Covid19_Detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Covid19-Detection.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "mount_file_id": "https://github.com/farkoo/Covid19-Detection-from-CTChest/blob/master/Covid19_Detection.ipynb",
      "authorship_tag": "ABX9TyN4xr4lMmQJFFPYMcRjeAAN",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/farkoo/Covid19-Detection-from-ChestXRayImages/blob/master/Covid19_Detection.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DaaiF9bHIDZh",
        "outputId": "32a9c50e-131f-4e19-d45d-4a98656cf821"
      },
      "source": [
        "%cd drive/MyDrive"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/MyDrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3iiciEZLILWS",
        "outputId": "37916697-6993-4e10-ce97-b65c3c1577b3"
      },
      "source": [
        "!git clone \"https://github.com/farkoo/Covid19-Detection-from-ChestXRayImages.git\""
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into 'Covid19-Detection-from-ChestXRayImages'...\n",
            "remote: Enumerating objects: 47, done.\u001b[K\n",
            "remote: Counting objects: 100% (47/47), done.\u001b[K\n",
            "remote: Compressing objects: 100% (42/42), done.\u001b[K\n",
            "remote: Total 47 (delta 15), reused 0 (delta 0), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (47/47), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hWgrRZJoIRBR",
        "outputId": "44073213-6875-4a83-865b-a13453d63d21"
      },
      "source": [
        "%cd 'Covid19-Detection-from-ChestXRayImages'"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/My Drive/Covid19-Detection-from-ChestXRayImages\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wjdi4yB-IhoX",
        "outputId": "eb6691cf-e6f8-4f03-a1c8-5b0396cbbfd2"
      },
      "source": [
        "!7z x dataset.7z -odataset"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "7-Zip [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21\n",
            "p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs Intel(R) Xeon(R) CPU @ 2.20GHz (406F0),ASM,AES-NI)\n",
            "\n",
            "Scanning the drive for archives:\n",
            "  0M Scan\b\b\b\b\b\b\b\b\b         \b\b\b\b\b\b\b\b\b1 file, 12417162 bytes (12 MiB)\n",
            "\n",
            "Extracting archive: dataset.7z\n",
            "--\n",
            "Path = dataset.7z\n",
            "Type = 7z\n",
            "Physical Size = 12417162\n",
            "Headers Size = 1053\n",
            "Method = LZMA:23\n",
            "Solid = +\n",
            "Blocks = 1\n",
            "\n",
            "  0%\b\b\b\b    \b\b\b\b 25% - dataset/covid/1.jpg\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b                          \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b 42% 16 - dataset/covid/17.jpeg\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b                               \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b 59% 20 - dataset/covid/21.jpeg\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b                               \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b 92% 26 - dataset/normal/2.jpeg\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b                               \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b 98% 48 - dataset/normal/18.jpeg\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b                                \b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\bEverything is Ok\n",
            "\n",
            "Folders: 3\n",
            "Files: 50\n",
            "Size:       12715586\n",
            "Compressed: 12417162\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xAwwPCVTIxPS"
      },
      "source": [
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras.applications import VGG16\n",
        "from tensorflow.keras.layers import AveragePooling2D, Dropout, Flatten, Dense, Input\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "from sklearn.preprocessing import LabelBinarizer\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import classification_report, confusion_matrix\n",
        "from imutils import paths\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import argparse\n",
        "import cv2\n",
        "import os\n",
        "import random"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EWbjYUErLrzd",
        "outputId": "0894688b-e756-43f3-c972-dfce2f3892ae"
      },
      "source": [
        "%cd dataset"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/My Drive/Covid19-Detection-from-ChestXRayImages/dataset/dataset\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zx4OR5DoaBqx",
        "outputId": "f37826b0-aa56-471f-f915-e64a3ff3096f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "%cd .."
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/My Drive/Covid19-Detection-from-ChestXRayImages/dataset\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KcbyPvrGIm-s"
      },
      "source": [
        "# Hyper parameters\n",
        "learning_rate = 1e-3\n",
        "epochs = 25\n",
        "batch_size = 8\n",
        "dir = 'dataset'\n",
        "width = 224\n",
        "height = 224\n",
        "classes = 2"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jtJFs3-hJVBV"
      },
      "source": [
        "def load_imgs_from_dir(dir, width, height):\n",
        "  img_data = []\n",
        "  img_label = []\n",
        "\n",
        "  categories = os.listdir(dir)\n",
        "  for category in categories:\n",
        "    for file in os.listdir(os.path.join(dir,category)):\n",
        "      img_path = os.path.join(dir,category,file)\n",
        "      image = cv2.imread(img_path)\n",
        "      image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
        "      image = cv2.resize(image,(width, height))\n",
        "      img_data.append(image)\n",
        "      img_label.append(category)\n",
        "  img_data = np.asarray(img_data)\n",
        "  img_label = np.asarray(img_label)\n",
        "  zip_list = list(zip(img_data, img_label))\n",
        "  random.shuffle(zip_list)\n",
        "  img_data, img_label = zip(*zip_list)\n",
        "  return img_data, img_label"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "64JQJ8u5LETD"
      },
      "source": [
        "data, label = load_imgs_from_dir(dir, width, height)"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O4K_chq2OyA5"
      },
      "source": [
        "lb = LabelBinarizer()\n",
        "one_hot = lb.fit_transform(label)\n",
        "one_hot = to_categorical(one_hot)"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5iTWUfd6QYYV"
      },
      "source": [
        "train_X, test_X, train_Y, test_Y = train_test_split(data, one_hot, test_size = 0.10,\n",
        "                                                    stratify = one_hot, random_state = 42)"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P3rOO-GzX_7k"
      },
      "source": [
        "train_X, valid_X, train_Y, valid_Y = train_test_split(train_X, train_Y, test_size = 0.10,\n",
        "                                                    stratify = train_Y, random_state = 13)"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4avGcDaMbHDk"
      },
      "source": [
        "train_X = np.asarray(train_X)\n",
        "test_X = np.asarray(test_X)\n",
        "valid_X = np.asarray(valid_X)\n",
        "valid_Y = np.asarray(valid_Y)\n",
        "train_Y = np.asarray(train_Y)\n",
        "test_Y = np.asarray(test_Y)"
      ],
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c5KlpiMjTXN3",
        "outputId": "6514124f-1f55-4592-fa6b-6796cacdf161",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "base_model = VGG16(weights = \"imagenet\", include_top = False,\n",
        "                   input_tensor = Input(shape = (224, 224, 3)))"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BA8eSNRyWQmw"
      },
      "source": [
        "head_model = base_model.output\n",
        "head_model = AveragePooling2D(pool_size = (4, 4))(head_model)\n",
        "head_model = Flatten(name = \"flatten\")(head_model)\n",
        "head_model = Dense(64, activation = \"relu\")(head_model)\n",
        "head_model = Dropout(0.5)(head_model)\n",
        "head_model = Dense(2, activation = \"softmax\")(head_model)"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PFOOZ9nKW7nd"
      },
      "source": [
        "model = Model(inputs = base_model.input, outputs = head_model)"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iYsZd6ThXDe7"
      },
      "source": [
        "for layer in base_model.layers:\n",
        "  layer.trainable = False"
      ],
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "f1loT6TdXVzw"
      },
      "source": [
        "opt = Adam(learning_rate = learning_rate, decay = learning_rate/epochs)"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Kn-h2n2XnAw"
      },
      "source": [
        "model.compile(loss = \"binary_crossentropy\", optimizer = opt, metrics = [\"accuracy\"])"
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cbGmo1ErX0el",
        "outputId": "b5627d32-c14a-439d-c6d7-330865779ee0"
      },
      "source": [
        "H = model.fit(\n",
        "    x = train_X,\n",
        "    y = train_Y,\n",
        "    steps_per_epoch = len(train_X)//batch_size,\n",
        "    validation_data = (valid_X, valid_Y),\n",
        "    validation_steps = len(valid_X)//batch_size,\n",
        "    epochs = epochs\n",
        "              )"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/25\n",
            "5/5 [==============================] - 32s 53ms/step - loss: 3.4919 - accuracy: 0.5250\n",
            "Epoch 2/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 2.9466 - accuracy: 0.6250\n",
            "Epoch 3/25\n",
            "5/5 [==============================] - 0s 47ms/step - loss: 1.5844 - accuracy: 0.6250\n",
            "Epoch 4/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 1.5972 - accuracy: 0.6750\n",
            "Epoch 5/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 1.0106 - accuracy: 0.8750\n",
            "Epoch 6/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.4193 - accuracy: 0.8750\n",
            "Epoch 7/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.5952 - accuracy: 0.9500\n",
            "Epoch 8/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.6369 - accuracy: 0.8750\n",
            "Epoch 9/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.2422 - accuracy: 0.9500\n",
            "Epoch 10/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.1695 - accuracy: 1.0000\n",
            "Epoch 11/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.2294 - accuracy: 0.9250\n",
            "Epoch 12/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.1380 - accuracy: 0.9750\n",
            "Epoch 13/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.1405 - accuracy: 0.9500\n",
            "Epoch 14/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.2700 - accuracy: 0.9500\n",
            "Epoch 15/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.1738 - accuracy: 0.9750\n",
            "Epoch 16/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.1303 - accuracy: 0.9750\n",
            "Epoch 17/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.1348 - accuracy: 0.9500\n",
            "Epoch 18/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.0730 - accuracy: 1.0000\n",
            "Epoch 19/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.0551 - accuracy: 1.0000\n",
            "Epoch 20/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.0468 - accuracy: 1.0000\n",
            "Epoch 21/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.1167 - accuracy: 0.9750\n",
            "Epoch 22/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.0575 - accuracy: 1.0000\n",
            "Epoch 23/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.1415 - accuracy: 0.9500\n",
            "Epoch 24/25\n",
            "5/5 [==============================] - 0s 49ms/step - loss: 0.0739 - accuracy: 0.9750\n",
            "Epoch 25/25\n",
            "5/5 [==============================] - 0s 48ms/step - loss: 0.0563 - accuracy: 1.0000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "c5VmixdTeYv-",
        "outputId": "bcf9d0de-15ec-49ac-d3e3-c7b9c928dfa5"
      },
      "source": [
        "predIdxs = model.predict(test_X, batch_size=batch_size)\n",
        "predIdxs = np.argmax(predIdxs, axis=1)\n",
        "print(classification_report(test_Y.argmax(axis=1), predIdxs, target_names=lb.classes_))"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "       covid       1.00      1.00      1.00         2\n",
            "      normal       1.00      1.00      1.00         3\n",
            "\n",
            "    accuracy                           1.00         5\n",
            "   macro avg       1.00      1.00      1.00         5\n",
            "weighted avg       1.00      1.00      1.00         5\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WgCOFj8dk8vs",
        "outputId": "3fcf0040-7842-4fae-988a-87e637c38360"
      },
      "source": [
        "cm = confusion_matrix(test_Y.argmax(axis=1), predIdxs)\n",
        "total = sum(sum(cm))\n",
        "acc = (cm[0, 0] + cm[1, 1]) / total\n",
        "sensitivity = cm[0, 0] / (cm[0, 0] + cm[0, 1])\n",
        "specificity = cm[1, 1] / (cm[1, 0] + cm[1, 1])\n",
        "\n",
        "print(\"acc: {:.4f}\".format(acc))\n",
        "print(\"sensitivity: {:.4f}\".format(sensitivity))\n",
        "print(\"specificity: {:.4f}\".format(specificity))"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "acc: 1.0000\n",
            "sensitivity: 1.0000\n",
            "specificity: 1.0000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 316
        },
        "id": "tWdwPBcolDeq",
        "outputId": "5775b9b9-d384-4c0f-d866-49e68bf57dca"
      },
      "source": [
        "N = epochs\n",
        "plt.style.use(\"ggplot\")\n",
        "plt.figure()\n",
        "plt.plot(np.arange(0, N), H.history[\"loss\"], label=\"train_loss\")\n",
        "plt.plot(np.arange(0, N), H.history[\"accuracy\"], label=\"train_acc\")\n",
        "plt.title(\"Training Loss and Accuracy on COVID-19 Dataset\")\n",
        "plt.xlabel(\"Epoch #\")\n",
        "plt.ylabel(\"Loss/Accuracy\")\n",
        "plt.legend(loc=\"lower left\")"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f9a2d9fc590>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 42
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAEaCAYAAAAG87ApAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeVxU5f4H8M+ZfYYdhn0xRQWFXBDFXRI01PRamW2apmVlt67V9V5Nixb1mqaZpmWuZd7ylv66t9JScsF9xX1BRJFkHRZBhm1mvr8/RiZHthkEBpzv+/XipXOW53yfw3C+55znOecRiIjAGGOMWUhk6wAYY4y1Lpw4GGOMWYUTB2OMMatw4mCMMWYVThyMMcaswomDMcaYVewucezevRuCIOCPP/6waj1BEPDNN980UVT2Kzo6Gi+88IKtw2CMWaHFJg5BEOr8eeCBBxpUbt++fZGZmQk/Pz+r1svMzMSYMWMatE1rcZKq2SuvvAKxWIzly5fbOpT7nlarxZw5c9ClSxeoVCq4u7sjKioKy5Ytg1arNS1XXFyMWbNmISQkBHK5HG5uboiLi8OuXbtMy/ztb3+Dn58fdDpdjdsKCwvDuHHjAAATJ05EbGysad57771n+psXi8Vwc3NDr1698O6770Kj0dRbj7KyMjz//PPo3r07ZDIZ2rdvX+Nye/fuRXR0NFxdXeHu7o7nnnsOeXl5dZa9fv16U2wikQguLi7o1q0b3nzzTaSlpdUb291iY2MxceJEq9drDO3bt8d7771n8fItNnFkZmaafjZv3gwAOHHihGna0aNHzZavqKiwqFyZTAYfHx+IRNZV3cfHBwqFwqp1WOMpKSnBxo0b8fbbb2PVqlW2DgeA5d+51qaoqAj9+vXDsmXL8Oqrr+LAgQM4fvw4/v73v+M///kPtm/fbrbcpk2bMGfOHCQnJ2PXrl3o2LEjYmNjsXbtWgDAlClTkJmZiV9++aXatvbv34/z589jypQptcbzwAMPIDMzE3/88QcOHDiAV199FZs3b0Z4eDguXbpUZ130ej1kMhmmTJmCp556qsZlzp49iyFDhqBXr144cuQItm3bhsuXL2P06NGo7/losViMzMxMZGRk4NixY5g1axYOHTqE8PBw7Nu3r851WzVqBXbt2kUAKD093TQNAH366af09NNPk7OzM40dO5aIiN5++20KDQ0lpVJJAQEB9NJLL1FhYWGtZVV93r59Ow0YMICUSiV16tSJtm7dahYDANqwYYPZ5+XLl9O4cePI0dGR/P39ad68eWbraDQaGjNmDKlUKvLy8qLZs2fTc889RzExMXXW9+5t3W39+vXUqVMnkkql5O/vT7NmzaLKykrT/L1791Lfvn3J0dGRHB0dqUuXLvTrr7+a5s+dO5fatm1LMpmM1Go1DR06lLRaba3b27hxI/Xq1YucnZ3Jw8ODhg8fTpcuXTLNv3r1KgGgTZs20YgRI0ipVFLbtm1p3bp1ZuVcu3aNHn74YVIoFBQQEEBLly6lQYMG0eTJk+vcH0REq1atooiICCorKyNXV1c6dOhQtWW+++47ioiIILlcTu7u7hQXF0f5+fmm+Z999hl16tSJZDIZeXp60mOPPWaa16ZNG/rwww/Nyps8eTINGjTI9HnQoEE0adIkmj17Nvn4+JC3t7dF+4eIKDs7myZOnEheXl4kl8upY8eOtGbNGjIYDNS2bVuaO3eu2fK3bt0iJycn+vrrr2vdJxcvXqThw4eTg4MDOTg40COPPEKXL182zV+3bh2JxWLat28fde/enZRKJUVERNCRI0fq2NNEf/3rX0mhUFBqamq1eQaDgQoKCoiI6LXXXiOFQkHXrl2rttzLL79MCoWCbty4QURE/fr1oxEjRlRbbsKECRQaGmr2+c6/j/j4eAoODq62XlFREQUHB1N0dHSddblTbWXNmjWLQkJCzKadOHGCANDOnTtrLa9q/96tsrKS+vbtS8HBwaTT6YiIKDU1lR599FHy9fUlpVJJ4eHhZr/bCRMmEACzn127dhFR/ce0mzdv0sSJE8nb25tkMhkFBATQG2+8YRbT0qVLKSQkhORyObVv357mzJljOmYMGjSo2ravXr1a575s1YnD3d2dli1bRikpKZScnExERB9++CElJibS1atXKSEhgUJCQui5556rtayqz126dKFt27ZRcnIyTZw4kZycnMwOOjUlDi8vL/ryyy8pJSWFPvvsMwJACQkJpmVGjhxJHTp0oJ07d9LZs2dp4sSJ5OzsfE+J4+effyaRSETz5s2jS5cu0XfffUeurq40e/ZsIjJ+ad3c3OiNN96g5ORkSk5Opi1btlBiYiIREW3evJmcnJzof//7H6WlpVFSUhJ98skndSaOtWvX0v/+9z9KSUmhEydO0MiRI6l9+/ZUXl5ORH8mjrZt29KmTZvo8uXLNHPmTBKLxaYDqMFgoO7du1NkZCQdOnSIkpKSKDY2lpycnCxKHJGRkbR06VIiMh6Unn/++WoxSiQS+uCDD+jcuXN06tQpWrJkCeXm5hIR0bvvvksODg60bNkyunTpEh0/fpzmzJljWt/SxOHo6EgvvfQSnTt3jk6fPm3R/tFqtRQaGkrdu3enHTt20JUrV+i3336jb7/9loiI5s2bR+3atSODwWDa1urVq8nNzY1KS0tr3B9arZaCgoJo8ODBdOzYMTp27BhFR0dTcHCwabvr1q0jQRBowIABlJiYSBcuXKC4uDh64IEHzE407qTX68nNza3e34nBYCB3d/dal7t+/brp5I6I6KuvviKxWGz2N1xYWEgqlYoWL15smmZp4iAi+vjjj0kQBMrJyakz1vrKeuutt6hr165m0y5cuEAA6L333qu1vNoSBxHRDz/8QADo6NGjRER0+vRpWrZsGZ08eZJSUlJo6dKlJBaLTYmpsLCQBgwYQGPHjqXMzEzKzMw0/R7rO6a99tpr1KVLFzp06BClpaXR/v376csvvzSrd1BQEG3ZsoVSU1Ppl19+ocDAQNMxIy8vjx544AF66623TNuuSni1adWJY9KkSfWuu2XLFpLJZKTX62ssq+rz5s2bTetkZWURALOz9JoSx2uvvWa2rdDQUJoxYwYRESUnJ1dLJBUVFRQQEHBPiaN///70xBNPmE1bsmQJKRQKKi8vp/z8fLOzlbstXryYOnToQBUVFXXGUJe8vDwCQPv27SOiPxPHokWLTMvodDpydHSkL774goiIduzYQQDMzsRzcnJIoVDUe5BKSkoimUxGGo2GiIgOHjxIKpXK7KwrMDCQXn311RrXv3XrFikUClq4cGGt27A0cXTo0MH0XarN3ftn9erVJJfLzb6/d8rKyiKpVEo7duwwTevduze9/vrrtW5j9erVpFQqTYmxqhyFQkFfffUVERkPbADo+PHjpmUOHTpEAOjixYs1lpudnV3td1nXcnce9O/m7OxMU6dOJSKi0tJScnNzo/fff980f8WKFSSXyykvL880zZrEsW3bNgJAhw8frjPW+spKSEggAPTFF19QRUUFaTQaGj16NAGgKVOm1FpeXYmjKvFs2rSp1vVHjRpFL7zwgulzTEwMTZgwod563H1MGzVqVK3rlZSUkFKppG3btplN/+qrr8jFxcX0OTg4mOLj4+vddpUW28ZhiV69elWbtmXLFgwcOBB+fn5wdHTEs88+i4qKCmRlZdVZVrdu3Uz/9/b2hlgsRnZ2tsXrAICfn59pnfPnzwMAevfubZovlUoRGRlZd6Xqce7cOQwcONBs2qBBg1BWVoYrV67Azc0NL7zwAh5++GEMGzYM8+fPN7sPPHbsWFRWVqJNmzaYOHEiNmzYgOLi4jq3efLkSTz66KNo27YtnJycEBQUBADVGgDv3B9isRheXl5m+0OtVqNjx46mZTw9PRESElJvnVeuXIlHHnkEHh4eAIz7NCAgwNSBICcnB+np6Rg6dGiN6587dw5lZWW1zrdGjx49qrWP1bd/jh8/js6dOyMgIKDGMr29vfGXv/zF1HZz9uxZHDp0CC+++GKtcZw7dw6dO3eGWq02KyckJATnzp0zTRMEAV27djV9ruoUUtt3m5ronacKhQLjx4/H2rVrYTAYAACrVq3CmDFj4O7u3qAyq2IVBAHXr1+Ho6Oj6efll1+2uJyYmBgsW7YMM2fOhFKphL+/P0JCQuDt7W11W2hNsQHGzgYzZsxAWFgY3N3d4ejoiK1bt1rUiF7fMW3q1Kn44YcfEB4ejr/97W/Ytm2baR+fO3cOpaWlePzxx832z0svvYSbN28iNze3QfVr1YnDwcHB7PPhw4fxxBNPYODAgfi///s/nDhxAl988QWA+hsyZTJZtWlVO9/SdQRBqLZO1RenOa1atQrHjx/HkCFDsGfPHoSHh2PlypUAAH9/f1y8eBFr166Fl5cXPvzwQ4SEhCA9Pb3GsrRaLYYOHQpBELBu3TocOXIER48ehSAI1fapJfvDWlWN4j/++CMkEonp5/Lly43aSC4SiaodNCsrK6std/d3zpr9U5eXX34ZP/74IzQaDVavXo0+ffogPDy8YZW5g0gkglgsNn2u+j7W9nvx9PSEm5ub6cSnNmq1Gm5ubjh79myN89PT01FUVGR2YjBlyhSkpaXht99+w/Hjx5GUlFRno3h9zp07B0EQ0LZtW/j5+eHkyZOmnw8++MCqsv76178iLy8P6enpyMvLw+zZs5Gbm4vg4OAGxwYA7dq1AwBMnz4d33zzDeLj47Fr1y6cPHkSw4cPr/c7Yskx7eGHH8b169cxa9YslJWVYdy4cRg8eDD0er3p9/z999+b7Z8zZ87g8uXLDU7arTpx3G3fvn1Qq9WYM2cOoqKi0LFjR6uf12gsnTt3BgAcPHjQNE2n0+H48eP3VG5YWBgSExPNpu3ZswdKpdLsSx4eHo4333wT27Ztw+TJk/Hll1+a5snlcsTFxWHBggU4c+YMtFotfvzxxxq3d+HCBeTm5mLu3LmIjo5Gp06dUFBQYPWZaefOnaHRaHD58mXTNI1GU2+vmG+//RYSicTsS3/y5Ens3r0bp0+fxuHDh+Hl5YWAgABTb5+atq1QKGqdDwBeXl7IyMgwm5aUlFRvvSzZPz169MD58+fr/C4OHjwYQUFBWLlyJTZs2FDn1QZg/B6cP3/erEtqdnY2Ll26dE8JRyQS4ZlnnsHGjRtx9erVavOJCDdv3jQt9+9//7vGs+Z58+ZBLpebdWEPCwtDv379sGrVKqxevRqhoaHVrp4tVVxcjM8//xzR0dFQq9WQSCRo37696cfLy8vqMgVBgK+vLxwcHPDdd98BAEaPHm11OTqdDosXL0b79u3RvXt3AEBiYiKeffZZjB07Fl27dkW7du2QnJxstp5MJoNerzebZukxzd3dHU8//TRWrlyJX375BXv27MH58+cRFhYGhUKB1NRUs/1T9VN1UlHTtusisXantGQhISHIzc3FmjVr8NBDD2Hfvn1YsWKFTWLp0KEDRo4ciVdffRUrV66Ep6cnFi1ahKKiIouuQq5fv46TJ0+aTfPz88PMmTMxcuRIzJ8/H4899hhOnjyJ9957D2+99RZkMhlSUlKwatUqjBw5EoGBgcjIyMDevXsREREBAFizZg0MBgN69eoFV1dX/P777yguLjYluru1adMGcrkcy5Ytw1tvvYVr165hxowZVl9JxcTEoGvXrhg3bhyWLVsGmUyGf/7zn5BKpXWut3LlSjz66KN48MEHq83r3bs3Vq5ciaioKMTHx+OVV16Bt7c3xowZA4PBgF27duGpp56CWq3GW2+9hffeew9KpRJDhgxBaWkptm7dipkzZwIw9qFfsWIFHn30UbRp0wZffPEF0tLS6j0js2T/PP3001iwYAFGjRqFBQsWIDg4GKmpqdBoNHjyyScBGA9aU6ZMwezZs6FUKk3Ta/PMM8/ggw8+wJNPPomFCxeCiPD3v/8d/v7+9a5bn7lz5yIxMRG9e/fGhx9+iKioKDg7O+PkyZP45JNP8Oabb2L06NGYM2cOdu3ahZiYGMyfPx+9evVCQUEB1q5diy+//BJffvllteelpkyZgsmTJ0OpVOL999+3KB69Xo+srCxT0jpy5Ag++ugjlJSU4PPPP693/fPnz5tu7VRUVJj+rjp37my6Sl64cCGGDh0KuVyO3377DTNmzMDbb79d63Mfd6q6ZVRcXGzaR2fOnMG2bdtMt7pCQkLw3//+13TLaPHixcjIyIC3t7epnLZt22LXrl24cuUKXFxc4OLiYtExbdasWejRowfCwsIgEomwceNGODo6IigoCI6Ojnj77bfx9ttvQxAExMbGQqfT4cyZM0hKSsJHH31k2vb+/ftx/fp103M7dd6ms7g1xIZqaxyvqQF59uzZ5OXlRSqVioYNG0b//ve/zbqX1dY4fnfDpVgsNutOevf2atr+3Y1bGo2GHn/8cVIqleTp6UnvvPMOjRkzhh555JE664u7usZV/fzrX/8iImN33NDQUJJKpeTn50dvv/22qZdMRkYGPfroo+Tv708ymYx8fX3phRdeMDUkb968mfr06UOurq6kVCopLCyMVq9eXWc833//PbVv357kcjl169aNdu/ebbZ/qhrH9+7da7be3Q1uV69epSFDhpBcLid/f39asmRJnd1xk5KSqnVSuNOSJUvMGsm/+eYb6tKlC8lkMnJ3d6fhw4ebuo4aDAZasmQJdezYkaRSKXl5edGYMWNMZRUVFdG4cePI1dWVPD09KT4+vsbG8ZpirW//EBFlZmbS+PHjycPDg+RyOYWEhFTrrpybm0tSqdTUoFyfixcv0rBhw0zdcUeMGFFjd9w7paen19l5osqtW7fo/fffp/DwcFIoFOTq6kq9evWizz77zKwH3s2bN2nGjBnUvn17kslk5OLiQg8//HCt3VirGsnvbhSvUlPjeNX3XyQSkYuLC0VGRtI777xj1jGgLm3atKnx7+nOLqdDhgwhV1dXkslk9OCDD5r1SqpNVecDACQIAjk5OVGXLl3ojTfeqNZF+fr16zR06FBSqVTk4+ND7777Lk2aNMns+3XlyhUaMGAAOTg4mP2O6jumffDBBxQWFkYODg7k7OxMAwcOrPa3uGrVKuratSvJ5XLT73LFihWm+UePHqXu3buTQqGwqDuuQMQjADYXvV6P0NBQjBo1CosWLbJ1OKyFOXfuHMLDw3Hy5EmzBm3GWpr76lZVS5OYmIicnBx0794dxcXF+OSTT3Dt2jWbvVaAtUzl5eXQaDSYOXMmHnroIU4arMXjxNGE9Ho95syZg5SUFEilUoSHh2PXrl013q9n9uvbb7/FpEmTEBYWhh9++MHW4TBWL75VxRhjzCr3VXdcxhhjTY8TB2OMMau0+jaOux/aspRarbboff73K3uuvz3XHbDv+nPdjXW3djyiu/EVB2OMMatw4mCMMWYVThyMMcaswomDMcaYVThxMMYYs0qz9KqqqKhAfHw8dDod9Ho9evfujbFjx5ots3v3bmzYsMH0NtK4uDjExMQ0R3iMMcas0CyJQyqVIj4+HgqFAjqdDu+++y66detmNhocAPTt2xeTJ09ujpAYY4w1ULPcqhIEAQqFAoDx/U16vd4mI+NVocw/ULxmCUhXfYQ3xhhjdWu2BwANBgP++c9/IisrCw8//DA6dOhQbZnDhw/jwoUL8PX1xYQJE8zGU66SkJCAhIQEAMD8+fNrXKY+5WnJKPz5P3AJ7QJFv8HWV+Y+IJFIGrTv7gf2XHfAvuvPdW+cujf7Sw5LSkrw8ccf4/nnn0dQUJBpenFxMRQKBaRSKXbs2IEDBw4gPj6+3vIa8uQ4GfTA7JdhUPtA/OaHVq9/P+AnaO2z7oB915/r3kqfHHdwcEBYWFi1YVGdnJxMw4jGxMQgNTW1yWIQRGIoh/wFuHAKlHWjybbDGGP3o2ZJHEVFRSgpKQFg7GF1+vRp+Pv7my1TUFBg+v+xY8cQEBDQpDEpYx4BxGLQ3t+adDuMMXa/aZY2joKCAixfvhwGgwFEhD59+qBHjx7YtGkTgoODERkZiW3btuHYsWMQi8VwdHTE1KlTmzQmsbsa6BYF2v87aPQ4CFJZk26PMcbuF61+IKd7eTtu7t7fYVj8DoTJb0DU+6FGjqxl43u99ll3wL7rz3VvpW0cLUrIg4CXL2gP365ijDFL2XXiEEQiCAPjgJTzoBtptg6HMcZaBbtOHAAg9I0BJBLQnl9tHQpjjLUKnDicnCH06Ac6tAtUXmbrcBhjrMWz+8QBwHi7qlQLOrrX1qEwxliLx4kDADp0BnwD+XYVY4xZgBMHjC9hFAYNA65dBqWl2Docxhhr0Thx3Cb0iQZkMr7qYIyxenDiuE1QOULoORB0JBFUqrV1OIwx1mJx4riDMCgOKC8DHd5t61AYY6zF4sRxpwc6AEHtQLu3oZW/iYUxxpoMJ447GBvJ44AbaUDqJVuHwxhjLRInjrsIvQYCCiVozzZbh8IYYy0SJ467CAoVhKhBoGP7QSXFtg6HMcZaHE4cNRAGDQMqK0AHd9o6FMYYa3E4cdRACGwLtAsB7fmVG8kZY+wunDhqIQyKA7JuAMlnbR0KY4y1KJw4aiFE9gdUDvwkOWOM3YUTRy0EmRxCn8GgEwdBRYW2DocxxloMSXNspKKiAvHx8dDpdNDr9ejduzfGjh1rtkxlZSU+++wzpKamwsnJCdOmTYOXl1dzhFcrYVAc6PefQPt/hzDscZvGwhhjLUWzXHFIpVLEx8dj4cKFWLBgAU6ePInk5GSzZXbu3AkHBwcsW7YMI0aMwMaNG5sjtDoJvoFAx3DQ3t9ABoOtw2GMsRahWRKHIAhQKBQAAL1eD71eD0EQzJY5duwYoqOjAQC9e/fG2bNnW0SPJmFQHJCbBVw4ZetQGGOsRWiWW1UAYDAY8M9//hNZWVl4+OGH0aFDB7P5+fn58PDwAACIxWKoVCoUFxfD2dnZbLmEhAQkJCQAAObPnw+1Wt2geCQSiUXr0pBHkLtpNaQHd8J10JAGbaslsrT+9yN7rjtg3/XnujdO3ZstcYhEIixcuBAlJSX4+OOPcf36dQQFBVldTmxsLGJjY02fNRpNg+JRq9WWr9tnMMp3/Ijcy5cguHk0aHstjVX1v8/Yc90B+64/191Ydz8/v3sqq9l7VTk4OCAsLAwnT540m+7u7o68vDwAxttZWq0WTk5OzR1ejYSBDwMGA2jfDluHwhhjNtcsiaOoqAglJSUAjD2sTp8+DX9/f7NlevTogd27dwMADh06hLCwsGrtILYiePkCnbqCjuyxdSiMMWZzzXKrqqCgAMuXL4fBYAARoU+fPujRowc2bdqE4OBgREZGYvDgwfjss8/w2muvwdHREdOmTWuO0CwmBHcC/fIfUGUFBKnM1uEwxpjNNEviaNOmDRYsWFBt+pNPPmn6v0wmw5tvvtkc4TSMbwBABiA7Awh4wNbRMMaYzfCT4xYS/AIBAJSZbuNIGGPMtjhxWMrbHxBEACcOxpid48RhIUEqAzy9gQxOHIwx+8aJwxq+gXyrijFm9zhxWEHwDQSyM0B6va1DYYwxm+HEYQ3fQECvA3IzbR0JY4zZDCcOKwi+xp5VyPzDtoEwxpgNceKwhq/xaXfKuG7jQBhjzHY4cVhBUKgAdzWQxVccjDH7xYnDWr6BIO6SyxizY5w4rCT4BgJZf/CIgIwxu8WJw1q+gUBFOZCfa+tIGGPMJjhxWIl7VjHG7B0nDmv5BgAAKJN7VjHG7BMnDisJjs6Asyu/s4oxZrc4cTSEbyCIu+QyxuwUJ44GEHwDgYx0EJGtQ2GMsWbHiaMhfAOA0hLgZoGtI2GMsWbHiaMB/uxZxe0cjDH7w4mjIXx5GFnGmP2SNMdGNBoNli9fjsLCQgiCgNjYWAwfPtxsmXPnzmHBggXw8vICAERFRWHMmDHNEZ71XNwAlQNfcTDG7FKzJA6xWIzx48ejXbt2KC0txYwZM9ClSxcEBASYLdepUyfMmDGjOUK6J4Ig3B4NkHtWMcbsT7PcqnJzc0O7du0AAEqlEv7+/sjPz2+OTTcZY88qfgiQMWZ/muWK4045OTm4evUq2rdvX21ecnIypk+fDjc3N4wfPx6BgYHVlklISEBCQgIAYP78+VCr1Q2KQyKRNHhdAChpH4Jb+3bAXSaFyNmlweXYyr3WvzWz57oD9l1/rnvj1F2gZnwYoaysDPHx8XjssccQFRVlNk+r1UIkEkGhUODEiRNYv349li5dWm+ZGRkZDYpFrVZDo9E0aF0AoDPHYVj6PkT/mA+hQ+cGl2Mr91r/1sye6w7Yd/257sa6+/n53VNZzdarSqfTYdGiRRgwYEC1pAEAKpUKCoUCABAREQG9Xo+ioqLmCs96pndWcQM5Y8y+WJw41q9fj2vXrjVoI0SEL774Av7+/njkkUdqXKawsND0JHZKSgoMBgOcnJwatL1m4e4JyOTcs4oxZncsbuMwGAyYO3cunJ2dMWDAAAwYMAAeHh4WrXvp0iUkJiYiKCgI06dPBwA8/fTTpsumoUOH4tChQ9i+fTvEYjFkMhmmTZtm7L3UQgkiEY8GyBizS1a1cRgMBiQlJWHv3r04ceIEOnTogIEDByIqKsp0m6m52aqNAwAMaxaDks9C/NHaeyrHFvher33WHbDv+nPdbdDGIRKJ0KNHD0ybNg1z585FUVERVqxYgRdffBFffPFFq+9iazXfQCBfAyrT2joSxhhrNlZ1x9VqtTh06BD27t2LtLQ0REVFYfLkyVCr1fj5558xb948fPzxx00Va4sj+AaCACDzBtC2g63DYYyxZmFx4li0aBFOnTqFTp06YciQIejZsyekUqlp/nPPPYeJEyc2RYwtl+mdVdchcOJgjNkJixNHhw4dMHnyZLi6utY4XyQSYdWqVY0WWKvg6QNIJDz+OGPMrljcxtGlSxfodDqzaRqNxqyLrlwub7TAWgNBLAa8/flZDsaYXbE4cSxbtgx6vd5smk6nw2effdboQbUmgk8AP8vBGLMrFicOjUYDb29vs2k+Pj7Izc1t9KBaFb9AIDcbVFlh60gYY6xZWJw43N3dkZqaajYtNTUVbm5ujR5Uq+IbCJAByL5h60gYY6xZWNw4PsVlSDIAACAASURBVGLECCxcuBCjRo2Ct7c3srOz8dNPP+Gxxx5ryvhavKouuZSRDiGgra3DYYyxJmdx4oiNjYWDgwN27tyJvLw8eHh44LnnnkPv3r2bMr6Wz9sfEETczsEYsxtWPQDYp08f9OnTp6liaZUEqRTw9OGeVYwxu2FV4igsLERKSgqKi4tx5yuuBg8e3OiBtSp+gQC/7JAxZicsThxHjhzBsmXL4Ovri/T0dAQGBiI9PR2hoaF2nzgE3wDQmeMgnQ6CpNkHVWSMsWZlca+qTZs2YerUqViwYAEUCgUWLFiAKVOmoG1bbhCGbxCg1wG5WbaOhDHGmpxVz3Hc3b4xaNAgJCYmNnpQrY1wezRAbiBnjNkDixOHs7MzCgsLAQCenp5ITk5GdnY2DAZDkwXXavjwMLKMMfth8Q35mJgYXLx4Eb1798aIESPw/vvvQxCEWoeCtSeCQmkcSpYTB2PMDlicOEaNGgWRyHiBMmjQIISFhaGsrAwBAQFNFlyr4hfIVxyMMbtg0a0qg8GA8ePHo7Ky0jRNrVZz0riD4BMIZP0B4lt3jLH7nEWJQyQSwc/PD8XFxU0dT+vlFwhUVAB5ObaOhDHGmpTFt6r69++Pjz76CMOGDYOHhwcEQTDNCw8Pr3NdjUaD5cuXo7CwEIIgIDY2FsOHDzdbhoiwbt06JCUlQS6XY+rUqWjXrp2V1bEdwTfAOIxs1h/GAZ4YY+w+ZXHi2L59OwDg+++/N5suCEK9Y3KIxWKMHz8e7dq1Q2lpKWbMmIEuXbqY3epKSkpCVlYWli5disuXL2P16tWYN2+eNXWxraphZDPSITwYaeNgGGOs6VicOJYvX97gjbi5uZlev65UKuHv74/8/HyzxHHs2DEMHDgQgiCgY8eOKCkpQUFBQat5bbvg4AS4uAGZ120dCmOMNalmfz9GTk4Orl69ivbt25tNz8/Ph1qtNn328PBAfn5+tcSRkJCAhIQEAMD8+fPN1rGGRCJp8Lq1yQ9sC+Rmwb2Ry20KTVH/1sKe6w7Yd/257o1Td4sTxyuvvFLrvM8//9yiMsrKyrBo0SJMnDgRKpXK0k2biY2NRWxsrOmzRqNpUDlqtbrB69bG4OkDOrQbubm5Zm1ALVFT1L+1sOe6A/Zdf667se5+fn73VJbFieO1114z+1xQUICtW7eiX79+Fq2v0+mwaNEiDBgwAFFRUdXmu7u7m/1C8/Ly4O7ubml4LYNvIFCqBW7mA64eto6GMcaahMWJo3PnztWmhYWFYe7cudV6SN2NiPDFF1/A39+/1ifNIyMj8euvv6Jfv364fPkyVCpVq2nfqFI1GiAy0jlxMMbuW/fUxiGRSJCTU/9zC5cuXUJiYiKCgoIwffp0AMDTTz9tusIYOnQounfvjhMnTuD111+HTCbD1KlT7yU026jqWZX5B4TO3WwcDGOMNQ2LE8emTZvMPpeXlyMpKQndu3evd93Q0FD85z//qXMZQRDwwgsvWBpOy+TsCqgcuWcVY+y+ZnHiyMvLM/ssl8vxyCOPYODAgY0eVGslCALgGwDK/MPWoTDGWJOxOHG0yltHNiD4BYFOHrZ1GIwx1mQsHo/jxx9/REpKitm0lJQU/Pe//230oFo1nwCg+CaouMjWkTDGWJOwOHFs3bq12ttwAwICsHXr1kYPqjUT/IwN5Dw2B2PsfmVx4tDpdJBIzO9sSSQSVFRUNHpQrZqpZxUnDsbY/cnixNGuXTv89ttvZtO2b9/eqt5g2yzc1IBcwVccjLH7lsWN4xMmTMCcOXOQmJgIb29vZGdno7CwEO+8805TxtfqCCIR4BPAVxyMsfuWxYkjMDAQn376KY4fP468vDxERUWhR48eUCgUTRlfqyT4BoIunbF1GIwx1iQsThz5+fmQyWRm76a6desW8vPzW987pZqaXyBwaBeoVAtB2bCXOTLGWEtlcRvHwoULkZ+fbzYtPz8fH3/8caMH1doJvrd7n2Xxg4CMsfuPxYkjIyMDQUFBZtOCgoJw48aNRg+q1fM17ifK4HYOxtj9x+LE4ezsjKysLLNpWVlZcHJyavSgWj21NyCRcM8qxth9yeI2joceegiLFi3CU089BW9vb2RlZWHTpk0YPHhwU8bXKgliMeDtzz2rGGP3JYsTx+jRoyGRSLBhwwbk5eXBw8MDgwcPxsiRI5syvlZL8A0EpaXUvyBjjLUyFicOkUiEUaNGYdSoUaZpBoMBSUlJiIiIaJLgWjXfQOD4flBFOQSZ3NbRMMZYo2nQQE5paWnYs2cP9u3bB71ejzVr1jR2XK2fbyBABGRnAIFtbR0NY4w1GosTx82bN7F3714kJiYiLS0NgiDg+eefx0MPPdSU8bVagp9xGFlKOQ+IxcaxyEu1oFItUFpi/FxmnAZtiXF6mRYQBIheeAuCs6utq8AYYzWqN3EcPHgQe/bswalTp+Dv74/+/ftj+vTpmDVrFnr37g2ZTNYccbY+Xn6AWAL690rjOOQ1EQRAoQKUt38USuDKRdD+3yEMe7w5o2WMMYvVmziWLFkCR0dHvPHGG+jVq1dzxHRfEKRSiP46C5SfCygdjE+QK1SA0sGYJFQqQKYwvtvqDvoFM0AHEkBxjxlHFGSMsRam3sTxyiuvYM+ePVi8eDGCg4PRv39/9O3b16qD2ooVK3DixAm4uLhg0aJF1eafO3cOCxYsgJeXFwAgKioKY8aMsaIaLZMQ3gPWHvqFvjGgr5YBqZeA4NAmiYsxxu5FvYkjOjoa0dHRyM3NxZ49e/Drr7/i66+/BgAkJSVh4MCBEInqfo4wOjoacXFxWL58ea3LdOrUCTNmzLAy/PuPENkP9O2XoAO/Q+DEwRhrgSx+ctzT0xNjxozBp59+ivj4eERHR+Orr77CK6+8Uu+6nTt3hqOj4z0Fai8EhQpCj76go3tBFeW2Docxxqqp94rj9OnT6Ny5s9nof6GhoQgNDcWkSZNw9OjRRgkkOTkZ06dPh5ubG8aPH4/AwMBGKbc1EvrGgA7uAiUdghA1yNbhMMaYGYGIau30AwBz585FamoqQkJCEBERgYiIiAa9Rj0nJwcfffRRjW0cWq0WIpEICoUCJ06cwPr167F06dIay0lISEBCQgIAYP78+Q0eulYikUCn0zVo3aZGBgM0rzwBiW8A3N77tEm20ZLr39Tsue6Afdef626s+732hq33imPWrFkoLy/HmTNnkJSUhC1btsDBwQHdu3dHREQEOnbsWG8bR31Uqj/HrIiIiMCaNWtQVFQEZ2fnasvGxsYiNjbW9Fmj0TRom2q1usHrNgeKGoSKnzch99IFCB6ejV5+S69/U7LnugP2XX+uu7Hufn5+91SWRQ8AyuVyREZGIjIyEgBw/fp1JCUl4bvvvsONGzcQFhaGESNGoEOHDg0KorCwEC4uLhAEASkpKTAYDHb/1l2hz2DQT9+BDu2CMGKsrcNhjDGTBr1yJCgoCEFBQfjLX/4CrVaLU6dOobS0tNbllyxZgvPnz6O4uBgvv/wyxo4da7pkGjp0KA4dOoTt27dDLBZDJpNh2rRpdv8Mg+DpA4Q8CDrwO2j4E3a/PxhjLYfFiePs2bPw8vKCl5cXCgoKsHHjRohEIjzzzDPo06dPnetOmzatzvlxcXGIi4uzNBS7IfSNAa1bAqRcADp0tnU4jDEGwIruuGvWrDG1ZXz99dfQ6/UQBAErV65ssuDsndCjLyBXgvYn2DoUxhgzsThx5OfnQ61WQ6/X49SpU3jppZfw4osvIjk5uSnjs2uCXGF8IPDYflB5ma3DYYwxAFYkDqVSicLCQpw/fx4BAQFQKBQAYLdd25qL0DcGKC8FHT9g61AYYwyAFW0ccXFxmDlzJnQ6HSZOnAgAuHjxIvz9/ZsqNgYY2zY8fUAHfgf68jC9jDHbs2ro2F69ekEkEsHHxwcA4O7ujpdffrnJgmOAIAjGRvL/bgTlZhl7WzHGmA1Z9eSen5+fKWmcPXsWhYWFCAoKapLA2J+EPoMBQQAd3GXrUBhjzPLEER8fj4sXLwIAfvzxR3z66af49NNPsWXLliYLjhkJHp5AaBfQwZ0gg8HW4TDG7JzFiSM9PR0dO3YEAPz++++Ij4/H3LlzsWPHjiYLjv1J6BsDaLKBy+dsHQpjzM5ZnDiq3oWYlZUFAAgICIBarUZJSUnTRMbMCN37AEoVaP/vtg6FMWbnLG4cDwkJwdq1a1FQUICePXsCMCYRe3+nVHMR5HIIkf1Bh/eAnpkCQaGqfyXGGGsCFl9xvPrqq1CpVGjTpg3GjjW+dC8jIwPDhw9vsuCYOaFvDFBRzs90MMZsyuIrDicnJzzzzDNm0yIiIho9IFaH4FDA29/4CpJ+sfUvzxhjTcDixKHT6bBlyxYkJiaioKAAbm5uGDhwIB577DGz0QFZ0zE+0zEY9H8bQDkZELzu7Z36jDHWEBYf8b/55htcuXIFL774Ijw9PZGbm4vNmzdDq9WaniRnTU/oMxj040bQgZ0QRo+zdTiMMTtkcRvHoUOH8I9//ANdu3aFn58funbtir///e84ePBgU8bH7iK4eQCdu/IzHYwxm7G6Oy6zPaFfLJCvAS6etnUojDE7ZPGtqj59+uCjjz7CmDFjTGPXbt68ud5BnFjjE7pFgVQOoAO/Q+jczdbhMMbsjMWJY9y4cdi8eTPWrFmDgoICuLu7o2/fvvxadRsQpDIIvQaC9v8O0pZAUDnYOiTGmB2xOHFIJBI8+eSTePLJJ03TKioqMH78eIwbx420zU3oGwPavQ10bB+EgQ/bOhzGmB2x6u24dxMEobHiYNZ6oAPgG2gcp4MxxppRszyAsWLFCpw4cQIuLi5YtGhRtflEhHXr1iEpKQlyuRxTp05Fu3btmiO0VksQBAj9YkA/rAdl/QHBJ8DWITHG7ES9iePs2bO1zrO0fSM6OhpxcXFYvnx5jfOTkpKQlZWFpUuX4vLly1i9ejXmzZtnUdn2TIiKBm352thI/tgEW4fDGLMT9SaOzz//vM75arW63o107twZOTk5tc4/duwYBg4cCEEQ0LFjR5SUlJieTme1E1zdgbAI0MFdoNHjIIjEtg6p2RSW6XAptxTXb5YjRK1EmJcKYhHfOrUlnYFwtaAMF3NLUVDa9J1mxCIB7koJ1Cop1A7Gfx1lonu6hV5SoYdGq4OmpBIarQ4FpTp4OkgQ6qmCn5PUZrfntZXmcWm0lQhVKxHh52iTeOpNHLVdJTSm/Px8swTk4eGB/Pz8GhNHQkICEhISAADz58+3KHHVRCKRNHjdlqQsbjRuLpwN5xtXIe/e2+L1WlP99QbC1TwtzmQW4WxmEc5mFuOPm2Vmy3iopIjuoEZMBzUe9HOGqI4/cGvrXq7TAwDkkpaVmLUVepRW6q1aRwAAkahRfveFpZU4m1mEM5nFOJtZhAvZt1CuMz6UKhEJaOpjrN5AMNz1eJlCIoKXkxxejnJ4O8ng5SiHl5Mc3renlRdVIEMrQU5xObJvlSOnuBw5t8qRXVyBnFvl0FbUvj9dFBKE+zrjQV8nPOjnjFAvRyik9/6dKNfpkXN7+9m348kprrjj/+W4dVdcAoDnegZiqBW/x8b8m291L5mKjY1FbOyfL/jTaDQNKqfqWZTWjtp2AlSOuPnrjxAFtrd4vZZc/5IKPZLzynAxV4uLuaW4pClD6e0DkotCjE6eSsS280SopxIBznKcyirBvrQi/HQ2C5tPZcJDJUG/ICf0b+OMjh6KameJ9dU9t6QSF3NLcVFTiou5pbhaUAY9AU5yMdSq22e4KgnUDsZ/PW+f8borpZCKG+doWaE3IE+rQ+4dZ5iaktv/3v5cUtGwNweIBMBNITGdpd9ZF+PZuxSuCrFZ8jUQIf1mxe39Yvy9ZBRXAgAkIqCtmwJD27ugk1qJEE8l1Cppo+yHuugNhMIyndmZeK5pP5XhiuYWCkt1qOvRZReFGGqVFN4qCcI8nU37wPP2PnFVSJB563a9c0txUXML+6/mAwDEgrHeoZ5KhKqVCPVUwtPBvN6VekJ+qTGm3Nu/u7yq3+HtmIvKqycrF7kYagfjd6uTh5Pp91IVn7tKAolIsOpv+M7vvZ/fvb3nrkUkDnd3d7MdkJeXB3d3dxtG1HoIUuntZzoSQNpbEFSNe+mqrdTjyB+3cD6nFKGeSkQFOMJB1rhn3gYiXMgtxf7rxTibpcX1m+UgGA9wbVzliG7rbPrj9Hasfrugfxtn9G/jbIp1X1oxtiYX4n8XC+DlIEX/NsYk0s5NXm3dO2+vXLidLPK0xtsscrGADmolHu3sAblEQN7tP/bckkpcyNXi1l0HbgGAq0IMtYMU7kqJ1UmkQk/QlFQiT6vDzVoOJh4qCbwdpQjzMh6cVVLrOkYaCKgQyXBdUwSNthJXC8px9MYtVOjND69iAfC4fZCSiAVcyStDSaXBFEeopxJDgl0R6qlEsLsCcsk9ddBsELFIgIdKCg+VFCFqZY3LVB2487TGBKNycITcUGpKmFJx/XEHucgR5CLH0PauAICiMh0uacpun1hosT2lED9fKgBg3GcPuMpRVK6HpqQShWX6aonLUSaCx+3td/BQmiVwTwcpPFQSyCyIy5ZaROKIjIzEr7/+in79+uHy5ctQqVTcvmEF4zMdW28/0xF3z+WV6Qy3D8BFOJFRgkoDQSYW8FtKIaQiARF+Dujfxhk9/R2htPLAVYWIkJxXhr1pRTiQVoy8Uh1kYgGdvVTo28YJoWolOqoVUFlxK0AlFSO6rQui27rgVoXeVIf/XsjHlvP58HWSon+QM7q2AZKu5eCiphSX88pMB00vBwk6eypvJykVHnCTQ1JHu0lppeHPs0et+VllZnEF9Fa+pUcqEuDRDAeTu6+4iAjFFYbbZ8DmZ8N5t69s+rcxJu9Onkr41JC8WyqpWIC3owzejjIAjXOl7ayQoGeAI3oGGE/SdAbCtYJy05XYH0UVcFFI0MZVbroarfpdeqikDf6baUkEaoaXUC1ZsgTnz59HcXExXFxcMHbsWFOPrKFDh4KIsGbNGpw6dQoymQxTp05FcHCwRWVnZGQ0KKaWfKvGWkQEQ/xfAZUDxDMWWLTO3fUv1xlwPMN4tl51BuquvOOWj1qBy7cP9PvTipF/+0Af6e+I/kFOiPR3rPesk4iQkl+GfWnF2J9WhFytDpKqRBTkhJ4BjlYlCksVletxKL0Y+9KKcCZbCwMZb6+0q7rNcPtqxqMZbq+0BPfTd99aXPfGuVXVLImjKXHiMDL8uhm0+SuIPvwcgo9/vcur1WpkZufgREYJ9qUV48iNYpTpCC4KMfoGOmFAG2d08lLW2MhcdWtpX1oRDlwvRmGZHgqJgJ7+jujfxhkRfg6ms2MiwrXCcuxLMx64s25VQiwA3XyNVy1NceurLoVlOpRLHOCKUpvcXmkJ7rfvvjW47vdRGwdruPxSHc5klYACe8PgfQbC3pMQIutu5zAQcOlEPhJTNNBWGuAkF2PgA8Z2gnALurWKBAFhXiqEeanwQg9vnMvRYl9aMQ6mF2NvWjGUEhGiAhyhdpDiYHoxbhRVQCQAXbxVGBPmgd6BTnCS26aHkqtCArXaCRpNuU22z9j9gBNHK5aaX4b3d6WjsOx2Q2qnp4AKAAcy613XSS5Gn0An9G/jhC4+DnXey6+LWCSgi48Duvg4YEpPb5zJ1mJfWhEOphdDW2FAuLcKo0Ld0CfQCS4K/roxdj/gv+RW6nRWCebtuQEHmQjzhwTBRSGB4cwx0KbVEE38G4T2nepcPzTIB0WF+Y0ak0QkoLuvA7r7OuDlnj4o1xvg2Iy3oRhjzYMTRyu0P60Iiw9kws9JivjBgaY+89QrEoZNyyCcSIAoomudZcia+P6+VCxAKuakwdj9iBNHK/PLpQKsOpaNUE8lZg8KgOMdbQXGcToGGIeVLdVCUKpsGClj7H5ln91KWiEiwsZTufjyWDZ6Bjji/cGBZkmjitA3BqioAB3bZ4MoGWP2gBNHK6A3EJYfzsJ/zuZhSLALZgzwr70raduOgI8/6MDO5g2SMWY3OHG0cOU6A+bvvYEdV25ibLgHXo3yqbO7rCAIxquOlPOgnIY948IYY3XhxNGC3SrX472d6Tj6xy1MifTGs109LXrVg9D7IUAQgQ7uaoYoGWP2hhNHC6XRVmLmjjQk55Vhen8/jAix/N1dgpsH0Kkr6MBOkKFhb1BljLHacOJogdJvluOfv6Uht0SH+IcC0K+Ns9VlCH0HA/m5wKUzTRAhY8yeceJoYS7mlmLm9jToDIR5Q4LQxcehQeUI3XsDShXoIDeSM8YaFz/HYWMVegOu5JXhwu1Bg5IyS+CulOD9wYHwcZI1uFxBJofQcwDo0G7QMy9BUPAzHYyxxsGJo5nlaStxSVNqGmHuSn4Zbg9uB18nKQY+4IzxXT3hqrz3X43QZzAo8TfQ8QMQ+sXWvwJjjFmAE4cV9HcPcFwPAwHXb5bfMeykFjklxnFIpCIBHTwUGBXqjtDbw226NvZLAINDAS8/4zMdzZA4qLICuJEGup4KpF8FVA4QRo9rNYP+MMYsw4mjHpV6Mo4idzEfVwsa/ipud6UEoZ5KPBJiHDionZui0canro3xmY7BoB+/AeVmQfD0abSy6VYRkH4VlG5MEnQ9Fcj6A6jqxSWRArpKCO1Cga49G227jDHb48RRi1sVemy/XIifLhUgv1SHQBcZnnzQA2Irz559nWQIVSvh6SCxyZm30Och0H83gg7uhDDqmQaVQQYDcOY46Nrl24kiFci/YzAcVw8gsC2E7r0hBLYDgtoBbh4wvP86DD+shSisOwQJf9UYu1/wX/Ndcm5V4qdL+diechNlOgO6+KjwWm8fdPd1aJW3XAR3TyC0C+jgLtAjT0EQWdeRjrS3YFjzCXD6KCCIAB9/CO07A0HtIAS2BQLbQXByqXFd0ZjnYfhsDmjvbxAeGtEY1WGMtQCcOG5LySvDjxfysP96MQBgQBtnjO7kjnbuChtHdu+EvjGgNYuBy+eBkHCL16Mb12FYMQ/Iy4bw1BQI/YdAkMst33CXnkDIg6D/fQuKGgRBVffIhIyx1uG+SxxEhLKyMhgMhjqvELKzs1FaVoZ8rQ7pReUoLNXjQXcxhrbxRICzDAqpGIABWq22+YJvIhTWA/TsVKC8HKLb9cnOzkZ5ee1tNqTJBg78DnlZKcRvzYXQobPV2xUEAaKxk2CY8yZo6/cQxjzf4DowxlqOZkscJ0+exLp162AwGBATE4PRo0ebzd+9ezc2bNgAd3d3AEBcXBxiYmKs3k5ZWRmkUikkddxTNxBBqwPyKiWoFIng7y5HZ4UEznJxveNtt1YU3h3Q3gIUCggiESQSCcQ1DLREREBhHmDQQdd/CCqGjIbKQ93g7QpBwcZuwb//BBo0rFEb6BljttEsicNgMGDNmjWYPXs2PDw8MHPmTERGRiIgIMBsub59+2Ly5Mn3vK26kgZgfHlgTkklZBIRvB2lcJSJW2X7hVUcnYFbRcbk4VjzK0xIrwc0WUCpFnBygcRNjYrS0nvetDB6HOjYPtD/bYAwZfo9l8cYs61meeVISkoKfHx84O3tDYlEgr59++Lo0aNNsi1LEoCjXIxAVyUCnWVwktumt1OzkyuMXWRvFdc4myrKgcx0oKwU8PCC4OEFQSRqlH0juHlAGPoo6Ohe0JWL91weY8y2muWKIz8/Hx4eHqbPHh4euHz5crXlDh8+jAsXLsDX1xcTJkyAWl39FklCQgISEhIAAPPnz6+2THZ2dr1XHADQ8Jd5tF56Z1cY8nMhJuOzFlX7yVB8E/qcLEAshti/DUQKpWkduVxe4+/BWoZnXkDe/gSIt3wFt/lf2jRZSySSRqlTa2XP9ee6N07dW0zjeI8ePdCvXz9IpVLs2LEDy5cvR3x8fLXlYmNjERv751PQGo3GbH55eXmN9+7vJpFIoNPp7j3wVoRUDkB+LnQ3CyFVe6GyshIoyAOKCgC5EvD0gUEigeGO/VJeXl5tHzd4+6OeRuVXy5D7638h6tm/UcpsCLVa3Wh1ao3suf5cd2Pd/fz87qmsZrlV5e7ujry8PNPnvLw8UyN4FScnJ0ilUgBATEwMUlNTmyO0Rnfz5k2sX7/e6vXGjx+PmzdvWr3etGnT8PPPP1u0rCCRAgoVcKsIpNcB2RnGpOHkCvj4NflDekLfwUDAA6AtXxlfT8IYa5Wa5YojODgYmZmZyMnJgbu7Ow4cOIDXX3/dbJmCggK4uRkHKzp27Fi1hvOGMHy3CpR+teZ5gmDsQWQlIbAtRE+9WOv8oqIifP3115g4caLZdJ1OV+cttA0bNlgdS4M4OgGabOiupxpfD6L2hlBLY3ljE0RiiJ6YBMMn74J2/gzh4ceaZbuMscbVLIlDLBZj0qRJmDt3LgwGAx566CEEBgZi06ZNCA4ORmRkJLZt24Zjx45BLBbD0dERU6dObY7QGt28efOQlpaGIUOGQCqVQi6Xw8XFBSkpKdi3bx8mTZqEjIwMlJeXY/LkyRg3bhwAICoqCtu2bUNJSQnGjRuHXr164dixY/Dx8cHatWuhVCrr2TKwd+9efPjhh9Dr9ejatSv+9a9/QS6XY968edi+fTskEgkGDhiAd54fj5937sYn67+GSCyBs7MztmzZ0tS7BgAgdO4GPBgJ+uV7UN9YCE7Nk7QYY41HoIacdrcgGRkZZp+1Wi1UqvrHnmiqNo709HRMmDABO3fuxIEDB/Dcc89h586dCAoKAvDnlVVpaSlGjBiBH374Ae7u7maJo1+/fti6dSvCw8Px0ksvYejQoXj88cdr3N60adNM7T79+/c3JePXX38dDz74IB5//HH85S9/QWJiIgRBwM2bN+GsUiE2Lg7ffPMNfH19cfPmTbi41PzaEEv3pzUoMx2G916DMGgYRM+81KhlW8Ke73MD9l1/rnsrauOwZ926dTMlDQBYu3YtYmNjMXLkSGRkZODqYpkPPgAAFPdJREFU1eq30gIDAxEebnw1SJcuXZCenl7vdq5cuYKgoCAEBwcDAJ544gkcPnwYzs7OkMvleOutt7B161YolUoIUil69eqFN954Axs3boRer2+k2lpG8A2EMDAOtGcbKPOPZt02Y+zeceJoYneerR84cAB79+7FTz/9hISEBISHh9f42g/5He+DEovF93Rgl0gk+OWXXzBixAgkJCTg2WefBQAsXLgQ//jHP5CRkYFhw4YhPz+/wdtoCGHU04BcAcPm9c26XcbYvePE0cgcHBxw69atGucVFxfDxcUFSqUSKSkpOHHiRKNtNzg4GOnp6aYrmM2bN6N3794oKSlBcXExYmJi8N577+H8+fMAgGvXriEiIgLTp0+Hh4dHtVt+TU1wcoEw7Ang1BHQhVPNum3G2L1pMc9x3C/c3d3Rs2dPDB48GAqFwuyBm+joaGzYsAGDBg1CcHAwIiIiGm27CoUCixcvxksvvWRqHB8/fjwKCwsxadIklJeXg4hMz8a8//77SE1NBRGhf//+CAsLa7RYLCXEjgTt2QbD92shmr0Ygqj+528YY7bHjeN2ytL6N0Xj+J0MRxJBqz6GMPFvEPWz/qWWDWHPDaSAfdef686N4+w+IPQcALTtCPpxA6i8zNbhMMYswImjlXj77bcxZMgQs59NmzbZOqx7ZhyzYzJQmA/67f9sHY5NEBHo5GEYflhvHLudsRaO2zhaiXnz5tk6hCYjtO8EoUc/0G9bQAOHQnD1qH+l+wRl/gHDplXAuSTj59+2ACEPQhQzEujak9t9WIvEiYO1CMLjE0CnDsPwr+kQIvtDiOgLtO1o9RjpdSEiICcTFZpMkLt3o5ZtdSzaEtDP34F2/gzIFBCeehFCr4GgAztBO382Dtmr9oYw+BEI/WIhqBxsFitjd+PGcTvVUhrH70RnT8Cw6xfgfBKg0wFuaggRfSD06AcEhzboQE8lxcCFU6DzJ0HnTwJ5OcYZXn4QoodB6BsDwaH5xkIngwF0cCdo81fArSLjOO6Pjofg9OeT+6TXAycPw5DwPyDlPCBXQugXY0wi3vfWqAlwAzHX/d4bxzlx2KmWmDiqkLYEdPoI6PgB4OwJQFcJuLj/mUQ6dKr1Fg7pKoErl0Dnk4yJIi0FIAKUKiCkC4TO3eDkoUbR1h+AKxcBmQxCz4EQHhoOoU37pq1X6iUYvlsFXE0GgkMhenpKvduktBRQwk+go3sBgx54MBKi2FFAaJcGj2nCB0+uOycOThwN0pITx52oTAs6fex2EjkGVFQATi5/JpGO4UBOBujc7USRfBb/3969B0VZ/wscfz+7IKDcERAVBRFvqIlCluloYc3vV5SeDmp1rJ9H1BotvIyUNZ1qUvOSjlraZB676NhEdaLb5M/fyPHSIKaGZll0FJRMUbm5XBTYZb/nj0dXUAgX111gP68Zht1l3f18nqeez/N8v9/n+6W2BgwG6NMfbeAwfWLF6H5oV9Zpufo/kPqjQJ/2ZP9uqKvV3zPu73pTWSevvw7MnhxM5agvtqD2ZekFMOUfaCPH2XXgV6Zy1O7tqD3bodIEPXqjJT2MljgGzbvlCTAbkoOn5C6Fo40VDpPJRGZm5g3TqrfkySefZP369c1ONuho7aVwNKRqa+CXH1E/7kMdPagXCA8PvVkL9OanuCuFot+QZvsFrj94qEvVqJxdqN3fwbk/oYuf3q8w9m9oYRGtj9diRv3vt6hvPgGzGe3+CWgPTULzbv32VOY61IHvUVlfw+mToBmgWw+03jHQK0b/HdkHzaf575CDp+QuheMvCsd/HzrPyfKm7w3QWrkeR3SQNzMSwpv9e8PZcRtqaT0OZ2uPhaMhVVsLx3JR//cLdO+FNmgYWtfm90tDzR08lFLw+89Yd38Hh/fr65UMHo5h3IMQM0B/fvVHWRs/t1pB1V97XFaMNXMrnDujNy9NmeGQ/olGsR7/FZX3kz6Et/AEXGww31hYd72I9I5B63WlqFzpy5GD5+3LXdXWwtlC6BmNdmVhurbCkYWj7RzJOghnr8exbds2tm3bRl1dHdHR0bz11lv4+PhQXFzMokWLKCwsBGDZsmUkJiby2WefsXHjRjRNY8CAAbz99ttO2zaOpHl5wfC70Ybf7bjP1DQYMBTjgKGoi6Wovf9Cfb8D6/olrfvAsAgMz/0X2tBEh8V4laZp0C8Ord+1qWKUqRz+yEcV5qP+yEfl58HB77GdHnUNh8hoTH7+WKur9VUg6+uh3qJftV193vBxvQU8PdH6D0EbPFzvW7mFK6aOSCkFBb+j9mXpfVGXL+lLMcfFow0biTZkhNMWS3OWDn3F8Vc6ynocZWVltmV4V6xYQWhoKNOnT+eZZ55hxIgRzJw5k/r6eqqrqykqKiI1NZWvv/6asLAwiouLbasuNqetXnHcCnvOOpXFAr8cQpWW6P0mDX+0a4+1Rq8bwdMTYuNcftapKiv0YvJHPhTmo86cwqhp1KOB0QhGD7257+pj228PvU/I6KGPTLvad2Q0QsxAtMHD0eLi9TNrFw5rvhmqohx15ADqyA94KiuWmAFoA4dBVKyt36tVn2sqR+3fhcrOgqLT+kCL4fdA3DA4/hvqpwNgKtP/O4kdiHbHSLRhd6KFOe7K0x5yxdGONLUex/bt2wFs63Fcv/66Petx/P7776xcuZKKigqqq6sZO3YsANnZ2axbtw7Qp2b39/fn888/Jzk52fZ9LRUNgb4O+7C7aN34JdfT/Pz1M9+4eNtrrWmu0Uer5aF+ydV/vtiC+mIL+Afqnx03HG1QfJtZ0VEVn0Md3o86vB/yf9NH1nUNR/kHoL76GPXVx+DTBfoP0Zs5B94B4d1bHLCgLGY4egjrviz4+ZDeLBkzAO2pZ/VBFVf7lu66F/Ufz+jF+qcfUEd+QH32Puqz9yEiEu2OO9GGjYTo2HZ5k6cUjtusufU4fHx8SElJuan1OGpqmp/Daf78+WzevJm4uDgyMjLIyclxbAJCAJqHp36Q7T8E/v0fqItlqF8Pw7HDqJ8PQc4ulKZB775ocfFoUbHgFwB+/uAbAD6dWz18+GYopeDMKVTulWLx55UF0npGoSVPQYu/G3pGERIaSvHJAlTeT9fu7zmyX2/OCw7VC8iVQtLo3po/T6Gys1A/7NZHtQUEoT3wb/p9QBE9m95mBoNeGKJjYeJUvZgdPagXkX9lov75P/oIwaGJerOjXyD4+l/ZZv7g5X1bt9mtkMLhYM5ej6Oqqorw8HDMZjOZmZl069YNgNGjR7Nly5ZGTVX33HMPqampzJo1i7CwMFuzmRD20gKD0UYlwagklLVeP7M+duVq5LvPUcra+B8YPa4dFP0C9DZ/X39bcdF8/cHbBzy9oJMXdOqk//a88ruTl37114Cy1uv37BzOQR35AYrPgabpVwCT/hMt/m600G43xu7nr0+umThGLzjFRXoB+e0n1OEcyN6pF5LIaLTYOL2vqPCEnsMdd+qzOMcNt7uZSwvthpb0MCQ9jKquQv3yo74eTe6+a9/ZkGcnfRv5+l23zfTfWlQsWlSsXTE4itMKx5EjR/jggw+wWq0kJSUxceLERn83m82sX7+egoIC/Pz8mDdvHmFhYc4Kz2GcvR5Heno6ycnJhISEEB8fbytar7/+Os8//zyffPIJBoOBZcuWkZCQQFpaGikpKRiNRuLi4li7du0txyDcm2Yw6vfARPeD5MdQ1VVQXASVFahKE1SZoLICqq4+r0AVnoCqCrhUDXDjQbMpBkPjYlJ7Gaoq9QP6wKFof3tU70cIuPmTIU3T9BFoYd1h3IPXiuCVmQbUnn9CRE+0KTP0e28c1BSndfFFGzkWRo7Vm79KLkB1JVSaUFX6trJts6oK/fXic3q+l/Vtxt9TcFXhcErnuNVqZe7cubz88suEhITw4osvMnfuXHr2vHaJt2PHDgoLC5k1axbZ2dkcOHCA+fPnt/jZba1zvL1o78Nxb4U7D0eFtpW/slj0g2RVhd75Xler3+RprkXV1V57XlcL5rrGzw0GvVlpSMJNz+Vlb+7KWg+aoU01GSmL2VYw7Slk7a5z/MSJE3Tr1o3wcH2c/ahRozh48GCjwnHo0CEmTZoEwF133cX777+PUqpN7TAhhGNpHh4QGKz/XP83F8RzvbbYca15eDa5vZzJKYWjrKyMkJBrU2WHhIRw/PjxZt9jNBrp3LkzlZWV+Pu3jVEarvbSSy9x8ODBRq/NmDGDKVOmuCgiIYS7aned4zt37mTnzp0ALF++vFEfAsD58+dv+g7ttnQnd0tWrlzp8M+8mfy9vLxu2MbtnYeHR4fLyR7unL/k7pjcnXLkDA4OprS01Pa8tLT0hnsXrr4nJCSE+vp6Ll26hJ+f3w2fNX78eMaPH297fn17ZV1dHUqpFg+K0sfRcv4WiwWz2dxm2sMdpS218buCO+cvubejPo6YmBiKioq4cOECwcHB7Nu3j7S0tEbvGTFiBLt376Zfv37s37+fuLi4VvVveHt7U1NTQ21t7V/+ey8vrybvoXAXLeWvlMJgMODt7e3EqIQQ7YFTCofRaGT69OksXboUq9XKvffeS2RkJBkZGcTExJCQkMB9993H+vXree655/D19WXevHmt+i5N05qd16khdz7zAMlfCNF6HW6uqpvl7gdOd87fnXMH985fcndMU1Xbnp1MCCFEmyOFQwghhF3afVOVEEII53LbK45Fixa5OgSXcuf83Tl3cO/8JXfHcNvCIYQQonWkcAghhLCL8bXXXnvN1UG4Sp8+fVwdgku5c/7unDu4d/6S+62TznEhhBB2kaYqIYQQdpHCIYQQwi7tZ15xB2ppGduObM6cOXh7e2MwGDAajSxfvtzVId1W77zzDrm5uQQEBLB69WpAX6d9zZo1FBcXExoayvz58/H19XVxpI7XVO6ffvopWVlZtnVuHn/8cYcsYdzWlJSUsGHDBi5evIimaYwfP54HH3zQbfZ9c/k7bP8rN1NfX6+effZZde7cOWU2m9XChQvV6dOnXR2W08yePVuZTCZXh+E0x44dU/n5+WrBggW217Zu3aoyMzOVUkplZmaqrVu3uiq826qp3DMyMtRXX33lwqico6ysTOXn5yullLp06ZJKS0tTp0+fdpt931z+jtr/btdU1XAZWw8PD9sytqJjGjRo0A1nlAcPHmTs2LEAjB07tsPu/6ZydxdBQUG2EUQ+Pj706NGDsrIyt9n3zeXvKG7XVHUzy9h2dEuXLgXg/vvvb7QolrswmUwEBQUBEBgYiMlkcnFEzrVjxw727t1Lnz59eOqppzp8cblw4QInT56kb9++brnvG+afl5fnkP3vdoXD3S1evJjg4GBMJhNLliyhe/fuDBo0yNVhuYymaa1aMKy9euCBB0hJSQEgIyODLVu2MHv2bBdHdfvU1NSwevVqpk2bRufOnRv9zR32/fX5O2r/u11T1c0sY9uRXc01ICCAxMRETpw44eKInC8gIIDy8nIAysvLbR2F7iAwMBCDwYDBYCApKYn8/HxXh3TbWCwWVq9ezZgxYxg5ciTgXvu+qfwdtf/drnA0XMbWYrGwb98+EhISXB2WU9TU1HD58mXb46NHj9KrVy8XR+V8CQkJ7NmzB4A9e/aQmJjo4oic5+pBE+DAgQNERka6MJrbRynFu+++S48ePUhOTra97i77vrn8HbX/3fLO8dzcXD766CPbMraPPvqoq0NyivPnz7Nq1SoA6uvrGT16dIfPfe3atfz6669UVlYSEBDA5MmTSUxMZM2aNZSUlHToIZlN5X7s2DFOnTqFpmmEhoYya9YsW5t/R5KXl8crr7xCr169bM1Rjz/+OLGxsW6x75vLPzs72yH73y0LhxBCiNZzu6YqIYQQt0YKhxBCCLtI4RBCCGEXKRxCCCHsIoVDCCGEXaRwCOEkkydP5ty5c64OQ4hbJlOOCLc0Z84cLl68iMFw7dxp3LhxpKamujCqpu3YsYPS0lKeeOIJXn31VaZPn07v3r1dHZZwY1I4hNt64YUXGDp0qKvDaFFBQQHDhw/HarVy5swZevbs6eqQhJuTwiHEdXbv3k1WVhZRUVHs3buXoKAgUlNTGTJkCKDPsLxp0yby8vLw9fVlwoQJtlmGrVYrX375Jbt27cJkMhEREUF6ejpdu3YF4OjRo7zxxhtUVFQwevRoUlNTW5xor6CggJSUFM6ePUtoaChGo/H2bgAhWiCFQ4gmHD9+nJEjR7J582YOHDjAqlWr2LBhA76+vqxbt47IyEg2btzI2bNnWbx4Md26dWPw4MF8++23ZGdn8+KLLxIREUFhYSFeXl62z83NzWXZsmVcvnyZF154gYSEBIYNG3bD95vNZmbOnIlSipqaGtLT07FYLFitVqZNm8YjjzzS4aeLEW2XFA7htt58881GZ+9Tp061XTkEBATw0EMPoWkao0aN4ptvviE3N5dBgwaRl5fHokWL6NSpE1FRUSQlJbFnzx4GDx5MVlYWU6dOpXv37gBERUU1+s6JEyfSpUsXunTpQlxcHKdOnWqycHh6evLhhx+SlZXF6dOnmTZtGkuWLOGxxx6jb9++t2+jCHETpHAIt5Went5sH0dwcHCjJqTQ0FDKysooLy/H19cXHx8f29+6du1qm566tLSU8PDwZr8zMDDQ9tjLy4uampom37d27VqOHDlCbW0tnp6e7Nq1i5qaGk6cOEFERATLli2zK1chHEkKhxBNKCsrQyllKx4lJSUkJCQQFBREVVUVly9fthWPkpIS2zonISEhnD9//panq583bx5Wq5VZs2bx3nvv8eOPP5KTk0NaWtqtJSaEA8h9HEI0wWQysX37diwWCzk5OZw5c4b4+Hi6du1K//79+fjjj6mrq6OwsJBdu3YxZswYAJKSksjIyKCoqAilFIWFhVRWVrYqhjNnzhAeHo7BYODkyZPExMQ4MkUhWk2uOITbWrFiRaP7OIYOHUp6ejoAsbGxFBUVkZqaSmBgIAsWLMDPzw+AuXPnsmnTJp5++ml8fX2ZNGmSrckrOTkZs9nMkiVLqKyspEePHixcuLBV8RUUFBAdHW17PGHChFtJVwiHkfU4hLjO1eG4ixcvdnUoQrRJ0lQlhBDCLlI4hBBC2EWaqoQQQthFrjiEEELYRQqHEEIIu0jhEEIIYRcpHEIIIewihUMIIYRd/h8W8EfyNrswpwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VY3n35qnlcEN"
      },
      "source": [
        "model.save('Covid19-model.h5')"
      ],
      "execution_count": 43,
      "outputs": []
    }
  ]
}