap+VzfAWXHqgr33jU7O+y4fZIPL7e65IoLMnbUxKxbt77kZNXn6KOH5t13l2XOnEfzgx+cXHacqrL8939Kj4F988as36TrHrukdbu2qemwWbbYpXtWLvpLenz1s1n+e+99bNo+cZJh/PjxGTNmTObOnZsbbrghe++9d4477ricc845zZWvybz66qLceOPtST48u7N8+Yp0796t5FTV5bpf3pyVH6zKPffdkIO+NiDPP/c7BUMT6NGjex6YfVNu+NWtmTHzjrLjVKWTTxmeB+c8lj36Dsjee309U665KJtt1q7sWFXjvNMuzLe/fETOvWhE2ndonyVvLs23vnxEbvnlHfn+uFP+6yfg7+I4bh7Ll72XOXMeTV1dXV59dWHWr9uQrl23LjtW1fls393Sa5eemfzjcZky7eL0/tSuOX/iqLJjVY1jjjksAwb0y+zZM7P77v8t1157SbbddtM+wfiP4tUZj6Ru1docePM56TFojyx7cVEeO/WqfGnS8Ox7zWl5/7UlWbf8g7JjtmiVSqVq/zSXTywZVq1alV69eiVJ+vbtm2effTaf+cxnsnLlymYJ15SOOeY7mTTp3CRJ9+7d0rlzbZYsebvkVNVljz13z1NP/ibfOOio3HPX7Lz++p/LjlR1unXbJr++91cZfc6E/OIXM8uOU7XeW/F+Vq788A1/xYr3UtO2Jm3atCk51abv60MPyHGn/EuSZN3adWloaMjFP5+QHXr1SJKsWbUmFcVkYRzHzeOJJ5/JoEFfSZL8U/du6dipQ5YtW1Fyqurz3LPz85W9Ds7gbwzL/xp+Zl55eUHOHeWyiaIMHHhoBg36Tvbf/7C8+OIfcvzxZ2Tp0nfKjlUVtum7c97+91cy69AL88Z9v8mqN97O9vt9NnOGXZS53700nXtumyWP/a7smNAon3i5RI8ePTJ27Nj0798/Dz/8cPr06ZPZs2enQ4cOzZWvyUyfPjNTp16cBx+8JZVKJSec8AOXShTstddez5hzz8gpp34377//Qf71pJFlR6o6I0acki5dtsjoUadn9KjTkyQHf/PorFtnSVORfnrFtPz0qsmZNXtm2rVrmx+Ouyhr1qwtO9Ym78F7H8m4S0fnZ7dfkZqamlw09vKsWPZexl86OnV1G7Nu7bqcf+bksmNWDcdx85h130PZ58tfyKOP3ZnWrVvn+2eMNcUHfGTlwr/kc2cPzW4nfC0b3l+TeWdNTY8BffO1O8dm47oNWXjbE3nvlcVlx4RGaVX5hLmJDRs25Oabb86CBQvSp0+fDBkyJPPnz0/Pnj2z5ZZb/s1/06FDzyYLS9KuTYu9IUizWlvn+s6m1r7GmHZT23Xz7cqO0CIsWPlW2RGqXn3FL+lNrbZt+7IjVL3316/5rz+JRrt66/5lR2gRjl18fdkRmsw2m/cuO0KTeXdl89y15BN/Y23Xrl2OOuqoj32sb9++TRoIAAAAytBQZXdVLMPffQtLAAAAgE+iZAAAAAAKoWQAAAAACmGLIAAAACT5hPsi8HcyyQAAAAAUQskAAAAAFELJAAAAABTCTgYAAABI0hA7GRrLJAMAAABQCCUDAAAAUAglAwAAAFAIJQMAAABQCIsfAQAAIEmlYvFjY5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACRpsJOh0UwyAAAAAIVQMgAAAACFUDIAAAAAhbCTAQAAAJJUYidDY5lkAAAAAAqhZAAAAAAKoWQAAAAACmEnAwAAACRpqNjJ0FgmGQAAAIBCKBkAAACAQigZAAAAgEIoGQAAAIBCWPwIAAAASSoWPzaaSQYAAACgEEoGAAAAoBBKBgAAAKAQdjIAAABAkkrsZGgskwwAAABAIZQMAAAAQCGUDAAAAEAh7GQAAACAJJWKnQyNZZIBAAAAKISSAQAAACiEkgEAAAAohJ0MAAAAEDsZimCSAQAAACiEkgEAAAAohJIBAAAAKISSAQAAACiExY8AAACQpCWtfWxoaMi4cePy8ssvp127drngggvSs2fPRj+vSQYAAABoYebMmZMNGzZk5syZOfPMMzNp0qRCnlfJAAAAAC3Mb3/72/Tr1y9J0rdv3/zud78r5HkLv1xi7do/Ff2UAAAA0OQ2blhcdoRms2rVqtTW1n70uE2bNtm4cWNqahpXE5hkAAAAgBamtrY2q1ev/uhxQ0NDowuGRMkAAAAALc4ee+yRRx99NEny/PPPp3fv3oU8b6tKpdKSFmgCAABAi/d/7i7xyiuvpFKpZMKECdlll10a/bwttmRoqtt18J+98MILueiii3LdddeVHaXq1NXVZfTo0Vm8eHE2bNiQE088MQMGDCg7VtWpr6/PmDFjsmjRorRp0yYTJ07MjjvuWHasqrRs2bIMHjw406ZNK+RNjo/71re+lc6dOydJevTokYkTJ5acqDpNmTIlDz30UOrq6nLEEUfk0EMPLTtSVbntttty++23J0nWr1+fl156KfPmzcvmm29ecrLqUVdXl5EjR2bx4sVp3bp1zj//fK/JBduwYUNGjRqVP//5z6mtrc3YsWOz0047lR0LClH44sdNxX+8Xcfzzz+fSZMm5aqrrio7VtWZOnVq7rrrrnTo0KHsKFXprrvuSpcuXfKjH/0oK1asyLe//W0lQxOYO3dukmTGjBl5+umnM3HiRK8XTaCuri5jx45N+/bty45SldavX58kCt8m9vTTT+e5557LjTfemLVr12batGllR6o6gwcPzuDBg5Mk48ePz5AhQxQMBXvkkUeycePGzJgxI/Pmzcull16an/zkJ2XHqio33XRTOnbsmJtuuikLFy7M+eefn2uvvbbsWFCIFruToalu18HH7bjjjt6UmtCBBx6Y00477aPHbdq0KTFN9Ro4cGDOP//8JMlbb72VbbbZpuRE1Wny5Mk5/PDD061bt7KjVKU//vGPWbt2bYYPH55hw4bl+eefLztSVXr88cfTu3fvnHzyyTnhhBOy7777lh2pas2fPz8LFizIYYcdVnaUqtOrV6/U19enoaEhq1atKmQRHB+3YMGC9O/fP0my884757XXXis5ERSnxb5iNNXtOvi4Aw44IG+++WbZMapWp06dknx4PJ966qk5/fTTS05UvWpqajJixIg88MADufzyy8uOU3Vuu+22bLXVVunXr1+uueaasuNUpfbt2+f444/PoYcemtdffz3f+973MmvWLO97BVuxYkXeeuutXH311XnzzTdz4oknZtasWWnVqlXZ0arOlClTcvLJJ5cdoyp17NgxixcvzkEHHZQVK1bk6quvLjtS1enTp0/mzp2bgQMH5oUXXsjSpUtTX1/vhBFVocVOMjTV7TqguS1ZsiTDhg3LIYcckoMPPrjsOFVt8uTJuf/++3PuuedmzZo1ZcepKrfeemueeOKJHH300XnppZcyYsSIvPPOO2XHqiq9evXKN7/5zbRq1Sq9evVKly5dfI+bQJcuXbLPPvukXbt22XnnnbPZZptl+fLlZceqOitXrszChQuz1157lR2lKk2fPj377LNP7r///tx5550ZOXLkR5dcUYwhQ4aktrY2w4YNy9y5c7PbbrspGKgaLbZkaKrbdUBzevfddzN8+PCcffbZGTp0aNlxqtYdd9yRKVOmJEk6dOiQVq1a+UGgYDfccEOuv/76XHfddenTp08mT56crl27lh2rqtxyyy2ZNGlSkmTp0qVZtWqV73ET2HPPPfPYY4+lUqlk6dKlWbt2bbp06VJ2rKrzzDPPZO+99y47RtXafPPNP1oSu8UWW2Tjxo2pr68vOVV1mT9/fvbcc89cd911GThwYHbYYYeyI0FhWuyp+0GDBmXevHk5/PDDP7pdB2xqrr766qxcuTJXXnllrrzyyiQfLtu0OK9Y+++/f0aNGpWjjjoqGzduzOjRo7PZZpuVHQv+nwwdOjSjRo3KEUcckVatWmXChAkm+JrAfvvtl2eeeSZDhw5NpVLJ2LFjlZJNYNGiRenRo0fZMarWsccem9GjR+fII49MXV1dzjhjfWpFAAAAaklEQVTjjHTs2LHsWFWlZ8+eueyyyzJt2rR07tw5F154YdmRoDAt9haWAAAAQLFa7OUSAAAAQLGUDAAAAEAhlAwAAABAIZQMAAAAQCGUDAAAAEAhlAwAAABAIZQMAAAAQCGUDAAAAEAh/jeF6v6fYhEKnwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1440x1440 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Saved.\n"
     ]
    }
   ],
   "source": [
    "if __name__ == '__main__':\n",
    "\n",
    "    optimizer = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)\n",
    "    scheduler = ReduceLROnPlateau(monitor='val_acc', patience=3, verbose=1, factor=0.5, min_lr=1e-5)\n",
    "\n",
    "    model = CNN((16,16,16), 10)\n",
    "\n",
    "    gen = ImageDataGenerator(rotation_range=10, zoom_range = 0.1, width_shift_range=0.1, height_shift_range=0.1)\n",
    "    gen.fit(X_train)\n",
    "    \n",
    "#     model = load_model()\n",
    "    train(optimizer, scheduler, gen)\n",
    "    evaluate()\n",
    "    save_model()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}