ocr_project.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Mini-Project OCR Code\n",
    "\n",
    "This is a walkthrough of the process we went through to develop an OCR model that recognizes handwriting.\n",
    "\n",
    "Our libraries used are listed in the README, we will utilize requirements.txt to load them all at once.\n",
    "\n",
    "**IMPORTANT** Make sure to use `python3.6` (the version we use for class projects) to run our program with little issues. Information can be found [here](https://courses.cs.vt.edu/cs4804/Spring24/projects/project0.html#python-installation) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "hide_output"
    ]
   },
   "outputs": [],
   "source": [
    "# Install libraries\n",
    "%pip install -r requirements.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load libraries\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "import keras\n",
    "\n",
    "from keras import layers, models\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.model_selection import KFold"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Helper functions for loading dataset\n",
    "\n",
    "We need to load our dataset. We will create a helper function for the model OCR. This function will load the English Handwritten Characters dataset that should be in given path."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_eng_dataset(datasetPath):\n",
    "\t # initialize the list of data and labels\n",
    "    data = []\n",
    "    labels = []\n",
    "\n",
    "    # loop over the rows of the A-Z handwritten digit dataset\n",
    "    for row in open(datasetPath):\n",
    "        # Skip the first row\n",
    "        if row == \"image,label\\n\":\n",
    "            continue\n",
    "\n",
    "        # parse the label and image from the row\n",
    "        row = row.split(\",\")\n",
    "        imagePath = \"eng_dataset/\" + row[0] # hardcode the path\n",
    "        try:\n",
    "            image = cv2.imread(imagePath)\n",
    "            image = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)\n",
    "        except cv2.error as e:\n",
    "            print(\"[ERROR] loading image \", row[0], \" fail\")\n",
    "            continue\n",
    "        \n",
    "        label = row[1][:-1] if len(row[1]) > 1 else row[1] # remove '\\n' at end\n",
    "\n",
    "        # update the list of data and labels\n",
    "        data.append(image)\n",
    "        labels.append(label)\n",
    "\n",
    "    # convert the data and labels to NumPy arrays\n",
    "    data = np.array(data)\n",
    "    labels = np.array(labels, dtype=\"U1\")\n",
    "\t# return a 2-tuple of the English Handwritten Characters data and labels\n",
    "    return (data, labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dataset Pre-Processing\n",
    "\n",
    "Next we will pre-process the dataset in order to train the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def process_dataset(data, labels):\n",
    "    \"\"\"\n",
    "    Help function to pre-process the dataset for ready to train model.\n",
    "    \"\"\"\n",
    "    # the architecture we're using is designed for 32x32 images,\n",
    "    # so we need to resize them to 32x32\n",
    "    data = [cv2.resize(image, (32, 32)) for image in data]\n",
    "    data = np.array(data, dtype=\"float32\")\n",
    "\n",
    "    # add a channel dimension to every image in the dataset and \n",
    "    # data = np.expand_dims(data, axis=-1)\n",
    "\n",
    "    # scale the pixel intensities of the images from [0, 255] down to [0, 1]\n",
    "    data /= 255.0\n",
    "\n",
    "    # convert the labels from integers to vectors\n",
    "    le = LabelBinarizer()\n",
    "    labels = le.fit_transform(labels)\n",
    "\n",
    "    # account for skew in the labeled data\n",
    "    classTotals = labels.sum(axis=0)\n",
    "    classWeight = {}\n",
    "    # loop over all classes and calculate the class weight\n",
    "    for i in range(0, len(classTotals)):\n",
    "        classWeight[i] = classTotals.max() / classTotals[i]\n",
    "\n",
    "    return data, labels, classWeight"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Verification\n",
    "\n",
    "To verify that the dataset looks correct, let's plot the first 25 images from the training set and display the class name below each image.\n",
    "\n",
    "We define the helper function here:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_train_data(train_images, train_labels):\n",
    "    class_names = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', \n",
    "                   'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'G', 'K', 'L', 'M', \n",
    "                   'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',\n",
    "                   'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'g', 'k', 'l', 'm', \n",
    "                   'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']\n",
    "\n",
    "    plt.figure(figsize=(10,10))\n",
    "    for i in range(25):\n",
    "        plt.subplot(5,5,i+1)\n",
    "        plt.xticks([])\n",
    "        plt.yticks([])\n",
    "        plt.grid(False)\n",
    "        plt.imshow(train_images[i])\n",
    "        # The CIFAR labels happen to be arrays, \n",
    "        # which is why you need the extra index\n",
    "        index = np.where(train_labels[i] == 1)[0][0]\n",
    "        plt.xlabel(class_names[index])\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load and Pre-Process\n",
    "\n",
    "First, we need to load and pre-process the data. We will use the functions we defined previously.\n",
    "\n",
    "Make sure the English Handwritten Characters dataset is loocated in `/eng_dataset` in the following format:\n",
    "\n",
    "```\n",
    ".\n",
    "├── ocr_project.ipynb\n",
    "└── eng_dataset\n",
    "    ├── english.csv\n",
    "    └── Img\n",
    "        ├── imgXXX-XXX.png\n",
    "        └── ...\n",
    "```\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define directories here\n",
    "datasetPath = \"eng_dataset/english.csv\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] loading datasets...\n",
      "[INFO] pre-processing datasets...\n"
     ]
    }
   ],
   "source": [
    "# load the English Handwritten Characters datasets\n",
    "print(\"[INFO] loading datasets...\")\n",
    "(data, labels) = load_eng_dataset(datasetPath)\n",
    "\n",
    "# pre-process the data and labels for training\n",
    "print(\"[INFO] pre-processing datasets...\")\n",
    "data, labels, classWeight = process_dataset(data, labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training\n",
    "\n",
    "Time to begin the training, we need to split the data for training and testing first. The training data will be shown here."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAj0AAAI8CAYAAAAazRqkAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAACYFklEQVR4nOz9eXxcd53n+7++59RepaW075tXeY+TOI6dlRCyQFhCOkADARqapYFpmtt9Z3q5w/Rv+t5e6JlhGoYGAiEQCJAOECA72ZyEJN4Sx7stW5Ila9+l2qvO+f7+kGJL8W5LKlXV5/l48ODho1PHH+XrU/Wu892U1hohhBBCiGxnpLsAIYQQQoj5IKFHCCGEEDlBQo8QQgghcoKEHiGEEELkBAk9QgghhMgJEnqEEEIIkRMcF3JySZGpG2qdc1WLOI32ziSDw5aa7etKW6bHzt3xQa116WxfV9pz/sm9mV3m4t6UtkyPs7XlBYWehlon256qnZ2qxHnZcEvnnFxX2jI9zMojx+biutKe80/uzewyF/emtGV6nK0tpXtLCCGEEDlBQo8QQgghcoKEHiGEEELkBAk9QgghhMgJEnqEEEIIkRMk9AghhBAiJ0joEUIIIUROkNAjhBBCiJwgoUcIIYQQOUFCjxBCCCFygoQeIYQQQuQECT1CCCGEyAkSeoQQQgiREyT0CCGEECInSOgRQgghRE6Q0COEEEKInCChRwghhBA5wZHuAoQQQmS+uE4yZieIaY1PKQoMD05lprssIWaQ0COEEOKSPRUp4O/2vY9QVz7F9SN8vflhbvDa6S5LiBmke0sIIcQle3p0Nb6HC2j+/zqxf1vM86HmdJckxCnkSY9YkOI6yYAVZ8I28Bs2JYYLn+FKd1lCiDMIWy7cozaprm7cY3WMJH1E7AROZUo3l1gwJPSIBenVmJs/33MP4aMFGNUR/n+X/Y4P542kuywhxHlwxGz2jFTxRH4J1Y4RVrliBAxPussSQrq3xML0SngJjicKWfpPRyn9pZfHh1enuyQhxHlyRG26Bgt5enQlO2MNjNqpdJckBCBPekQaJLVFjxVlwDpzd9XuiWo8IxprYADPcB0to6XsrEic8fxCI0GNw41bOeeiZHGJInaCHivBqH1+XZTSnpnhrW7oiFb0RPJRlgbASNgkx9wcHisjYMaJ+w+kuVIhJknoEfPuSDLOnx68h75d5agzTO7w9iuqDo1gA+5jw4w+UsFH6v/89CcrcC0b51trfyazRRaol2N+vrLrT0gdzjv3ydKeGeONuMF/a/8QR3tLMY56aeoZm7pnh6h6rpLR3dX8al0Z1994kEXOSLrLFUJCj5h/rakihl+uYMl3WyAeP+052rLQUz+zWjso6+mn3DzDYEhD0fehFby+tIEbvK1zVba4BNsii/A9mUfZw/vB1mc/WdozY+yK1XF0Wx0lb2q8Q0mM3iFsIHXsOPk9feSbJmZyDYc2VYL/aLrLFUJCj5h/Se3ATIA9Pn4i2JyVbWGHw6ccNnw+jKIg2uMi6Vc4lTUH1YoLEddJWpNJ+qwAFurE8R0j9bjHbKzRsfO6jntMs22sgRd9rZQaEZqcTunqWoAithtHVOEeTeGYSKKTyckf2BZ2bPJ+NJJgaxk+KhYGCT0iY9mrFtF6R4BkXZyltce41ncYcKe7rJz2ZgI+veszpHYVzui69AxoKvYPc76xtODAGAceWs6fli3Ds3KU7675CRtl8s+CM2Z58QyA78gQKhrHDksXlljYJPSIjBVq8HPDLbv4p8pncSoDr5J1fNLtYLwS9VIhjfftQyenzdixLKxE8vwvdKiN6t4hlMtFx4frOLi8ko2e/tkvWFySsZQX75CNdfQYaBv0ObouhUgzCT1i3vlUnHiRRi1vwkie/O6vYgn04DDW+PiM8w2PB6O8FO33zjgeqjSo9wwTNH3zUrc4NxtjsjsjFEanTj9N+UztqcJR7L4B7Fhs8rWhMLiSGAlIaHmrWqiUBmzpWhaZQd5JxLxb7hrh+ht382LjIvS0KcxWbxENvyvG8dzOmS9Y2kDrnUVYy2aO61le1co78/YCMtYjo5yhPc2DJTQ+7IN9h9CWNTnI3bIwUmBrdYaLCSHE+ZPQI+ZdnSPAt2peIFn97Izj/za8ll/tewfFz6sZj8ljlQEW3dDGzxf/esb5TmXiQJa3zzRnas+7yj9I7A8VOPcBWqNTKbStUeea7SWEEOdpQYSepLY4kEyyJ16Nic1qdzfLnW5MJSP+s5VbnTobx2fG4XRf6JXCYyZlGfsMY3g8GFUVWAX+GcfHG5ys9A+f0p4eM0XCOM0/AMk8QqRNTyrE64kSBlL5NDgHucwdpsDwnvuFC9SCCD2DVpQ/O3gPo89XoA2of1c7P138sIzVECKDqdoq2j5cgbl+FKVOJpemYCt3F29NY2VCiPP1k/G1fOfpm8lrMxhbm+BbN/yEd/ti6S7roi2I0DNqG/QcKGP5j1vB7WJ/YzWxRbISqxCZzCoOENg4wAtrf4qpTj7BMTBk120hMsSrw03UPW3hfmYHznuuZO9Vtbzb15Lusi5a2kJPUlvsTljsjDWwN1yNp99Ax+Io0wRbBi0KkSlCdoytcT/7YzU80b8S96hG2xqtFKZh41aO8+qqrvKN0bGoiZKr1+IYDqM7u7Fj57F4pZhzI1aErfEgLfGKGYtOvtTdhC+lMfPz0ZaFHY2BbWEGg+jaciy/m1CNosQxkcbqxXTT27LUMc4mbyd1jsCMc3pSIV6JVdGZLGLP8WoaJ5LoVOqM2wZlkrSFnjE7xl8c/hijT1XiGdbU7J/AjkQw/dKlJUQmOZQ0+NzWj5P/khf3qKZ41yCWvvB3x48Wv8Lox7y03lHM8K5SFv/EgIOydcFCsDUe5AvP30PRdgfTeipxRjXOUAprRQPmRByzswdrfJzkmgaO3u2ivGmQd5Qc5FrfUcB/xuuL+TO9LSca4CPvfpG/L90345yHJ1byv5+4jcJDiqpeC2drB6dfgCLzpC30RLSm82gpKx46Rqqre3K2RrqKEUJctK5UIZ43fZT/8HXseBzrIheo2+wx2NjwLDaamx0fJPl4ITKVYWFoiVdQtN1B6X3bJ5cTmGKWlJBcXkO0woPbaeLsn1wRPVzp5gMbt/EvFTsmz1MSeBaK6W1ZeN1atm5sgLeFntcn6qh6ycb72+0ApLJo0cm0julRempqchb9BxWzzzmR5I1jtXwtf+WM44vcfdzsa6fybY9mxfyyMVAa9Czcy6YyMGHGwGcjBccTRXSkDpGnDPINj8zsnAcjVoTnohXsjdbwRNcKPCP2ZOCZ3sbJBGYogdNpYEYSMG1BSkNpaacFyEJN3q+WhdIafYY1sJRFVn42L4iBzEKcjeNIN3UP1PFk2XUzjg+t0yRvf4RPF/SmqTIx57SNM6x5dbCRIkeYFZ4urvaMUqAyd8pspng9kcdfbvkQpS878IxZ5O3uO+Ubvx2OYnT04On3oONx7NCpGwMLsZBI6BELntXXj+upft6+s5aR2sjBGytBQk9Wc8Q1feN57AlU4zPiXOEeTndJOaEzWUzBbidFP92GTqVOO6ZDJxNYQ9IeInNI6BFpNWiFeTxcz+vhBrYcX4x/+PwHwKrse/IqhBDzxjGR4NCRSv5TwZU0eIZ4T94eljqze/yVhB6RVq/HC/na83dSucWgYMzCv787qwbNCSHEQmW0dbP4wXp2PrOe3zebjH3Qe8pMrmwjoUekVXcqSMF+B/kPn/kRuhBCiNlnDQ1jbBkmALhuvZJ9N1eeMpMr20joEfNuxIrwq1ATr4wtZkdvLYE+G32em0qapaUkV9QQDzoZWq1Y5Omf42pFulkuRZE/whJfP1XOEZwyI2jOnOneNPPzsVY0EK2YuV+aazSF+2AXqd6+NFUsxIWR0CPm3f6kh3/Y8l5qnlKUjKXwtBwnZVvnfiGQXFHD0U8rbl6+hzv9PdweOATIlPWspQwSfsW1ZW3cVfA6hQb4lGw8O1fOeG/WVtLyER83XrUXh3HyXn22ZTn199XgkNAjMoSEHjHvBqx8Aq0O/I/vRMfjF9SlFQ86uXn5Hr5b8+rUEQk82c52QqN7IOsHWC4EZ7o3U/keypcO8L3aF2esvfMVM8n24ivkLhQZI22hxwC000bn+zHDBehYHDuWuTu3ivNXYY4RWp5g9I8uwz1mE9jXT6q1HTMYJLmmgXCl+4yvHVqtuNPfM4/VivmyM57ggeFNtIZK6NhTydLRUbJgq5+c5O1P8vCb62mZKOOywk7uKdxGo1OikUi/tIUeEzB8KeKV+bgNA2NoFLs/ma5yxDxqdiX412seYttlTbzc18TwjyrIb+9E15Zz9G4XH9i4DeMM89EXefqlSytLPTC8ied+uoGSvQmWDEzAsa50lyQukntPB0u/W8l4fi0PXtdE4139NDoH012WEOnt3jIdFimfE4ffhSP09qXnRLYqMLx8MDDOBwO7uN/bzTdK7qLAUFh+N+VNg/xLxY5zLF8vgWchMc7yPEZrhY3GPI/rtIeKKdmbwPn0DjRM7sVnnM8rxUJjDQzAwABOoKDqanqSQUBCT0ZTp9+uItOkLfR4lMGa6m52X7ME10geZW+4cfcNpKsckSYNzkHG1iZw3nMl4SrFO0oOprskcYEqHGNEVkcZ/Nh6PGM2+bsHsQ4fxTEcZnhXKTc7Pkhj/hCfKdvCZo/MvBIikzjqa5lYV0ksaDJ4hc1id2YPWk9b6Mk3PPzX2kfZ9f4a9kVreNTcRP122U8n16x1hfjf1z/I3g21lDgnuMHXIjsyZ5gVTov/c/WDvL62gWf7lzF8fxUFR9rRx3tY/FOT5BNBdq2u4v5P2Gyu/UO6yxVCXIDwygpGPzXBh5teZ7Gnl+u9PUDmvkenb0yPMljj8rDGNcgBTwe/LN4IDplMlmuCpo/3+iO8139o6kjm3ky5KmB4uNUX51bfIapcI3wjeBeFhsIOh2HfIQyg2LGe9lARSX3mpQlS2oDp6zUphTJNtJEdj9WFWOhsrUhqi6R9sls5kW9yS91B/qYkO96jJWUIIeaccyDC0a01bIx8hDNlmLG9xSwamJyxZS5pYvTyMqJFBqNXxlnilk1lhZhL7r4Ie15dzKbxjzByqIjFvSGycUMgCT1CiLnX1sni+5PYBb4znlI6PoA+PrkcwfiaUlx/0ss9NdtZ4u7lCncIkO5vIeaKOtLBkvvKsfO9lIaG0Md7JfQIIcTZGNhoA9RUV7W2LNB6sqvr8NGzvnZ6x1ciz+COyn18vvCtaesSeISYDSYa2wTlcoFlnbxHJybg0AQw817MNhJ6hBCzZpGrn8hVEbrM9XiGNSXbhrD2H053WUKIKcvc3YxuimN5L8czqCne2o91ji8k2URCjxBi1qx1JXhg4w9ov7yEB3uvYiDRQP4BBTobH5QLkXk2eSZ44Jrv035VKfd1bmYiXIlPQk+a2TBsm+TZMdzKiVPJAmVCZIKA4WGjBzZ6RhhIHeC+wiYKfT6wLuCBuWFgO8BQsgnFQmRpm7hOkcQinHKj7PMPtEltEddJLDROTHyGLEo73wKGh80e2OwZ4nDpUZ70V3PKSDulUA4nyjSwnAqnyp4OrwUXenQ4QvFOk/d4vkwgGOFTS17ji8FDuJUz3aUJIS7Aak8n4WvDdBSsveDXxldFWevpmIOqxKXalUjxtWPv42BXBWa7h8Zj5z/L56FQGd9ouYmRcR8bG9r5++pHWSR7ci04jsZ6Bq6rJFypiC6Pcbm/Pd0lzZoFF3qssXHKfnuEshfziNcF+f4XNvGpq/biNiX0CJFJNrhjPLTxe/RemY/F+a+1Y6IpNSdY5rQBz9wVKC7KlvByOn/TyNLH+1GxQeyBofMKPZa2uf/4Jjz3BVl6aJQdd69gzx/vYJEzNOc1iwsTWVqK68N9fG3x41SY4yx2WmTLZIIFF3qwrRP7trh1I7FQEVZWTpwTIrv5DBfr3ADxi3i1dHssJMqyiSWc9FsR2mPF+PpsrENHzu+1NkRsFyEdZzDkp7wzgn3oKO6RYsK2G5DQky5OZWG5wMjLm3E8XmCyvriLd/tiZNu9uPBCjxBCiAXFMTCO/UI1mwa+gqvHSUNr+Lxf6x6zeKRtDWMpL2MdBVRGRs+yRa2YT1f42vjxzRsI166ecTzeGOOa/OycdSmhRwghxFlZx3uofjBKjc8LiST2yPkHF89AnL59hfxmcB35R0xUKDKntYrzd51ngl9v/C5DG2Z2XRUaMZockI3dyxJ6hBBCAOBUKSwXmMFCdCSKHY2hkwl0MoHV13/qCwwTw+uZXOhO22hrKgolk9jxOGiNEU/hHFdYLifOkIZU9swEynQ+w8XK0/ZeZV/YeYuEHiGEEAA0OYYpubaHQwVNeIYUVVvCqNd2n3GdJXNRPb3vLGeiHswEuMYUZhwC3Rb5WztJdXVj9o9R+qaPRJ6BvzuODp9/15gQs01CjxBCCAAWO9082PwAw8sc/HRkI88MX03JVgP06Z/OxOuC+N7Xx33LfsbhZBmPDq2lK1xI++4q/G1B6OrG6unFMzSMxzQhmcSKx1GmrL0m0mNBhB5TaWyPhpJCzGlbMFtFAUyXhXkB012FEEJcHKcyqXEEqAH2+Dp5LLgJR2U5OhbDDoXR8TjK4cAI+MHtJlTsZFHBIOvcbvKMLo7lleA1k7QGytFOAwCdSqFTqfT+YkJMWRChp8iA6y/fzwt/3owRKz5x3A6kuLt5Bz5D1ugRQoj5tNrdRclN3RyorsXTZ1L7bAhe241ZV0P37VWML7ZxVoX5ctEeAIoMg6t8R1ni7uMPJU3YbhdGmn8HId5uQYSeEtPP/6x+iuHKJ2YsYuZEU2SauFV2LIokhBCZotnp5KHmnzKxTPPtwet4qesqglsVieogztsGeGLV/fiUpshwAS6Cpo+rDQubEb4XHCHmrpTQIxacBRF6AIKmj6B08wohxILgVCZlpp8yE5b5evl9scJRV8NYuYu6/BGWOv2nfQ2YuAyLmHHqsATD58PIzwOvh2QAPCo5D7+JECctmNAjhBBiYbrc007xrV3sX1KJrzTEF8reuPCLKEXq8mV0vMtDoixF89J2Vrh64dTtLoWYMxJ6hBBCnNU6l4NfLf8ZsWU2TqXIM1zABY61VAYjyz18+L0v8vngVnyGSUBl73owYmGS0COEEOKsTGUQNM//iUy+K0pvqYOixY0nDxoG0RLFIncflQ7ZWV2kh4QeIYQQs+qukh38wx+Xc/DmopMHlWZNw1Eu93SSLTt2i8wjoUcIIcSsus03wfVrf0xSn9yhy0ThVAZe5U5jZSLXSegRQggxq5zKpECWGhELkCyjIIQQQoicIKFHCCGEEDlBQo8QQgghcoKEHiGEEELkBAk9QgghhMgJEnqEEEIIkRMk9AghhBAiJ0joEUIIIUROkNAjhBBCiJwgoUcIIYQQOUFCjxBCCCFygoQeIYQQQuQECT1CCCGEyAkSeoQQQgiREyT0CCGEECInKK31+Z+s1ABwbO7KEadRr7Uune2LSlumjbRn9pC2zC6z3p7Slmlzxra8oNAjhBBCCJGppHtLCCGEEDlBQo8QQgghckJOhB6lVCjdNYjZI+2ZHZRStyqlDimljiil/ku66xGzQyn135RSf5nuOoQ4nZwIPUKIhUUpZQL/B7gNWAF8RCm1Ir1VCSGynYQeIUQ6bACOaK1btdYJ4OfA+9Jck7hISqm/VUodVkq9DCxLdz3i4iml/kkp9cVpf86qJ3cSeoQQ6VANdE778/GpYyLDKKUuBz4MrANuB65Ma0HiUv0CuHvan++eOpYVHOkuQAghREa7Fvi11joCoJT6bZrrEZdAa/2GUqpMKVUFlAIjWuvOc70uU0joEUKkQxdQO+3PNVPHhBDp9x/AXUAFWfSUB6R7SwiRHtuBJUqpRqWUi8nuEXlCkJleBN6vlPIqpfKAO9JdkLhkv2DynryLyQCUNeRJjxBi3mmtU0qpLwFPASZwn9Z6X5rLEhdBa/26UuoXwJtAP5OBVmQwrfW+qQDbpbXuSXc9s0m2oRBCCCFETpDuLSGEEELkBAk9QgghhMgJEnqEEEIIkRMk9AghhBAiJ0joEUIIIUROkNAjhBBCiJxwQev0lBSZuqHWOVe1iNNo70wyOGyp2b6utGV67NwdH9Ral872daU955/cm9llLu5Nacv0OFtbXlDoaah1su2p2nOfKGbNhlvmZssTacv0MCuPHJuL60p7zj+5N7PLXNyb0pbpcba2lO4tIYQQQuQECT1CCCGEyAkSeoQQQgiREyT0CCGEECInSOgRQgghRE6Q0COEEEKInCChRwghhBA5QUKPEEIIIXKChB4hhBBC5AQJPUIIIYTICRJ6hBBCCJETJPQIIYQQIidI6BFCCCFETrigXdYXIkvbhHScuLbPeI5bGfiUC6cy57EycaGkLXNDUltEdOKUdnaicCoDAwNTKdzKmaYKhRDZKuNDz75kgv+79S4OtVeCPv05FVUj/MPSR7jJa81vceKCSFvmhlfjJn9z+KN0dRXNOO4tiLGusotq7ygrfV3c7m+jzPSnqUohRDbK+NDzWrSJrifrWfFwN+jTf1L231jF419cy03e1+e5OnEhpC1zw9Pjq4n+upwVT3fPOB5eXsaOa5fzanmSXQ01rG46Tpk80BNCzKIFHXosbTNuxwhrmzN1eByNleEZ0KTajp3xg9K3spy2cDEdqRA+pSgwPNI9kkZvtevE27o3LrQtZxyXdl3QktpizI4R0ZrWcAm+AXuynafx5vvwDAWxXQ56gvkcTZZSanbhU4qg4cVUMgQxU8R1kjE7Qext97G05dw602emEygwXPgMV7pKWzAWdOjpsiJ89dj72bmvCWWp057j7jepPRQ+44ckgK99nMOPLeGGuv+L4voRvt78MDd4zzxuRMyt9lSEP2/7I/YdqEXZJ9v1QttyuvKmQf7XsofY6JmzssUleDVu8n8f/AT9bcV4j5vUtY6e8kXGGByjZI+feKeDSGsBf9vyx6TyLBqW9PG/Fj/EOrc7LbWLC/frUBn/fe/tRHsDJ45ppaUt59iZPjONYJwvrX2BLxe25nzgXNChpz0VYPcfltD8/V5UNH7ac3QiiZ6YOOOTIAB9qJW63gGU203vu+t5vraZG7z75qZocU4tyWKOvtDIivuPQzJ14viFtuV0XXc28FrdIjZ62uemaHFJXg0vIf5oGc2/aUfH49gToVPOSfX04R4ZxW2aFDgcKLcLnE46PlzHrtpa1rn701C5uBi/HVxH4S/yaPhDx8mDhiFtOcfO9JkZba7k51+9gi8WHiXXn4UvuNBjaZshO8qoDXtiy3GNKHR3H1YkctHX1MkE1uAQAO7xOiKWPOKbb2+165Cl2BNbhGsErK5edDJxQdeZ3pbTuSbqidsy22chsbRNXKdIYtGTKMA9ZpPq6j7zC2wL++33uWHinKglJm274EXsBIN2grBtcGwiiHcgMbO9pS3nXFKbOMIK3dOPFQ6fOO4qCzKYcDLZ6XXu2DO9LafLM2xKTXdGz6xccKGn34rwxfYPsGtXE64Rk6qdcXTiwj4YxcLTkYrwxda7Obi7DvewQfXuGNqSGVjZbMiO8mS4nj2RWp45toySUWnvbPZEpIS/3fU+Up1+8loNCjp7kBbPTNPbcrr8JSN8c9XP2JzBwwgWXOjps5zs3raI5m/3osfG0eEIdip17heKBa3TCnD0lXqav3ccHQpjh8JoW94Ss9mAZfDo4Fp2ddagO324RqPpLknMoWfHVhD8tZ/gM0chHj/1qZ3IGDPacpqeP1rC1qbFbPa0pqmyS7fgQg+ANkG7nSivF+X1XtSy0TqZxB4dQ8dPPxZIzD1L2/RYEQYtJzsizbhGFfbA0Iw3Q+VwYBTko7zei/+LlCKRr3AbyVmoWsyWmDbpj+SRHHXjHVcYCQsNKKcLo7BgcszOGehYbPL+tSzMOOwN17DT20apmaDS9MosvQUiYic4biWZsJ0cGC3HM5zCGhhId1k5y6kskn6Nqq7AMRHGHp/AntbNdb6ilhPXhH1KWzoji0nqzL73FlzoqXKk2LxxPy/lLYVU8KKv4+twUPvEGLwhA5bTpceK8Nmjd3N4Rz3uEUXljtgpXZVmRTk976ljZI11SZui1Db2sMnXwuTkTLEQdKaK6DhQQdUfwD2awOwZJgUYi+vpfE8poUVnfoKbf9BBzW+7SbUdo3hviC2/uJynS9ZTuHqQ7698gDWuzH7jzRbPRAv5y513Qasf/3FFZVufdGmlUYMjxOXXHmJr8SKcQ2XUPJfE+cwb6S5rQVlwoafM9PPt2qeZqH78kq7zV8ffQ/uhZfilvdNm0HLSsq2eZd/qnPzWHouj39ZVaRfnE74xzHMbv43n9KsSnBePMggYMg12IelMFFNwyCD/0V1oyyI1FXhjtQVU3NrJD5f87LRDKi3gztpPkdxWgGrV8MYBag56UW4XHZ9exsEl5axxjc3r7yJO76WJZQQf91P82/3oZAorGkt3STmt2vTx3frHiNRa/Ca0jG/3v4+q5+ULwnQLLvQABAwPgUtcSqDEHaLNvIRPUXHJLBRGCvTYOPbExInjyuHACAZReX7CNQFKC4aocUiXRbaxUJhJjR2Nzlh7SRuKAleUGkfgjK/1uxJow48CdCqFnphARR2YcUjqBfm2lZOS2sQR01ijEkIXAlMZFCgvBQZUOMfQDsBQqKRFaNTP89EARWaIJkeCkrdt8RKyYxxJKoZsH4dHS3EnJhcPMXw+jKIg2uMiVqTwGZk9sUjePcS8M0uK6flAEyNXJAkUhfly/WsYSEAVQoi5oPqGqHy8kK+2fJpYpcWfXLeFvys5OOOcpyJl/NVrd+Fp8eDr1hQe6SEF2KsW0XpHgGRdnKW1x7jWdxjI3KfqEnrEvNPBfMY2xdhy/TcpMEx8yoUpT3mEEGJOWIOD5P1mnHzTJLlhOU8uXsFfF++fsTrzKxOLKXvKRfCRN8GySE1NAgo1+Lnhll38U+WzOJWBV2X2Onc5EXoMnw+jrATt8xAuNwg6ZCrlXElqi7ZUjJZkMTvCK3COK7TWKIcDs6QYHcwn0lhIQcE4JbIXjBBZz/D7MUqL0T4P8SKFz5AZtfNOa3Q8jgaMuIWlJ5+sj9lRDiUddKeCbB+sxz1mnzLbSxuQ74gRNH1pKHz25UTo0c2NHLkzH90UYXlVK+/M24vM8pkbg1aUL7T8Md0v1uAehcrXI+hYHCMYpOcDTYxtilFQMM5nF7+MW+XEPz8hcpq1djFH3+MjWZlgZWMbl3s6gUtYokLMmi3RYr7yyofx7/Pg7dOUHMr+2Xc58akTqfFz5Y0H+E7dkziViSPndx+ZO6O2QfveKpbfdwxrYBCdTKFtC5XnZ+SKJM9d903KTRdOZUqXlhA5YKLey8Yb9/GlimcpNeJUOTJ3PEi2eSPSQPELboof3Am2xkpl/1pnWRV6RqwIryfy6EwW80pvI77QVGZV4HfECRgZvHZ2BlEW6HhixsKQ2lAoU5NnKOnSEiJLlDhDhKoN8i5biRGOobt6z9A9EqfUiOM3FMalLMglzlueESVaqtFrl2KGE9A7gDU0jBlN0tdXyP3jVTzftxTPqH3KIr5GXh6qsgzb7yFUbVDiPHWD4EyVVaHnuWgFf7nlQxTsduIbsPEfmBx9LoQQYvbdlrebvXdVsf/6cqKHy1j0Cw/snLkgrDYVhc4IpaYDpzJlaYp5ssw5xvvesZXnli5ltDef+t/m4358BNU1QO0vG/nWjjtxj9gU7e0/pUtLL2+g5e4AviWjrCo9wC2BvWTyjK3psir07I3WUPqyg6KfbkPbmpTs7SSEEHPmcreLBxp+j91g8+mym+l8cSlvf56uFeSZMXnSPs9qHAH+pWIHqYqt/HCsge/ueR/lpok1MIDnsWG8hkLbGus0n5ORKi/XXruX79Q+h4GBU2VH4IEsCz02CmVzyqq/QoiFwxFOsburim8UNZzxnGPdxSyNTs42EQvb5JMbkzrvMHubHLivXotjLIru7MGemMA9YfO7rtX4jARL3L1c7RmlwJCBzPPBVAYmBh4jiZ6+FJptoe2Z556Y5ez3Eqo0KXNP4FbZN+Enq0KPEGLhcx3pofKndfyk4rYznlPXlcLoOPWxu1i43l+4k+MfKaTltlIG95Sx+EEH7NpP3u5+xu4r5/7g7QxvSPLd63/Eu3zZP2A206iaSrpuLWei0cZdO8HmwOF0lzQnJPQIIeZVqqcXz6N9p3SDvJ2l5TlPJtngdvLDuhegDt7nfTeRp6twAtaRNgJH28kzTbRxJa1Xl4GvK93lirexivyMrUyyeXULS/z9LHcNAP5zvi7TSOgRc0Y5XRgNNSSrCpiodFNcMoxTycyNXFHhGGO8EQpuXI9zIoHR3os1MDD5Qwk0WemtFX5dRoqQ00A5HGhbT3anWBaeYc0DHVfRlzzIWl8HN3oHpKtrHhSaESJVGnvDSsxQHNXVjzU0POMcbRoYvhQNviFqXMN4VHbeoxJ6xJwxCvI4fkcFee/qpSYwxofKtuPL8CXMxfnb5O3kI+9+ka0bGzjUXknjg3U4nxlId1liHrhMC8trYBQWoOMJ7HAEbIvCnX2E4qU8VnA991+X5Ec3fp/rZHzznFvh6mPdtYfZWVuH0R2k/nEf5gszQ4/lMaksHeGugh3kGUlKsnRpEQk9Ys4oj4eJpSkeX/HjaTtqy3TVXFHnCPD3pfugdB9fK17JU89dS0G6ixLzwkBjORXK7QatUVGFtie7urxH2vA5XcSDV9B5TTF4htJdbtYrNw3+pOJlNhZW89uyNUR2VpCn1IwnrrbToMI/zjq3m2yZnn46EnrEnJvNmHM0GeJ3oVW0x0ousAabq/KOcouvVx6np0G1a4SR5QrPbVfiGkvgaOk+2dU1xSwtJbWkimS+C/dgFONwB9b4eJoqFpeiyT/IG8ub0UYd/t4Ert3tM7tT3j51SMypYcvi0ZF1/KG7kdHufBpHcneGs4QekVEeHr+MH/7qZor3Xdibpu2AX918GQ03fIcN2fslZsG62XeY/vc+z+4bqtne0sCiH9diPj8z9MTX1NF2j2Z9Uxtv7FzM0vuqYLeEnkz0kcJtmHfatEWKeXlHM8uGy+FtY0jE/NmfLOGJly+j9hmLkvEkriM9pHJ0XJ2EHpFRDoYrKN+ewvPotgt6neHxMF6/noFr84DY3BQnzqjRGeDvSg5CyUG+4r2C7aVXEHjbOdFSJx9cvZWvV7zBjTE/qfygbFiQoVa6vJNdm8CNoSCpPGnLdBpI5ZPXauB5agc6lcrpnQok9Agh5tUizwBPrFaYiQ0zjk/UGDzTuYx74nkcO1TBsolxbMBRW0N0WTmJAgfjS1NUOEbTUneu6bfC/GpiKdvHGyl3j/P+wp1scF/4YnXNhX28fFU1hSUb8HZHMQ8ewwqF8ffa/OvBm3mhopPNBS28z99O0PTNwW+Sm0asCL8KNfHK2GJ29NYS6LMnZ9LlOAk9Qoh59Z7APpIfMDl0S/mM41vaF+N4qohjHfksGYyijnWDUoxtqGb84+NcU93GpwLtrHWNk43rhyw0r8eL+Jdn3kPNc5qddQ6Of6SQH9a9cGJa+vn6fOkLlN8zTke0iBdeXcWyeytg/2GCr3bhHSjjYMEqnrl1Bcvf+R02yjyHWbM/6eEftryXmqcUJWMpPC3HZWsmJPSIeSC3mZiu0Rngq0WtUNQ64/j7o/mMH3HjfHoHMPXvxjCJlBr82dIX+XzhWwvaSeCZD72pAvJbTHyPbsNz5UpabiuFugu/zhqXhzWl+0lqi02DVdj+yUF1qWOdOI514vJ48C1bz5DtR7qeZ8+AlU+g1YH/8Z3oeDynu7Smy9rQY5aWEl9TR7TUydBqxZ3+nnSXlHN0LEbgiIOPH/ootYER7irZwW2+iUvaZXlFoJtXrlpFcWAj3v4k7j0dp8wCEpljXyLK/cObaJkoY/fuBpYOh2W/rQXIMRZlcE8Z7/O+m6bAIB8rfuWiurpE+pnBIMk1DYQr3bjGLPx7e0h1Hsc9FOP13Yt4v22yJK+fTxa9wkpX9s10zdrQk1pSRds9mg+u3soiTz+3Bw7BKUMnxVyyxyao/V0/8ddLOFxZyT/8cTnXr/0xBerib6S78t7E98EE7e8u4eE317P0u5UgoSdjPTR2BU/84mrK3kiwdDiMcaRTngwuQLqzh8UPOog8XcWWlY2kPmayofq1dJclLoKuLefo3S4+sHEbTx1rhh9U4u48jnHkOEt/WM14US2PX76I4EcirCw5lO5yZ11WhR5LG6ipaXjJfBfrm9r4esUbUz+VwDPfdDKBdegIjkNQtLiRgzcXkbzE9TkanQG+WNgJdNIyUcZ4fi3n9X3TmByHYMpzhAWlI1pE8b4Uzqd3oDlDV6g63UExn+yJCdi1HydQaq3n6HgJVpV9weN7ZBxt+ll+N+VNg/xLxQ4CjjhPFl2HG7BGRmDHCE6g2LeBY7GidJc6JzI+9LQlQ9w7vImtQw20tpfR1CO79+aKywo7efC6JgoqrybQk8TzehvW4LTVXZXC0VhPeHkp8QKTyJIERWYIzi8miTRyNNYzvq6CWKHByLoUDa7BdJckpjiHIrS8Xss77Q/MON5c2MfnS19gjWvmvhIvxuD7fdfRMVFE6M1iysf65WneHBm0wvxwbA3P9DXTNlBEyTEbLAuzsIDEZYsYr3MTqlW8o+RguktNm4wPPa/E6vnlY5upfSbOsokoxrFeuaFyxD2F26i/a5DjiSJ+sP0alg1VwIzQYzB2WTl9H4yzqKKbO8sO0uRIIKFn4RtfV0H8M8N8oHY3K7xdXOUeAWQ684LQ1sWSH0Hqt0VMX3znpSurKf/4OGtK9884/ft917Hv/pUU7Y+xaGgIfVzGV86V1pSLb790E42/tmkcT+Do7CCVSkFlGa13Ovn4dS9S7RrhBl8LpsrNCQEZH3oGUnnkHQNzyxtorSXwLFDK1mgNNmDpC38sfjqNzgCNzn6gn2frlmF5Zy6ApgxFrNDgyvpjvLdkF02ufnxKAk+6WNrGflv3YtJ+26B2pUAZxAoN7qrbxX8ubpn6gQSetFCAaYJlndinyZ6YgL0HT1lsMBjcQEe0iKSe+S7cMVFE0b4oxsu7Tn1/Vmry+kjX86V4694atgJ4uxy4X96FHYmcmLFle1wEasf5m5JduJWTGTMgFWCYk1uDTFulOWWbJLWFgZqV9+uFIuNDj8gQ8QTeVjefbb2TOv8w7wu+zg2e5JzfTNoBlZ4xap1DFBtxnJcwiFpcvH2JKN/oeye7BqpnHB85VMTi3hCayS6t0SsriRYrRtalaPZ0nf5iYl40ufoZvyKGSl2OZ9QmuHMQ69CRM57v7Y7ywqur2DRYdeKYrSH0ZjGLhgZnBh6lMFcsZfiyILGgQWpdiApznGze6HKujFgR/n3kMn7XtYr+oXzKj9ho6/y+/q/0HufHmyyS/g14B20Kt/eQau+Y0ZarSnr4i4rfn9Jtmakk9Ih5YY9PULE1wbGRRRwsa2LiXR6uqX0Oc44Xp7fciiXePta6EjiV+5Kmy4uL91x4Oa88spaaZydQ077Ql4aG0Md70UBoZTnJTwzx2cZXaHANSpdWmq13xbh3849ouaKCX/asZ9SuJu/w0RlPA6YzDx5j2b0VJ9bheUv5WP8pXVrKNBnYUETtp45wa+k+lru7WeqUEesXo9tS3PvqdSz6ucXSkRhGbzepROK8XvsObzf//s4f0X5dCfce3Uw4XI67veNEW1oBNztuWM1T93SzpujoHP8m80NCj5h12gDlMFEOx+Q3Dq0hmcTdHyHfbWAkTbrDBZf891jaJoWFpTWWbZz2H7M2IM+METCy41tKpklqCxub44kgeR02bNszoxPjxPdRpUjkGVxfeWTaIoQSeNIpYHi4yWtxk7eLpHbww8Ia8l0usKyT9/U01vg47D91g9i3P+FRpolyuYgHFXeX7+DDeSNTP3TN2e+SzWLaxDXowLXzENb45NYtJxgmylDgMDDVqU9/Skw/t/ri4OviYFUr2wNXTM7kmtaWeU0b6U/kz8vvMh8k9IhZVWjYVK/oo/2TTbhHGil5M4Kx4wDasjFHQ3gdBpbLz1jcg6X1JU1HfiHm5OvH7qBzpBB7dwGNAzIrZCE5kIjwL723sL27juixPJo64qecY/h8GMVFaK+baIlBgSOahkrFuSxzdzO6KY7lvRzPoKZ4az/W4Qv/5m+uWMrAhiLiQUVsQ4gG5yAgT1/ngllYQOTqpYwucRKp0NxavQOH/LeW0CNmV4np5d+XPci+xgp2hht5/MFN1Ox1Y09MYPX0oQaGCOgGjoUvfWzNzwc3MvjTOup2jGKMd2H39M3CbyBmy9ZYA1sfXU39o6OoyAD0DpwSSlVegPjiMuJBJ5EqTYljIi21irPb5JnggWu+T/tVpdzXuZmJcCW+Cw09SjF8WZDaTx3h7vIdNDgHWeOykNAzR0qK6Hi3wX+56REqHKOsc/djKlmvLrtCj1IohxNlGthOhcO4tIXwxIVzKpOVLi8rXWMscm7j18VXY/h96EQCnUyh43HMSIJEPMCwnSA57WPQROFWzrOOu7G0TVynSGLREQpS0J7AfvMA0tILz6jlw9ensd88MLMrZNp9qnxeEgUOYgUGllfjMWSdrYUoYHjY7IHNniGOlB7h0YJqAp4L7DI2TWJBg1tL9011aZlI4Ll4cZ0kplOM2gGm91wph2NyaEHAi6sswsfy2vEZLmSB3klZFXoc1VUMX1fLeL1BpD7FR4LZt4R2Jik1E5St66P1c4twj0DZjjDGtn2o0QnyXyvlOvs/YZgnPwyLCkN8dfEz0/r4T/VSzMHft36Azv4iXAe8NPQOSZdWhnHU1zJ4bTWhGkXKp0mUWhj+GIurBlju6kE+CBe2K/yt/PgdVxGpWH/Br02tC7Hc3T0HVeWWMTvKPw9cxa9a1hIf8VC1V6MTCcz8fCLXLmNopZNYseYdDbtk8sbbZFXoSdaVMHZniK+ve5hiI8wyZxwZDJk+laaX+5ofoH1xIS9MNPOYeQ1Vu1xYA4NUPaypfDowuU7HlPHVxdz/Z5v4o2W/O+NU9kdG1zPxiyqWvjSACvdjT1+MUGSEeGMJibtG+JeVj+BRSfwqgVNZFBoJKk0XEnoWtus9ozx0zXfpvfrCBreaaCrM8alZWjJo+VIMWhY/23EVS38QxxwchJExrHgcR0UZHe8y+W+3PESdc5jFznGc0qU1Q0aGHkvbRHWCiLYYSflPPNqz3Sa1wSHe7Ysx+cYpgSednMpkqdPPUmcSU+3jkcA1oBQ6lSLV2we9M8fg+Au89E3k0W9FMNXpRzh3hIMEulNnXS9EORwolwvldmO7wJTOr7RwKgvLrTDz8tDTurcihU6aS/qm7lM42c0hH4SZIGB4uNwNEDvXqach6/BcioidIKKTdKbycQw5MI+0zth6RztMKEhyu/8YJaafC+nScioLywVGXh4kk9iJJNjZ9xw9I0PP4WSMvz72Ad5srcHZ66K+JXbGtSNE5nAMjGO/UM2mga+ccVaXt81F/fHhs8eYtcvovqaAeLHGu3aY5a5e5M12/q32dBK/YZzW0lUz1uaJ1SX4ZNGB9BUmRAYK2TH++8BGfnlgHfaIm4o3NDp6McHz9DbntfDr29Yxtng1/i5N2fO9WEfaZu36C0VGhp49iUoOP72I5od6UdE49siofJfPAtbxHqofjFLjO8vMrlgca/jMY35QiuHV+Wz44zf5ZOlLlJpR6h3yBCEdrnAl+OUV32Ng/cwnroVGjHqHBmR1bCHO16id4qFdl7P03xM4+rrRY+NYkcisXf+d3kF+s/nbDF3t5e9aPkC0oxiXhJ6FIWK7cY2D3daBTqXO/QKRdpNdHRqjsADlmhZCLAs7GkMnE+hkAquv/8IuPL0bTGtQBikPXJ7XzmaPwYw9ZsS88hkuml0umk/5iSwUKcRFSRqYEzH02Dg6Fj/Rw2F4PCiPGzvPh+GyMC9iAbSA4WHl1Ftzc7CXPYXleINBkl6F28iez9mMDD0i89SaIRqu7uSQpw6VOnlDukcVVVvCqNd2X3AXpXK6MPIDKKcTHYthh8JoW7o5hRDZJ88wuWHNQbZ8YQWukVLKt6fwPbcXTJOJd61g4DKDRJHFHUvfwK0u7aP9psL9PPe+ZQxcthzn4gmuCRyepd8i/ST0iHlR5/Bx/5KfM7DIga1Php6fjV7FM8NXU7LVAH1hg+aUxw3BAmyPGyMUQcXi6ISs8yKEyD4Fhpd/rn6Sgfc+zZ54FV9zfJhFr3pQHg891yj+7Y4f0uAYody08RmX9oT7Pb4BLt/8LSaudlJgJCk3XWTLRAMJPWJemMqg0hGg8m3H9/k6eSy4CUdl+QV3VSqXC+2Uf8JCiNxQZvopM8GpjpMstKGsGNvlQAeTrHcNUumYnenpPsPFIuOtkJNdk0DkE0Ok1Wp3FyU3dXOguhZlX1g/tL/ToPKlCRwd/eh4QsZ3CSFyQpEB11++nxf+vBmU5o5Vu8gz5OP8fMh/JZFWzU4nDzX/lIllGusCB9/9yYGPk9oVhP5B0PbkmCBDFrYTQmS3EtPP/6x+iuHKJzCAItMkYMhsyPOR8aFHORwYBfkor5dwsZMK1+ytWyDmnlOZJx7ZXqgSb5hxsygrF9ASQoizCZo+gvId74JlfOgxa6rofk8NoytT+MpCfKHsjXSXJIQQQogFKONDT6qsAOudI2xZfy8+pcgzXIAz3WUJIYQQYoHJmNCT1BZ9VpRBy8m+SDVmXKNtjTYN/O4E1abvjJtUCiGEEEJkTOg5noryt8fv4LWjjRh9buqPJiYHrwohhBBCnIeMCT19lpdXDy6i4lkH7rEUnrYhUrLJqBBCCCHOU8aEHqeycPqSxIIuLJfCjBbj9rgYq/US9AykuzwhhBBCLHAZE3pqHUn+dPXLPF+xjGjKyUDYRyzqpbhwgE9W/0HG8wghhBDirDIm9JSZfr4SPMwXCw8AYDM5nsfAmNpcTUKPEEIIIc4sY0IPTC5k51SyGpOYVOoJcbzOScllKzHHwtjdvdjxOK4JzeMDq/EZcRa5+lnrShAwPOkuVwghRJplVOgRYrqPlrxK38fzaH9PkMSbVTQ96IBDRyh6Y5i+HzTyP4JNRK6K8MDGH7BRMo8QQuQ8CT0iY93gtdm8+DFsbG723k3y9wUYh8A60ELhwaMYLidd5nraLy9ho2ck3eUKIYRIMwk9IqNNdneamMa0NZu0Bm2hUwplgyXjvYQQQiCjf4UQQgiRIyT0CCGEECInSPeWyAoNecO8ubqKIuf6E8dsUxGqsyk2Q2msTAghxEIhoUdkhc+UbeG+T2iOhwtPHHMZNp8uPspa1xAQSFttQgghFgYJPSIrbPYYbK57+Qw/lcAjhBBCxvQIIYQQIkdI6BFCCCFETpDQI4QQQoicIKFHCCGEEDlBaa3P/2SlBoBjc1eOOI16rXXpbF9U2jJtpD2zh7Rldpn19pS2TJsztuUFhR4hhBBCiEwl3VtCCCGEyAkSeoQQQgiRE7I69Cillimldk3737hS6ivprktcHKVUuVLqQaVUq1Jqp1LqVaXUB9Jdl7g4Sqk/V0rtVUrtk/sysyml/naqHXdPvddele6axIVTStUqpdqUUkVTfw5O/bkhzaXNmqxekVlrfQhYB6CUMoEu4NfprElcHKWUAh4BfqS1/uOpY/XAe9NZl7g4SqlVwJ8CG4AE8KRS6lGt9ZH0ViYulFLqauA9wHqtdVwpVQK40lyWuAha606l1L8D/wR8dur/v6e1bk9rYbMoq5/0vM1NwFGttYykz0zvABJa6++8dUBrfUxr/c001iQuXjOwVWsd0VqngC3AnWmuSVycSmBQax0H0FoPaq2701yTuHj/C9g49fT1GuBf01vO7Mql0PNh4GfpLkJctJXA6+kuQsyavcC1SqlipZQPuB2oTXNN4uI8DdQqpQ4rpb6tlLo+3QWJi6e1TgJ/xWT4+crUn7NGToQepZSLyW6Q/0h3LWJ2KKX+j1LqTaXU9nTXIi6c1voA8M9MfmA+CewCrHTWJC6O1joEXM5kd8gA8Aul1CfTWpS4VLcBPcCqdBcy23Ii9DDZgK9rrfvSXYi4aPuA9W/9QWv9RSa7LGd9cTgxP7TWP9BaX661vg4YAQ6nuyZxcbTWltb6Ba3114AvAR9Md03i4iil1gE3AxuBv1BKVaa3otmVK6HnI0jXVqZ7DvAopb4w7ZgvXcWIS6eUKpv6/zomx/M8mN6KxMWYmiW7ZNqhdcgqxBlpasLIvzPZrdUBfJ0sG9OT9SsyK6X8QAfQpLUeS3c94uJNfeP4X8BVTD5GDwPf0Vr/Iq2FiYuilHoJKAaSwFe11s+muSRxEZRSlwPfBAqBFHAE+KzWejCddYkLp5T6LHCT1vpDU382ge3AX2itt6S1uFmS9aFHCCGEEAJyp3tLCCGEEDlOQo8QQgghcoKEHiGEEELkBAk9QgghhMgJEnqEEEIIkRMk9AghhBAiJ1zQLuslRaZuqHXOVS3iNNo7kwwOW2q2ryttmR47d8cHtdazvoq0tOf8k3szu8zFvSltmR5na8sLCj0NtU62PSV7As6nDbd0zsl1pS3Tw6w8Micr1Up7zj+5N7PLXNyb0pbpcba2lO4tIYQQQuQECT1CCCGEyAkSeoQQQgiREyT0CCGEECInSOgRQgghRE6Q0COEEEKInCChRwghhBA5QUKPEEIIIXKChB4hhBBC5AQJPUIIIYTICRJ6hBBCCJETJPQIIYQQIidI6BFCCCFETpDQI4QQQoic4Eh3ASL3WNomqhPEtIV9CddxK4OAcmMqye5CCCHOTUKPmHcdqQh/ffy9bD3cCLa66OvU1w7y/y7+FZs9s1icEEKIrCWhR8y7Q8li3nh+Gc0/7kclUxd9ne7bqnnmc6vY7Nk/i9UJIYTIVhJ6xLyLaSeucYV97Dg6Hr/o63iHKwlZ7lmsTAghRDaTwRBCCCGEyAkSeoQQQgiRExZ095albYbsKEOWwqVsSkyTAsM7J39XyI4xYKWI6Zk5sNjUFBtemSE0iwxstAJlmmjDBNua+oGJmR8A91m6rBJJ7IkJdCqFkdJ0RoMcSETIM2xKTTdu5ZyfX0IIIUTGWdChpyMV4Yutd3Nwdx3an+KjV2zla6W7cCpzVv8eS9t8b3QF39p5A8ag68RxbcLi1cf59uKfs8gZmNW/M5e5lIXlBaO0GBWJoScmsGMxHNWV9N5ey+hyDWeY1OXvMKh5YgDrQAt5LePs+/Vy7qhcSv6SEb656mcyk0sIIcQZLejQ02kFOPpKPc3fO06yuohfB9fwNyU7Zz302Gh+dXwdDQ8oPG8ePXFcedwc/UwdR+uDLHImZ/XvzGVOlSLl1Vgl+RghNyqVglgMqyJI/JZxHl1/L6bSp33tV1r/iMiBatwHQO8/Su0xLzhd9PzRErY2LWazp3WefxshhBCZYkGHnqQ2MWMKe3AYh9tFZCTItriHUjNMlakJmr5Z+7viKQeBsQTWwMCJY4bHgxmrI4kJSOiZLU5lYQVsYmU+HHluXA4TR8BPqNxLeX4vS52eM3YnlnsnaDUnHwPpZAJrNDF5zchiknp2w7AQ4vyF7BjdlsWEfeYu5iIjQZVDuqFF+izo0DPD8CiVz5Tyud7PkSix+OTVL/N3JXtlrE0GanCEuOmKvbxYsggrZWLF8iBZgLsoylerd0ibCpGBHg1X8nc73o9x7Ax9zAr8K0b47poH2CArTYg0yZjQYw0NU/DILgofd5Fc08Rvq1fxNyV7kO/2mafa9PG/qp8lUvU0AJae7MoylaLAcAHyLVCITPP82HLKf+Om4Ml9pz/B6aDr48vZtbSeDe7u+S1OiCkZE3rQGjsWg1gMx2iMkaE8no36KDJDNDkSlJj+dFcozpOpDALKw6UODTd8PoyiINrjIlak8BmJWalPzL6QHaM1BQOWn0IjSpMjNavd0yI9QnaMI0nFkO1j73Al7tEU1vj46U82TMyYxtYXv/WMuHhxnSSpLWLa4njKwZB98fefiabUDNPoMPEZrnO/YAHJnNAzjeobovLxQr7a8mlilRZ/ct0W/q7kYLrLEvPMXrWI1jsCJOviLK09xrW+w4A8N1+IXo4V8KVtH8F1wEe0LslfX/MYny2Qb/uZ7qlIGX/12l14Wjz4ujWFR3q4+I1lxFyJ2AkOJzVdViE7wk38aNdGPEcufqqrVqDWjHPv+h9n3IzZjAw91uAgeb8ZJ980SW5YzpOLV/DXxftlLEiOCTX4ueGWXfxT5bM4lYFXZdY3jlyyI9JI8BkvJb94k/DNK3lhxTIJPVnglYnFlD3lIvjIm2BZpC5hWxkxd+I6xdFkBXujNTzds5zyp1wU/m4P2PZFXU+ZJt2fXMW+lTVs9mTWfZyRoQet0fE4GnCMxekZLOCRcCGljnGWO8OUSVdXVhmzoxxIuOi1Ctg3VEEgNrmYoTYg3xGTbpIFKmTHOJQ06EoV8vLgItxjNnY4jGs0xb6BCh4pC1BqjrPCKW2YqSwMzCTY4fAZzzHy8jBKi9FeN7ESRZ4Zm8cKBUBY2+yN1vDKYBM9/YXUjVnYoRDo0y8Nck5K4R7VvDC8jArHKHWOEZY6VUZ0dWVm6JnG7Bmk8pEG/tubHyNca/GFdzzDXxUdPfcLRcZ4IlzFX7/8QQKHXPh6NN6j3fIIPQNsjfv53NaP43nTh3dAU7p/EAtwtw3g/Y9q/p9X7iG0KMnfX/9r7skfTHe5Yo6k1i2m7f1ujOoIV9QeYL27E5CQO5+OJPO5f8cmil9xUjVm4zvUf2ICyUXRmqJdIxz60XL+pqgZ+4px7r/8hxkxKy/jQ0+qt4/AI0METJPE9at5fvUyCT1ZZnuokYpnHRT8cifYmlRK1kzKBPtjNeS/5KX8R7vQyRTWVLulOo5T+FAfhYYidMc6Xlu/WEJPFhtv9PDhm17mv5Rsx6lM3EoCz3zrSBaRv8dF6c92oRMJLMu65Gvaew5ResCJ8rjp+uxqWleXscE9MgvVzq2MDz0AOpWCVAojaWPZMq4n21gYGCnQMl5gwQvZMd5MuGhJVPBE/0rcoxo7Gp35GF1rdHJypp2R1MQtB0ltYaBkXF4GGLEivJ7IozNZzCu9jfhCkx+gRl4eqrIM2z9tZKsBoWpFnXuIgJFhI16ziMdIksgH1VCDEYlhD41gT0zMOMfw+1FV5diBM+9vaY6Fsbt7J2dSv3Ufaxv3iOaRwctI6N0sdfWxxmUt2K6urAg9QoiF4c2Ei0+8+mn827x4hjXFuwbP+hhdWTCRcjNoRfEog4DhnvVtZsTsei5awV9u+RAFu534Bmz8ByZnbOnlDbTcHcC3ZHTG+Rsq9nCt9wjSpZU+S539FG3q5XBBGe5Bg5pnC2DbnpknLannyIcK8DSPnvE6iTeraHrQAYeOnDimLYuSHSO02sv4H8HlJDeP8/Mrvs+ahZl5JPQIIWZPS6IC/zYvVd99HTuRxNJnnx2itCZmOZnQCgsbHzbIkqML2t5oDaUvOyj66Ta0rUnZk096IlVerr12L9+pfW7G+QYGTunSSqsqR4qP121lZ7CBV443EDkYwKfUjCewsQo/Kza18ovFj57xOjd77yb5+wKMQ9MOao295xDBvQaG30eHdzXd6wpY41qYT+YzMvQYHg+qtgqrOIARTaK6+rEGhzDDSY50l/KNigaqnCNs8nRR45Dd0YWYT8qe6nK2T44bMPx+VF0VqUIf5kQcdbwHa3QMR9hif1cF3w5cz3JvD+8JHKLGIStyL2Q26mQbT6NsiFpOJuwEHmXiVS7prlwgnChKHRPUeEbI91Vgn+Fz0WFYZ90XrS5vhIPLKihOrcEcDqM7uk50daEtSCZBg8XCXYAyM0NPVQVtH64gsHGAgZ4CGh5uwvXUMOaxPmp/Xs9PXryNscXwmduf4T8Xt6S7XCFEYy1HPhakaM0A/a3FLPq5B+PlXbhbeql+sIZXSq7kd+s0nluTfDK/P93ViotgJGzax4rYEq2kwjHKKlecAnXm8SFi/gQMN1d6ulnkHKAvkc8OX9lFXedT5S/x75900DEeZGJ7OY0/1dDSOsvVzq2MjOFWgR9z/SgvrP0pX974HBO1k9nN6uvH/fgOir//GtVbUuwcq0tzpUIIgGSpj/orj/PS2l9w+1W7iFROzm1NdXXjeWw7wR+9RtlWOBCtSnOl4mIZSc1Y2MuhWCWdyWIi9qXPEBKzw6lM6hwB1rndLPX1Yl/keJubvBa/aHqal9Y9iO/yQezCzFsTLyOf9AAopTGVwlBvGzMw1UfpGkuys72Or+WtpNE9wDt9rdLVlSWUwwHKwDbBUJew1oSYU4bHg1FVgVXgZ6zBTbNvDKcycRgWqGmPv6fuWaWRfZkyQKVzlPEmRd6N63FMJDDaurGGhnGEEiQ6/TzsXkfQF+Xlgj6KXaFzXs9Ac4W/les9ozLDKwOYygAN9QUjdKxdTKF/Pa7+ELq1I92lnZeMDT3n4jjSTd0DdTxZdh0jK2Hijsf5cvBYussSl0g5HBg+H7icpDwKp5JvkwuVUV5K921VjDVbuMtDvDO4P90liVnwDv9h2t7/Cm9eX82hlioWP1iPsWUYs62Xpl/XkCgsIOEsZKenCts8d4i1HfDjd1zFQ9d8l8szYHE7MRl8vlj9HD/5TIyOUJC2l2to+lEcu6cv3aWdU9aGHquvH9dT/bgA9/s3sPcdVSChJ/MpA9xulNuFdoD59id9YsHQAR/ji2w2rT9Eg2+INe4uZEPYzLfU6eefy3dB+S7+U/6V7HxmPQHAGhjAGBjgQp/VGB4PkYr19F6dD8gWFZniJq/FTXUvEbETXDn2Kex8H/Sku6pzy/jQU+UcYWwxBG67AtdYEseRbqw+GQiZ6Y6nQjweXsr+SBWPH15JzfDkar5mSRGRNTXEihyMLYY611CaKxVnJb1VWa3BM8Tvm01ct16JazSO2XIca2j4gq6htcbXp/nXtlt4obiDzXktvNM7KF1dGcJUmXWTZ3zo2eTp4jO3P8POzXXsbK+j7oE6XE9J6Ml0z0Ua+JdH30f5dk3NcBLvvi5SQGJxJcc+orl++T7uDHTzDt8RQMZqCZEO78nbw9gHvey7uZKdBxpZcn8t6pULDD2JBKUv9xMeKOelgkp+fds6frP526xcoIvbicyW8aGnxhGYnJZe3MLX8lbyZNl1yL2S+Y7GyyneDYGHXgM4scFovNjJzSv28N2aV6eOSOARIl2WOv38fek+KN3Hp5zX0lrUfMHdW2iNdfgonsNH8eXlMbZ4NUNXy1R3MTcyPvRM1+geYGTl5Bge12gK98EuUr19uIeS/H7/Cj5lO1gR6OauvDdpdMqHZTbZFY/z8NgVdMUKuSK/nTsDB6iU2XpppSIxAh0Gf8hfzFZvA1vLG6jxj/Jiy2IaBmXT2GyzxNfP1stWU+LYcO6TNfi6IqgDbdjh8NwXJ2bdtniSX45eSWckSLIlHxXuJRPm0mZV6Hmnr5WJOx5n7zuqeLZlOfX31eDo7cN58DiLvl9Na2Ezr1y1Ct8HE3yxsDPd5YpZ9O8DN/Dazy6joDXFi9evpPqOEd7vOPd0WTF37L4Bqp90Ub7Nj3YoUt4KWh2VNI0lcR7uRObdZZe7Cl7HvNum9Y6Sc56bsB28vGUVS+4tgyNt81CdmE2WtvlG97vY//NmCluTNPWOo3syY1hJVoWeGkdgclp68BhfMZNsL77ilFkFxYGNtL+7BJDQk00Oj5VRtj2C8coeCio30JssACT0pJMdicCBFhST45mndztL4Mk+S53+qRXwz70KfsROcOXiOrRfBitnIhtNy0gpFa9NoLfvQQOayZl4C11WhZ5L4XUmiRcX4KsoR8diWOPygbkQeQYSPL1rFe+P5p84ZqPo2FPJ0tFRZAK7EOmxLxHl/uFNtEzM3OLgssJO7incRqMzwIsx+Ong1XSGg6T252NM9Mg9m0Fei1n8ZGgT7eEixvYWUzo+kHFfYCT0AAaKEm+IjpoKzFg1rv4QRmxh7hCb6xwHO1j6/WrGC2tnHF8yMAHHutJUlRDiobEreOIXV1P2RmLG8Qeva6Lxrn7qHP38W9e7aPvJEgqOJmjqH8HuzYwuETHp3v7ref0nawgejLNoYBR9PAMW5nkbCT1TfI4kSb8iUejAEXajTBOsTMuw2cPSBkqfOizOGhqGoWHevg/wW49XMcx5qE4I8XYd0SKK96VwPr1jxvGCyqs5nijGpo+20SJKXw+ht++RJzwZxtI2rRPFlOyKYLy86/TtZ0xu52ku4CHNWRt6Vvi6efQqSPquxjeQwv96J6meXny9CR7efgWvL6plfVEnny9+iQaHL93lCqAtGeLe4U1sHWqgtb2Mph6Z4ZNpGpyDjK1N4LznSjyjNnm7+7FkoGpOC/QkuXfHtTxd00zozWLKx/pndokohbliKaOrgsQLFHZziFIjAsj78kLwWszi3v7raZ0o5vjrVSwZGT61/ZYuYmxNMfECg/iqKKXmBCzQxWOyNvTc4T+M/z3/QdvNpTxwYAO14SrMnl5ce9pZNlaJlV/MozfWsPxDPXwyvzvd5QrglVg9v3xsM7XPxFk2EcU41ptx/cW5bq0rxP++/kH2bqjlse6VjP2gjMDR9hObiorc43m9jWXDlVjeIIuGhk7pElEuF/2biqj5WCtXF7Vyha+VekfWfjRlnJ8MbWLnT9dQuivK4pERdPvxGT9XpsnQVaUUfqKT95YdYq23g2XOhfscL2v/ZVU6Anw0bwjyhjhaX0proBmTk90jBpDXeDWDqbx0l5qTLG1jv+0RaF+ygLxjYG55A631+QUepSb343rrj6aJNjJrWfRsEjR9vNcf4b3+Q+SZMe4vup0800Tb09pa22cMQbZWJPXMljdQkzs7i4xkDQ7B4BAGp5m1pxTK4SBaovh89Qvc6ntrLOXCfEqQS956j24PF1F0II7x0hundmkpBaZJrFjxqZo/8OG8kakfLNxZXFkbesTC1ZMK8Y2ha3i+ewlanwwog10FNHScf5eWo7Ge0SsriRadvIY2FKNXxlni7p3VmsWFW+LuZXhDEm1cyfR86x2yKdzeQ6q9Y8b5/q44v37lSl5sXDzj+NUVbfx56fMskgVFs8dUl9bwZUFiQYPUuhAV5jiyIe3CsC2e5Bvd76JlpJSxvcUsGjh1ZqzZvISRy4qJBQ1Cl0epdQ4BC//LiYQeMe/2Jwv45XMbafpVDCNx8rtfSXQC1dOPdZ5dIaGV5SQ/McRnG1+ZcXyJu5cr3CFAlrJPp6s9o3z3+h/RevXMKcz3Ht1MOFyO+22hx7mnleXDFdi+kx982lQ8ffsVbPrQERY5RxDZQZkmAxuKqP3UEW4t3cdydzdLnfKEdqH45eiV7P95MxWvTVA6PnDqLC2lGLmsmILPdPKJil0sd/ew1pVgIT/heUtOhB4DjXaAcjgmH7Pbkx+0yoaY7SSqE6S0AQbYpkKbCmUYMntrliW1hY3NQCqIt8fA2LYfnTw5vfWsUccwUdO7rZRBvMDgxqoWPl94uqnqEnjSrcDw8i5fEnwz2+dgVSvbA1ec8p3eGh2D0bGZBw0T37oNjFo+QELPQjb9ffZclMtFPKi4u3zHtC4R6dJKt5RtEtdJOiNBCluT6O17Thm0rExzsksraPCJil3T3n8XfuCBHAk9VxW08vw7V5LXsIFAl0Xhq8dJHe8iryPOj7Zcy0PVlxGPO7GWJ5lYZBDc7afimB97ZDTdpWeNfivM1weu4amO5UwMBKg9kpoc23EeHLU1jF5dQ6hq5qPT8RVJLvMdm4tyhRAXaPr77LloA2IbQjQ4BwFZZmIh8PSEOPByE+sHP0myJX9ya4m3nWMuXcTQVaXEihWhy6Msd8s6PQvSnYEWFt32fXpThfzD7tvx9ZRiHO/C9foRlneXon1ujr/Lx413b+eWwj18yfcxyl/OBwk9s6Y16eFXL17F4p9HqBkfg/4hrFTq3C8E4ovKGPvwBH+98skZx6sdI6x1hZCprUKk3/T32XMxsWlwDrLGZSGhZ4FoOcbiH0Sx87yTm4f29J8SesbWFBP8ZAefrH6FWudQxnRpTZcToafE9HOT1wKG+I+yfsZ9tZMzCcbHYXwcDBPnxg2s8HXzTu8EvoIo2ik34myI6yQxnaLXKsXTb2DsPoIViZz7hUqhHE6UaRDLc7CqvH1yNt4pJPBkJaVQLhfK4cB2KgwlU94Xuunvs+fHRALP3Hrr/ddE4VZOnOrM/73tSAS7tf3UH7x1L5om8QKD95cdnOqSNMi0wAM5EnpEegxaYf6h/3oeb1lBcthDzQELnTy/pzuO6iqGr6tlvN4gUp/iI8FDc1ytWEjMFUvp21xErFiRWhdiuVvW0hLiQnSkQnyt+zZebl1EXiDKf1r6PB/Pu/BZrebSRfRfU0q0TBFfFWWtp+PcL1rAJPSIOdNtmfz2lctZen8IY2wAhkaxkolzvxBI1pUwdmeIr697mGIjzDJnHHmqkztGVwWp/9gRPlv1IhXm+NTMHhnoKsT5OpgI8ofnV7H4pyOEFxVw359t5qOrHr7g64Saiwj+8XH+of5pSs2JqYUHM+8Jz1tyLvQ4DAvLbeDOy4NkEjs+uRiWYcFgKo8+K04y4QAtG45eqqQ2cI4bGC2dk12Jb5nWdYHW6EQC/bbxPbbbpDY4xLt9MSYfgUvgySUpr+KqYPvUYnWydosQFyqi3bhHFLqlDb9jEW1jAXqsKCMpP+p0E5OndWO99WcMg1iByTVFx6buxcz/4pFzoeedxQf45/c14Fm/mvw2TclzHaS6uilsifPjx2/kB8XXUrjLiTE0IBvizRGzKMj4jUsYWmHiGofKP0zAjr2yVYEQQswBY2gc74t1XD/2FzgHHdS3xE55v3VUVzG2sYZQpYnthJQXtBOSSyNsDBxJU+WzL+dCz0fyj7Dxpv/DuHbzhTc+SuHhYjjehXP7IZYcKQSHiQ5HsUZGzmu9CXERioN03Wzz7zfdz6Mj63gttJ7inQZoWRdJCCFmm9XTS+Uv4lQ96oWUhT1y6grLqZpiut5lc+3qA5S7J7gy0EqFY4xiI0q9Q5HJXVrT5dyneoHhZd3U0/Ky/BC2I4gB2OEwdjg841zlcGAkoStZxPHUYfIMk4Byyz5A58lEYztBBQswzZOzBqxCH97iKNd5JujNa+Nl7+UoQ53vsj0iSymHA+X1olxOUh5wnvYZvBDiQulUCmtgAAbe9gPDxPB6UC4XkaAbb1GUK/OPUesaYpOnjzLTT7Yt9JpzoedC6FSK0t0J/vWR9/GPQYsrVx3lf9b9hhqH7AF0PsrNJM1XtbHP0YCROLmacqrQ4u5F2846fVLkHrViMd3vKCJSqfEuG+UKX2u6SxIiqznqa+i9uYrxJrAqEvzRor2s97ZRaMTxZen7s4Ses9CpFO6X97P4zQDkB3jjT5dyvMpLjfxXOy9lpo/vNj7MQL0De9rGom5lUWGCU8ngZHHSxJIClnzwMP+19lHyVIoqhxtwprssIbJWojqIfs8w/7H6x/iMFKWGwmc4MXBl7ZdS+fg+BzsSgUgEIxLFEakgIYtpnTdTGVQ6AlS+7XhSWwxaUQ4kIhyJlWMm9OSeaIaJmR8At5togYMix/nvuC4yh9tIkQgozPKZG5HGggYr83tY48qOsQNCpJNTpUh5wSwrRSdOv1RItMRFXUEP69xucmWWpIQeMe/aUjG+0PLHtO+pwj1oULM3AtrGUVNN7+21jC7XmBVR/qT8jXSXKubAjXkHeOK9zRxa0zTjeH7jCNcFDqapKiGyyxLnEHXXdXCgpBplnX4He7M8995nJfSIedeSLKb7xRqW/6AdHYliRyJorbEqgsRvGefR9ffiN2xKDBfZsC6EmOlGb4jH199L5LKZb8R5SlNkSpeWELNhkcPLA0t/wcRijcXpQ49H6Zx7n83p0JPnijNa4iKvphodjWKPjZ+ySJ6YfUntwBEDa3AIHT+5CKTtMMjzRljq9MgMuSzmVk5qHBJshJhLpjIoM/2UyYiMGXL6k+Wu8h0MfizC/r+tpufDyzErK9JdkhBCCCHmSE4/6flQXg+3X/VdJmzNe8o+h7W1EDqPp7ssIYQQQsyBjAk9yuHACAZReX7CVT7yPUMYl/igyq2cuE0nBYaFz50ElRuj14UQQohclDGhxywppucDTYxckSRQFObL9a9hnGFwlhBCCCHE22VM6NHBfMY2xdhy/TcpMEx8yoWZpYsnCSGEEGL2ZUzoQSlMh02R4SBgzO7iZQaKQm+UUH2Q/NAS1HgYq38QnTz9gk7i0vhUnHiRRi1vwkie3F9prNZL0PP2zWGEEEKI2ZE5oWcOmcrg0zUv840/uYnDowH820upedggJYOa58Ry1wjX37ibFxsXoe2T60MUFw7wyeo/yHR1IYQQc0JCz5Q/Cgzx3jUPEtJJrnN8HvvZAHSmu6rsVOcI8K2aF0hWPzvjuIGBWznI8ZUUhBBCzJEFHXr8KkGsxMZau5houYfC/OFLnrF1JqYyJscJaYVp2qBkkPRccisnbiUL1AkhhJg/Czr0NDljfPi6V3iysZmAe5wv1b809SRACCGEEOLCLOgEUWb6+e9lu/ha2U4MDAyUjPcQQgghxEVZ0KEHJrudzHke41GWH2J0RSX53jU4+sexOrvn9e8XQgghxOyTxyZv48Dkzxuepfhzxxj4mzgdd1VhlpWkuywhhBBCXCIJPW9jKoP3+0P8ZsljPH3ZfUTWRNF+b7rLEkIIIcQlWvDdW+liKgOPMqkoGWPkijLcI0XEqpLkqzgge3QJIYQQmUZCz1l4lYu/WvQ0P//iBkJJN18t3U+9wzr3C4UQQgix4EjoOYu3urre3/jctKO+tNUjhBBCiIsnY3qEEEIIkRMk9AghhBAiJ0joEUIIIUROkNAjhBBCiJygtNbnf7JSA8CxuStHnEa91rp0ti8qbZk20p7ZQ9oyu8x6e0pbps0Z2/KCQo8QQgghRKaS7i0hhBBC5AQJPUIIIYTICVkdepRSWin1P6b9+S+VUv8tjSWJS6CUWqaU2jXtf+NKqa+kuy4hhMgGSilr6r11n1LqTaXU/6WUyqqckO0rMseBO5VS/6i1Hkx3MeLSaK0PAesAlFIm0AX8Op01CSFEFolqrdcBKKXKgAeBfOBr6SxqNmVVgjuNFPA94C/SXYiYdTcBR7XWMjMiQyml/lYpdVgp9bJS6mdKqb9Md03i4iil/Eqpx6aeDuxVSn0o3TWJS6O17gc+C3xJKaXSXc9syfYnPQD/B9itlPqXdBciZtWHgZ+luwhxcZRSlzPZhuuYfB96HdiZzprEJbkV6NZavxtAKVWQ5nrELNBat049VS8D+tJdz2zI9ic9aK3HgR8D/yndtYjZoZRyAe8F/iPdtYiLdi3wa611ZOoe/W26CxKXZA9ws1Lqn5VS12qtx9JdkBCnk/WhZ8o3gE8D/jTXIWbHbcDrWuus+OYhRKbTWh8G1jMZfv5BKfVf01ySmAVKqSbAAvrTXctsyYnQo7UeBh5iMviIzPcRpGsr070IvF8p5VVK5QF3pLsgcfGUUlVARGv9E+DrTAYgkcGUUqXAd4Bv6SxaxTgXxvS85X8AX0p3EeLSKKX8wM3A59Jdi7h4WuvXlVK/AN5k8lvk9jSXJC7NauDrSikbSAJfSHM94uJ4lVK7ACeTE4EeAP5nWiuaZbINhRAi7abWzwpprf813bUIIbJXTnRvCSGEEELIkx4hhBBC5AR50iOEEEKInCChRwghhBA5QUKPEEIIIXKChB4hhBBC5IQLWqenpMjUDbXOuapFnEZ7Z5LBYWvWN3uTtkyPnbvjg1rr0tm+rrTn/JN7M7vMxb0pbZkeZ2vLCwo9DbVOtj1VOztVifOy4ZbOObmutGV6mJVH5mRXeGnP+Sf3ZnaZi3tT2jI9ztaW0r0lhBBCiJwgoUcIIYQQOUFCjxBCCCFygoQeIYQQQuQECT1CCCGEyAkSeoQQQgiREyT0CCGEECInSOgRQgghRE6Q0COEEEKInCChRwghhBA54YK2oRBCCCFE7rC0TVQnsNAzjjsxcSsHpsqsZycSeoQQQghxWk9GffzjkbvoHSo4eVBp1td18vc1v6PZ5UtfcRdBQo8QQgghTuvhwStJPljO0q0DJw8ain0fWMb2e96k2TWYvuIugoQeMW9CdoyYtgDIM1y4lfOCr5HUFhGdIK5t3MogoNwZ93hVCCHm2lvdUjFtYV/CdTpCQQJdCaxDR04eNEzcwyVEbPcl1znfJPSIeXE8FeJvum7n5cOLcXmTfHbFH/hysAWnMi/oOs9HPfzXwx+ir6eQ+tpB/t/Fv2KzZ46KFkKIDNWRivDXx9/L1sONYKuLvo6/xUVd9wDWLNaWThJ6xLw4lvLx6ksrWX7/EPHKPH705av4/BUHLzj0PDq6DvuhUlZs6ab7tmqe+dwqNnv2z1HVQgiRmQ4li3nj+WU0/7gflUxd/IWiMayR0VmrK90k9Ig5NWZHmbAtWhJNuIcVuv04bquKibF82lIWRUaIAsOFz3Cd8RpJbTFmx4hoTXu4GF9filTbMbzDlYSszHu8KoQQc8HSNuN2jLC2OZqoxTOksNs60clEuktbMCT0iDnTb4X5Ssd7eHXvYpwjDqreTKKTKdToOMVbSrlj7CsYJXG+tPYFvlzYesaxOc9HPfz1gY8w0hHE325Se2wwax61CiHEbOmyInz12PvZua8J15BJ9Z442pJ3y+kk9Ig5051ysHXrMpq/O4gam8CeCGEnE1gDQ5T+Kk7ZE16iK6v5+Vev4IuFRzlTR9cLE82oR4ppfvIYOh7HHpuY199DCCEyQXsqwO4/LKH5+70wEUZPhLBtCT3TSegRc0ppwLLQqRToqcWtbAtrdAxGx3BVFDOYcGJjwxliT8R24Z7QpLq6561uIXKFpW2G7ChDlsKlbEpMkwLDOyd/V8iOMWCliGmDQsOmxPRe8Lg+cQ42KMuefMJjn2belmFi5gfA7YZUCns8dEr3l3K6MAJ+cJ1+hq0yDFJ+hVNlXqCS0CPmTLmZZP1VLWz3LsI1XEHVS0ncz+2W/mUhFpCOVIQvtt7Nwd11aH+Kj16xla+V7pr1MGJpm++NruBbO29ADbuoXtHHvy97kJWuuQlYuajBEWL55jb25tdPdm9tieN4YRdMe9rjqK6k9/ZaRpdrPP0GtU+Pwc59M65jLG3k+K3FhGtPP9ldK6hc1ss6Twdw5vGYC5GEHjFnykwf367/LWO1mmcjS/m3ifdT+5JDQo8QC0inFeDoK/U0f+84yeoifh1cw9+U7Jz10GOj+dXxdTQ8oPDua6f9k03sa6xgpWtsVv+eXFZt+riv6ZeMNWh+M7GGH43cSuVLJnpa6LEqgsRvGefR9ffyr73v4kDbKvJ2zrxOpCGfhve28q8Nvzzt32OiyTMUQSPz1guR0CPmjKkMSkw/JSZ0ubuJBzWqphLHRBh7fAI7HEZZFuGQlx1xk0IjSo2DOXu0LoSYZGmbHivCoOVkR6QZ16jCHhjC4feSTHinuptnXzzlIDCWwBoYxD3SyM5wI4uc2yg1E1RKV9clm/6eu8LTRTwIZm0VJJInzgmVeynP72Wp00ONd4S9zlPX8LGdikrvGEud/vksf15I6BHzosER4vJrD7G1eBHOoTJqnkvifOYNjP4RSp8s5E+OfZFEWYo/u/o5/qroaLrLFSKr9VgRPnv0bg7vqMc9oqjcEUMnElz8EnYXRlsWJbsjPPbzTfyq6Goq1/Xy/eU/ycoP2XRZ4hxi+Q1HebOibrI/aoq7KMpXq3fk7Er2EnrEvKg2fXy3/jEitRa/CS3j2/3vo+p5k1RfP8FfTVDkcBC/cgm/q1vDV4MtOXtDCjEfBi0nLdvqWfatTuzRMXQsPjnZYL5ojbH9ALV73Bh+H62fW0T7okKWOpPnfq04Lw0OHz9a9AiRxpmDjU2lKDBcwIVvA5QNJPSIeWEqgwLlpcCACucY2gEYCrTGjkQAcI7G6Bgq4NmoG1OdfLy+b7QSMz75Z8PjwSguQnvdRIsNChzRdPw6Yo4NWmFaUy4m7HOPGTDR1DrGqXOcuXskqS06UlE6U/k4VYp6R4QaR2C2y84YFgojBXpsHHsiPUtA6GRicnxfMomRgCQmIKFntkx/zxUnSegRC4bZPUTJow18Zf+fMv05u69XU3Zwau+X5U20vq+QeGOcxprj3BjYD8hdnW1+OLaGb790E97uc79F2U5NyYY+Hmj+MY3O0weZPivKZw5/lO7XqrA8cP21e/hmzXNnXQlcCJF9JPSIBSPV20fBL0coNE1Q01KPZWFNDcSL1Aa47F0H+Le6R/EoE6+SD61s9ExfM42/tHC/uu+c56qAn6PGIrqX+mg8wxP7PsvF8Z1VLP12G3ZxIS9ULiFe/TS+DJtuK4S4NBJ6xLzzqTjxIo1a3oQRjsHg8ORihVqj43H0WV6rTchzxigxZcBjNhixIuxPehiw8mccbxsoonE8gR0On/MaBuAeUTw2to5R+zBNjmEWOyf3ZGtLxWhJFrMjvAL3iMIOhTG8HuyknyQaS9syfkzkrIAZI16ocDTUQTyBPTKKHYulu6w5JaFHzLvlrhGuv3E3LzYuItmXT/1jxbie3nlyxWaRM34VauIftryXQOvMt6KSYzaOznbOZ2itjsepeDXME/FreKTwGkqu7eHB5gcwgS+0/DHdL9bgHoXK1yPoWHzyRZZi1Aa3iuPDJVOlRU5a6+ngB9dOcKSsGl+vovL5Qdh/ON1lzSkJPWLe1TkCfKvmBZLVz/LDsWXcd+jdlD9rzu/sEbEgvDK2mJqnFf7H3rY6mmWROs+NEnUqhXptL+U7HZjBQg4VNDG8zIETm/a9VSy/7xjWwCA6mZpcpE1rsBUR20FEJXEaloQekZOWOEf42PLt7K6qZntLA8X7A5j7013V3JLQI9LCrZy4lZNa1xCRSg2XNWOG49AzgDUyku7yxBwatMK8Hi+kOxVkR28tJaMpdDx+Xq81i4ugohTtNDFGJrC6+yZnANkWOm6hI1E8Q4qfjmzEqSzcAwY6HJl5/anxYk5lY6r5WplGiIXHo6DSOcqYz8ubvmpsp+uMGz9nCwk9Iq3Wunq56ZY3+MOqRsZ7Cmn8dT7O30tXVzZ7PFzP156/k4L9DgJ9Np6W4+fVjYVhEr56MR3v03iDUZx/qKXm50lSPb0nTrGjMaq2hHlm+GpQULM3cmJJhBkcmjzDosBw48j6t3khTq/AcLHJ28pydzf7yioJeWuyfvUeCT0irRqdAb5Z9Qqpqpf4t5Hl/GLnuyhOd1FiTr0ebqByi0H+w9vQtiZln183ljIUo4scfO3ah7nd38bV0S+hH/NDz8lzdDKBem03JVunBidrG/22AK0NhTI0PqVwq2x/ixfizHyGi2bX5AzGxXkD7HTVprmiuSehR6SdqQxMDJzKYt7WwRfzatAK80qslNZEGVuOL6ZgzDrvMVxmSTG6ppxUwEWkSlPqGMetDJTBzKUN3qI16JlBSjkcmDVVpMoKmKj2URAcwymztuaVgaI8MMFAczGF5hrMoRB2R1e6yxJTDKVPfz9lGQk9Qog591y0iv/8zIco2mXiH7Lx7+8+vy4tpYhsaOLYBzVV1cPcWb6Lta5BuMAuKaMgn647arBvGqEyr4s/rX4Nn6zxNK9MZfCFmue599PX0zVRQPjVChp/msLu7U93aSKHSOgRQsy5fdEayl4zKfzJNtA2qfMds6UMxuscfHXj4/xZYRsApgowZl/Y9iPK62VsZYot6++l2vRNrc0jY3nm27t9MW5d/ARRnWBD7DPYj/qh99yvE2K2SOgRGcU5bvFC2xK+5h5nkbuPm33tVObwHkoL2aAV5rloFfuiNTzStobCEQvOc/zOCdrGM6L5eeflDKZOtnMo5UZ1elDxoTO+VDkcmHU1JKqDjJW7CJRP4FNKFiNMM1MZODFRSiYriPknoUdkFM+BLip/VMuTxdcxtE6TvP0RPl0gXxUXoldipfznZz40+YRnxCKwu+f8urSm05rCnX2E4qU8mVd+4rCyNY3tUeyBM4ceI+Cn+/YqnLcNUJc/wpfKdpIne20JkdMk9IiMkurpxd3TixswUhs5eGMlSOhZkFoTZRTtmurSsq0LDzxTrCNteI+04T3Nz+yzvdDtZnyxzROr7mep861tS2S2lhC5bEGEnrhO8nw0wPMTzcTtkyUFzDjvyt/DZrfsjyNOJU/HFz6lAX3maGKWFJNaUkOyYOYTGPdgFONwB9b4+BxXKITIJQsi9PRZcf5y9yfx/i4fV/jkJ1m0WPGHO5v4XfMvCChPGisUQsyF5PJajnzSweXL2mYcf2PnYpbeVwW7JfQIIWbPggg9Ydsg1pZH3aNHsAYGThwvWNzIwatKiC23kKGqQmSfeImLd63Zw3drXp1x/MaYn1R+EHm+K4SYTWkLPXGd5KlIAU+PrqZlvBR/pwHJRLrKEWnSb4X51cRSto838urxBkoGJmf3mMEgqeY6YmXuGee7h5I4Dx7HGhjAM5zi1wfWMZgIsDavk7vy9lIjM7nSKqktBq0oE1rRHivBSF7cdZoL+3j5qmoKSzac81xlga9jAg61YcdiF/cXCiFyQtpCz5id4O/2vQ/fwwV4By2qjw1gh8LpKkekyevxIv7lmfdQ85ymaiyFu6WLlNbYTVW03OPmnZftw1Anx4T8fv8KFn2/GmNgAO/uThq/U0VrYTMvb15F6Qcm+GjemWfziLkXsuNsjVewN1rD9oE6nOGzDjU+o8+XvkD5PeN0xwvOee5Y0svuJ5fT+P0R7B4Z1C6EOLO0hZ6Y1oS68ql9voNUVzcXuHqHyBK9qQLyW0x8j25Dp1InZvik8t0sXdLNvbV/mHH+p2wHrYXNeIBUbx9Gbx8eIFh8NccTRYCEnnRKoulKBjkaKWUk5KMkeXGjzde4PKwp3X9e5/ZbYa5qaAKP+9wnCyFy2oIY0zOdcrth9RImGv1Eykwaa47jUbJyqhCZwK0MGlwDRAIujuaVYLlO7W40S0uJr6kjWupkaLXiTn/Paa4khBCzb8GFHqOwgPbb8rn63btp9A1yY2A/XtkjR4iMEFBurnIPscI5SH9pPi/4K045J7WkirZ7NB9cvZVFnn5uDxwCmaoghJgHaQs9lgY0YM/s81cuF9GaJP9Q9eTU9gIyf0O8zek2AtZgafm3km6mMigx/ZSYUO0eQZ/mHSaZ72J9Uxtfr3hj6ogEnlxjne/eayJt7Cx9P53X0JPUFr8MlfDz3g10TRSQf8hEJxInurRGlwWIFRvUN3bjkcUIxWmszevk5c2rCBZdjb8vhfeNY1h9/QR6kty741qerVvG1SVtfLboVepkJlfWsrTNI+FCftp7Fd2hAgIHXOjIhW1CKubfzniCR8fX0REtInE0HxXpQ+LPwuPpi/PsjlXcNFHM6mA3nyt+kWaXL91lzYp5DT0hO84/HriVvJ/kE+yJ4+zpxhobxyjI59jN+VzzgTdY5uvjGv8hAoYMShSnuitvL0XvD9Fxewk/fPNqFo1Xovr68bzexrLhSlKBEh66pZo1H+ygLjCW7nLFHInqBP969GbUj0rJOx6jqLcHe3g03WWJs7C0zUOjG/jlsxvx9RhUHkmBtNmC5DjQzrJ7q0kVFPP7zTXU/fEwzUWt6S5rVszvkx4040N+anb2kmo7dnIvHoeDWLnN/1X+zNQeOSf3x7G0jX2W7wIGsmtyJnqrXRPaMXM7CaVAGWhDnXYX5hpHgHvyB4FBdtTVMx6oxQlYg0MwOIRpmASWb2A4FQAk9GQrC83AcD5LXx/AOnx05uzPqX9DyjDQCkx5lrBgdEUL8XcaFLSn8HVF0LF4uksS02gFGCbWeAh2H8IA8quvpC+Zn+7SZs2CG8g83aAV5pvDG3iqqxnLPn2wKQ9M8IWa53m3TxYlyxQ9qRDfGLqG57uXMNBbQF17Cm1rzJJiIhuaGK9zMNEAHytuO+e1RG4ZsSJ0W4peK4AVcpw6JtDpQl+2jOFVAeJBReWyXvKM0w0CE+kwmvDiGdZ4e6IYo2HsVAqlpH0WgtW+4/zu6itIBDaAAm1OhqCRNRZrfJ3pLm/WLOjQ05py8cCL17DooQRm7PR7NA80F3Pvp6/n1sVPyBOfDLE/WcAvn9tI069iFE+EUN0DWLaFrinn2Ac1X934OFXOEa70dCODXMV0nZbBc+HldMaKcIw4UKmZK3wZXg8dN+bxzru3sdp3nHWeDoKG7Nu3ENhohqI+8jrjmAePoROJk2M6Rdrd7D+CdctvabmuHLeRosgRxmfEqXUNcZV7CPCnu8RZsXBCjw0xbRLXJ9etH7Dy8PSaOHYewo5ETvuyAsda+iLywZhJJmwv3j4DY8cB7PjJx9uW10lF1Qh/Vtg2FWDP3q6GsrFNhXI40LYGe/IDUGmI2G7iOomBgVPWeUoLE41tTs7IxLLQlgVTs3ZStnne7WNpmxQWltYMWz664kE6o0HMBGBNPelRCmWa4HYTL9b8SfHLrHF5AFnuYiFJWSaOcBJrXDaSXWjKTTe3+w8T8bXgU5pS041bvTXUJDsCDyyU0BONUbRXcWf5F3B7Tu6/FRnzUnV46s3ydJQ6/fRlkROuLmrlO7csIrBsA/kdFvmvtJPq6ye/Pcm3ttzMD6s2cm1NK39V9gyNTgnG822Zu5vRTXEs7+V4BjXFW/uxDh/F0xPiwMtNrB/8JEtKBvkvtY+z0XPm4PNIuJBvtr+DgQk/0QkPasSJGVMUH9DoSGQy9K5vZuCyyS6tktV9FBmnfzIshDg9AwO/MjCx8SgDI0uXi1kQoccKhSl7vJWSbUEwT6YYlRyHgRGsxGk2IjVMlKGwDYVxmgGvIvt9omA3a9/TwYCVz9d23oG/swx6+/BuO8ry9mLsgJtn77yMW+/eTaPz9E8KxdzZ5JnggWu+T/tVpdzXuZmJcCW+w0eh5RiLfxDFzvNy7JpF/PYzl7HRs/u017C0zb3HryXxgwrqDo6hUiFIpibH8oyFsMfGUS4XPZvzuPFj29iU18JyVx/lpneef1shMptTmeQbHgLorJ4gNK+hxwCUqdEeF4bHg06l0KkU2Bap3j7o7Tvr65XDgXI4wDBOPM5OukxMCT05qcz08y5fEhji+6UjWO4gBmCNjMDICMrhwHPtBiZsLyChZ74FDA+bPbDZM8Th0qM86a/GB9iRCHZrOwD5dRvoiBYxZp9+jZ2ktukez6fqaAj7zQMzf2iYKKcD5fcTK9L8UXAbmz0GIGN4zoc2Qfm8qHgcnUyd6B4WuctUBtk+GGBeQ49HmWxc0sqOjy/HPVxK6e4E7pf3n3G8znRmMEjo+iUML3OgDU50a0UrLT5RcShrU6kQ2cx7PMz2Z5u5vKnxtD/XGny7vJhDXczosDJM2LCS3isDxINQdEUfFWYEGfh+fkrNBGXr+mj93CLcI1C2I4yxbV+6yxJizs1r6PEqF/9U+ztaPvQiRxPl/Osj72PxmwE4j9BDSZDOW+Cfb/op+cbJ6el5RpQlzijZNNBKiFyhDrWx+PvF6DPtkK41KtyH1T8483VOB71XBrj5U69yU/5+6h0j1DmkS+t8VZpe7mt+gPbFhbww0cxj5jVU7ZJB3yL7zWvoMZVBnSNAncNimfMw/xi0ID+AcR7Lx9sFPjwlUd7jG8BnTL85DSTwZDClUC4XyuUi6TFxGva5X3MaLsMi6XPgyMuDZHLGrDCRfk5lYbnAyMsDy8KOxcG2Jru6jp1/16NyOCb/vfj9xINwU/5+bvXFgexYIn++OJXJUqefpc4kptrHI4FrJieGWDappEmflSKpI/gM57QZPEJkvrQNZM4zTK5cdZQ3/nQpjsipOzG/XaLQ5o6mHTL9OMuYZaWM3tjEyFKDWGWKP63cf1FdlbeU7+ebd1bhuXo1hS02Rc+1Ta7SLBaEK3xt/PjmDYRrV+Pt1VRsGcQ60HJhF1EKvb6Zns15xIo0RVf0Ue8YQQLP7FFjEwT/UMo7Y3+BrzjCl5tf4E8LOmX4gMgaaQs9AeXmf9b9huNVXhLnMXTKo5I0OBI4lTzVySa6tIje2xN8Z9MDFJth6h1JLubJ3acK9vKOWw4wZPv43NaPU3CgBCT0LBjXeSb49cbvMrTBy//X/m5CPdV4D5z7ddMp02TgsgA3fmwbfxTcRoUZkS6tWZbqH6T8lykqnvITXlHOA//pKj69uiPrB7eK3JG20GMqgxpHgJrzrsDJ9D25ROZyqhSWC8xgIYmgl+KiEDd6YziVi4tdTC5o+giakNQJyovGSQaDuINBLM/k3yfSy2e4WDnVtL8r7GZLYR2BYPDCLuJwEA8qNuW1TM3SkkHLs862Tuxj5wnmMRxzY2PDJcSeuE4ybMWJaIjEnRRPbR2inC4Mrwfl92G7wInMHhNzb0Gs0yNyS5NjmJJrezhU0IRVmOLTdbsxZmmVSQPFe6v38J2PXo9jaDElq/tY7upDpjEvHDfkH+B3717FSPPyC3uhgbRnBvrZRDX/+OatJAe8FO4zMPs7SSmFvWEFndf5iBfb1K7tosE5iozPFHNNQo+Yd4udbh5sfoDhZQ6c2FQ5FKaanW4KUxl8PribD7xzFzFtUmSkZKG6BeZGzzhPXPXvjF15YU9uDaWlPTPQw72XU/6gh7xtx9CxONbExGRX5RovH/jQS9xduF3aVcwbCT1i3jmVOdm1OUfXLzC8FMi4ywXLZ7hoNGR69ELhVBYpr8YoK0FNeNChMHYshrJtohE3BxI2BUaIUtNB4CI2bw0nXbiHEqR6ek8cUw4HllvR7O2e2iNNiPkhoUcIIXJYgyPE8s1t7M2vxzVkUr0ljuOFXRj9IxT+vpG7ur6CXR7nLy5/li8Hj6W7XCEuiYQeIYTIYdWmj/uafslYg+Y3E2v40citVL5kkuoboOThMKVuF9HLG/lV5Tr+rLBNpq+LjCahRwghcpipDEpMPyUmNLn7sV2AoSBpYU9MwAS4hyppH8lne1zjVCc3gPapFLUOg4DhIWTH6LYsJuyTY7UsFIMhP5XW5P6Ihs+HkZ8HXg/JwORSJELMJwk9QgghzsrsHSH/8To+0fIl9LSJllZtjP/nysf4ZH4/j4Yr+bsd78c4NnOMTsERcB6fnLGVunwZHe/ykChL0by0nRWuXmRxSTGfJPQIIYQ4q1RXN8UPjVDimPmRMfHOZp5uWskn8/t5fmw55b9xU/DkzI1LdSpFKhYHZTCy3MOH3/sinw9uxWeYBJQMYhbzS0KPEEIIAHwqTqJAYzTVoSIx7OHRyS4urbFPszG0azTFoeFSnq0w2TtciXs0hTU+fvqLGya2CZXOUSodsrCkSA8JPUIIIQBY4hzhyhsP8EptE0Z/MfVPlOF47nXQ+rTne9oGMX9ZxZe3fQ5ft6bwSA+y/rlYyCT0CCGEAKDRGeB7dU8Tr03xw7FVPNB6K+VbTHTq9FEm1d5BUXcvxQ4HWBapeHyeKxbiwkjoEUIIcYLPcOHDRb1rkFgZqOWLMab2ywJQ0Tj24PCJbi8dj6PfFnYMvx+jrATtdZ88qBSxEkWeGZuvX0WIU0joEUIIcYp17m6uvGUvry1vYPqULft4MU2PBFF/2HXG11prFnPkfT5U/f+/vTuPjrO+7z3+/j3PLNJo3yzJWi3v2BjbBBuzhZgkkBNSIFs5TUuT25M0vfe0oTntbXvbnjbt7b1tki73tCVr23DSJmlCILQQEgI4rMUGvIBZvUm2JVn7PprteX73DwmwjAReJD2amc/rHB08j58Zvjo/PzOf+W3PafOAjOVdTS+zNXoCrdiSoCj0iIjIW6wMF/ON5odJN828+/kXenfw+P7tlD0593PHmwu5/rq9/GX9ozOOh41L1CjwSHAUekREZFZREyZqZt4YtiE6xHijQ/mWDXM+b7zRoalg8Lzu1SWykBR6RETkrO0seoW9Nzezf0fDnOdsrHmZ64sPAtE5zxEJgkKPiIictU2RAv6peRd+sz/nOQ4OYaPAI0uPQo+IiJyTsHEBN+gyRM6ZbpcrIiIieUGhR0RERPKCQo+IiIjkBYUeERERyQsKPSIiIpIXFHpEREQkLyj0iIiISF5Q6BEREZG8oNAjIiIiecFYa8/+ZGP6gI6FK0dm0WKtrZnvF1VbBkbtmTvUlrll3ttTbRmYOdvynEKPiIiISLbS8JaIiIjkBYUeERERyQsKPZIVjDFNxphjxpjK6ccV049bAy5NJK8ZY1qNMQfPOPanxpjfCaomOT/GmCpjzP7pn1PGmM7THkeCrm8+hIIuQORsWGtPGGO+Avwl8Jnp/37dWtseaGEiIjnCWjsAbIap4AqMW2u/HGRN8y3ne3qMMbcZY543xhwwxnw76HrkgvwtcLkx5nbgKiCnLsZ8Y4z5vDHm4PTP7UHXIyK5L6d7eowxG4A/Aq6w1va/PjQi2clamzbG/C7wE+D91tp00DXJ+THGXAp8CtgOGGC3MeZRa+2+YCsTkVyW6z09O4EfWGv7Aay1gwHXIxfuA0A3sDHoQuSCXAXcY62dsNaOA3cDVwdck5yfufY90X4osuTkeuiRHGKM2Qy8D7gc+G1jTH2wFYkIMABUnHGsEugPoBaRt5XroecR4GPGmCoADW9lL2OMAb4C3G6tPQ58Cc3pyWaPAzcbY2LGmCLgluljkmWme+q6jTE74Y332RuAJwItTGQWOR16rLUvAn8BPGqMOQD8TcAlyfn7NHDcWvuz6cd3AOuNMe8OsCY5T9bavcC3gD3AbuCbms+T1W4D/tgYs5+pL5tfsNYeCbYkkbfSbShEREQkL+R0T4+IiIjI6xR6REREJC8o9IiIiEheUOgRERGRvKDQIyIiInlBoUdERETywjnde6u60rWtTeGFqkVm0X4iTf+gZ+b7ddWWwXju+WS/tbZmvl9X7bn4dG3mloW4NtWWwXi7tjyn0NPaFGbPT5vmpyo5K9uuP7Egr6u2DIZbf7hjIV5X7bn4dG3mloW4NtWWwXi7ttTwloiIiOQFhR4RERHJCwo9IiIikhcUekRERCQvKPSIiIhIXlDoERERkbyg0CMiIiJ5QaFHRERE8oJCj4iIiOQFhR4RERHJCwo9IiIikhcUekRERCQvKPSIiIhIXlDoERERkbyg0CMiIiJ5QaFHRERE8oJCj4iIiOQFhR4RERHJCwo9IiIikhcUekRERCQvhIIuYDHE/RT9fooJf2bGK3F8atwoURMOqDIRERFZLHkReh6IV/OH+28ic6JoxvHS1UP8/cbvcmVBQIWJiIjIosmL0PPwyEVU3FNExUNHZhzv/thqdret4sqCowFVJiIiIotlyYUez/p0e3H6vTAe5o3jYeOz3PWodove5tmzm/TCRMZ8vL6+GcfD8VWkrXvBNcv8SNo0XZkkg36EmMmwPGQocwqDLkveQdp6dHuT9HmROc9Re4rIUrDkQk+3F+czRz7Oa8+24KTfPJ4u9/jo5c/wf2qfJWwUVHLRs0mX//78p5l8uZx0XYrf2/4TPlveGXRZ8g4Op5N8+pXb6Nlfi/FnPyddr/YUkeAtudDT74U5tKeFtf9wAjsy+sZxf20L99Vs5AvLdiv05Kj9iRbMwxWs+teXiV+xih+3XawPySxwNFPJ4BN1rP7aIUgmZz0nfvVataeIBG7JhR4Pg5MBOzKKN/pm6HEnkniZQnzm+CopWS9tXUIJizc8TCjukcwsuX+eMou0DeGmwB8dxc4ReiLDaY4OVPFwo0u5M0lbKEOFG1vkSmW+xf0UHZkMI36UMidJU8ih2Jl7ZUi/N8HRTIS4H6XWHactHNbqWVlU+lQRkQUXPt5P0T1N/Oa+X2eyOc0fXHU/nynrCrosuUCPJUr4nwc/zER7GYXNY/zVph/ywVhi1nM96/P3g9v49mNXEe13CW8Z4l8uuZNLo4tctOQ1bU4oIgsuc7KTiu/vpeWv99N0v+Hng2uDLknmwd54K/bJClb++yTuY2XsmVg557k+lp92rmfl91Os+H8vYp+s4FCqdhGrFcmmnp50hvRQlHsnGqgLDbM2PEJjqDjoquQCxf0Ur6UtxzMVPDG4ilDcBl2SXAATjuAuq8aWFmGSaWz/4NQwtbXYZBILuAlLyte8vGw14k/yajpEV6aCJwZWEh2yhEYmcRMxEv7bD1V5voObyOCPTxAdsvxsaAMxZy+t4UHWhl0NdS0RQ16cl9IF9Hml1LkjrI+kcmblZfaEnv4hmh+o5i+P/iKTtZabdu7mi3XP4hp1VmWz51Mun3z2k4T2llAwYKk5MIRvFXyylbu8lhMfbSJ+aRzbU0Drf1YReuS5oMuSefToZBW3P3UrRS8WEB2wVO8dhd4BImOVTHpnF1qs51H93Cj7QhfzdMUmQtuH+M7mf2ZDRKFnKbh7vI3//egvUHw0xPi6FF++6vt8pHj0nZ+YBbIm9HhDQxQ+sJeGB13slrU8smYNmbrduBqhy2qvpWqJPlXC8m8cwKbS+Jn0Oz9Jliy/ogTvihGe3vZ1vj60lbtf3EnVLgMKsjljX7yVqp9HqfrOc+BbrOfh+R7h+CqS/ll+pFiLPfAKyw66OEWFtDsbOLGxnA2R2SfCy+J6amQVjQ8aiu5/juGPbWHPljY+Urw/6LLmRdaEHgCbyUAmg0l6eNa88xNkyfNxMB5TQx+ZTNDlyHkocSaZrLHYS9Yw0RAjlUxy9/hqnhhYSXhiKuw4sRhmeS1+SYzRlhAbCsYCrlrOl//6CtszVuqZcw22vof1PWzCwfjM2IxWFl+/N8HeZDldmQqePdVE9XAGm0ziZMDPoc/brAo9IrL0rA2PcNPO3TyyZg3jE0lMeyF3PH0z0SFL1f5+PGsxK5o4+ouVRDYN01ZxlI9X7Q66bBE5zY8nWviTXR+m7KUQxT0+BYdOkotfQxV6ROSCNIaK+WLds2TqdnPnaAt3PH0z9V/fi59K49mpfbVSNUU0XNHJj9f/EAdHG4yKLDF7J1qpf9Sh9K49WN+S8b2gS1oQSy70FBiPVE2G5GWrCQ8ncLsGyJzqOec5Ab3eBE8lamlPVbP7ZAvLJ6Yyq1tRgW2qxSuKMt5oqA6pm32pO70t57I6eoorooPa8C4A436C3ckiXkqs4IHeDUSHLH4qDae9abrxNIe7q/hK/WrK3ThtkV5qnDhFjk+1EyHmzH3fLhE5d29el40zhg5rQmNcXtDByvBbVz8bn5yfZrDkQk9jCD674+fc37KR4wNlVN/XStkPh+bc6XUu/zm+kr/42U1UvuBQ3ecRPdRJBkhvauXIxyPUtvWzs/oVro4dAc79JqayeE5vy9lYA4NbPf72uu9wc9H4Ilcnr6Ydfn33r1D6eCHR4deHtGbunO529ND0vRb+9bEPEF9mMFtHuLium7XFPXyifA9rFHpE5tXp16VzWo4ZbzTccOMe/rpuT16ufl5yoafMKeT3qg7xO5Wv8vBklNtf+jTlrsu5rv14Id5I3ZOGku8/DUBmuqdooj7KLZfv4Yt1zwLgGgWepe7MtnwL4+D86jZevbIeig4tbnFCZ6acggMxav9lL34yiTdLr6zX00v0x31EgZpN6zgWLWdfxmV0WQHvL3mBNVqpLDKvzrwuX1e5YxO7t7VA3Z4AqwvOkgs9r3ONQ5U7wWRjhsQ1GwiPpggf7yfTefZb1xvLrMNijrF5mXCz2VxtCYD1pv5eAlHuxok3eCSv2UhoLE24vYdM96m3njjdfiaepLDXMFZYxFFj6WssBdRDl29C9XWkW2tJlIaZbPQod+LoJgHzx8fBWLDWznzvzPP3yiX9L6wtlOFzVz9I0e+f5MRv+wxd1YRxNQFSZClZF57gN3Y+RMEfdHHsNw2jl7eA8zbXaU8/dT8fZMW9k0R2l3Ag3rx4xcrS4LiMXNlC++cshb/fxW+8+yHWhieDrkrywJLt6QGocGPcXtHO7RXt/KihmD9+5jZKFXpElpRlbhG/W3mE3608wrfql/F3//VRihzDGdN63uCNjsLBURygvHobXcmyRa1Xgmccw1ijy59suZ9PlAxMH9VUA1l4Szr0zMWJJ5k4Ws1vN1xNY3SIG0sPsClSEHRZMo9CYylePVzPb5Vdxk8OXUTjoHZqzgY1oVHG2nzKr9+COW3Fa3gsTehwF15Pb3DFiQjuWJKeI9V8rnoHIefNi/THr22Y9X02OpThPw5dTNq6XBTr4kNFr1Gfxfe9zMrQQ3cvK+8q4sB/XcLjrS7HPlrNVxsf1zydHOIc62LVd1p47qGtNA5lKHyxMyc3yso1WyP9/I/3PcjTl62YsYvrvo4mmr/dTOSnCj0iQTInu1n5vQL2PboZzJvXaONgetb32djBLhr+aTnPlL+L+7ZD6Y3/zq0lQ4ta83zKytDjjY5injpAMVC04xIO7qyDxqCrkvnkDQziPDrI698nFHiyQ32omM9XHoXKozOO/0npBn6y7Bq0MF0kWN7wCM4T+5mtr2a299lMZxfhzi7CQDq2g473VwMKPSIiIm8R91P0eCni1mUsHqXKn2Oyl8yrGneU8ZVpxj+0mcioR+GrPWQ6TgRdVuAUekREZMGc9NI8MrGW7nQ5yf5CTHI031dNL4qLwgm+8O57eHrrKp7ubiHxbw2UnOiasVN6PtIkGBERWTBxP8TxZBVHJmpw4w7Gy+8P3cVS4ca4rbSfOxqe5rfW7CK+zME4uXO39POVFz09blUl3soGUhVRhtY6NEcHgy5J5kGosYH4xuUkKlwGLrGsiGqSrMhS83yygR8d2cRkX4zyYw4mngi6pLzTFB5gZEOGgl98FwXDHkUv9pBpPz7jHLemhuSmZiZrwhQMZih8/sTUfS9zTF6EHr+1nvabirFtcdbUd7I9dhjQvvfZLr5xOad+LcGNKw+yIdbJtYVdaK8PkaXlJwMXU3R/Cc0Hx3GGxvF7+oIuKe9sjYzxf9/zA/Ztb+Ghk2vIfKueoo4TM3ZqzqxezrHbLB+5eDf3vLyZFV9djqPQExyHc5v85lvzRoN6xREyzQneu/I11hd1U+cmUejJfokKlxtXHuRLdfumjyjwLFWedTBz3UZEclp3vJTyQ5PYZ17g9YEtE8qaj56cUOHGuLVkiFtLhvi7yDDfrvoARcYB++ZQY6oswta2Y3ypbh89yRI6ytaRi7vfZc2/vGXuGONrUozesoXIiEfRK71kjnXMOOe19ARf67+a/UONHD1cx6pTKQC8sENRSZxNxSdpjfQRMxrXzFahhuWMb2lksnpqSGtDrDPokvJSvzfBv4xs4qGe9ZREEvxy3dN8KDZKtxfnG0PbeaJv5Yx9eto7amjrntr4zC0txa5oIFNewNCqEMujI0H9GiJ5Z3X0FINbPZxf3TbjnoWTNYYTHU28J3ETHa/Wsbo/N28LkjWhpy2c4LPbH2VX2xoOn6qh/gd1xDpOzjjnZxPreOBHl9Pw2CRrRydw2rvwAK/AZU1VHzcVv0yBMZQ5uZhf80NibT29vzrJr6zdQ1u0l3cXnoBZd5yQhXQ0E+GOx69jxT0+ndUhvnxbKe/f+D32JpfxnZ9eQ8sDSUL+m++oa8cncTpOTX3Tr6uh+5oKxlt9ws1jbIm1B/VriOSd7dEBvrzze7x6RT0+b34xeaDrIkL31RF6oZK1I2OY9k5yccp51oSeYhPmPcUv0RgZ5N7IZtrL1hADjLV4vkMGj85kBaXHfJxH92HhjQazIaiKTtCYxVtny5R0qcs1LYf5X9WvTh9RmwZhzC+gsCtE9MkDFDQv59CNpSRthlOZcoqPG0KPP4/NvLnV2YzrsSBMosbiNsZZUT1AlTuOFpLmhozvkrYzPyqtVc/6UlLtFvGR4lEoHgXAsz4+llPJMvb11OA8uu8cJ5Nkl6wJPWHjUuVMkg73UVcwytHpyt3BCcaeqeUK/xMMHa+grTMVbKEiecaMxSnYX8O1sV9jdKCI5vYM1p97/o4fi5Bc5nFl03FWF/VS406i+VjZr/BUgp8/vZErB+qBqbDjWxg/UMXKgf6c7DXIdi+mJvm7nveyv6+BgSOVrOrK/ZV1WRV6GkNRat0Uewt7eTxqMI7BHu9kxb9Z/PuLqI0PQ1evLi6RReT19NH8gxDerlKWJ8cxXX14b7MBWro4TEtbL3+4/McUOT61bnQRq5WF4r7UzppvLscrjYEF4/kYC7UjvdiT3UGXJ7N4ZGIdT/3oEhofHqNmfBC6enL+8zNrQg9A1ISJmjAl7iR2ujfcTyTg0NR9fnK9sXJJ2nr4+CT8MOZc+lItpPwQSZvGwSFs3AWrUc6OTaemFhUc46077TruWzZE86MOy2JjrI/EFq1GmT8OFuuACUewngfWB2un7on4ShwHpnr6fG/GsCYAxmBcFxOJYB1wc3ogZel6/f33ZKqCkuM+7JleWee4Uz9nyqFdnLMq9EhuOJkZ54u972HXidWM9xTTcjQ99eZ5FmKdcZ54dCNbVzWzofYUv9fwAJdGdRvLpcitXcb4jlZGWkKcNl+SsTaPmyuOBFeYXJBLYsf51jVpkhXvorDPp+q/TpE52g4wdR0bZyoIzcK9aA192ypJVhgS28ZpDfcD+uKymF5Oxfniqet5pquZyY4S2o4nganrNXFxE5M1U9u5vL6yq7A3TfSF43h9ubG/kkKPLLpX02U88PC7WPmDMZyJITjVj3eWe7iYl4+x+hvLsEUFvHLdWh761HEujR5a4IrlfPjLazh5S4Y/3X437mlrY2vcUTZHh9E8nuz0nsI+7nzPNzlxVRX/cPRaJoZriE6HHqydsffLDMYwuKWCpk8d5uO1z9Ia7mdTxEOhZ3HtTrSy+76LablvGBPvg1N9eExdryfeH6H8ogF8OzUny1pD/4sVrByoBoUekfMz5hdSMGDghUN4yeTsJxmDCYUxroO1FptKgbX4ExNw+BgAxesupyNRxYj/PGFcoiaEa7QKaDG4WKwLprAAZ45eukxphOW1w3yipHeWdlHgyVZlTiHXFAAFA+xddoynymooLCjAej42k56xyy8wdS1HIphQiESFww01L3JryRBTYUeBZ7ENezFiPRb/wMsz2sovCJGpTrO9tgMfQ9p38a3hkVMl+NHciQq585tITgk1LGfwmiZGWxxiPZaaJ3rxXps5JFLSEeehh7bwYPN6Lmrs5s+a72VzVJNiF8Py0Bgll/XR/htrMZnZz5ms8/mV+lcVRHPY9pIj3P2+LYy2bKWoy1LzWNdbNo11L1pDz5WVJKoMmc3jrIt2BVStvB0bcYiVTbKj9DCeNfg4eNbhifI2/IibM5tKKPTIkpRurmbkw+N8afNdfPnY9Uz01VJwRuhxXjjC6u5KbEGUYx9q48n/torN0RMBVZxfVoQKuHPDnXSsrcBj9n1YSpwE68ITqFcnd10fO0XrtV+l7+oS/vy1G5nsrCZ8RugZ3lhByy8f5jPLH6POHWVN2ACah7fUZApcWiqHuLFoatNff7oX6HvVQ/jRZQo9QXLwsSEwhYWQSmHTmbfOLndcnEgYwmEyUYfwOS0RkkAYgxONQjjMZFmEtuqTfDCW4OdVx3m8rJ5YSQmk0/ipNPge/sTE1HAXENtWQ3uiml7vFaLGodhE1cOwgMLGZX0kxvrIHMOTb1DgyWVlTiHbogAJflTVzcvlVURLSmackywz7Kg8yg2xJKCe2CXLGGKhFGVO4YzDRaEUo27ubDCZlaGnOTzI2CVJen9pIwVDlrJ9vXjTy9ZfF2pppGfnciaaDMkVCa4qfS2gauVshepqGbiulZGVDomGNJ+veQmAK0sOcc8HNjOy6mKKOi3Ldp3Cm57X87qyI3H+44HLuav+Uja0dfJXrXezIVI42/9GRBbAteWv8PCN64huvHjGcX/9OFsL24MpSuQMWRl62sKj3HrJMzxRv5ITnVWEx6qIHj42Y1JWYkU1zof7+dr6H1DuJGgLATl5z9jc4TVUM/jBSb6+7duUO5O0hDwgxnsL+7n3yjsY2FHIHx26hcnjVUTOCD3u/kOs6ijHFkZp/+gKnvlkCxsivcH8IiJ56Maik1x87T8y7M98n61x4rSEQmhIS5aCrAw9MWNoLehnoKKIwYkYXsFb77/kRxyaS4emVhko7GQFP+xSVT7O1QUZXPNmN3ixU8CG6ffL9RWneKG8lsKKCmwqhT+ZmBrqisfx43FwXCLDdST8cEC/hUh+KnMK2TRrrtEmlNkmadOM+CkS1jKcLMTNnN2WItkgS0NPmM0FHZS7E/RMltJTWPLOT5KccF35Szxy01r6tqyjpB3qH+x+Y2M0ERG5cD+Nl/FHL97EeGcpJYddGk7mzu0psjP0OBEuiaTZGO7jcEU7d8fagi5JFsmNsT4uvfIfGNsR5rMvf4Lkq5W4Cj0iIvPmweGLid1VRtOu49hEAn90POiS5k1Whh6Yug8XBmJukjlWzEoOijkRVjpTfeitZYOcrKqitHbZG39vHIdMkSFscuV7iYjIwjMZn954CS+n4hwaraGw3yPTmXt7KmVt6BH5UPUB/uxjjXRf9WZPnzVQv/YUmwuOo4mTIiJnp+D4MF331fOhg5+n6IRDQ0dfzgxpnU6hR7LWLUXdvHvHHcTtm119LpYSx1DhaPK6iMjZ8o92sPzb/RCOQDqFPz4RdEkLQqFHFl2BSZMuArehHlLpN46PV0WJhYfO+nViToSYo94cEZGzFXNSpEoNoeZGbCKJHRnFTySwmQze8EjQ5S04hR5ZdKvDA6y79ggH6pqnxqOmRSsn+XzDs9pJWURkgVxa0E7VDZ28tLqegq4QLQ+MwTMvBF3WolHokUXXGopx58ofEV8xc8TYNYYyJwJojx0RkYWwORLi7nXfJbHW58973su+Q5spfSboqhZP1oeeEidBotLgrloxY0fm+LIQqyKTAVYmc3GNQ5kppEwdOiIii8o1DhXu1IaRKwr7eHKZQ8WqFWf13ESlIeakFrK8BZf1oeeywnZaP3CMly6qnzFUUlvby81VzwVYmYiIyNL17qJXeOgX1vPKltp3PtlY1jR1cHXsNbL5xrFZH3o2hCN8b9U9pFf6eLzZ0xM2DjETAdzgihMREVmiLo24/HDtXaTXzPz8nI2LIWwcCk12Lx7J+tDjGodio+XJIiIi5yIfPz81q0JERETygkKPiIiI5AWFHhEREckLCj0iIiKSFxR6REREJC8o9IiIiEheUOgRERGRvKDQIyIiInlBoUdERETygkKPiIiI5AWFHhEREckLCj0iIiKSFxR6REREJC8o9IiIiEheUOgRERGRvGCstWd/sjF9QMfClSOzaLHW1sz3i6otA6P2zB1qy9wy7+2ptgzMnG15TqFHREREJFtpeEtERETygkKPiIiI5IWcDT3GmF3GmOvPOHa7MeYrQdUkF8YY02qMORh0HXLhjDEFxpg9xpgDxpgXjTFfCLomOT+zXZfGmD81xvxOUDWJzCVnQw/wXeDWM47dOn1cRIKVBHZaay8BNgM3GGMuD7YkEcl1uRx67gI+aIyJwNS3EWA58HiQRcn8MMa0GWP2GWMuC7oWOXd2yvj0w/D0j1ZViATMGPNnxpjbT3v8F8aYzwVY0rzK2dBjrR0E9gAfmD50K/B9q+VqWc8Ysxb4IfBJa+0zQdcj58cY4xpj9gO9wM+stbsDLklE4J+B2wCMMQ5Tn53/GmhF8yhnQ8+004e4NLSVG2qAe4FPWGsPBF2MnD9rrWet3Qw0AtuMMRsDLknOz1xfJPUFMwtZa9uBAWPMFuD9wD5r7UCwVc2fXA899wLXGWO2AjFr7XNBFyQXbAQ4DlwVdCEyP6y1w8Au4IaAS5HzMwBUnHGsEugPoBaZH98EPgl8iqmen5yR06Fnes7ALqYaTb08uSEF3ALcZoz5paCLkfNjjKkxxpRP/7kQeB/wSqBFyXmZfp/tNsbsBDDGVDIVYJ8ItDC5EPcw1YaXAT8NuJZ5FQq6gEXwXaYa8MyVXJKlrLUTxpgbgZ8ZY8attf8RdE1yzuqBO40xLlNfvr5vrb0v4Jrk/N0G/KMx5m+mH3/BWnskyILk/FlrU8aYXcCwtdYLup75pNtQiIiIyBumJzDvBT5mrT0UdD3zKaeHt0REROTsGWMuAg4DD+da4AH19IiIiEieUE+PiIiI5AWFHhEREckLCj0iIiKSFxR6REREJC8o9IiIiEheUOgRERGRvPD/AdY0PdA9XBIyAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 720x720 with 25 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Training image Shape: (3069, 32, 32)'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'Testing image Shape: (341, 32, 32)'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# partition the data into training and testing splits using 90% of\n",
    "# the data for training and the remaining 10% for testing\n",
    "(train_images, test_images, train_labels, test_labels) = train_test_split(data,\n",
    "        labels, test_size=0.10, stratify=labels, random_state=42)\n",
    "    \n",
    "# show train data in plot\n",
    "show_train_data(train_images, train_labels)\n",
    "\n",
    "# Show shapes\n",
    "display(f\"Training image Shape: {train_images.shape}\")\n",
    "display(f\"Testing image Shape: {test_images.shape}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks great!\n",
    "\n",
    "### Modeling\n",
    "\n",
    "We need to define some parameters for our model. Here is what they mean:\n",
    "\n",
    "- `EPOCHS` - amount of iterations to fit the model\n",
    "- `BATCH_SIZE` - size of slices of the dataset\n",
    "\n",
    "Info about model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "EPOCHS = 80\n",
    "BATCH_SIZE = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] compiling model...\n"
     ]
    }
   ],
   "source": [
    "# initialize and compile our deep neural network\n",
    "print(\"[INFO] compiling model...\")\n",
    "model = models.Sequential([keras.Input(shape=(32, 32, 1))])\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(128, activation='relu'))\n",
    "model.add(layers.Dense(62, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_66 (Conv2D)           (None, 30, 30, 32)        320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_46 (MaxPooling (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_67 (Conv2D)           (None, 13, 13, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_47 (MaxPooling (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_68 (Conv2D)           (None, 4, 4, 64)          36928     \n",
      "_________________________________________________________________\n",
      "flatten_23 (Flatten)         (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_44 (Dense)             (None, 128)               131200    \n",
      "_________________________________________________________________\n",
      "dense_45 (Dense)             (None, 62)                7998      \n",
      "=================================================================\n",
      "Total params: 194,942\n",
      "Trainable params: 194,942\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Use categorical_crossentropy for one-hot coding labels\n",
    "model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),\n",
    "                loss=keras.losses.categorical_crossentropy,\n",
    "                metrics=['accuracy'])\n",
    "\n",
    "display(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since we are using a CNN, we need to add the channel parameter to our training and test data shapes. This is so our data is represented as `(batch_size, new_height, new_width, filters)`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(341, 32, 32, 1)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Keras needs a channel dimension for the model\n",
    "# Since the images are greyscale, the channel can be 1\n",
    "train_images = train_images.reshape(-1, 32, 32, 1)\n",
    "test_images = test_images.reshape(-1, 32, 32, 1)\n",
    "display(test_images.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] training model...\n",
      "Epoch 1/80\n",
      "62/62 [==============================] - 3s 27ms/step - loss: 4.1115 - accuracy: 0.0257 - val_loss: 4.0033 - val_accuracy: 0.0557\n",
      "Epoch 2/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 3.4961 - accuracy: 0.1310 - val_loss: 3.1457 - val_accuracy: 0.2082\n",
      "Epoch 3/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 2.4779 - accuracy: 0.3506 - val_loss: 2.2735 - val_accuracy: 0.3930\n",
      "Epoch 4/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 1.8061 - accuracy: 0.5122 - val_loss: 1.8761 - val_accuracy: 0.4663\n",
      "Epoch 5/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 1.4414 - accuracy: 0.6002 - val_loss: 1.5583 - val_accuracy: 0.5689\n",
      "Epoch 6/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 1.1454 - accuracy: 0.6693 - val_loss: 1.3188 - val_accuracy: 0.6334\n",
      "Epoch 7/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.9367 - accuracy: 0.7247 - val_loss: 1.2408 - val_accuracy: 0.6540\n",
      "Epoch 8/80\n",
      "62/62 [==============================] - 2s 28ms/step - loss: 0.7973 - accuracy: 0.7517 - val_loss: 1.2723 - val_accuracy: 0.6510\n",
      "Epoch 9/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.6792 - accuracy: 0.7937 - val_loss: 1.1206 - val_accuracy: 0.6716\n",
      "Epoch 10/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.5712 - accuracy: 0.8130 - val_loss: 1.1829 - val_accuracy: 0.6569\n",
      "Epoch 11/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.4888 - accuracy: 0.8436 - val_loss: 1.1345 - val_accuracy: 0.6774\n",
      "Epoch 12/80\n",
      "62/62 [==============================] - 2s 24ms/step - loss: 0.4121 - accuracy: 0.8729 - val_loss: 1.2073 - val_accuracy: 0.6921\n",
      "Epoch 13/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.3276 - accuracy: 0.8996 - val_loss: 1.2323 - val_accuracy: 0.6804\n",
      "Epoch 14/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.2886 - accuracy: 0.9075 - val_loss: 1.2447 - val_accuracy: 0.7067\n",
      "Epoch 15/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.2318 - accuracy: 0.9312 - val_loss: 1.4181 - val_accuracy: 0.6774\n",
      "Epoch 16/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.2303 - accuracy: 0.9260 - val_loss: 1.2899 - val_accuracy: 0.7067\n",
      "Epoch 17/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.1555 - accuracy: 0.9537 - val_loss: 1.3898 - val_accuracy: 0.6774\n",
      "Epoch 18/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.1170 - accuracy: 0.9635 - val_loss: 1.4132 - val_accuracy: 0.7097\n",
      "Epoch 19/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.1247 - accuracy: 0.9612 - val_loss: 1.4860 - val_accuracy: 0.7009\n",
      "Epoch 20/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.1290 - accuracy: 0.9563 - val_loss: 1.6237 - val_accuracy: 0.6774\n",
      "Epoch 21/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1078 - accuracy: 0.9651 - val_loss: 1.5645 - val_accuracy: 0.7038\n",
      "Epoch 22/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.1179 - accuracy: 0.9645 - val_loss: 1.6006 - val_accuracy: 0.6950\n",
      "Epoch 23/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0931 - accuracy: 0.9752 - val_loss: 1.6176 - val_accuracy: 0.6862\n",
      "Epoch 24/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0616 - accuracy: 0.9795 - val_loss: 1.5915 - val_accuracy: 0.7067\n",
      "Epoch 25/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0510 - accuracy: 0.9860 - val_loss: 1.6762 - val_accuracy: 0.6862\n",
      "Epoch 26/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0519 - accuracy: 0.9840 - val_loss: 1.7727 - val_accuracy: 0.7009\n",
      "Epoch 27/80\n",
      "62/62 [==============================] - 2s 24ms/step - loss: 0.0439 - accuracy: 0.9883 - val_loss: 1.8385 - val_accuracy: 0.6891\n",
      "Epoch 28/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0572 - accuracy: 0.9818 - val_loss: 1.8223 - val_accuracy: 0.7009\n",
      "Epoch 29/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0627 - accuracy: 0.9814 - val_loss: 1.7373 - val_accuracy: 0.7067\n",
      "Epoch 30/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0512 - accuracy: 0.9863 - val_loss: 1.7541 - val_accuracy: 0.6891\n",
      "Epoch 31/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0436 - accuracy: 0.9889 - val_loss: 1.7925 - val_accuracy: 0.6921\n",
      "Epoch 32/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0364 - accuracy: 0.9906 - val_loss: 1.8122 - val_accuracy: 0.6950\n",
      "Epoch 33/80\n",
      "62/62 [==============================] - 2s 24ms/step - loss: 0.0525 - accuracy: 0.9824 - val_loss: 1.8509 - val_accuracy: 0.6950\n",
      "Epoch 34/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0816 - accuracy: 0.9775 - val_loss: 1.7586 - val_accuracy: 0.7038\n",
      "Epoch 35/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0999 - accuracy: 0.9681 - val_loss: 1.8649 - val_accuracy: 0.7009\n",
      "Epoch 36/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0824 - accuracy: 0.9713 - val_loss: 1.9866 - val_accuracy: 0.6745\n",
      "Epoch 37/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.1054 - accuracy: 0.9658 - val_loss: 1.8624 - val_accuracy: 0.6774\n",
      "Epoch 38/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0755 - accuracy: 0.9723 - val_loss: 1.7141 - val_accuracy: 0.7097\n",
      "Epoch 39/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0477 - accuracy: 0.9831 - val_loss: 1.9349 - val_accuracy: 0.6833\n",
      "Epoch 40/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0275 - accuracy: 0.9919 - val_loss: 1.9551 - val_accuracy: 0.6891\n",
      "Epoch 41/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0823 - accuracy: 0.9765 - val_loss: 1.9577 - val_accuracy: 0.6950\n",
      "Epoch 42/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0375 - accuracy: 0.9892 - val_loss: 1.9938 - val_accuracy: 0.6891\n",
      "Epoch 43/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0176 - accuracy: 0.9951 - val_loss: 1.9272 - val_accuracy: 0.7009\n",
      "Epoch 44/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0136 - accuracy: 0.9961 - val_loss: 1.9688 - val_accuracy: 0.7155\n",
      "Epoch 45/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0074 - accuracy: 0.9984 - val_loss: 2.0054 - val_accuracy: 0.7126\n",
      "Epoch 46/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0178 - accuracy: 0.9938 - val_loss: 1.9492 - val_accuracy: 0.7067\n",
      "Epoch 47/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0086 - accuracy: 0.9971 - val_loss: 2.0478 - val_accuracy: 0.6979\n",
      "Epoch 48/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0080 - accuracy: 0.9974 - val_loss: 2.0697 - val_accuracy: 0.7097\n",
      "Epoch 49/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0104 - accuracy: 0.9958 - val_loss: 2.0860 - val_accuracy: 0.7155\n",
      "Epoch 50/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0089 - accuracy: 0.9967 - val_loss: 2.0863 - val_accuracy: 0.7155\n",
      "Epoch 51/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0044 - accuracy: 0.9987 - val_loss: 2.1547 - val_accuracy: 0.6979\n",
      "Epoch 52/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0063 - accuracy: 0.9977 - val_loss: 2.1304 - val_accuracy: 0.7214\n",
      "Epoch 53/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0130 - accuracy: 0.9945 - val_loss: 2.1268 - val_accuracy: 0.7126\n",
      "Epoch 54/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0153 - accuracy: 0.9948 - val_loss: 2.0379 - val_accuracy: 0.7126\n",
      "Epoch 55/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0222 - accuracy: 0.9912 - val_loss: 2.4216 - val_accuracy: 0.6657\n",
      "Epoch 56/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0900 - accuracy: 0.9756 - val_loss: 2.1243 - val_accuracy: 0.6950\n",
      "Epoch 57/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1326 - accuracy: 0.9514 - val_loss: 2.0786 - val_accuracy: 0.6891\n",
      "Epoch 58/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0921 - accuracy: 0.9668 - val_loss: 1.9326 - val_accuracy: 0.7067\n",
      "Epoch 59/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0635 - accuracy: 0.9778 - val_loss: 2.2042 - val_accuracy: 0.7067\n",
      "Epoch 60/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0336 - accuracy: 0.9876 - val_loss: 2.1325 - val_accuracy: 0.7009\n",
      "Epoch 61/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0282 - accuracy: 0.9909 - val_loss: 2.3317 - val_accuracy: 0.6862\n",
      "Epoch 62/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0510 - accuracy: 0.9827 - val_loss: 2.4462 - val_accuracy: 0.6745\n",
      "Epoch 63/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0515 - accuracy: 0.9811 - val_loss: 2.1255 - val_accuracy: 0.6862\n",
      "Epoch 64/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0338 - accuracy: 0.9906 - val_loss: 2.0464 - val_accuracy: 0.7038\n",
      "Epoch 65/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0315 - accuracy: 0.9902 - val_loss: 2.1223 - val_accuracy: 0.7097\n",
      "Epoch 66/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0337 - accuracy: 0.9915 - val_loss: 2.1998 - val_accuracy: 0.6774\n",
      "Epoch 67/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0144 - accuracy: 0.9951 - val_loss: 2.1812 - val_accuracy: 0.7067\n",
      "Epoch 68/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0174 - accuracy: 0.9958 - val_loss: 2.1777 - val_accuracy: 0.7243\n",
      "Epoch 69/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0125 - accuracy: 0.9971 - val_loss: 2.1572 - val_accuracy: 0.7038\n",
      "Epoch 70/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0073 - accuracy: 0.9977 - val_loss: 2.1636 - val_accuracy: 0.7067\n",
      "Epoch 71/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0064 - accuracy: 0.9974 - val_loss: 2.1590 - val_accuracy: 0.7214\n",
      "Epoch 72/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0096 - accuracy: 0.9971 - val_loss: 2.1166 - val_accuracy: 0.7155\n",
      "Epoch 73/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0116 - accuracy: 0.9967 - val_loss: 2.2153 - val_accuracy: 0.6950\n",
      "Epoch 74/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0042 - accuracy: 0.9984 - val_loss: 2.1617 - val_accuracy: 0.7067\n",
      "Epoch 75/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0026 - accuracy: 0.9990 - val_loss: 2.2007 - val_accuracy: 0.7273\n",
      "Epoch 76/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0042 - accuracy: 0.9984 - val_loss: 2.2867 - val_accuracy: 0.7009\n",
      "Epoch 77/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0030 - accuracy: 0.9993 - val_loss: 2.2717 - val_accuracy: 0.7126\n",
      "Epoch 78/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0023 - accuracy: 0.9993 - val_loss: 2.3106 - val_accuracy: 0.7126\n",
      "Epoch 79/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0038 - accuracy: 0.9987 - val_loss: 2.3152 - val_accuracy: 0.7067\n",
      "Epoch 80/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0026 - accuracy: 0.9987 - val_loss: 2.3278 - val_accuracy: 0.7067\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# train the network\n",
    "print(\"[INFO] training model...\")\n",
    "history = model.fit(x=train_images, \n",
    "                    y=train_labels, \n",
    "                    validation_data=(test_images, test_labels), \n",
    "                    batch_size=BATCH_SIZE,\n",
    "                    epochs=EPOCHS, \n",
    "                    class_weight=classWeight)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "------------------------------------------------------------------------\n",
      "Training for fold 1 ...\n",
      "Epoch 1/80\n",
      "62/62 [==============================] - 3s 30ms/step - loss: 4.1090 - accuracy: 0.0283 - val_loss: 3.9846 - val_accuracy: 0.0616\n",
      "Epoch 2/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 3.4496 - accuracy: 0.1466 - val_loss: 3.0225 - val_accuracy: 0.2199\n",
      "Epoch 3/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 2.6281 - accuracy: 0.3073 - val_loss: 2.5281 - val_accuracy: 0.3548\n",
      "Epoch 4/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 2.0361 - accuracy: 0.4630 - val_loss: 2.1338 - val_accuracy: 0.4252\n",
      "Epoch 5/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 1.5790 - accuracy: 0.5666 - val_loss: 1.8311 - val_accuracy: 0.5279\n",
      "Epoch 6/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 1.2539 - accuracy: 0.6383 - val_loss: 1.5830 - val_accuracy: 0.5777\n",
      "Epoch 7/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 1.0528 - accuracy: 0.6898 - val_loss: 1.4818 - val_accuracy: 0.5777\n",
      "Epoch 8/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.8619 - accuracy: 0.7387 - val_loss: 1.3882 - val_accuracy: 0.6158\n",
      "Epoch 9/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.7394 - accuracy: 0.7696 - val_loss: 1.3694 - val_accuracy: 0.6246\n",
      "Epoch 10/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.6509 - accuracy: 0.7937 - val_loss: 1.2895 - val_accuracy: 0.6540\n",
      "Epoch 11/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.5439 - accuracy: 0.8254 - val_loss: 1.4169 - val_accuracy: 0.6070\n",
      "Epoch 12/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.4559 - accuracy: 0.8495 - val_loss: 1.3238 - val_accuracy: 0.6745\n",
      "Epoch 13/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.4011 - accuracy: 0.8742 - val_loss: 1.4445 - val_accuracy: 0.6422\n",
      "Epoch 14/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.3292 - accuracy: 0.8935 - val_loss: 1.5683 - val_accuracy: 0.6276\n",
      "Epoch 15/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.3216 - accuracy: 0.8957 - val_loss: 1.5567 - val_accuracy: 0.6041\n",
      "Epoch 16/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.2594 - accuracy: 0.9195 - val_loss: 1.5756 - val_accuracy: 0.6246\n",
      "Epoch 17/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.2395 - accuracy: 0.9215 - val_loss: 1.6202 - val_accuracy: 0.6569\n",
      "Epoch 18/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.2460 - accuracy: 0.9202 - val_loss: 1.5781 - val_accuracy: 0.6452\n",
      "Epoch 19/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1743 - accuracy: 0.9446 - val_loss: 1.5596 - val_accuracy: 0.6657\n",
      "Epoch 20/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.1538 - accuracy: 0.9514 - val_loss: 1.7793 - val_accuracy: 0.6569\n",
      "Epoch 21/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.1143 - accuracy: 0.9671 - val_loss: 1.6882 - val_accuracy: 0.6745\n",
      "Epoch 22/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0964 - accuracy: 0.9743 - val_loss: 1.8612 - val_accuracy: 0.6393\n",
      "Epoch 23/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0946 - accuracy: 0.9756 - val_loss: 1.7344 - val_accuracy: 0.6393\n",
      "Epoch 24/80\n",
      "62/62 [==============================] - 1s 18ms/step - loss: 0.0764 - accuracy: 0.9759 - val_loss: 1.8458 - val_accuracy: 0.6569\n",
      "Epoch 25/80\n",
      "62/62 [==============================] - 1s 17ms/step - loss: 0.0875 - accuracy: 0.9749 - val_loss: 1.9101 - val_accuracy: 0.6158\n",
      "Epoch 26/80\n",
      "62/62 [==============================] - 1s 18ms/step - loss: 0.0638 - accuracy: 0.9811 - val_loss: 1.8817 - val_accuracy: 0.6510\n",
      "Epoch 27/80\n",
      "62/62 [==============================] - 1s 18ms/step - loss: 0.0552 - accuracy: 0.9834 - val_loss: 1.9406 - val_accuracy: 0.6804\n",
      "Epoch 28/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0737 - accuracy: 0.9821 - val_loss: 2.0013 - val_accuracy: 0.6598\n",
      "Epoch 29/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0518 - accuracy: 0.9834 - val_loss: 2.1261 - val_accuracy: 0.6452\n",
      "Epoch 30/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0652 - accuracy: 0.9788 - val_loss: 2.0143 - val_accuracy: 0.6422\n",
      "Epoch 31/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0499 - accuracy: 0.9883 - val_loss: 2.0202 - val_accuracy: 0.6569\n",
      "Epoch 32/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0390 - accuracy: 0.9899 - val_loss: 2.1174 - val_accuracy: 0.6422\n",
      "Epoch 33/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0271 - accuracy: 0.9928 - val_loss: 2.1187 - val_accuracy: 0.6686\n",
      "Epoch 34/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0209 - accuracy: 0.9951 - val_loss: 2.1488 - val_accuracy: 0.6628\n",
      "Epoch 35/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0276 - accuracy: 0.9925 - val_loss: 2.1829 - val_accuracy: 0.6628\n",
      "Epoch 36/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0193 - accuracy: 0.9948 - val_loss: 2.2642 - val_accuracy: 0.6716\n",
      "Epoch 37/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0203 - accuracy: 0.9932 - val_loss: 2.2034 - val_accuracy: 0.6686\n",
      "Epoch 38/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0278 - accuracy: 0.9915 - val_loss: 2.2266 - val_accuracy: 0.6657\n",
      "Epoch 39/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0394 - accuracy: 0.9889 - val_loss: 2.1084 - val_accuracy: 0.6657\n",
      "Epoch 40/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.1516 - accuracy: 0.9531 - val_loss: 2.4824 - val_accuracy: 0.6100\n",
      "Epoch 41/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.2175 - accuracy: 0.9260 - val_loss: 2.2106 - val_accuracy: 0.6041\n",
      "Epoch 42/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.1473 - accuracy: 0.9505 - val_loss: 2.2446 - val_accuracy: 0.6422\n",
      "Epoch 43/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0707 - accuracy: 0.9785 - val_loss: 2.2119 - val_accuracy: 0.6129\n",
      "Epoch 44/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0458 - accuracy: 0.9866 - val_loss: 2.1403 - val_accuracy: 0.6393\n",
      "Epoch 45/80\n",
      "62/62 [==============================] - 2s 24ms/step - loss: 0.0258 - accuracy: 0.9919 - val_loss: 2.3236 - val_accuracy: 0.6393\n",
      "Epoch 46/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0209 - accuracy: 0.9922 - val_loss: 2.3172 - val_accuracy: 0.6569\n",
      "Epoch 47/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0195 - accuracy: 0.9935 - val_loss: 2.3468 - val_accuracy: 0.6510\n",
      "Epoch 48/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0202 - accuracy: 0.9932 - val_loss: 2.3622 - val_accuracy: 0.6540\n",
      "Epoch 49/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0253 - accuracy: 0.9938 - val_loss: 2.2941 - val_accuracy: 0.6569\n",
      "Epoch 50/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0998 - accuracy: 0.9671 - val_loss: 2.4474 - val_accuracy: 0.6305\n",
      "Epoch 51/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.1169 - accuracy: 0.9583 - val_loss: 2.1563 - val_accuracy: 0.6393\n",
      "Epoch 52/80\n",
      "62/62 [==============================] - 2s 26ms/step - loss: 0.0904 - accuracy: 0.9710 - val_loss: 2.1488 - val_accuracy: 0.6510\n",
      "Epoch 53/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0526 - accuracy: 0.9834 - val_loss: 2.2409 - val_accuracy: 0.6657\n",
      "Epoch 54/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0336 - accuracy: 0.9902 - val_loss: 2.2674 - val_accuracy: 0.6540\n",
      "Epoch 55/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0148 - accuracy: 0.9951 - val_loss: 2.3492 - val_accuracy: 0.6364\n",
      "Epoch 56/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0126 - accuracy: 0.9964 - val_loss: 2.3339 - val_accuracy: 0.6540\n",
      "Epoch 57/80\n",
      "62/62 [==============================] - 2s 26ms/step - loss: 0.0068 - accuracy: 0.9980 - val_loss: 2.3669 - val_accuracy: 0.6628\n",
      "Epoch 58/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0121 - accuracy: 0.9948 - val_loss: 2.4540 - val_accuracy: 0.6569\n",
      "Epoch 59/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0155 - accuracy: 0.9951 - val_loss: 2.3950 - val_accuracy: 0.6452\n",
      "Epoch 60/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0105 - accuracy: 0.9967 - val_loss: 2.3566 - val_accuracy: 0.6628\n",
      "Epoch 61/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0060 - accuracy: 0.9984 - val_loss: 2.4758 - val_accuracy: 0.6569\n",
      "Epoch 62/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0067 - accuracy: 0.9980 - val_loss: 2.4640 - val_accuracy: 0.6510\n",
      "Epoch 63/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0087 - accuracy: 0.9961 - val_loss: 2.5688 - val_accuracy: 0.6569\n",
      "Epoch 64/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0059 - accuracy: 0.9990 - val_loss: 2.5526 - val_accuracy: 0.6540\n",
      "Epoch 65/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0064 - accuracy: 0.9971 - val_loss: 2.6648 - val_accuracy: 0.6481\n",
      "Epoch 66/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0076 - accuracy: 0.9974 - val_loss: 2.5262 - val_accuracy: 0.6481\n",
      "Epoch 67/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0040 - accuracy: 0.9987 - val_loss: 2.5773 - val_accuracy: 0.6598\n",
      "Epoch 68/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0047 - accuracy: 0.9980 - val_loss: 2.6086 - val_accuracy: 0.6598\n",
      "Epoch 69/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0050 - accuracy: 0.9974 - val_loss: 2.6261 - val_accuracy: 0.6569\n",
      "Epoch 70/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0064 - accuracy: 0.9974 - val_loss: 2.7467 - val_accuracy: 0.6510\n",
      "Epoch 71/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0073 - accuracy: 0.9974 - val_loss: 2.7125 - val_accuracy: 0.6452\n",
      "Epoch 72/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0050 - accuracy: 0.9984 - val_loss: 2.6094 - val_accuracy: 0.6598\n",
      "Epoch 73/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0033 - accuracy: 0.9993 - val_loss: 2.6774 - val_accuracy: 0.6628\n",
      "Epoch 74/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0060 - accuracy: 0.9974 - val_loss: 2.7463 - val_accuracy: 0.6481\n",
      "Epoch 75/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0053 - accuracy: 0.9980 - val_loss: 2.7188 - val_accuracy: 0.6569\n",
      "Epoch 76/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0121 - accuracy: 0.9967 - val_loss: 2.7309 - val_accuracy: 0.6510\n",
      "Epoch 77/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0080 - accuracy: 0.9967 - val_loss: 2.5759 - val_accuracy: 0.6481\n",
      "Epoch 78/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0031 - accuracy: 0.9990 - val_loss: 2.7343 - val_accuracy: 0.6598\n",
      "Epoch 79/80\n",
      "62/62 [==============================] - 2s 26ms/step - loss: 0.0055 - accuracy: 0.9974 - val_loss: 2.9240 - val_accuracy: 0.6393\n",
      "Epoch 80/80\n",
      "62/62 [==============================] - 2s 26ms/step - loss: 0.0149 - accuracy: 0.9961 - val_loss: 2.8269 - val_accuracy: 0.6364\n",
      "Score for fold 1: loss of 2.8269495964050293; accuracy of 63.63636255264282%\n",
      "------------------------------------------------------------------------\n",
      "Training for fold 2 ...\n",
      "Epoch 1/80\n",
      "62/62 [==============================] - 3s 32ms/step - loss: 4.1196 - accuracy: 0.0212 - val_loss: 4.0523 - val_accuracy: 0.0381\n",
      "Epoch 2/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 3.5801 - accuracy: 0.1274 - val_loss: 3.2976 - val_accuracy: 0.1554\n",
      "Epoch 3/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 2.6998 - accuracy: 0.3095 - val_loss: 2.6077 - val_accuracy: 0.3284\n",
      "Epoch 4/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 2.0276 - accuracy: 0.4646 - val_loss: 2.1007 - val_accuracy: 0.4692\n",
      "Epoch 5/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 1.5628 - accuracy: 0.5696 - val_loss: 1.7295 - val_accuracy: 0.5572\n",
      "Epoch 6/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 1.2572 - accuracy: 0.6429 - val_loss: 1.5298 - val_accuracy: 0.5748\n",
      "Epoch 7/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 1.0024 - accuracy: 0.7116 - val_loss: 1.4188 - val_accuracy: 0.6217\n",
      "Epoch 8/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.8527 - accuracy: 0.7331 - val_loss: 1.4124 - val_accuracy: 0.6188\n",
      "Epoch 9/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.7340 - accuracy: 0.7748 - val_loss: 1.4814 - val_accuracy: 0.5953\n",
      "Epoch 10/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.6120 - accuracy: 0.8071 - val_loss: 1.4258 - val_accuracy: 0.6364\n",
      "Epoch 11/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.5608 - accuracy: 0.8260 - val_loss: 1.3703 - val_accuracy: 0.6774\n",
      "Epoch 12/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.4628 - accuracy: 0.8511 - val_loss: 1.3545 - val_accuracy: 0.6833\n",
      "Epoch 13/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.4080 - accuracy: 0.8697 - val_loss: 1.3929 - val_accuracy: 0.6628\n",
      "Epoch 14/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.3363 - accuracy: 0.8921 - val_loss: 1.4738 - val_accuracy: 0.6804\n",
      "Epoch 15/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.3132 - accuracy: 0.9000 - val_loss: 1.4183 - val_accuracy: 0.6686\n",
      "Epoch 16/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.2528 - accuracy: 0.9225 - val_loss: 1.4681 - val_accuracy: 0.6862\n",
      "Epoch 17/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1976 - accuracy: 0.9368 - val_loss: 1.5874 - val_accuracy: 0.6950\n",
      "Epoch 18/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.1966 - accuracy: 0.9394 - val_loss: 1.4784 - val_accuracy: 0.6891\n",
      "Epoch 19/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1672 - accuracy: 0.9446 - val_loss: 1.6273 - val_accuracy: 0.6481\n",
      "Epoch 20/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1697 - accuracy: 0.9482 - val_loss: 1.5722 - val_accuracy: 0.6921\n",
      "Epoch 21/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.1342 - accuracy: 0.9593 - val_loss: 1.6102 - val_accuracy: 0.7126\n",
      "Epoch 22/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.1189 - accuracy: 0.9616 - val_loss: 1.7595 - val_accuracy: 0.6745\n",
      "Epoch 23/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.1313 - accuracy: 0.9557 - val_loss: 1.6686 - val_accuracy: 0.7009\n",
      "Epoch 24/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0879 - accuracy: 0.9769 - val_loss: 1.7893 - val_accuracy: 0.6774\n",
      "Epoch 25/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0929 - accuracy: 0.9733 - val_loss: 1.7226 - val_accuracy: 0.6569\n",
      "Epoch 26/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0851 - accuracy: 0.9752 - val_loss: 1.8638 - val_accuracy: 0.6716\n",
      "Epoch 27/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0835 - accuracy: 0.9739 - val_loss: 1.7895 - val_accuracy: 0.6833\n",
      "Epoch 28/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0668 - accuracy: 0.9788 - val_loss: 1.8133 - val_accuracy: 0.7067\n",
      "Epoch 29/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0426 - accuracy: 0.9866 - val_loss: 1.8173 - val_accuracy: 0.6950\n",
      "Epoch 30/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0692 - accuracy: 0.9775 - val_loss: 1.8854 - val_accuracy: 0.6833\n",
      "Epoch 31/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0655 - accuracy: 0.9762 - val_loss: 1.8763 - val_accuracy: 0.6979\n",
      "Epoch 32/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0418 - accuracy: 0.9863 - val_loss: 2.0026 - val_accuracy: 0.6891\n",
      "Epoch 33/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0560 - accuracy: 0.9837 - val_loss: 1.8753 - val_accuracy: 0.7067\n",
      "Epoch 34/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0429 - accuracy: 0.9850 - val_loss: 1.9350 - val_accuracy: 0.6774\n",
      "Epoch 35/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0504 - accuracy: 0.9879 - val_loss: 1.9904 - val_accuracy: 0.6950\n",
      "Epoch 36/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0426 - accuracy: 0.9896 - val_loss: 1.9476 - val_accuracy: 0.7067\n",
      "Epoch 37/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0356 - accuracy: 0.9892 - val_loss: 2.1378 - val_accuracy: 0.6745\n",
      "Epoch 38/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0324 - accuracy: 0.9886 - val_loss: 2.1462 - val_accuracy: 0.6950\n",
      "Epoch 39/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0358 - accuracy: 0.9896 - val_loss: 2.0163 - val_accuracy: 0.7126\n",
      "Epoch 40/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0328 - accuracy: 0.9922 - val_loss: 2.0598 - val_accuracy: 0.6979\n",
      "Epoch 41/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0269 - accuracy: 0.9925 - val_loss: 2.1684 - val_accuracy: 0.6979\n",
      "Epoch 42/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0538 - accuracy: 0.9853 - val_loss: 2.1001 - val_accuracy: 0.6804\n",
      "Epoch 43/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1490 - accuracy: 0.9518 - val_loss: 2.0953 - val_accuracy: 0.6686\n",
      "Epoch 44/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.1028 - accuracy: 0.9616 - val_loss: 2.2777 - val_accuracy: 0.6569\n",
      "Epoch 45/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0985 - accuracy: 0.9690 - val_loss: 2.0599 - val_accuracy: 0.6745\n",
      "Epoch 46/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0525 - accuracy: 0.9814 - val_loss: 2.1140 - val_accuracy: 0.6804\n",
      "Epoch 47/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0364 - accuracy: 0.9892 - val_loss: 2.0247 - val_accuracy: 0.6804\n",
      "Epoch 48/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0197 - accuracy: 0.9941 - val_loss: 2.0582 - val_accuracy: 0.6979\n",
      "Epoch 49/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0136 - accuracy: 0.9951 - val_loss: 2.1951 - val_accuracy: 0.6979\n",
      "Epoch 50/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0159 - accuracy: 0.9948 - val_loss: 2.2076 - val_accuracy: 0.6891\n",
      "Epoch 51/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0091 - accuracy: 0.9977 - val_loss: 2.2009 - val_accuracy: 0.6950\n",
      "Epoch 52/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0090 - accuracy: 0.9971 - val_loss: 2.2378 - val_accuracy: 0.6950\n",
      "Epoch 53/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0110 - accuracy: 0.9958 - val_loss: 2.2394 - val_accuracy: 0.7038\n",
      "Epoch 54/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0136 - accuracy: 0.9948 - val_loss: 2.2383 - val_accuracy: 0.7038\n",
      "Epoch 55/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0080 - accuracy: 0.9977 - val_loss: 2.2580 - val_accuracy: 0.6979\n",
      "Epoch 56/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0059 - accuracy: 0.9977 - val_loss: 2.2853 - val_accuracy: 0.6891\n",
      "Epoch 57/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0158 - accuracy: 0.9945 - val_loss: 2.3802 - val_accuracy: 0.6862\n",
      "Epoch 58/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0074 - accuracy: 0.9977 - val_loss: 2.3188 - val_accuracy: 0.6950\n",
      "Epoch 59/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0068 - accuracy: 0.9984 - val_loss: 2.3324 - val_accuracy: 0.6950\n",
      "Epoch 60/80\n",
      "62/62 [==============================] - 2s 28ms/step - loss: 0.0043 - accuracy: 0.9990 - val_loss: 2.3576 - val_accuracy: 0.7038\n",
      "Epoch 61/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0075 - accuracy: 0.9964 - val_loss: 2.3535 - val_accuracy: 0.6804\n",
      "Epoch 62/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0061 - accuracy: 0.9984 - val_loss: 2.3606 - val_accuracy: 0.6921\n",
      "Epoch 63/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0064 - accuracy: 0.9980 - val_loss: 2.3529 - val_accuracy: 0.6921\n",
      "Epoch 64/80\n",
      "62/62 [==============================] - 2s 29ms/step - loss: 0.0083 - accuracy: 0.9971 - val_loss: 2.3187 - val_accuracy: 0.6862\n",
      "Epoch 65/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0034 - accuracy: 0.9993 - val_loss: 2.3468 - val_accuracy: 0.6921\n",
      "Epoch 66/80\n",
      "62/62 [==============================] - 2s 25ms/step - loss: 0.0037 - accuracy: 0.9987 - val_loss: 2.3879 - val_accuracy: 0.6950\n",
      "Epoch 67/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0058 - accuracy: 0.9980 - val_loss: 2.3879 - val_accuracy: 0.6950\n",
      "Epoch 68/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0036 - accuracy: 0.9987 - val_loss: 2.4120 - val_accuracy: 0.6950\n",
      "Epoch 69/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0032 - accuracy: 0.9993 - val_loss: 2.4616 - val_accuracy: 0.6979\n",
      "Epoch 70/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0029 - accuracy: 0.9993 - val_loss: 2.4736 - val_accuracy: 0.6891\n",
      "Epoch 71/80\n",
      "62/62 [==============================] - 2s 24ms/step - loss: 0.0114 - accuracy: 0.9967 - val_loss: 2.4131 - val_accuracy: 0.6833\n",
      "Epoch 72/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.0116 - accuracy: 0.9964 - val_loss: 2.3520 - val_accuracy: 0.6950\n",
      "Epoch 73/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0079 - accuracy: 0.9980 - val_loss: 2.3576 - val_accuracy: 0.7038\n",
      "Epoch 74/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0039 - accuracy: 0.9990 - val_loss: 2.3837 - val_accuracy: 0.7009\n",
      "Epoch 75/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0043 - accuracy: 0.9984 - val_loss: 2.4302 - val_accuracy: 0.7009\n",
      "Epoch 76/80\n",
      "62/62 [==============================] - 1s 19ms/step - loss: 0.0035 - accuracy: 0.9987 - val_loss: 2.4356 - val_accuracy: 0.6862\n",
      "Epoch 77/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 0.0440 - accuracy: 0.9896 - val_loss: 2.7332 - val_accuracy: 0.6364\n",
      "Epoch 78/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.4275 - accuracy: 0.8713 - val_loss: 1.7967 - val_accuracy: 0.6452\n",
      "Epoch 79/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.2304 - accuracy: 0.9221 - val_loss: 1.8456 - val_accuracy: 0.6598\n",
      "Epoch 80/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.0906 - accuracy: 0.9733 - val_loss: 1.9607 - val_accuracy: 0.6862\n",
      "Score for fold 2: loss of 1.9607480764389038; accuracy of 68.62170100212097%\n",
      "------------------------------------------------------------------------\n",
      "Training for fold 3 ...\n",
      "Epoch 1/80\n",
      "62/62 [==============================] - 3s 31ms/step - loss: 4.1058 - accuracy: 0.0342 - val_loss: 4.0144 - val_accuracy: 0.0381\n",
      "Epoch 2/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 3.3727 - accuracy: 0.1541 - val_loss: 2.9004 - val_accuracy: 0.2522\n",
      "Epoch 3/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 2.2790 - accuracy: 0.3874 - val_loss: 2.1670 - val_accuracy: 0.4194\n",
      "Epoch 4/80\n",
      "62/62 [==============================] - 1s 21ms/step - loss: 1.6338 - accuracy: 0.5455 - val_loss: 1.7109 - val_accuracy: 0.5367\n",
      "Epoch 5/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 1.2646 - accuracy: 0.6445 - val_loss: 1.5148 - val_accuracy: 0.5806\n",
      "Epoch 6/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 1.0503 - accuracy: 0.6895 - val_loss: 1.3928 - val_accuracy: 0.6041\n",
      "Epoch 7/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.8583 - accuracy: 0.7374 - val_loss: 1.3870 - val_accuracy: 0.5953\n",
      "Epoch 8/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.7073 - accuracy: 0.7794 - val_loss: 1.3297 - val_accuracy: 0.6334\n",
      "Epoch 9/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.6045 - accuracy: 0.8166 - val_loss: 1.2579 - val_accuracy: 0.6686\n",
      "Epoch 10/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.4967 - accuracy: 0.8426 - val_loss: 1.2434 - val_accuracy: 0.6452\n",
      "Epoch 11/80\n",
      "62/62 [==============================] - 1s 20ms/step - loss: 0.4295 - accuracy: 0.8589 - val_loss: 1.3035 - val_accuracy: 0.6481\n",
      "Epoch 12/80\n",
      "62/62 [==============================] - 2s 26ms/step - loss: 0.3649 - accuracy: 0.8843 - val_loss: 1.3667 - val_accuracy: 0.6716\n",
      "Epoch 13/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.3028 - accuracy: 0.9104 - val_loss: 1.4333 - val_accuracy: 0.6598\n",
      "Epoch 14/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.2520 - accuracy: 0.9225 - val_loss: 1.4903 - val_accuracy: 0.6979\n",
      "Epoch 15/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.1981 - accuracy: 0.9381 - val_loss: 1.3998 - val_accuracy: 0.6891\n",
      "Epoch 16/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.1895 - accuracy: 0.9440 - val_loss: 1.5507 - val_accuracy: 0.6657\n",
      "Epoch 17/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.1510 - accuracy: 0.9514 - val_loss: 1.5462 - val_accuracy: 0.6686\n",
      "Epoch 18/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.1564 - accuracy: 0.9518 - val_loss: 1.5418 - val_accuracy: 0.6716\n",
      "Epoch 19/80\n",
      "62/62 [==============================] - 2s 27ms/step - loss: 0.1160 - accuracy: 0.9658 - val_loss: 1.7675 - val_accuracy: 0.6657\n",
      "Epoch 20/80\n",
      "62/62 [==============================] - 1s 24ms/step - loss: 0.1052 - accuracy: 0.9671 - val_loss: 1.6862 - val_accuracy: 0.6716\n",
      "Epoch 21/80\n",
      "62/62 [==============================] - 2s 27ms/step - loss: 0.1174 - accuracy: 0.9684 - val_loss: 1.6607 - val_accuracy: 0.6921\n",
      "Epoch 22/80\n",
      "62/62 [==============================] - 1s 23ms/step - loss: 0.0856 - accuracy: 0.9756 - val_loss: 1.8643 - val_accuracy: 0.6862\n",
      "Epoch 23/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0727 - accuracy: 0.9762 - val_loss: 1.6741 - val_accuracy: 0.6891\n",
      "Epoch 24/80\n",
      "62/62 [==============================] - 1s 22ms/step - loss: 0.0729 - accuracy: 0.9756 - val_loss: 1.7043 - val_accuracy: 0.6745\n",
      "Epoch 25/80\n",