CS4287_Assign2_21217009_21147469_21147418.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9rMoHxgyIGPU"
      },
      "source": [
        "**Team:**\n",
        "\n",
        "  MaoLin Wei   21217009\n",
        "\n",
        "  YongHeng Yin 21147469\n",
        "\n",
        "  GuiYang Fan  21147418\n",
        "\n",
        "\n",
        "Code can execute to the end without an error.\n",
        "\n",
        "\n",
        "\n",
        "**Dataset from:**\n",
        "\n",
        "  https://www.kaggle.com/datasets/pratik2901/multiclass-weather-dataset\n",
        "\n",
        "**Third party code:**\n",
        "\n",
        "https://www.kaggle.com/code/utkarshsaxenadn/weather-classification-resnet-acc-91\n",
        "\n",
        "https://www.kaggle.com/code/iganarendra/using-alexnet-via-tensorflow\n",
        "\n",
        "https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator\n",
        "\n",
        "https://www.tensorflow.org/api_docs/python/tf/keras/applications\n",
        "\n",
        "https://blog.csdn.net/u010397980/article/details/88088025\n",
        "\n",
        "https://blog.csdn.net/liuz_notes/article/details/104293440?spm=1001.2101.3001.6661.1&utm_medium=distribute.pc_relevant_t0.none-task-blog-2%7Edefault%7ECTRLIST%7ERate-1-104293440-blog-88088025.pc_relevant_multi_platform_whitelistv4&depth_1-utm_source=distribute.pc_relevant_t0.none-task-blog-2%7Edefault%7ECTRLIST%7ERate-1-104293440-blog-88088025.pc_relevant_multi_platform_whitelistv4&utm_relevant_index=1\n",
        "\n",
        "https://blog.csdn.net/liujingwei8610/article/details/123470276\n",
        "\n",
        "https://mtyjkh.blog.csdn.net/article/details/117587326"
      ],
      "id": "9rMoHxgyIGPU"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FqcJ0WU25-lg"
      },
      "source": [
        "# 1.Import Data"
      ],
      "id": "FqcJ0WU25-lg"
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "4O6U123OxWEx",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e49e8896-bf5c-4c35-c4b2-7c1bf4d44066"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n",
            "/content/drive/My Drive/Colab Notebooks/Dataset\n",
            " data7.csv    eye\t       'Multi-class Weather Dataset'   SH600519.csv\n",
            " esport.txt  'kaggle_Dog&Cat'   resnet.h5\n"
          ]
        }
      ],
      "source": [
        "# Import data, data was stored in \"/content/drive/My Drive/Colab Notebooks/Dataset\", name as \"./Multi-class Weather Dataset\"\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "%cd /content/drive/My Drive/Colab Notebooks/Dataset\n",
        "# Lists all files and folders in the path\n",
        "!ls"
      ],
      "id": "4O6U123OxWEx"
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "p3nXrWY1yAOq"
      },
      "outputs": [],
      "source": [
        "dataset = \"./Multi-class Weather Dataset\""
      ],
      "id": "p3nXrWY1yAOq"
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "h3f_zQagDI7y"
      },
      "outputs": [],
      "source": [
        "# Some of the packages that will be used later\n",
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import seaborn as sns"
      ],
      "id": "h3f_zQagDI7y"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gpu8d-6y57uC"
      },
      "source": [
        "# 2.Data Pre-processing"
      ],
      "id": "gpu8d-6y57uC"
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "vfgdzOI_emIH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5a557358-a07e-45c1-a22d-3e3a06b61bc5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['Cloudy', 'Rain', 'Shine', 'Sunrise'] 4\n"
          ]
        }
      ],
      "source": [
        "# Classify the class of dataset\n",
        "import os\n",
        "class_names = sorted(os.listdir(dataset))\n",
        "num_classes = len(class_names)\n",
        "print(class_names,num_classes)"
      ],
      "id": "vfgdzOI_emIH"
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "I1oPA7Bkf8SQ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "29db6125-a3cc-4dfb-924e-03da16ebb3dd"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[300, 215, 253, 357]"
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ],
      "source": [
        "# Check the number of different class\n",
        "class_dis = [len(os.listdir(dataset + \"/\" + name)) for name in class_names]\n",
        "class_dis"
      ],
      "id": "I1oPA7Bkf8SQ"
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "_Za4efyc4qgj",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 173
        },
        "outputId": "8d3f90e3-36c8-48fe-ebf1-5040cc1c2351"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 288x144 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "# Visualisation of the key attributes\n",
        "# 1.Bar chart\n",
        "plt.figure(figsize=(4,2)) # Adjust size\n",
        "sns.barplot(\n",
        "    x=class_names, \n",
        "    y=class_dis\n",
        ")\n",
        "plt.axhline(np.mean(class_dis), alpha=1, color='k', label=\"Mean\") # alpha: the thickness of the line\n",
        "plt.title(\"Class Distribution\")\n",
        "plt.legend() # Displays the content of the label\n",
        "plt.show()"
      ],
      "id": "_Za4efyc4qgj"
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "EcHBeN1-Cw2L",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 542
        },
        "outputId": "91e4bb4b-ef82-4eac-96f6-534612295eb7"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<html>\n",
              "<head><meta charset=\"utf-8\" /></head>\n",
              "<body>\n",
              "    <div>            <script src=\"https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js?config=TeX-AMS-MML_SVG\"></script><script type=\"text/javascript\">if (window.MathJax) {MathJax.Hub.Config({SVG: {font: \"STIX-Web\"}});}</script>                <script type=\"text/javascript\">window.PlotlyConfig = {MathJaxConfig: 'local'};</script>\n",
              "        <script src=\"https://cdn.plot.ly/plotly-2.8.3.min.js\"></script>                <div id=\"98563ccc-318a-428e-aef7-ac6b08afbf80\" class=\"plotly-graph-div\" style=\"height:525px; width:100%;\"></div>            <script type=\"text/javascript\">                                    window.PLOTLYENV=window.PLOTLYENV || {};                                    if (document.getElementById(\"98563ccc-318a-428e-aef7-ac6b08afbf80\")) {                    Plotly.newPlot(                        \"98563ccc-318a-428e-aef7-ac6b08afbf80\",                        [{\"domain\":{\"x\":[0.0,1.0],\"y\":[0.0,1.0]},\"hovertemplate\":\"label=%{label}<br>value=%{value}<extra></extra>\",\"labels\":[\"Cloudy\",\"Rain\",\"Shine\",\"Sunrise\"],\"legendgroup\":\"\",\"name\":\"\",\"showlegend\":true,\"values\":[300,215,253,357],\"type\":\"pie\",\"textinfo\":\"percent+label\",\"textposition\":\"inside\"}],                        {\"template\":{\"data\":{\"bar\":[{\"error_x\":{\"color\":\"#2a3f5f\"},\"error_y\":{\"color\":\"#2a3f5f\"},\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"bar\"}],\"barpolar\":[{\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"barpolar\"}],\"carpet\":[{\"aaxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"baxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"type\":\"carpet\"}],\"choropleth\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"choropleth\"}],\"contour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"contour\"}],\"contourcarpet\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"contourcarpet\"}],\"heatmap\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmap\"}],\"heatmapgl\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmapgl\"}],\"histogram\":[{\"marker\":{\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"histogram\"}],\"histogram2d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2d\"}],\"histogram2dcontour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2dcontour\"}],\"mesh3d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"mesh3d\"}],\"parcoords\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"parcoords\"}],\"pie\":[{\"automargin\":true,\"type\":\"pie\"}],\"scatter\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter\"}],\"scatter3d\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter3d\"}],\"scattercarpet\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattercarpet\"}],\"scattergeo\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergeo\"}],\"scattergl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergl\"}],\"scattermapbox\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattermapbox\"}],\"scatterpolar\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolar\"}],\"scatterpolargl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolargl\"}],\"scatterternary\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterternary\"}],\"surface\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"surface\"}],\"table\":[{\"cells\":{\"fill\":{\"color\":\"#EBF0F8\"},\"line\":{\"color\":\"white\"}},\"header\":{\"fill\":{\"color\":\"#C8D4E3\"},\"line\":{\"color\":\"white\"}},\"type\":\"table\"}]},\"layout\":{\"annotationdefaults\":{\"arrowcolor\":\"#2a3f5f\",\"arrowhead\":0,\"arrowwidth\":1},\"autotypenumbers\":\"strict\",\"coloraxis\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"colorscale\":{\"diverging\":[[0,\"#8e0152\"],[0.1,\"#c51b7d\"],[0.2,\"#de77ae\"],[0.3,\"#f1b6da\"],[0.4,\"#fde0ef\"],[0.5,\"#f7f7f7\"],[0.6,\"#e6f5d0\"],[0.7,\"#b8e186\"],[0.8,\"#7fbc41\"],[0.9,\"#4d9221\"],[1,\"#276419\"]],\"sequential\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"sequentialminus\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]]},\"colorway\":[\"#636efa\",\"#EF553B\",\"#00cc96\",\"#ab63fa\",\"#FFA15A\",\"#19d3f3\",\"#FF6692\",\"#B6E880\",\"#FF97FF\",\"#FECB52\"],\"font\":{\"color\":\"#2a3f5f\"},\"geo\":{\"bgcolor\":\"white\",\"lakecolor\":\"white\",\"landcolor\":\"#E5ECF6\",\"showlakes\":true,\"showland\":true,\"subunitcolor\":\"white\"},\"hoverlabel\":{\"align\":\"left\"},\"hovermode\":\"closest\",\"mapbox\":{\"style\":\"light\"},\"paper_bgcolor\":\"white\",\"plot_bgcolor\":\"#E5ECF6\",\"polar\":{\"angularaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"radialaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"scene\":{\"xaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"yaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"zaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"}},\"shapedefaults\":{\"line\":{\"color\":\"#2a3f5f\"}},\"ternary\":{\"aaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"baxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"caxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"title\":{\"x\":0.05},\"xaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2},\"yaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2}}},\"legend\":{\"tracegroupgap\":0},\"title\":{\"text\":\"Class Distribution\",\"x\":0.5},\"piecolorway\":[\"rgb(247, 254, 174)\",\"rgb(183, 230, 165)\",\"rgb(124, 203, 162)\",\"rgb(70, 174, 160)\",\"rgb(8, 144, 153)\",\"rgb(0, 113, 139)\",\"rgb(4, 82, 117)\"]},                        {\"responsive\": true}                    ).then(function(){\n",
              "                            \n",
              "var gd = document.getElementById('98563ccc-318a-428e-aef7-ac6b08afbf80');\n",
              "var x = new MutationObserver(function (mutations, observer) {{\n",
              "        var display = window.getComputedStyle(gd).display;\n",
              "        if (!display || display === 'none') {{\n",
              "            console.log([gd, 'removed!']);\n",
              "            Plotly.purge(gd);\n",
              "            observer.disconnect();\n",
              "        }}\n",
              "}});\n",
              "\n",
              "// Listen for the removal of the full notebook cells\n",
              "var notebookContainer = gd.closest('#notebook-container');\n",
              "if (notebookContainer) {{\n",
              "    x.observe(notebookContainer, {childList: true});\n",
              "}}\n",
              "\n",
              "// Listen for the clearing of the current output cell\n",
              "var outputEl = gd.closest('.output');\n",
              "if (outputEl) {{\n",
              "    x.observe(outputEl, {childList: true});\n",
              "}}\n",
              "\n",
              "                        })                };                            </script>        </div>\n",
              "</body>\n",
              "</html>"
            ]
          },
          "metadata": {}
        }
      ],
      "source": [
        "# 2.Pie chart\n",
        "import plotly.express as px\n",
        "fig = px.pie(names=class_names, values=class_dis, title=\"Class Distribution\",color_discrete_sequence=px.colors.sequential.Bluyl)\n",
        "fig.update_traces(textposition='inside', textinfo='percent+label') # Display percentages and labels directly in each section of the pie chart\n",
        "fig.update_layout({'title':{'x':0.5}}) # Center the title\n",
        "fig.show()"
      ],
      "id": "EcHBeN1-Cw2L"
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "CqwExPnF4A4E",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "outputId": "619910ea-e28d-49f7-8625-3881c013c5c4"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAD8CAYAAABXe05zAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOy9Xah1W3rn9Z9r773W2l/nPadSpCgrIeVFvNAGFSS50ItIUOymofCm6AiatMHKRYIIfZGyb1qUhrrQDoGG4JEOnQK7k4A2CRJs20BoBKOxg6DdrVK0CVWVqpxK1an3fffH2l9rerH3f67f/O9nzLXec+rY+5A9YLHWmh/j4xnP83/+4xljzNn1fa/n9Jye03N6Th+fNPunXYHn9Jye03N6Tm+WnoH7OT2n5/ScPmbpGbif03N6Ts/pY5aegfs5Pafn9Jw+ZukZuJ/Tc3pOz+ljlp6B+zk9p+f0nD5m6SMD7q7r/q2u6/7vruu+0nXdFz+qcp7Tc3pOz+lPW+o+inXcXdftSfp/JP0bkr4m6fck/UTf9//oe17Yc3pOz+k5/SlLHxXj/hFJX+n7/p/0fX8t6Vclfe4jKus5Pafn9Jz+VKX9jyjfz0j6Kv5/TdKPti4+Ojrq33nnneF/13WTmft863sqj67r1Pe9cqTR971ub2/V973W6/VwntfxHM9X+fGeD5O6rhu1bzabDd/+vV6v1XWd9vf3tbe3N9zjYymLlpxcV+eX5/nfbbY8fSzzY/t5b+bp9rhN/jDPzJv9xGP+VLLMtra+M2U/V+3K31mWJM1ms1F+vid1q9LP6ner7G02xHxafUzdy1Tdk8daMs32Vf+rOrTqknpX9TPzyrz5oX212kQ74bFWm3ieMnc51fXr9Vrf+c53dHZ2Vjb6owLuranrui9I+oIkvXjxQj/zMz+jvb29oSEP15Tg7Gv29/c1m820t7c3fCxw/l6v17q7u5Mk7e3tDUJar9e6vb1V13W6ubnR2dmZVqvVAFzX19cDmN/e3urq6krn5+e6u7vT7e2tbm9vdXd3N7puNpup7/uhPP+uOrICZZ8nGO/v7+vg4EDL5VLL5VKLxUJHR0daLBaaz+e6u7vTzc2Nlsulvu/7vk9vvfXWANjvvPOO5vP5ICvXz3WZzWY6ODjQ3t6eJOn6+nqQwWw2G/IhECbQSBslTsfotrtsX2u53N3dDW07PT3V0dHR0La9vb0BhAnuknR7e6ubm5sRUF9fX+vy8lJXV1e6uLjQarUadKDvex0cHIzaY7ldX1+P6puOa71eD9e5z9n2dDjuQ8va9/R9r/l8PtJLy8r65TKsT+mMUu57e3tDmywb63/aDh2t86HuZTssLx5nm52nbdFto467j2kDloePuS8sV8vfbbON87fb5HuoqwmUJDHz+XxUz/39fS0WCx0cHGg+n+vw8FCHh4fDMZeVbbm+vtbV1dXQFusjZexjt7e3ur6+1vX1tW5uboZy9/b2dHBwMOizy7Ae/MIv/MIj3HT6qID765J+EP9/4OHYkPq+f1fSu5L0mc98pq88ZnrDPE+AI9OkAvtaX0dlZadI0snJiQ4ODgZD3tvbGxTj5uZmpJRXV1darVa6vr4enIHLTgNhGcnYKzZNkExDpeL6Hjut+Xyug4MDLRYL7e/vj8q3InZdNyhaKnzXdYMjsOMh8yV4VOCdjsEyZjtzlEADNBATKGikzN+K776g0djI3V7nbzDd379Xe5dnJ+/77+7uHhEIOxjnb71wH1UMysfsgGyUFchQtlVy3nQqeb5FBHZNFTut8qYMKQfmkU6MRGYb4077dpkpZ/enpJGeZDm8h7rl+6wr+SH5y3ofHBwMwOwPsYj1t06nTG5ubkZyMc6sVqtBn1vpowLu35P0w13X/bO6B+y/IOnf2eXGSvkqRc3jBI9UMgKNDdjnCCQEMLM1AsV8Ptfe3p6ur69HYG4lTqfBTvExGzCNNIGJwFAxrmwvmYSVxr8PDw+H88vlUpKGUYZBi7KnY3NdCX63t7fDecqT7abBVaEby4XydZuyH6s8aIw3Nzfa29sbhblsKIvFYqhjhlBcFzqKJAFsx8HBwWjE4Ov9n7pKQsE8CXA0ZsqfgESQZp50DklSEqQoWxII/6f9VPcbvOgErHMJthWRoj0mEFWjUNpDRdYoP9clQdvHk+yYsbs8jmb9MfkxGfL9OeI0W/YI6fr6ejhuUtQahdKRMV8zeePLVPpIgLvv+9uu635O0t+VtCfpl/u+/4dT9xBs81h2YHW8lQ+FzGHuQz0HAfq4Qc7CM4guFgtJGsIrvk4ag0wqfzJ8hlkq4PbQKcMATmlMZthWtr7vdX19rcViMTBvD/0IgG5z1td5Z9lW4gT2HG4TPFM2VShsNpuNnEMF3gmKlFtluO43GhFHIOv1egD1ZL28333l+nkUZsBk2I0pQZH1ZLt8zI6U+WVogqDNUQ/llvqX17N+WWfWl3JPXSGg2358DetFAkL50on6HpKUyoHQCbC+qbPMjyBpR+NRGvvEwG0b4jyR78920aHSNq0rtAPLKMlD1XfbRl1MH1mMu+/735L0W7ten+CXKRUyPwQMDlUSnJgSdDJPe2h3GkHMgHhwcKCrq6uRN7chMmxAY3XZZHvO10zZjsYfDtsM2IeHhzo6OtJ8Ph/i3QQHg3kyDbct2avvIwu/uroaGS0BnIaWTsi/3S82dB9neInXOmRye3s7GhGlbrCPDKg05kq//M0wjPvIxuX8yMavr69H99L4ExATNNz3ZKn+TXAm0LUY6lRqhQaozxWAV78J0FkX/k8WWTHMBG3rM0eTBHcCpEGURMVycn9Tv2nvJA60hWwf2bgZd5KIJAquM0Mkfd8P7aINmEg5LOIP+yf1lA6vlf6pTU4yVUrm4y0l5HknDu2nWCVB3kZPppkMQNoothkxJ8729vZ0c3MzhEwcc81JSedDZaYBW8HM5qkc2QaHPjxZyQkVaaOwlVzolDj6sEzpJDgJa0OqhuzZRoMeGRnBk4aeTvvu7k6Xl5eSNJqkZLvSEXLi0ceo/HQ8ZF4+R0fm/LquGwz+6upqVOfUX8qwFfe37O2gDeSWrdvuUWDKNsvJVDHWypZajiFBn0w288n+y3ZTD+yI+Z3OLevOBQe0EQOs+5AM2fZop8v8CO6pDwTuHJVWMrddeHKSbTF4c5R2dXU1clwkgOkgiUNT6UkAt1NL2baBuv9LY9BwSmaXrJDJM7rVMHo2u4+busPdGWRnBnKHWgy+rkeymDQOKxBBz0qZQzjnR0XmTDUNzIqcDNYpr8khLw1LqiciM7SRw33mRcYlaXB6rsvl5eVIPgkorL/7wHFKSYPjzNikZcc252okaTOqotyT0XHIy7zp+OlAna+Bm3MO2T8J9uzPCnRTL/J4i+FRHtW5tC/ek0N+hj/82/Ff9gf7v7JN90cCKvXbNuLzLDfDfLYR19vlLJfLYdTsUEolK9Y3/6c9OZTKVUF2WD5nuTmskpOgU33l9CSAmx6XDdg2lKuAnPnx+jRgMjPel0DlcjK0YaPuum4YXhmk6cV9TYIEh800/sVioeVyqdlsNoRg7ATMsF2e20ildsjEx9lOtr8yVsqSQ7qqrpRPyn8+n49k5Wtz0o/GZZbpsJCV2vMJHNZWYR8vKbRBmQ1l/5IFcjhOoPQ9BwcHwz25BLTSSeoVDTCH+NI4lJUsrYqd09nYqWcieFTMO0cMlZ1RTpSHk8GncixZT8v5+vr60eiHOkg9orwIzAz7cTKRo8xqySFlTT0gcCfbTmdLudJhk0TkSNo6Q/1hrNuJOpgynEpPArilWlmoaJXnT/BtXZOJxsX/s9lsFLe2whkcGJelchko6U0Zg5vNZoMCk7VnqKbrumEds8HHs9ZWUteBzMOTkGQiyXqr8EAyUEmDseVwn3WkYfg42Sj7hI7P9+U59lPGmq+vr0frblNfpI3jWi6Xwz2sR4ITDS+deBqs+5Cg4XoSvOg83d80UIe/3E7G/DlxxklU6mf+d9mVU2a7nfI6y4FlVSOKitmzjziflPdxrXb2QwK+60iQ5koPr/bwf4YHTZDI+G27lG+OAJ0fz7tubC/lZ/bM0QPj9rlu3Y6Obc8wrfWJsqlkz/QkgNuNtzImU2Tow40h46iGGvw4padzfull9/f3dXJyotevX48217A8gpikgRXTkF0+45dUTioQHcfh4aH6vtdisRgMxJtqHFu7u7sbNgyQTTgmzhAKwxCUg8sk6/XQ1opERkvWRqOtwj+cCPKH8wc5FKWs2DdXV1eDcbL/87edCyd3XZ/WWmP/tgNaLpdDCIzhC7fD7XL9STQI8nTkyYwpK25EoV6lHlVD/8yP/ZksP0mN+4HtT9bofs7JPraZbc/QFEkX28x2sE4MjXCi0DrMDWhmyp6YT+dvG/CxFmmhzSceZP3oiLxJjSEgtp/gzbh2Mu+W/FsjIaYnA9yMfZHJEbwJwgny0ngI1mLaObyjgRjELMiDg4PRBCGZJ4GbM8bSJkSRSrpcLkdKxIlIK6TL8PJD/t/f39f5+flQb64ttjysJBn3ziEuryFzuLy8HOLLBv8qjMA2JovMcJPr5zXcZCvstxwpOHmTk2PCCbwECzoa1zuXYrmO1C32hyRdXl4OTjvvTUacIGUAyonZZL2sN+9J5uVPtfSOgJOxYTpc9iH7zGQi49VuE50UHYrl5FBW2h/1nqNbtiv7n/soyH4N2gbr4+PjYXet77GMXSe3i7bB0AlHo9SpHBWkLDgyMYh7MQL7kyvCKv1zHyauVfWo0pMAbkkjxaGHz+OVR5Q264ytuHme+RK8K4ZOAJQ0rPl1fawAZKoGcE5sOnlbLWO+VioD0osXL3R8fDwwTq6QoAysxM7X9WGcnaMIt4VM2OVTAc0gKnAiA7N8CEQ0ZsvJwM7QCEHVZZNlVKORNB7Lnv2VxzxiMeuq1tEyLMFzJAOUXSt04EQZ0WGScbO+ObS3bClD973bTbKS7JZ9wviw9Yn9TlBzfdivSZDoLAg0tCk6c06qU898DVml82mNDrzs9eTkRMfHx8OIdLlcjrbks184t0S7JfNn/bNPk/S5TSRntk9Oviajzri2r0nHnw6zRTyZngxwW5kY+6UipcDJXmwwVBIyECoXO4oshKxCGq+zZuczL3YA6+dk1sc4IK9lnfb390dDPxqg6+WJTw/FmVI+rr/jvQSp3AXIOtnJUPYVaLvP0iByiRMNlcabbFjarGPn7kouW/T6ePcd5USwsHw4EkrQZ/8b4Fer1cAgCcT5sW7ZKeWoMGOakkYOzH3JOGgyUs6RpAOh3iaRsUzYRo4+ZrNN/D1HAOmQPcxnHxLE02EyH+pMysj/M2SX+uO9Em+99ZaOj4+HWLdBk2uzM1Wslks+yWyZKju2XOgkfJ9H43zWCJ+j4/8cZVKvWGYlw1Z6MsBdMTwfZ4fyWnYCh47pzX1PNSxJZbRR0YtyeJQrAzg8dR4GArLHXL/JxJUNdDZkIG4fnzniY/6fjM319W86KbbZRjCbzUbgaBBieyhLJis2h+YE/WpobFmw7ZyAZf9VjJf19HDdgGTASqOiDCyXZFQeeXCom/Mc7h8nGmU1p0DW7LyTNSexqPSYrJt1oZ4znEf5ku1bT2gPFbnJ/qPMfd46kyOXBG0fy35MckbQPj091fHx8RBmNHhnmISyoM0mwUsdbDHc/E/n0vf9ECLxFnUCNB9Gxhi3271t5LZLenLATS9EFuUGMbYqbTw3r6Oxu/PS6GzI/u88r6+vdXFxMSyadz0MYNxd5bI4ceb4MEMX9tQMDRhwzSA4IUujIqskA3O9OcQkOLn9NFp/LFcyG4ZnnDIu6yGoz/E446p5nxW9cqbuU64k4LyFVIc6KgPl+nfnv7e3Vz4+gMPpVhnsd47S6Bx8HXWXseipUVml80lO6PBTZuwn6iOdqPuEebvuGZrIetHxVgDXIltJnvJ6/86RQtbdS1vdr3wqJslH5pOkIskZ/2fbtyXW1Y6dE5RTgN1i0ZTJxwq4qcAJxBS4DTAZTTI8enYPfX3O7MDnXaYn8K6uroaJqWQpLovGYxbAvOwA3I75fP5IaT0hySVJFcNKpc7VGgRtMinp8SQMJ1WdB8MWkkahCgMiQY7O1OUacAkQnLBy/1WKyXolGFlWnOg0a2UbCN5k3naq/s1hrq9jXJaONYe17Lssd5vBEZwJyBlSYZunJnHZFpbh45z0ZmiBbJ1hv8oxMATGPqn0kiSAIZEqbzpdytG67XCZR5Y+vlwudXJyMiwNTWeYziudIvXN1+fIIPusuod6miOyXEmSRI15s57sv13TkwDuVIbqvJWMKzeSBVF5qUzMw/mQBVLZPPSxgpMBk7k6FsxNGr7Hm2bsca+urkYTKQcHBzo6OhomGRlz5gYe1j1ZM4/RSBLcc6ae+SbrIqMn02RohkyScwlmdu4DjiD6/j7Ew/5zH1ZGX/U/l8mxHum8fJ5Gk3pE0GDs30DJNjoRwBlyqUI4vKdi3q4fwye+rrUCIQGUjiBjpZJGS+roRB0+oxOkY+OIpmK0LcefI0KOWAiwCebUBdsGR6F2QEdHR6NQXUVwUvYpd17HkWnqRjVSyJHYarUa5kQy9JXxbOrOVJjkTdKTAG6pXqIj1Yvhk70QAHwNwS8nRghMZMnuANen67rRemCu4nBKQ+ayQo8SXCczICuhFZGTYVZ2gnkCduXgkiW6Db4/ZcsJ0wRMGqCvNdNmWCj7phodGLhttB5Wsr8ZzkkgynKq8wRbhisY4iAgVeEm1oVGRr2oALViVGSzNFzXjcbNB5JNMa6WvFMePM82OZlo0KHYVrhcM59eWZGrjKGTFLiv/dt1yf4iqJttm1VbfgZx26vDaVWZ1Sio+k7nwVFdOlvrkkfj5+fnQzg1WXXaYHWuSm/CtqUnBNzJKqXpmE8yqvS8OaS3ITFfKhcFS5AzyNBDJ6tzvnQSaVDz+Vynp6cDcDt8QqZDQGsZM+VTOSI6kIODg6ZRJ/h72CdpZMiUUzKtnG9g/k50Hl6nznBVxtbdlgQclkFH6XP5O/vIbXT/2gG5nPl8Ptq4UZGIXN6VLC0N3wCYj/FtrTBI3a9kmn3HtreIjz/WLz6ILOvgOtPxZH9m6I36wkl6zgNwdFfJy6TAG6jo6G9ubnRxcTHUjw9UI8AbZC2/CijTKadsKfuccFytVjo/P9f5+fnw0DE6yATxVh9WTvhN05MB7mTFTsnipMcNT4/q31z7asWgUHNDCI0nFTeZFpftZWyUdedw7/T0VKenp8PkihfuG2RdlvPh6pWUh2VFgPLH9UrGxHs5RDfz8+SKRwM+53wdDiHQGhCqFRu8hmDJOYp0IFx6xnZXeVZsM5f+uW2ULZ+x3XXdiMFRj6h3GT+vmGOChcNkGatuAUoSjOy3yuGyrxP4OUKgfK2PlJX7nvnmyMb9QGLAEAfnaZJlV/FhghtHx8naV6vVSA/oXHyvy83RgstOZ87lmEkeaF9eXebX4lX7HSyPXJFFWVKGrWveJD0Z4E6lTm/mayrmkcrs89lhZAsG8lyS5W8f53Z0MnJfl8PfjJ3NZvfPMjk+Ptbx8bFevHgxUuaMlXOzDMGCcqISsG1cFug6EshdH5aT93Xd400FPs86JQuzoVuuGTt3fav2UF7Zv9luOiTeX4V7pM2bgDIGmU7LYTBOVnIb81SYhL8rJp16kWzdiY5gCrSp1yQYDBkyzwz/uQ9NZBiu8bPHE2iSHdsBcCI0nSvJhDR+VATXpftc6reB8+zsTLe3t8PuYjtEP0qZz6F3SMM64T7lSMDXeBkf65XEyOGZfGgVw3vpcJL8Zb4txv0m4P1kgLsKbaQxZEwrExWbw7dkRfTy3KXHITBBnvfRuKXxzHIOD91Bi8VCx8fHo8dT0thcR3by1dXVyAg5FEyH5O34llEOmclgCD7Oj+1PhTKocShMeTp/yjFj1s4rY8utIaXP8zoC1tS9dDCsXzpZ50lH6f70eYY6KpacToTHcsvz1GQp781z2W4eI5imjLJ/02bcZusiwxNJTqhrCUQO/S2Xy5EsfX3ahs9l/S0Hhvu67v4ZKOfn58M9ZLVe603ZcYmvHS8XF7hMP/MnN0Zl/X08R3K0LdoNbZZY0gLr1PXqfJWeBHBbkLkkzOeSXTqlh+NvgnwCbzJiDv8cX/UKEOfpfDK2y/yS0dsxWLHJAMx42M7ZbDaavHNsjeXbGGis0ngnm/9LGsWLyX7cPj86dj6fj1h412221/d9P6ov48gun/2WgERZ53AxWSrPMT/KKRPv5fCefZ/5cXIwRyUJgtYjyrdyhpk3vzMuT6fGVUBJXCoZVOWn/kvjcBhl5bpaH/3bQJw243Ky/w4ODnR8fKy33npLi8VCV1dXurq6Go1a3H6G2GgzljN10+fpMJ2f5yDm87lOTk4e1cnrvjnKMLmxnDnKljbP4Kc+kKQkYFdYQ7tOW5wiGUxT4J7pQwF313V/IOm1pDtJt33f/ytd131C0q9J+qykP5D0+b7v35/KJ5lfDit8jOCLOoyMzsf47Xwy3keBVuU4kWUki++6+9AAmZU70UDPhxeRwWZM1YqTRk8j8HWSHikXjcLy5ONkWSbrxEkYOy3ma6daMYeUXdV3VZ8Q+Fug1ALydOSuZ5532xIoXHaLRXHkVT0LIxk/46qVbrYcWZ6j02/JjN8c3dFxVDKnnXCjlHXVvzNPtikJj59O6Y0ylC2BN8MVlDEJCoHa99uGvPzOOupzDqFUj2dlWIaT726Hl/VdXV0NedN+GF6hjvi8yyA+0EYoj9SBioD4+Da2LX1vGPe/3vf9n+D/FyX9dt/3X+q67osP/39+KgMKwx2VTCFTDgUZo6Jht4anBAcyqzTqDD2kAlrJ0zD8ZMHWMJfto1evhpFS++H4Vnznmdf4XhsGGbiNj8z/8PBQs9lsWHctbR6okzHMdKTZruyvCvBbfUMnlOX5d04yZr+6H3MyqgKidHj+sC/yWRUEejptD/c5Omqx/0o2lMfUMcoqQZqrPfKYX9bhujtZTtZpxvdz8jnZpXXQm9FyopN6wdVK/nBOJWXCidP1evMQr8vLy2Hyki8gMdEhFlifXN+rqytdXFwMa7G9f4M6RFJAh8z/tnc7p7yWfZwhVNerBeJT6aMIlXxO0o89/P4VSb+jLcBdDXUNxoyptpgXBWyhU/hWkjR8K7SBicaXwiTIuaP4IBmCvL01QwpksgYdOgEaYIZcCFz5TA4aFPOnQ7Hiexu/lyLyNW3r9VrHx8fD0jhJw3JCx68ZoyQYZT+2mHbFoMmaadw5Ako2l8DIEU2OYAis7N8EOwJJZUwc1bGPyLq5N4AgR11kmUk0KJ+sg48TjJm/5UKAZgybgMtHEzgvT9plbJ+yyzr5WpfJuQ6WR12ns0jgYn/lKJHhPzpY689isRiWrdpWqn68ubkZdkfzIVEkgbSTbLfrwbmqdKCpp28Czrtc+2GBu5f0P3Rd10v6L/u+f1fSp/q+/8bD+W9K+tROGTUAoGKQFYBnWCG3upNpSeP4uDvAcbCMAZJFz2abZ2J4Qb4VjUZMhmMF9Cfb507PnYdOXdcNIOr6UbkJ6C7LQ0TKlYbAWLfr4Jg2jT0ng+kYE6Sr2GX2V6XcU7pQAVs6Yhoc60FAZTijNZJrDWfJvAmabCfB34Bo0PJ1BIBsd+pC1oejOYNuEhvqDvWVIzKOEkxGuFrIcyEGNO5zILhRP6xrfITDYrF49HgJ2intMYHc5fk57Lnqict42Z9JHGiTlDvxwHrBEVeluwxFpf1WpKD6TOl06ty29GGB+1/r+/7rXdd9v6S/13Xd/xWV6B9A/VHquu4Lkr4gSaenpz7WNOop4Ob1ZGEEfWk8y8uOsXKbdSRrt7FwuZUZjfMwgLvONgY/Q5isOzvf9WdbXRbDRmRBFRs303d7PGveil36mDdk7O3tjZbAcTi8beTjfOk4qmvzPx1AZYQMAdHYaCiWiwGNwOAPl3NxSMtJMeqZ+4iyY7+k82C/W1+oDxkGyz6nbDL0ljqcYQDWj06D/cB+SiaYOs1QD8t0O/jbm2P29+/fGsXX6LkcOwCyY3+46ati4rye/cO6+17O5bCMBOeWDNJxV3ZDx1NNPpO4TKXW+f9fgLvv+68/fL/Xdd3fkfQjkv6467pP933/ja7rPi3pvca970p6V5I+9alP9QmWFVBbwAnIPpfXZow5E5WdbIkTEdWmE6dqx5/vnc1mWi6X+uQnP6nlcjna6ZX1TOXkLHwVC5c02orPj+vACRdfx1ghnUjGNM3C9/f3h1go5WVmS4Mh6+W2+IppVH1IQ+J/50nmzNfJZWyawJ2gypGDNN56zsTQAwGUgJBsMUGTwJ1A3Yq1c6RHmTFUQLDIdfsZBkn5pkPgNbniouqXBCrHibmSw+EZL33lrlHrH8mC9Y0gzn5nSMd97XAIAdWhwOwHX+NEhkzgd79Tlr6e5IHgTQfgaxii9D0kAu5TfrNuPpa6mukDA3fXdceSZn3fv374/W9K+k8l/aakn5T0pYfv39ghr0eARkFIjzee8Dfv828KJw2xBeYEHH+TbaeX9jVmpE7uyBcvXuidd955VF7FQt1OxpCl8TAyQwCUD+Ot6/XmYVnJNJJtMyboaz3MzXIMFDZK9olj5fP5fBjiWqY5usm2U25uJwHEhlY9L4QfTqYRFPP6Sr+q+LPrX+kKATB1wfl6tJKOwf1Kx0sZOI+US+ofQYkgxBVDTJYnQy3cpm6HxLXXjtvbObs9HplJ9zFjPzKAq0toT5zboT7QobCtkkbAzREHR4DsW+uI9a8KjaYTZb38n+y7BaC0S4bR3HepH9Zjp8QD9vsurPvDMO5PSfo7D4XvS/pbfd//913X/Z6kX++67qcl/aGkz++SWYJadiRZmo/ntTZ4hj7SIzL5OIWcLChjgwR/afOyAna2/3PYSGVj52T5+SICJ4KXyzSjIAjT2Lw910NIl8d7fa3L8wQmwydWbv92HPzm5kZHR0cjxt513bD7zuBlFsa2ViMgttWTZOx7GiLzoPFSD8iuMiVQM5803DSy1EXGPw3WqVvcOEKwImiRjBCkKyOnTKhbBGOODvidbSLpcPjDTNmgza37nMyTNsoWZoMAACAASURBVJuWrJ8JUrPZbLQ7mPLy/XxmOm2IH8qjCgNZb5Ktuw4cYXOy1NdX/UOdSuJDkuljHh2mDjG/FolLMjqVPjBw933/TyT9i8Xxb0v68Q+ab5Hf8F0ZfQX0VFQbAMMUvo7XUplbZeW9BFs+e8EGzFBETmpVTiQZXxpyzqYnmzw/Px+tCHCohEPL2ez+TfK5+cfhGW7Cke5fmmuAN6M2K3fZvpYPouK68NPT09GLfqdGSm6T652KzH5m6IZAkUPaPGfdaMmfx5jSuFv9lmRhNrufyK7KmjJU6mbKwPqQAOwVHtZBEpacwOaoiZPUzt+65NUmBujcTUpbs1w5MW4g91tsrE/ZV7428007TPnyOtoJl75SHpZFhhszVp2k0bKjo/R1Ztu0M9a3hSVJFHdNT2LnZKaqk6ZYGpWbxs+1n3ktGTPZIllsgnd2sJU/H64jbR7hSo/PtmVbEsD4oTHQWJPx9X0/rG3la9fcTsZFzRJub2+HJ695hYHzcRv4oJ/Dw0NdXFzo7bffHlj+0dGR9vY2j+SU7kHEoZrz83Ot1+vhumq0QWNJJWYcmnLK2KNlUG1Rz9BK5kVGlUyJfZfO3fqQTplO2/WrwL4F4tmm7PMEV8qDMdokDBWD58oN6hlX6ZCJM2RCnWVoiHWqdJdM1m0lI6ZMeI7HuOIpMYKyYazbZVtG6/Xj17glsUiZu97sP5MH5+vyq1FTtr3Sx23pyQA3BcdjU/+lx0u4Mg6cCprL7Oz9GVrhtt+qbgRUd7y0eWaHpOFt1DSILDcN1fnmskGfT6Dyta6LlTgn73w/ldTnLy4udH5+Ptql5s/JyYnW6/XwDGLGLO/u7vTixQt13f2KmsViocPDw4FdMcbLlxyTAVGh3dYEXR/j41YJrGRo7v8q7lmFS9wXLV3j6gnnTXBJg8zwluuWS8UY0mN4jmUncGWbnZiX5UTnaZnTUXPVEMNmvs8PcOJboFgXtoVyaTknziMkoCfQElwzL8qZYQ7KO9nuer0eYt45MqK9kRj4f8vRsww65cw3nS3ryb7OcFnKr0pPBrirlOyKHi89IQ3YSlp1PJfw5XCHrMTGkMyi7/shHkeWnaDnuG4yDLKiTAnmqah8A4jPc5UFn3rGZX1U0L7fbLy5vb3Vq1evRrvl7Li8LMzDS28+4uYdv1/T956eno4UmjJ32QQvy8wA33WbJxNyizLDJmSUFWtj2QwbWRa+l2DS0jkaUY7AkkFVdeNx60WyMt+fYFh9Ux+yfj7PJ/YtFotB9gajXPnEfnF4b7FY6O7ufju4ddg7gXMkx3rQuVq2lbO7u7t7RI4YjksbJ7hZZxwG4f3uS/YVAZhOI20zQd/3cbexy8qJzyRSGZpL/aqwiakisZmeDHAnEEvjdZS8zqkC5mTUvt4TF1ZuDpnyOufJvAgOGQem0tPBGBBzpUh64dZxaQzmVniXSbDmG6fJ8iwbM+3VaqWLi4tHa1EJoC7XEz0nJyc6Pj7W7e3tYHDX19d69eqVrq+vhye0mXW7vQYRsjpJo2dOsC+q4TgZs6/JOGOyWsuM97aW5tHwp3SL+SYb5jXstxYrJ6PLkE7q7hRrY4zfgHt4eDjsHfBrwLzSw2vdXQ9JgzNm//C5H5zko+PmCh/Xw6DOkAlHiOwfO4McRbK/UhYkZa4PCQ1j+iRl0vipjyRjrhf7bDbbPPTK+ka9sC1TJ+lskmSyHXToPN/SgVZ6UsCd8WD/TuWthtkJwMyTSuEJNHpvlk0DZ/iBw++K9dtInBwiSHbU+s06WDE4/GJd6PE55LYRE7hYFmOTZIK8l7vfPFy+ubkZngNxeXk5TGIaPA8PD0fslgBAYHHbyIzN0txGx+cT2MieOMeQMVSCBcGlAlv2QWVErGsaMK/z/blypDJoMlyGGDIRAFJHyOANYPlZLpfDyImx3LQlAx+fB0/dM1hXk4kEQPePr6E8M7zHvrR+MLRHAOdIxnlZbtYdh/WslwRgOgguj801466v62jwdpkV6KY9uW2sb8v5tgCc90ylJwHc7rxkJsnA+TsFQyHwGBm0FZNelQrt/NLoUsBZL3ZYAgwnZMi2chRB4KaDYHttAL7WiurlhhXzM7gmi7HyOn87NLMzM7j1ej1MeLp9CcJe47tarbReb94JyDXClIm0eUGyRwzuf4d+bEAZaySQctRDlp6G5bpyJGe5t8CYxpTAn/MHrg+H4ll36zj7r2WcaezUaeqRPwZehrscrnMf+j4CLuuQoGn27HzIUHN1E50AwZ926USy4ONcQmnnzXXj1nXrC8G/kiUZOZ9DROZP4Oa3Qza0tap/pkA3Q0MkZEkQKlxJ/avSkwBuqV5hQKWqwJtG4GRmRIVn53g4TkW3Ajglc6hANu/jqo00EF5no6/WmNJAqVwG3mQ5OfxnHaksDDu4femYXO5qtRrCFV6j3febDTZmZmbg0n3og4blh+s7sY4cBdCQXEeuiKGcCcYEYIeAmFeCNQ2EcnGqNl6lMaWxU0ddNvuLoZJKDtSlipGxPqn7XM3EMIflZxnx1WKum+XG/FyWl5D6Rbh0oPlgKu7aTMaacqzqnySAzi51nfkk6bK8+VwTh3Msn67bPC7CIVPKyTrO58pUzs7nEnzZd3TodlKWC/XB17X6OAlFpicD3GkMPlax22xUi7mQ6bAzbGR8tkUleP/OzQDS4x1W0marOtf7tjqAys5vd6aVjA+AJxNyvnwYEHezud00AgIkDcT52ZDIwhmnzhDMYrFQ32+e+W2ZmnkdHh6O3s/nZGPyva4Pn9RWMRM6HwJmxiA5JE+HzLzoJKvhsJMNbrVaPWLBLLMKefi8gSNjuBVwV2yLw/+cM6icUrJptyHnAejEvdGGT81jrJfg6X62jqVe5kgx28VQBll8AiRttiUrt4f7Kay3fMIfSZf1jksBq7Csy0jAbvUd5UOHSR1zX6QeMc9t6ckAd8Vik1HTO2bjqlhyTlTy+QgJ0CzPv2nkVDQDXCokWTdj56kQldfOYb4Nis93yGcW52Qega+ajPE1nFShQpP5e3XHbDYb2PVsNhsmwLwhwyzHb8F2W6+urnR6ejqa8Oy6bvQWbxq8Qy7euVkpbwKPZVCF1dLB58jIIGM50NmmTrg+OZ/AfiTjZH3phDlZmiwujT/BhKGRrCvv42Qik68xaNkBUL+tO+4H/2f4IuucI0WXQZCqSJN0b5/emERHQaZNW6SN0tFSN6zPDNOxH62nXqhQrXdPDMk+pv7kQ+MYF7fs6DCYT4UF2V+t9GSA2yk9aRokr0vWVcUNc1hIIVVhEubjTuDKC4I2FcN1djlksNV1ZEUcatNhMCZNYCYAJCjY8KoJPg6lc4RDeXJST9LoBbIOgTgWPZvNhvXbZjp9v1kuuFwuR3FtP/ObDscMxG/SrvrTis95A67l5bUcirvtbF8yO4IMGSJHY9QbAmsF2HTC+T8/rEM1bCZwc5VHEoEMy1hXqjbQadMZUHfcf9QBkoK0vwRZfhh7d584tEHCRt1muIH9SCLidjM0xWTC47zPz8/16tUrnZ2djcKGZOXUHfcJyyUukZz5frY3QyEVGLdwbVt6MsBdDUfYiOq8lbISToKS9HiXVl7DDnMenNw0mFuRqDBkFaw/z1MBXJ9keWQcnOxiDNjL6Xg9V2Nk/JpsO8Mp6RhzSNx140fdelu937Yt3Rs1VyaQzfKFrN4u7/IM3AZgPsjfhpHsNmPNBLyKZXLpXToot5UGyK3f7CeWxWFuC7TyHp+fStVoMgGaQJsOiw7cq3+4rth55OMOzKTPz8/1+vXr4fvVq1dD/lxfT4BLMLUjpm2xX6yvnAD0vQl2dKRmsMQA6kkuXWT4xbp1fX2ti4uLEbP37+vr69EaeJKCqn7UGbJ9yiFHJ/xNzGD/5PWt9CSAe4p5UAEyhljFztKICJrVsCRBPz03gVYab//tuvHLVZOFpwGyXtVkpoHXy5y4XdeGmkPKBAsCBmfSvdabm4ecL+XsNhIweZ3rdXFxMSiwWbonfRn7ZLtp1NLGmMhwXT6dZjpktpusjeyLKdlh9m0OZ72sM51qArLrmnqZ+pXMmAZLY6WuZ/39TWfGc66Xn4+dIGj52pG5PLPom5sbvXz5Ui9fvhyA++LiYriXIEcHYkC1Xvla9hnLNfjmDln2KcE+wYz55j4BvgjEOmnwvry81NnZmS4uLkbzDX3fD8/j8QsgvIzSTHybw2W/EmOyvtVv6lLVr630JIBbaoMvUxofBZQTWj6eDI/sJMubErCvI1szE/Vnb2+zxTg7LRlwxvoYn/bGl77fhB6s+GxPDt8oGyuTN+VwlYCv4cqA7Atfx5UL1Ww/h9zelUpZ0hDNBF2+2+trM85YjZqcJ1lxFSbzfEbqEWXEIS7rzNhl1oejtSQTrkf2CetvULHM7MxyvsH5UQ9zCJ/hB+u4V4VYp7i+m2um6UAuLy/16tUrvXz5UmdnZ8MSUIZZ0hkwxMJVKzlic+Loh0yTjoRtTVZP3eJ5jiRsfw7HOORzeXmp169fj7bz29k4r4ODA11dXWk+n2u5XA7P1qGc3DaPGCo9T+eY8qCztROk/iV2VOnJALcVWKpBnInDVSpyMphqwikFnQrjurizCLjMyx3p+O9isRjlxTJZVhp2buu+ubkZ3qZjw/Y1jD1T6QnGrG9OhFWM0Xm4/QYtabw5g4woDc3nuB6X8qQc+HIHAnfGKisHW9XVTIn30xA4wiHIpu6w35KF55I6GzuBk3Xmb4bD/HKFBO000Mrxk4W6XNeHwJ1D+fV6PaznTgfsdH19rdevX+v169cDK3XIiqHC7EuGLirwJdFJJp158T9HfDxu2biP3S9ePWL2LWl4p+TZ2ZnOz8+HF4twhzHlZ7ClbJx3jvKc6Aypj4kTzov1JwGo0seCcecQNgGwAoIUCA3N13HJX95Xgcp6vXkhQg6NeU86C75wNxWXRkSlI3hlnQg0Hs5mbN3HCcoEbZ9n/pWRkPFmTDGXnLGOBNSu26w7p7PMFRBmKl5JQpbp69MA2A8ZCqIRmGF7lEFHnu32PQRm9gGH9EkECJpk/rwml+vR0VJeBjNfnyEGXk+9Y39MERtumjIx8FJMJ4fRzEo9J5Erk5g39ZsyS5Cmg/M5O1lfl590UFW8PNevm1z4nuvra52fn2u1Wmm1Wun8/HxYTeJ2MWRovaTT8+iQj+PlSCcXKlg21IUcJZGls0+39WeVngRwSxsQYcNyqNlS0hxiMnEo7evdOT6eneHr/J2ATEdg8KGnNvvkkJLgQQXPWfrscCq1Fc3Ka4BJwGYoJleI5LDMoJf3Spu3v1DBOblEObIvpHunmSyIw0iuiHF9zICSxRC0sy0Mr1iedozcMEIjs+FSZ1xOxuHpHCwLjl6SZZMBut0OfyXT5XyIJ26ZJ/Uw62KQSL1kPdjHrDMnN6+urnR2dqazs7NHoTTqSdpBnnM51EXaLG2msrOKULFf6WQZ+rEc3Oer1WpolzcTnZ+f6/Ly8tHu2qqddvxSvdQ0nYpH2mwL6+p8+HAun6MsaQe+Zio9GeCWxsMk/ydrail0Mm0eZ6fn8I2Azms4zGwxAwKd68NOlzaGaQXLthAIpMevQMp6Vcw/44AEXy77q9pKWdIwOfLIvvGH68ldPld77O1tnsfNySO3IydX2We5q40A1JK/5U2gt4PjlnC2N5lkOkj+rpyI28mQEgkBnS+By3nRiCuikn3j/Nzm1soF9wl3O7qtvsdk4+LiQmdnZwMbTeCuUhIF6iaPUU8pb8sjHTHlzA/7lzrP652vCQVXkThEksttaRuWK8MnaavWQTLoy8vL4V4SuBw9pI5Td6lvbP9UejLATWWxQJPF0oiTvfJYpvTijDERoNxx3HGVLIIrNSh8PhHNq0Jo3FMdmYbCOCTvI9Azb7dvf39/GNp59OLdZG57lsW2cPmhh54MAzFxpJLDZCfmZ/nwBQOWpQ2KCstRVMs5UXZkXjTs6iH56QAr587rCC6Ul/NfLBbDw7kScHIUlW2hTpAtE/QToHMuJ5mgZcCRC0cMzsNhhIuLi5Jtu4+qlIBL2aY8k3gQuBPQUmZsG3WW9yRh8xxNrqSi3VSEIcN23FnNfnP7uSeCwE/CVK1ykjYrxpI8uOyPBeNO4002aqEmc8lUscj0+j6XXk/agABBI8+n4RJIOJFI46QR0yuTAVBx+GAcf5xPsjS2lU6I61cT8LL9dJQVgLoPcpeYy3ZZDAkl2yfjIyutHCGdlHWjaq+TmSKB0b+TMbGfExSZdxqUh7sEG4aAXF4FNHSEvt/trnSF7UiwsQxJDjKs5tVNCRq0J7NRrkDh8rltzJt9xvq3rpUez/vQcaWO5rHKoVonGWe23XD3J2Wd97M82yRlzmep8Lnxvr/rutFKKcuOpLDSw2xvBdRT4L0VuLuu+2VJf17Se33f/5mHY5+Q9GuSPivpDyR9vu/797v7kn5R0p+TdCHpp/q+//1tZbjBabhOFfOhR5bqV4NlJ1WsyvdS6exN0/OlQTIv3+MOzrZ13eZh9H7yWQInyyIIG7xYBzKWZOC5gcAOIhlOpTRsi/8z7utruYyNBk4nWxmzQb7aksyVRdVwshpxUTdcto2GISiCKgGbRs0yXR7rnOGhDH/wdwU8roPBjkBKY88+5OoO6lMyQOdVxZBpF465e/KNzyapRmSVI6uuqWRRkaWsUzrOLDflSPlzdMydw9zIxZFDpT+sW450rU+c/CZJ6vt+kB91hc6z7/vh/iQQuQs1ca2VdmHcf1PSX5f0ZRz7oqTf7vv+S13XffHh/89L+rOSfvjh86OSfunhezIRQGisPic9Btdtw6fqXEsoOeSTxh2QQ9fMg0bh8lLRs365TK8aZbDeLp+Ti3wIUHp8X+fnauewM51gLv0im/KxFsg53pf9Q4XMtb6OPbPNLiNl6XalwdPoKD8Oby0zO0u2n0bKOjtf5+H6cf0zyyfY06DTqTF/6hj7weBN+XDU5ev9SF+OBKhvSQh8LNloi2m3Qgotfa/Os20VWFd2RLBMGWf7SIi4KS0nIZk3HSrrkOVJGnSGGOBRJ1ehUI50wNY978ZMp8PRcMpgG3hvBe6+7/9+13WfjcOfk/RjD79/RdLv6B64Pyfpy/19TX6367q3u677dN/339hSxsjQKu9eMQECGhlSJguhGspSufO4O7YVs6zqweV1NAKfkzYMMEMr/hAckln5Wp8joBIgWc8cynM1h+uRm2Hcfj4wi/3B+qbSJxiSWbgf9/f3RxO5yWCzr9PQWBcahOVOZtt13RD75xA2+7Nie6xDgqfb4R2jGb8nczeT9r2um51e122G/blhRhqzboI1R27+TbChLlrfEqxzUi3lPiWXqZQkzPJ2XXJkbT3neesP+9nOh3bHlRscaaXuUz8rrMn+9THP9XillOvhunoy1Gycox+SLdqT781XErKtrfRBY9yfAhh/U9KnHn5/RtJXcd3XHo49Au6u674g6QuSdHx8/EjYOdxKZclhk1Sv5X4o6xFL4LkMs2Q5NMDMP/+v1+thoofDXtffBpXrfG3I1VMHkzUlmyNr9XkyAoZvKsZYsU+OLqZGG5RbtofHCBJeJ+s0m80Gg6CMs//ToVEvbLysCxlRxbwsF07iWZbJignuHK04/GU5ux3e/ep+djmWDz/WFz5vg+vB6eRdDtuVfeh6Ox87Th/3mubVajV6eFSCVcVyeb6VqLeVnjC/dMRJctLJsl8I2HRw1hPLkv3KERnLrJgv68pwo/NIskTCxFGS7dWPQXaedtYuLx3nh2Lc21Lf933Xdbu54PF970p6V5I++clP9m6cND3JGHk0lai6np7W1+TkDYEjwwp0KAl+9NQ2RgK3DdzLxhizZJyTcTnp8bZsK4/DH65LxQjpDLm93eUZDPiuSjJEypLMIZcrVo4ljYHMkA7FxyxbrylPR5H5pw5knyebTX2ogJvrwtk3vJ6GZ9kRlJk/n3liPWDfULft0FwXMsV0iDZ4t4+T3AmuBnPOq3DnIOXJPKjXrWMt8E7gngJ5lsvrWF6GsLjs1O3zVvecFK8ISwJvlk/dY118LuPSVSgrf6ejZbsSS6aIpNMHBe4/7h5CIF3XfVrSew/Hvy7pB3HdDzwc25oy9jQVw5YeM6Bqooidzc7KfFsdlr85/EnGR5ZMAHZIwHE4xi7d6e5Q7/SksXJXmMHbS/38yclAA6AN046AzJ/DOy6XIrAbUKSNAhJYqdjJhsj8s5+5eYfHLL/lcvnIubqdWZ8WcKejTd2RxmEiOthcw26ZELzZPj/bmU6L+S0WiyHfBFfqpcGesqR+ect8grflQiC2fN2HftSAdZPA7WN0rrSNbQBXpSlwpy1l/2RYxP2U+mV5MVa8t7c3xLoZnkhHOVUvAmeOHDmiy7kMArWvp+Olfrj/GF1I0kTsqdIHBe7flPSTkr708P0bOP5zXdf9qu4nJV/2W+Lb0uP1m9IYqFuNmDpuxeAxX89vd4b/E9xtgFRmsuEqT9afQOPtxvv7+8ObVNz5ZE18PGrXdVoulzo+PtZ8Ph+9h4/leQJEGu/8qiZcqDw5nKPcCfg5IpDGDwxKh8cYOkdRyVhpfB5FsN848uGxdFRpNE4JAuyfdPrVCII65v/VfIeZN9u/v7+vw8PD4a3rZt/JcD0Ko9FatlVfuzzKmPkytu46rdfrYQOKzzM8QmDLkVILtPObdpY61wJJ1judbPYTgc+JE34+Z903086t+ynXKkxCXcj8k5j5HHWPfWUnSXKZo3q2MfWrlXZZDvi3dT8R+cmu674m6a/oHrB/veu6n5b0h5I+/3D5b+l+KeBXdL8c8C9uy5+NzcpWzGPqXicCqqRHgDHFFAwK1cYFnyd4sw4GUSqEy2QH+jifHczOZ3zPLKz1jGAqANl3rmjo+/EjYVtyS+BkP7BMAnrFIpM9pHG0mFzXdcOT6RwvTmbG8jOvKtGQ0lCyzjSwyjFwoisnHO0I2R8Z/qiYeYa1sj+rcAYdn4GfoxGyQIbEDGIVQ285w23ynUpVn9Oephyo7/E11E+eZzu9LM9yNRmw/FsTsVV/VyMC21Y1mnO/Mdy2Xm9WIzHe7pSjVerohwLuvu9/onHqx4tre0k/uy3PKlVgwe+WZ6+8NJWa19JYqzxSoD5GtuiO4GQWkw2Jntl1YdggQdLXpHKu1/fvOlwsFqPty1ZQl2ml9ct+M85mQ62WffHalHeGoXwuV6D42mSo2bZkrjTKdJAOCTmkkY6DsXaWxbawPhkqY7k2SO5mS7lUYQquVOKohDI3EfCHIJTD+GwLZZIAzr7gA6X8YRl06Nn/lcxSRhVot0jVLv9JfhK8KX/qB/OyPWVYjjrCCVfaDp1cReKqtlbOhon6zMRRUZ6v2rQNsJ2exM5JG0KV2KhkZryGrK8F9mQy7mjp8WoUesucpOO9yeQcsrBxenmQ8yKr4nCLZbKOBguCNBN3hTFeXW3WYRkMkVSOz795PSd6psDS5WV4Iw11ag6DMUZPwnKSKfPcBixsW8uxsM8NIJl/EoCcPCTIWB9z5560GUlRN6p+SaCmHEwO7GxSD51yiE/gqth2RUQ4smn1eSX31nmfS/acTDq/+du2aTnRvuiw/J/x+5ZjIobwHGVim009Wq/Xo9CN68WQIpk68cfExrImLrXSkwBuabyhhkLJjmsxk5biJQun4Vcx1wrAq7Jyiz47yEbFWCVDFwS2HBonMEmb4RaBwpMyXIqUDwkim6zaUjHRVO4Eo0rm28CQ374+5xbYJ2x31SeVo2YZGQpq9WOLXVbMquUYmE9FAMh2WY8cUksajYjc/qpNLjd3WdL4U39yLoNtqmyram/FNCuZvEmq9CyP81rmz41lHNlWbSJjzry3OX0ea/W183M5vI5L/rhyqZJnFcKp0pMF7jw+5Rml8QPrpcfeO8EswYTlUgnoDclWOOHgDkm2TEXLIau0YV4expFlOU+yTy8xpAc3qHOddAU8ua6bMVjnX8mR+VBhq2F7yjv7Iw3T8kqWk/e5jZRJBTa+L9l8pnRoCQ6VQ/A5yiHbz37j0JyhCepLtRGIMW/GZ8ksCVZ7e/dbq/maLcuLupYOuPqwnflNWVXXMlG2LdIw1SfMsyJelXNJ/W3hRNYx68Cyqjrmb5IfxrMZ9uFIiNjgBQUZ4kn9qtKTAG43jP+pLE5uUMaGq/zyPt7P4U5ew+s4w0tBMobpvHI4nG2qFIntoydmx6XBcWs5l4FV23ydfA0nzhjLpcxTHukAU6lS1ml82a/ZT1WZWVcnTvalQ0+5t4yPdaYDq2KU2fZk0xmCshP2Lko6/Co8lWXR+SfbdriM4Mm+9GvjCPhTI6eU+zbgftOU4D2Vqut8rAVmVf1bDqYiDFMOpdKdlpN24ioTtz9DlbT9DJURG2znH4tQiVQv10rF8+9dUwqcRrbNuDO5IxKkq7hYKlFLgf2yXO8a5EoT1o+TLFRgMnnWi+3NFQQErb7fPKmtYjoJUhXrSSeaeeUkMeVBGfkaskqep0yqbfhZZ55jXfu+3uRQsTv2n9uSq0nIoPkShRyap4FW5fq/dWJvb+/RaiS2kWXn5CQBj3XJfqr6NGVHGe4K5GlD2xxpVT/mk22o9LFVx5YeVO3L63h8CoO2OUBPIEubyVWOwm2TnNRupScD3JUg0phpuOk1pxSP3wkiea9TsmXGZLN+CWzsDDN3lpmGxZ1vNvpkGbe3t8PSODodTsxk3C3Bn/m1WIjrl8fYFubD9li+zDv/V6ybqXJ21bEqnrwtb/aBQxDsv5xQ5blKZzL0wQ036SA8R8HHraYeb3NE1EluwqqcVBVinEot8GY9qnym6szjlY5lHtLjlTyZx1Rdp9qY7SF2tOQwBf5TbagIKMOYxAoSrsSQKYL6JIC7ZZA+N+WpK+ESLHlcGk94tVYM+LrW8LmaYEgGN6VUNMRcmsWNOmTeObnkFSTVOt8Ej/X16AAAIABJREFUuRaj2RXo+LE8Mn4qPX6CY5bnayu2Wzm1qi5OXNUz9ZCeCnSzXaxXgiTrknMUdCp2AgZvaTN89pb4+Xw+vLR3yijpsPzfu14zBOJ6cY6DdcsdkVNgVNlBpSPbnEDlZLMsXpvlcxK/qmsLVFt1ru6vbCPzmsKdKu/ML8mXryNR43nev22e5skAt7fzchhJ4zND4nAzvXgqc4JoTjolKLcAvGJ2OWFpUHO4g5NMbhfvYxkV+zajyjZcXV0Nw6x86DsNO1lklp0yzu3OlTxaToDAxnhxS/bu06oeLeaS1/H/FMhM1Zdts9y4Pjuv9XW56sPr+r0RxteTGJiJ812UVcr2WJ9cDsNiXAJKRlfNfWR4L9u/DYyn+mBK5tkmlpupFR7MtGt9s++yDtVIZRentGtZiU85UmY9stxccZTpyQB369nDFSvblnwdF78nICeoZTnJenIYnWDJ2BVjVZztTgUh4GfH2+BYx1zyRLaerIpMLx2O25CgRgdSyTQdTUu2lF2ynMpYKtBupVZ+eT7zdZ/lMTJTG1aOtNLp+Hq21azbm2Gk8TPHZ7P7F8tWoD0FBqmHLt+g7d2QWT86GU5cb0tsD8ub0g2mFhDxPL+dto0Gsm1VHhW5aNWvut/1yDIzVccrIlhNfGfdK6fxsWHcmSog8P+M/VSTLj7eEpJBkEOyFrtjRxNAadRkV07O0+yN4OplXM6T13ZdNzI0AwQdAcG7GjoT/Ak6rGPK1gyfedghcWY/ZZ2gSzaRIZFk41OpxaLZNym3BBvWO/uwciKcxMs4tDR+8UEao58LYsaeL4jmEwSnRgmsYybXj8/UzhBagn3OgbTYYJaffZD93HIoridTy4nz2taKkRZxm2LRrbblKCDbOOWUqvbzHIlR9ckRHMlftq2SYaYnBdyVUlSCTbCulIcAQkX1fwvO+VWgnQaeRpGszKDr81yHnbFsMzTfa4BkCIYOxvWtYvQZ3mCYwrFgy8V1TrC3ciU7NROt4pQVO6O82c5cfliBN/uXhlCldArsq8rA6Ty2MTJflztFWR8Ctx3szc3N8P7Bg4MDLZfLYZ21XxHGZ3/vAhRkzdSflDP1omofbaWym6k6ZH5V3vxdjVAqp5+kI2UyxVBbNl+B9y4gXeVXtTWPtQif9Z4j3gqrLK+0zyq0xfRkgNuGnROG7AwCVYsBtrxyertWR/l8GkDGb3N22Ncwz0oRfa4aOSQLZJt3+bQMKmOHTgwN8AFWrAcZd0teVZkJ3JT7LqBRtYdyrPqsCjnxd2U82ceZF/PLSelKXnysaDL0rut0eno6WjnUSnQIuT3drDtHZi1gTjlM6X6mapI2Uyv/yj75P4kM217Vr2W31bFWnpWNpM4kwFfkoJITyQZ3sxKcc9RDWWxz5kxPArg5xCCAVAbOBpIJ+h5+p6fl9xRryKFPJcx86h0VxspOQ/c6bTImbshozaD7XCten4ZSdX4VSmFbPaxPoPf5SgFb8spjHHU4tTYrEVhbbJ5tyHrQeFjXyoAJpmTRBBTufpQ02viTIxDWh09gdB38kDDrN2Pfrfb5Wd8Oi9zc3Ojq6mp4kBhZuOuZDmEKvFvASjmxL6ccTVWeZZn5US7Uueo61qsiRtV1zDfxZEo2rbyZWgw8Q6N8FjrLyd3NWbcMa7XSkwFuvsGCT7cjcBGsbRjumJzprzz1lCCyEzkMttImoLUYpOvJ55l41UG1PGub4VgOLC8dSksJfSzBPQ3TjpPLzlJ2dEotxaKR5y7HKsaaMrPsnTgJ2DLqZMDUm8oR8OP+zZh1zn/kCIv3s200VE5Svn79ehQyOjw8HG2aybL9GFY/Rztj22TbbGsyP+ZPVp4MvtKZCrxT5wgyGbOdiuGyr1LPWgSE+UydY50re61si3XOc8yjkgX1lpuhWA/3H0Omlf06ZWgx05MDbk+ScU2q9Ph1WfymArJzKvBm2gbk2clkNqwfQZTlcHLQ25JpVOkMaEwVy8mObg0dU7atRGP3f4IHy8/HuLLsKTmmYbPsjFNPMZ1t51Mn8lhlxJmvnSu3q/MRvb63Mu7URRKNm5sbnZ+fj3Tn5uZmWB7oN7b7fgN09fxsLu+rhuGeDDV4GEi4vrsC7bQVtk/arElnOzOEM/Vp9WeC4q7nss5VG7Ls7KMPYzu8JokPnxBatTlDW1X5HwvGLW2Ycw47aASpOE6VACpvua0jtnWk9Hhdt6TRcixODJIRdV03rDzI7ckMn/B3ho6cKnDaBqItxpmz364DH+qULKgl+/xvEMrJl9azzD9sSoAm0Ge7eS31jeutF4uFFovFYGiMx7ZAKR3Uer0eQhz5LO7Dw0Mtl8vRI419zkv9uCbbOtV1929GWq/Xw0uJmS8dUT4aIK9Lp1Q5wEqeOeRnPtVIsgLfbaGPFvBln0/pfnV96/yUg0k7SwLGY4kDFRHLPtjmdDI9GeCm8XBVBVmEr5M2jSMr53NspwTyJqlSWk5U8hyH5gRfX2OWlM+6IBuqAI0sj2Xs0qY0iAq4CeoObxBo+75/BLbMM1fMMIbvNhogyQSren2YlIZRATQ/bDf1z7scvSKA1zJsUa30SNn6Wm+csky7bvMKMq/vto7wkxN4fAFxzqdY5n59l//nxCVHQdsAIp2U/9OZ8XwFyM6HKa+jXrUcYoLsLqRlCowr0E57yetZT+pRBcjEqNwIlY6uatPHCrg9KeSdlI4Tkr2SnTP2KG2G9AREpxTKNm/tVHVmxrt9LMty29br9SOwIvPOejmPZMq7MD7WoWLnbBPbYJDmSwsYBqoeRWr5c6jPUUS2ifWq6tcC36x/lVzfSn4sIx+jy4lxyrTvN28aosOkIVaMlU/5Y/J919fXury8HMnG+l5tW5c02IQnOHnvbDYbHlZmti7VYSq2LWPb1TG2IVliTqTlZ1tfsU+ybruA8C7ltO5tpUpfqmsStCudpoPPXa/p9Kq6fijg7rrulyX9eUnv9X3/Zx6O/SeS/gNJ33q47C/3ff9bD+f+Y0k/LelO0n/Y9/3f3aGMR0MMG0CyE2nMzqsJMzeYIEehvAkLrxhDXs88KuaVBpHhD7O8zJNtIoCzHWlMCfS7GEAucyOIuw6WeQXc6SCr4XO2qwJzyo3H8jzbkEDb949HB2x3zvo7zwpsnR9DYRVgSeO3emcftEJRfN0Yn/pI2Vg3PBKi06Fuui+SDLAtCbKUI6/jN0N9vK8aZVUMNlMy2ynWPHWMqUVOKv1PYrRLSp1Mlm39oWOVNDjpavTUcnS71mkXxv03Jf11SV+O47/Q9/1/Hg385yX9BUn/gqR/RtL/2HXdP9f3/Z12SBXg+L9XZKTgq51j+QZs591KuyhPsukpICeAuqNovNxF57wTMJNtZ/4VaFftnWo3J5xYN7LRrHs6HcuZq1HsUFvycXtTntVnm6FSDtX1+TIKtpcrXtIYc3esWSzbb2OkM+CoJJ2D68dnazsPT4L6GudJpr1YLIZlpbkTz/HuSi/T2UyB9y665PwSjHZJ24D5g6TKRqrzLVuqfjtNOQXqEm3I8vBzZKbCSgncu8pil5cF//2u6z67U27S5yT9at/3V5L+367rviLpRyT9z9tuTCGQeRhEOFwlMFLJOZEzVRa/p1hpJdCKmbpO7jjXjw7E58laWWfWvQVclsUUcG9TgjRuAjhj0D6f7DEZmO8nq2T+ybKkDZPMfFPGVTuq/qM8uLLCZRFEvXrEdSBrJnAzVm/g5hwM28/whfPxUwDZR1yJ5DcaSePnemc7Z7P7x8EeHh4OK5Nms83KKz/kyrIwo+fQfAq4U4Y8ln3N0EtruL8rOLYcc5XPtuu2gfe21LLvrEfavrTRL67yybh2ay4gbfV7ybhb6ee6rvv3JP1vkv5S3/fvS/qMpN/FNV97OPYodV33BUlfkKRPfOITo4kgei0bgoeBnKjJJUoMK6QAWv+nFCmVucoj2WM6mMrb55Kt7EznSea6DaxYZ39Xitfoi9H/BAGWx/am8m2Tue8hu09nwDa0DLdisaxb5SC4eqPruqFsTuplHfi4gK7bvDU8DduATFB3Ojg4GCYn3bde/uclgX4MbI4GvG3en8PDwzKkljaQgF05+QSMjIFnnxGwqacJ2lN5ZN9uIxa7ss+plDbQGpG07t1mQ5Xsq1h21Q+sx5TMqvRBgfuXJP1nkvqH7/9C0r//Jhn0ff+upHcl6Yd+6Id6siN/bCgcFppNOPGZ1Al6LY/5UH7zmAGzUvIEhvzPejqvnJTMuLDBjI+t9XVub7WaJo2ToFE5L7aV3zlBylEMwYsslMrKiRoDGPPm9WanzpPPc2D7KGvKye3MlKMchnYWi8UI9GazzRuHfC/nTygDAzb7w/U2o+a91ls+58QbadznNzc3Q+jDa6zpEOxo/Dadatkgk+tzcHAwrCFOh1b1fcUCeZ4yZ9unQLtVvwTJJFiVDWUeU/pcHZ8Cbebbuj/zymiAscnJYdsplp15f1Dn9IGAu+/7P0aj/itJ/93D369L+kFc+gMPx7amZF40AhsrGRY3J1hYNOqcwIz6l+WTQVcMt5BD2Rbunss125IG1uW2EOR5DWWRa2apGBzST7HWFutgHmTUXdcNjJFLmTKskUCczJNGSrCkk2UelAfzqEYWFRjRmAxoh4eHms/nuru7G9Zquw8MrJQBy6Sx+jViHgq7nuv1ejTZaDl5Pbbv4/Wso6SBfS8WixFwk5UzGeQ9IVZtd69AowW4FVhW4ZVWX1R58FjljKcSr20RkarMSuer9GEYfbY/nWE6Nut5S4Zvmj4QcHdd9+m+77/x8PfflvR/Pvz+TUl/q+u6v6b7yckflvS/7pKnDS4ndAze0mNgs7F4AoBsSaqfU5CdNcXKq+Fja1iYQ1bfb8bqZ1MkeFWdzI7N0FHLi5OlZ7v4qdrfMm6CMIeAdKzZHwaTbBfB246n68Yrh6qJwmTgGZrxfZxUrT5myH2/eamB9c27dj16sx55dMC17dy9yFGOZdP3m9fPecMViQX79/r6emT0s9lsGBnM5/NRPZNxuuzlcjnc6yWGuWGNo9gKLCpy4jbxreW8dyqf/J16NQXuLRCugLtyGLuw7Lz3gyTiDPWT/1lv4s+HAWynXZYD/m1JPybpk13XfU3SX5H0Y13X/Uu6D5X8gaSfeajwP+y67tcl/SNJt5J+tt9xRYkVI+PbZDK+jqGRqOvwO0H8oX7DdQkIPs6hfaVoCZS+NhUhY92+h/Ev5pl1orF0XVcOvVogTllkiMV1Y1t53HUhy8/RDkMzBlPKrwXcbC9Zq+thp80lVQTsajSRKy98nC+RsA4dHh7q6OhocKIkBFyO53wljXYszudzrVarQVZk45S1JwunnCbl4zYsFouRbDhySD307729PR0eHqrrOl1dXeno6Gik+wkimSrdTT1NXWQfZH1a5UyV5+MJztvAt8pjl/s+LGgzUSZTdfG136u0y6qSnygO/42J6/+qpL/6JpVwhxnM/Jtrt6mIXNRuNittjKwC5qnOymsqZc98835/5/A+Y6Y5CeYyGPLJPMky0/ir5ZCV53e5CTK+no6QMeQEd+fBaygbgjdZaMo2nZPbWIUEWG46AdcxH2iVT2hzOOTw8HAIYTjW7OscorDDcpjFy/ccP84RkxPbwXoRxK2/lhON+eDgQEdHRzo7O9PNzY1evHgxPDuFywUzOWbvsArj63xeTupHtoH9kTv86ASqOmwDQ+pe6j7rVI2Wqj7PvLeVsStYb3MSlcNK8tfKh7acujMF/lV6Ejsn3RAavgXCIakn76hUZlc2iBbITf2vGHWmXRUhyzZz4vbvqq2+x7LI+vnaimG3FLtlaK2RCZkz2SSH+Rmvr2TDOrdkW4GCQXabESSIm7UTVDmK8AObVquVXr9+PWyKcP0cmmC/GuwNvJwQd7kZ/mDd/NsOhcv+MqRk/V6tVrq4uNDd3Z0uLy+HlSdvvfWWTk5OHj3ThH2VztCbP0xwEjRyVOs8eS5DAVWqCE7VZ1N6kiRCUmkXrbKrMlo6P0XktoHuLs6D19thVyHO1r27jjKeFHDnagGCtZOZOEMr1Wxv5QVZnlMCSwX6Ffhs8+JUFIM21/By1QPBhvWnkiWoZ515fbKVKSWonBn7ogrppKG1QLySBWXM4XgqeTo1lstRieVxc3MzyJqMk3mu1+vhWdaOD1v+XKLn/L0SxPMsfsQqVw1Qhtk+9h1DDhV4eALz4uJCV1dXo/qY/R8fH49kz80d19fXevnypd5//329evVKFxcXown8ZHoJ0OyT1MFtep62MqVvPs+4fa48qvQqw1KUd6ucqfq+CQv39Qx/VnY2VY9qWWBVz8y3lZ4EcEuP2Vir83JiiGCfw2MrRSveXIH2VAekQe4C3jZaxsIIShVwVx1LAOBxnyOIUHYM3aScnU/l2HhdDrV5X6W4Uwyixbrp1OgsKBu2iUDYdZs11jk6qFa4XF1dDXU6PDwc3s5ucHZ/XV1dDatNXEb1Yt5sC9uXjJjA5TJdVwN3121W83g0wdeg8YFUL1++HMD+W9/6lt577z29evVqVO+c76kczbZP1beVHvhYBeb8MH5fTSrzXtt6RawqvarqyT7JvHdJdGxVm9y3fI5Sy0Em+co256qtKj0J4LYS2iCtmDlk87XS+J1/vifX1VYxSOfRUjKmN+nYTDmRk53hujDeyfa1PHIqbwXKBIjKQaWyV2Xzd07A7MrgqZCsU+ZD4K76bGrYXNWPDJ7P+SBgOhzhZXcOpxiszWLNWl0OtzGzPgRutjUn0jnKoiF3XTdMfPq36+F13X3fD6tNHH47Pz/Xt7/9ba1WK333u98dwkDS+GH8rGMFpJWep/7wfvYfVwO1dIQrh5JlczloC7inRgIt0pfnsm4t8lXZSEseTtTTqRH0NodGcvKxYdzJniSNQFwar4GloLiygMuwpDHrS9bh4y0PXjGMCgx9bhf2meDUuq4qZ5vSVnJNGeT5zI9toPKRvdGgMq8qnFO1k4l96utcDmPEu+Rn+XLFSS6r4/dqtRoA+fLyUqvVagjJ8Y0zjpdTFtI4nJTAXc1XZN+5LD9b2ymftX15eTl60YMknZ2dDYzb333fj+LhKasEkJRpyr8FYtSpFmNmHiRbPMZlpwlYlqHlWi1WSKfU0ouqfen0p+5N20hbz3q0wLqSX84dfSwYt7SZRHLy7rN8qFF2DuPbVgAv3drb23tkeGZb6cGd3xTTZbkVSDpVIFzlx+/q+iyzmnjNNmT+Vd4ZfiDwJKhNfXJOQhrvxsuJxzTqShZT4J9hnzSCar6Da7AZauu6bsRczbQvLi60Wq1G+pL6w7bTWVd1z6G1mRmPW2ddro9zdOBVOgyTuO9sO3Y+Dq14nTjLoryZR9Z7Sg/ZnikAZ14VcEsarQYigFNPLYME7UrntqXUv9SpvGabrbs+Lb3maGtKhjniqGyL6ckAN8MbfB60hZadQzbG2XkbBw2UKwMYI7ZAuZLBefuaFgBSoRNkeE12RBX6STDK1SMEBtavpawtRUqFYDta92YZVMbW/awfY9dTLKJyPAQYgmV1bQ5TpU24yjHsXLHiOLakgfE6NMKX8ea6+6n6ps5mHVN+lWwtr77vhyf+zefzAbip2+4vjwS8nNJA6HwN6HZmWW7uXK0YI4mS/1dsked4XxUuyacqtnYaO6RoR5P5t8jLlM5Vdsv68zuPZzvZxgyVtO5jOy3/DBu10pMA7r7vhzgin1PMiTsqqRvKN5X0fT96NViCi8MpFgZj6jSYBO0MDVTemAaY7IQdSkZWGYa0CQVVeU45FCp5pooF+Xg6CYMD+yadyC4sJO/ddk/LCVJOrWFt5cCS7RBwPNHnmHUFFAZrLjF1HcnS2Mfse15fyYNtMQs3oPo7V0+ZeXsFDcu03bCOdlicT6FNUAf8O3WT9a4mETOv1FOunqr+c7kkgZz5cs7CcsgNaTkftg20t53PNuVxhssolyrUM5VvyuRjB9xcXsUdcClgK1l2Ng2GHroaXqfRkSFVwpryoBUTznNULk7C+btiP7y/BdoJ/BWYTzGHFmvhOnnXO8uYKjOHeZWssn6WRTo0XpOhkJR5vifz+vpaZ2dno1GcgcHt5CokGn0y7ZZsXZ8cgZgYpNxJEpzPlF5Z9hcXF5I0bMRJvfBkKkHa50xaUvYGQfY7Q4isZ8W0E3xYd5ZBJp3ATfDmXBX7OYH75uZm5NzYXyY+DE9UrDdtLjEhU+p0jh7yPvZRRYQS3CtZfSyAm7HQZH/pmfK/8/DETQ4FraC8L9lMxZgTLNPo8npfy2/f11rHyXY7VXViMlBUgFrJdkoBWFfek6s8qvxbDqsaMbTqV9U9+y71Iq8huFI/HD7wDkkDA8MMHoVxMtLl5PK/ahItl6B6niY3g021mwzcsicouC6r1WqkG3ZWDPlNOZ98/rnj+wwj8TknLIcOOcG6ChGkzBKQKEP3ja/z3AP7N98C5P6yLSQhapEdksEcObFtTNneJBV5T5aX92X/pwN0JGEqzv0kgFt6zE79vAdv9+UT0ChwroN1PhwmSuNZf0mDUOitpXHc2//5O40mQZ7HeI7sOsG25TBaIY9KVj7G63YBy2w36+nUAp9KCStFpuOqYqG8v6p/TuxUv/PD8pwHDV4av7EngZrx7Sq0kMSBk2tsS45csp2Vw/O6bYKZ2+MNQPkcFoMugdr1tvxctwRsxsv9sZ6nLWS704HlaIjMmkza+WX4xPWknbuvDN7GgZubm6E+3NXKNrT0NHWN+lS1gU7F/VSRKTrE7GPfx3szGkB942OHq/RkgJvDNmnMQsygKUAaAlmClcRPTaMxSxs26OVhFg4VnsJN9p3frvfUNU4VmBLkyWp4D6+byov19TUVm858psC5akfmk84nmVdVbmsUQDbj65LJ5/3VsNXt4soM18urL8wwnXcCH9uTepB1zk1hCXyUT/7OvqXe0/lw2SDndkhQ+Ml8c7TKmDfr7TwrO+C1jFPn8J4g5j0WZtLpTPw7l33mhKnJHB0GdaTSL4Z+tiU6IeeVE6cZR2dfJsBXdpYj2GTdrvO2t3g9GeCmwkmPY0oJahQQDY0e3hty3Kmcmabi+ZyXXLWGuVZqskGylW3glyygAicydrad37y/AmL+ToeQrLTlALY5nZZSOd8EJbOgBPZkyky5KohGzHzIrnmeZbte7jOzynwuOg2MQJYf1ttypON3/ap4tjRmdrlKIkEj60UQq3TDbXQ9c6UGddF1pPOjXmd/V7FeAhbDH1xAYOCmHpAZ+0UR2ce8br3erALK/qCcKHe3gf3IfFMvkwBw9ONjrSXKVbw7yQxHcuxX3uvR1ZSzeVLATWETWCmUVEILNo1KGse9rUx+FgU7gatNfA+Zl48zf8bVmFqsOxWN5yo2vQsjnfLIVX0q58D2tVilwSXr3CqPgCmNXwpR/Z+6vyWzBBK3g6ws25lyS9ByqsAwjd36ROaaLGmKCTrf7HcDT4YiODLgg6PcDtoEdZNPP6ycdDon2sYucyist2XiJxX692KxGP6zrwyAvo/6QcfiY6wLSRhX49Dpst6pk+yfBGxiDCezKxvKvKjXzJ9hNc+75LJOaRP6teNtpScB3KlQBGQCb2VAyca4TtuAbqF56Rdn32mA0kZRPNFEppM75Fz3ilWxMyrG5mt9vkpT58kus+xWHgn6CYgtI50CzqpeFVMi68hrM9aX9c6RDc9Lj58k5/PpGFL2CVYEMdaJ8WEaeW7uIVN0GTm6cJoKRVC3+T5K3yeNH7aWbDcB3UDXcnoERzoJhgFTftkHBGO/Qo0xaYZNnLdDJ7Zfzi8kCLoMxsoZNrXsMw5NnUs9TiypRg4kARWYtvSMzj3zq8Il1Qj/yQO39HizghWTkzDpZZOZZCglPbM7suu60ZPf0hvTITDuySFSrv9mGdtAu2KlzqNKFWhWTmCqo/P+ZGBTwM2UTD2PZ/6t+rutVSw4jZbsqzK2NI5koE4JbFWdGGLwt4f4VTsTJJmv7yezNBiyDhk6yOtdZgKa9bDaIGQANRASxJIFsk2MQVcsk862AqwMA6T8E9BNzsg0c3Iv2+e2ZUiM/xm/Z79ln1VMu4qf+zqOBqg3VUpHkICeWEAZbrPnJwHcCQRuDDuCE4kc4iYopOFyudd8Ph9CJZyJZ6dkp2YogQLlpBdZ2i4AONXhPD4F2rvmV7GEKp+pem87V52fYtlVvyUwsE0ZRpPGLypwP7eMLuuVTrQ65jJcju91XTzkzWF9JaOcvMshcgIPWZ4ZrMkMnz+eYJntY5kJ1iynAqRkgeyrdK4kV07sw3Sw+e3r8zEXDickmHGHKB1SLmPNfqlwgvVgfhxFuK0ZPvV36i/llNcYz6p5BLLxJw/c0mNgcvIsOt8CYkP1Oa8QIFgzxFIpq4dbVogccqUhJCgmw07jT1aVStOSQYsR5u8pGVJxpvJPwKVDSqXbVo9kkCy/Am+fb7WpklXlxDJEkYCSAJeOoCWfSgYZGmG7+Jxu6kGrXuko2F4TAk5QOW6cuyMdM6UcJI3q4nwZ3kmHxhFkS6cToFInWN5sNhsWB2Q4ZL1eD2/r8Vb+2WymxWIxIj/O32vjqQsZvqpAmqvR3EeUHUE7lypSZnwER5I5AjWXJhOPpuw+7W6bjTvt8s7JH5T0ZUmfktRLerfv+1/suu4Tkn5N0md1/97Jz/d9/353X7tflPTnJF1I+qm+739/Wznp7f3hzLm0YdsZqqBX5puxeT3z8ItjuYzLRskHU1mQXPtaAQiVPFeaVOyllfKaXTtyl3taSpQA0qrDm9SNCr3r9dX9rfg4k0GKhpcMqnIsvjYdM8tKRilptMaWoTT3PdtgUCJL5zle73qYyVM/ObLgi6dzJVaSD7bBIEVWl88FSsaXIMW2JXBVI2C+C9Rt8gPgTL4cu7Zzct1tgw6NVKMn90Pa8Xw+H7F/h6CoT7mOnPqTrNn3cb4tw2qUB/PNPmD5Qu4kAAAgAElEQVT9K1K4S9qFcd9K+kt93/9+13Wnkv5B13V/T9JPSfrtvu+/1HXdFyV9UdLPS/qzun+7+w9L+lFJv/TwPZlSKSwoPjHQRsDhlJkOh5Dz+Xx4BGbOqNMDSxpYgI09mbQ0FjaHY1PM1sc/SKdUQFkx/m1slUpUXVN9knFlPaqyW8Deqte241Wdp2Sakz3SmHmRfTIkVjkvlsFwhcFT0rC0za8Hs+54o0iGHbKeJB5OuemESzUZFpjP5zo5ORk9K9wf1zs3llVyJiPM+GvWj7KpHHLKjKBFNmu9sfzsoNx2v/TYoH15eTl6A1DaJJ0mHZDlmE8EdZiUOMCRuPPmvFoCsu+pwiWV46JsmCeJXO5LyfurtMvLgr8h6RsPv193XfePJX1G0ud0//Z3SfoVSb+je+D+nKQv9/cl/27XdW93Xffph3xaZTSXIfE4O5/sxr+tvARidg4X0rNsaWNUZlN93z9aM+pOyNgdU7I5f+fi/Wx/gmQy420dmdfk/a1jPFcBYOZL0K7K21bXqbyrulUONOcnpI1T58oEsktpzP7SeF32bDYbMWo6NIZc+Foxy6QySpMJXlPNpVBP3Z4EA27bn8/nQ5jQj3RN+5kCEbeBJMWyZf1SfyoCkXFiLwE8Pj7WycnJ8IJmk57Dw8Ohb2ibDpVYZn2/iXFnmXt7eyMgpn543bgdmLHBzpajc8rceXNejLIkDkmbiVxel04vGX21wSvxiOW10hvFuLuu+6ykf1nS/yLpUwDjb+o+lCLdg/pXcdvXHo6NgLvrui9I+oIknZycPFrhUXk0D7HsSStPZcFRwBSeh2RmMTQkLxO0YnMoy0kc5s0hcDI5dh47vgK2KTba6sAWqO/KbKkkU+UlWLfAm/9ZtypezLxaDDvPE9T4NEn3l7QBQj9hkuXM53NJj7e7cxLK+pGbLKgX1hPOrRjI+c5KabNsL2WTbXI+HPVJ9+x+uVxquVwOj2qwXi6XS52fn49kkkCS/Zl9nSDD8EpOnmUfsn8oo9ns/nngfskxn8bo+LakYfcy4/YZBrKtujz/9z3uf9eJQJxxcIZnXB6dP0fn2V6fXywWj+THUVY+TdK4RcJJcE8daJG7TDsDd9d1J5L+G0n/Ud/3r0IZ+q7r3igY2/f9u5LelaTv//7v72ngBBM2guCeuxtppIx7U8HMtquF777PDyRy3pxFhiyGbdPstCpO7995/5RBFbJqnqsAb1serBfrmUo0xagrgN+Vdft85eDyXNYz+4uJjpFM2v/ZLjpegqhZI4f3ZFYG6Iwtk8lR76pNWhxBkHVKGjHP2Wymo6MjvfXWW3r77bcH4O66TpeXl7q4uBixbTqjBJ0E3IzFuz60v+znyvGyT/b27re2Hx4e6ujoaPRtkmWHaud4cHAwPJ6Ck5hmpwZqxv19bfXsE4KvZWvgzHBJpSvVNQyNuByDrK+rXrbh/HKlCzElQ7mcL/nQjLvrugPdg/Z/3ff9f/tw+I+7hxBI13WflvTew/GvS/pB3P4DD8cmUxollcn/LdyceaanpfL6d3pVKzuHpwT39J58QpyvNbsiaFfgxrY4j6rtHzS1nIMTgSkNkfduA+wq/8wrr2n9bwH2VGrFpZ1yEpIkwP3KbeKMVfb9+FG6XOJp4Oi6zdp/huik8eMacmjvMjJ84TqwLpJGI8LDw0O99dZbOj091enp6SPW5vYQbAgwldxpIz6XNlD1ZeUI/DGbXSwWOjo60vHxsY6Ojkaru9zGfNGDAdivlzs4OBjem5ltMFgS5Aiuln86T7N+hmB4jR2i7dl5tdgwQZ4hM9ez6uvEAI4GKsJSOXymXVaVdJL+hqR/3Pf9X8Op35T0k5K+9PD9Gzj+c13X/aruJyVf9hPxbTfErEUax/fIjPkAeS7fs8ezIGgoORSk0tN46ElzaZQTOzwnPipg8T3Z1hYw7gJgVaocHs9VBpD/U7GqOk6VXR3LfPJYC1xa5bTuc5+xD8i2eZ2v5ejLwG09MsA4VippBPiuC7dwJ7BV9a9ixwRug5eB5vT0VEdHR6NVGMvlUuv1WmdnZ1qtViMA9G/rZ5IH1yH1nCSJ/ZSrMDL0VIUfF4vFENrx8eVy+eh+SSPWzYe+8cXNDD2xL5O10nG6LSR1dGicRE4yxtVknCwm6SEOMERnXWBdWXeGwVi/dMhk3a20C+P+VyX9u5L+j67r/veHY39Z94D9613X/bSkP5T0+Ydzv6X7pYBf0f1ywL+4rQAqPRlyBUgUQAorBVWFOTKuSdZt5kJFp3L7PgJ2tURwiqWyzVOgt2uqGG+C7rY8WyBdAXCyZf+mAbTaQlaX56t7qRO+JoftbKONNNtstmQQINi5T7uuG21s4XIyx5kljVaTZF2tK62RWKvdBN3FYqHT09MhPuwVVGajjrF6V7GdjJlqRRxcHtliyjDncAha/J/3pOPwtdyazo03dASeBOYDqG5uboaRjeXnFTQc2RCo3R+r1WpE7qiTDrvwmSiWCfur67pBxpZJ121ef0d95DnqT7JoEwI6PI4euAqIjxaeinXvsqrkf5LUohI/XlzfS/rZbflmSqWomOObXC+1DZ3XE8QTfDPmnh6RwL1LyuFoi32+aZpizNW5VqqAlsdb9+bxvK8FXDksz/5KGUnjbe9VGcmw/Zv9ydEV8zcTN5PmpKMnAw3cHuqSeS0WixFhSJCjnA2iuVksJzMPDg50eno6WhGRQ+3FYqHVajXUJ3U1dZ27QKf6UdIASlwZ4Wsz7OQyGBoxGLktDGd51HtzczOMKvi2ejNvfwxmBmmGU7gskgDJ+nOfhp0NQ6KWJ0NYdDRuH0c0TowAGJRzxJ6roRJPGCXgGvQqPbmdk9WxiklWBkxj9TXVbwIxZ6V9jkKvjD8FvY3RVkDaYtxT8tiWL+/blv+bpm1Ocpeysn8qh0dgaOXROk+jyDr5WE5KMmW/7+/v6/r6egBkx2dtVAYK5+9hP9k428XJNAOoQwuMsRu8XJ8XL16MVrucn58PwO+1zy9fvnw0Eqx01DZT2U0Fzvzv65lHXt/3/RDm4DJAX2PHaCD29Zbv3d3dwJyvrq4GAPdLJAjmOdLzbzsrOxaHUtwvdsxdt3kJM+cm0iFk/qlT7P+0g5Q/dT7n0sjovxehko88kfVKbZCg96NwK4Fa2JzQ5HDQno1l0wtSkMnEqxBJ1abs7ApUK5B5E7lVrOl7CdhVmdvSm9ahYiXOh9fkcfZxJee8T3q8Hd+/c9LKw28P5f0GeDPfly9farVajVaBGLi9fI1G7/ALwXF/f3+YzHO83WEFb3fvuk6r1WpYh7y3t6dXr16NVpScn59rtVoNIFfpZcondZtOj4+CZX0JQCQ/0sa+7u7udHZ2JklD2KfrNm90cZ52QKvVSqvVagSwZrNcAMAVO3Z6tEXLmM6X9faoiZvy1uv16OUOuUGHIxO3kw6WbSc2pXwTw4g/OUJK7GmlJwXcFRDldWxksh0qRrKvZAY+x4nGrE/GrqXNelnG4NJxtEDrTQDtgwBwBXSZR5XnNpl/kESQ2CabrCdlvosc6KCrPDKvqm7UKxrb3t6eLi8vJUmXl5daLBa6vr7WxcWF7u7uhuG3h/oenmfdCfCe9Dw8PNTh4aFOTk50dHQ06JWB5eDgQG+//bZevHgxgPR3v/vdYWLy4uJCZ2dnevnypS4vL4edlGaOFRtNWyPoZH/YrrhxKVfM0Alxsvb29lbn5+c6OzvT8fGxFovFsIbaSxnff/99XVxcjJwv65BLfj3KcAgrJ/FcX9/DjVRcUdL3vRaLxVCnLD/lU+lnOkfiTBViTXCu5s4S5KfSkwDuamiQQqEB5uQA42YE6CqO6fxy1tlgXnm9FCw7LQGDZVTG4PzSuNie/F3JYyp9lGGTXRxUq64fdFSxSz5V6KMF2gSyBDT2x83NjS4vL7Ve3795xWyULHqxWAxszR/rFGPlBuyu63R0dKR33nlHb7/99rDD8MWLF1qv1wP7fPvtt3V0dDSsyri5uRmA0N8XFxe6vLzU5eXlwMopL+t4paO+jgyWskh7NAvn41JJhCjD29vbIeY+n8+HteheintxcaFXr14NIR/Lif1MUmQ5d103knE6Z7fX4O1jDEl4BOP15BxZcFetnQR3eFpvqvi177PsqoeO5Wo1fiqMmkpPArg9lKmW3uR1eZygm0tspLEAWpsheF/LC1afD9rWKdb7YfL2/VXaBby/FwC/i/JlGR8W3KXHb5afKr9VFwK6+96g7Ly4u3KxWAznHKbo+/5RqMF5ciPG4eGhFovFsEnl5ORkFHI5PT0dliK6brPZTH/yJ3+i5XI5gIMn6ao13EkEqvb6euefKyIIegY3hiLo/Eh26BA8wnjrrbeGEFSCIFeoMN/b29shxORVXzxPp0QcYNilWkTgvri+vn7kOBwOS8fOGDjlaTkxnMPyCPSJKal7TFP6+2SAmwJMlkUWxCFIGjy9dCotY2dTQNFyFpmyPlVeU+3NslpscZd8po61GHxek3V6UwBthTuqulXD9yptyy9HNb6OAJZO3sdbieUwFsk8ZrPZsJ469xlI4wdGebhOwLu5udHr16+H4wYz7440w7SMXD7z9yTexcXFwLY5Usy1zxyRtEAiQyNc0kewzlUptD3OUcxmM7169WpwMF4ds1wutVgsdHV1NcSq/Vyg5XI5LIM0eHJViRk1QyJ0OBxJ3N1tHo3LzTp2CD7ODU92ih5p+Z5qVZAdSQXQ1B/+TvCm/D92jFsaT0pRwBUI87g/NhQKJJfwUJkrVt6aKKhSMo0E76kRg/NvAXcFni0HUbGqVn13PT/lqHiefZDtnConWR0/lcJWMk0HXo1Y2EdVvSqHxeuSdUmbOKtXKJAxkol5opITYGTtXnPtGOvV1ZVevXo1Wqvtcy9fvtR3v/td/dEf/ZG+/e1v6+XLl3r//ff1/vvv6/Lyclh9wU03BJ2WgyNr5ccsmc9IsawYvkhZcZKQa5HPzs70+vXrIdb94sULHR8fq+/70VMACdJceZLPgLFdJxi7Lbe3t6MHcOXmLAO3Qd/fdirc+MPFDbl80PMOlkkFyC2S1iKEeW0rPSnglmoWJemRx7dQqVAEZuZBtk1WkOVwuENPSyCXHr9RYxfgYh5khbzeBrGNiVblpDN7k7QNsCs28CYjAsuZy7Do9KT6eewJuC3dqO5PR8R5kKxf/s5y02Hk6iQDEPuVOuTYbtd1Q3jk+Ph42Al5fHys5XKpi4uL4dr33ntvYNvf+ta39NWvflXf+MY3dHFxMQCaJyRTb9mGbEflvPjbwMvJU7/od71e6/Ly8lEYIEkUN7TYKRkIDw8PB/CTNMTnLZ+Us8ug/XGbv2PkroOPs47sK8aqPRLw+dlsNqzxpr4Qd7zSyO3knEYCd8t2KPN0KNknrfQkgNsCTxbA78rIGEfzcbLtZHKpABVwJ6gyJcvj8db12U4qdALD1HBpFxY8xVpZf947Bcjb2vgmDiL7KJU7H7fKPubx1mTwNudZ1ZWTkSmXHO3xvPuRz7uxDhMs2K8+7u3W1m0TkJOTE33nO98ZWPQ3v/lNvXr1aiAd/197Xxtr3VGd94zf1zZ2HcUNocjFVkkEVUSrxrQRARGpKRUViaqSSlQFVQltrbqVEjWR0o+QSk2/+IGU4lIpQnVFG6jSUkqCgiwqSgApipRA+XDAQGlMi2RbEGMX7Ncfr23w9Mfd6/q5z11rzZp9zr333FdnSUdn79kza9bMrHnWmrVn7/3oo4/i4YcfPnyC0Or3dkSxF83kGXaeeyYfGwDzuDlmrG3kuK7x490t7LVyHz/77LNHPGPePsk7xoz0BqGFQXgMjK8aFG8FYuBr5Q3E9Sas9fX1119/qMdqHPgFWCx35Nx5K0S9ltHOADe/txg4usWPY1jcYI49epOE42BczgNmBVLPI/bOPZnUqpqS8HLa458BYQa0+ovATXkpn+iaR54BUPn5ZhN7pRF465LXeHvj7MmS9Z9NcC2nAD0iGzt+1w3vKWZgAXAYUuEdBtddd91h3qeffhqXLl3ChQsX8NBDD+G+++7DQw89dLjFz/rv8ccfx+OPP35kn7aG/4wiz0373XOGeu+H8WXejscPrzAQ2jUOjfDNOR1LK8shEH6Zm/UdA6GCrfFRo255eZzV2Jijp7ppdfINZOPF98bsASvrH9VNdQq5nzxDog5XdTW7M8D99NNPH4kD8tYeI90jqctRA0bg6FZAK+stgYHjAMgehxLzMx6Rx8b/2d1k7QsuF/WX8vCAM+OlYOUpTAYAniFR48qK6HmjPA468dg7Y4OXef4jhWfw9vpGPcLMS+KbktouXTa31o48Jn/99dfjiSeeOLz5d/311+Mb3/gG7rnnHjz00EO4dOnSkfjqk08+iUuXLh3uIPFWht5Sm+Wycx1zvddiMvNDSLYVz9uyy3nZqNi4cV9ZGdvaqG/fNBn5ZmIE3gyQlp9j9FYX8/KcEutjfp2AXbP9+FaXPsXJISo2ADoG3jiw0eExOHfArftDPc/SOkgV01sWmTLZHejIS7PjEWhHS2o2MPowgBoYrVfl4QE0ANDrUaiHlVz7yatrlJYZjgj0PePIS2j1LlRuPudVE1/zvBI91n5h8iZJ5HWrUda4ZzTJPDkZJO3BGdsD/uijj+L+++/H17/+dVy6dAlPPfXUIQ9bpluIhPXAM6y6AuQ+zWT3vFQDY05nsOJjHmOVwea3gbi+P5x1mkOJXsyYXwym7eP7X9wfTIwd3vZgu8arfeN/+fLlQ0PDsulWQ+5f7+c5cfo/Wv3tBHAzRWDFE0H3a7NVZ8trnpr3YYQIPCLQ9shk8uLwqpiWpjdceAD1XQfa1swI6HEUGmDZvAmeebQKEB5xuwAcUWqdhBkfBq7IE/ba7xkG4PiDUroq8Or3jLDW5QGDB16mI/a4PPBcPPny5ct4+OGH8eijjx55DweAQ6+XH65RgGQj7zkoChIqO4MM66ftznjyySeP3QDU92tEO4U8j5Nl8AA4+gHPhc28ewnq7fN8Y4+d5WQDaDxZZy00pCsd72Eaz2hqn3igzWOmzltEOwHc1sHmHVuaN1G93QF2LQIvViAFaU9prP5IVpWP22B16ZvKOB97kZ4RUf5al7aTz/kGjsdX+9PjqZ5iBlBcVj2viqHxJtSojZqWrWiMrxqy6k3OqH6PtC3sZave2j7uRx55BM8888zhu0ZsOW7EsVQeVwYKrodlicBP+8DKshdqT4radfswAj9Wr94295WCNnD0RjM7Nt5Ye0ZSQ106XxmsuQ5rq+cUcl+xMbC22W4X1W1uP4+H5ywpWI9CsRV92wngVuJBAPxYtJ1nHkZkyUZLSZaD83GaAmvv/cjbyzT0w6CuPPSuuycDx/e99rIcXh7798DR69Noian5I2MYAYa3ymG+Oka8msl+3B6uS/tG2x61j0HL0yfuKy2nuqH9YiBgL6kygNM48WgLo5cWrSYjMIgcJM5rstoThTYmfAOSgct7jwjLprJrfZHRVkOg7dD5xbwUlD1jyrrCsho/bpOuLJWHGhz21nWMtA1cZ0Q7A9xs8fStXB7weIPDk00nuxoDLsdlR5bOk9ue0HriiSeO7Jxg5TAvij0AzwBlwOkt9bz2R0Cj7WMg4XqisIKV0WUqj5HunIkMoo6N134eM+tnliGb8B5weoY8Wplkk9KI+8nre62P+fDHGBgYOCSiu6kyA+UZxJH3pstyjltzLNm2zPELmazt3jZPjltr7Nrk8kIdSp5R9sZY26J9orzsuhlNjnV7zpOGSDT0k40P94PqbeY8ePflmHYCuHnAVZG8iWfH3mBoJwJHl4IRMGYKrvXbuT31Fb0vQh/UiPjyv94kUpk9T1B5cZs9pYqURwEVwLF3R3j18rjwjauoDk9ers9k5FWGet7m2UUemBr7aCugyq/9pHLzP/cx95+V5yW7jYlXxnjxUlx3vyh/r60ReKhh9/pBnR0jA27+Ig5vpfPu77ABUOOuoBfpho6H5tVzL+zDPLVt2mZvR5S2h42SpVvdHq4w2Gs/RG1RPY9oJ4AbeG4nBA+wtzPCA+YK8URSfvwfkTfwly9fxmOPPXZsSWjtMc8lWvJGHpwaLa4/s9hWlmXQtnnWX/tC22FgyhNC28Dj5snJ/e+1z/joQxumB/zSJku3MmpMIgPPk7QaizR+bIg9oOb6vLHgcw+g2DOzuhQQtL913KLr3nlE2h4LK+gzFZ4TxEYoi32P+sdL8+at6rr1FwNrNL+j1Zf3U732wh1qaFmHGOyj9qgsmbMH1D4WfAuA9wB4IYAO4M7e+ztaa/8MwN8B8I0l6y/23j+0lHkLgNsAfAfA3++9fzirwwMRBjvdhsX/JOdheuRVezFiLjtD9oIf3v/Jg6ieWzYI7JEp0HBbWVkYxDxrrpPJAxku6xlDlsUmhqXz5FDAiDzVrA6WhUk9KUuzpftVV111ZBcGt415s3fE4BNNNO1L9cy4bV5bFEwzsOf6FHQYwJky0Kscs8x8XYGKZea+0zZx33JIYaRzmWxaxupivdR54vUfyxg5J3qd+z6TT+tV2XR8uf2c36jqlFY87m8D+Pne+2daa98F4NOttY8s1+7ovf8yZ26tvQzAGwH8KQB/HMBvtdb+ZO89/d68Noy9LE6LwNfyeArugXiUzte8SWh3+fmdvpzfBorB2COdyPbP8TZviWVlWCF5MunTZLo0ZzKF4gcPPNmMDED0AYdocio/9rg1f0YeT2956oFBBA5sJL2lsGc4PbBg3gp2Khv3hbZJy6vByPolS6/0r5bl9ukmAV4Zq0HR0KY3bhl4K6mXyqQAadhg19TgW10RT9V/L9bt6W5kULwVZ6VfmFdGlY8Ffw3A15bjS621LwF4UVLk9QDe23t/CsD/ba3dC+AVAH53UA+A4/sz2dJzoyIPxOt09hi0oz2yhwT4tZZ2Z52/Nq2TkevyBmdE0bLM48Ned+/9SChBl/jqoTM/T7k84np0ZVABCs8wqBycxoqvXq56z/YtSK9+nrBqdIxGcXLOp+31gIrLRjfEIxDw5M/uzXAbM8rGlsszyHG/WT42vnr/Q9uXUSZvBo4M+F4+9ZiVr/6yPuU6PT52rCtuxjKVX/l7zoKGfDyainG31l4M4OUAPgHg1QB+prX2UwA+hQOv/Js4APXfo2L3Iwf6wwYwcEUDyx6u5lHPywNUrUsnXu/9cE8tgCMv2LE3o1l+byJrm6rEMulgRkDB7dMlnQ0+f6CWFUN3fozkVc9dl6AsXzaxlKdXp5bjieEZJwNw715D1LcqFxtInYAql94A88J6kYMwAu6sL5RUh9eQx99rv84dO+YQGpdVIFMQylYSViZzCDww9K5H9428laKOk+q8zksGWw9kR/NfjQg7RCMqA3dr7QYAvw7g53rvj7bW3gngXwLoy/+/BvC3J/jdDuB2AIcvjfcapaTAlIG3V47TOfRiAGEPQ9hAXL58+RAU7BWaVo+nkFYHK7Anj6dwHJ/j3RmcXz0vrw+uvvrqw6+weHtVW2vHwiOcx2uTd5/Bi716fRF5Sd5qwgN87UN7Otau2WojMnLqTeqKgW+kGT8bc24Dy6AeGXuomk8Nx6hftO88igxkRmt5qdMTye7pqJdvBMrZHK6uLNiosAebvX6BZWc59LUL3i+TxzMWHmhHuuJRCbhba1fjALR/rff+G0vFf0jX/z2Au5bTBwDcQsVvXtK0QXcCuBMAbrjhhu5NBAbGaBIqkCmfLI2B4Omnnz78krd+u8/y6H5P7nDpr9DTMvJu4Fk+3VLH13gJqBOttYZrrrnmyIv7ub3cloiqXgJ7m1wuUswoj9bL/2osvPbwCoBvOPIYRKsr9Z6sD42XfjWc22J89WVmqjdWNrqRPCKWW/Uk8rhVhipVANQDLe1r7kNvdeOBlieDtxur2g7mrR6zgje31caU+VTi4lm4LdtRoqBtvDYG7nYg3bsAfKn3/nZKv6kfxL8B4K8CuGc5/iCA/9xaezsObk6+FMAnB3UcS2Nl5wayZxPFDyOL5aW11g7fVmYv94luUGnnslxM3nYzJb2Rpe2MBs+us+Hg8vq1FW4nK65+QTvrH7um56qUWj4CbjuOxt3q0Vi9toX70lYY7DnrODAf62PdsqYgaeCjAK95uK3R7ptMLyMj74G1l88jr9+jfNlYqCwqn+exevekFKTs3LabKt/RfI5k5fw8ZvyAUzQG0TX+j3SR/yOs8OTmlXUFtIGax/1qAD8J4POttbuXtF8E8KbW2q04CJV8FcDfXQT5QmvtfQC+iIMdKT/dBztKgPHNCu8GTeSBRLEiz0Ly9iWbxPqmsxG4eUozuqGkSyPl71lqr91azuqz2DYDvIZgIuCOwNrrN56UkeKOAETz6qqLyQMu7isDDN665tXl9YHVqUaVwdvrJ64XeO6bkzweXjtHFLVVr3n5tVx0PePB+ubt4rE0BWndRqf5GdDUG85i4d5czIjnstWXOSdcB9/H8BwrT+/5Outk5rGrIdO5E1FlV8nvAPBG/ENJmbcCeOuId6HuI8e6HOPOGSmlpwDcwaZw/NmlSBZPTg/wWGZP4dSr57RMwbgeVXzguSfb+Jq9B4Pf6hYpzGiye33hGYFKOe4/Ph+BtqYxDw1fmHx87K1q1PhH2xf1x5ObJ6MajqhP2FABx2/Ae3lUb06KeL4xUPMxt13np8rIeqmOhM4VPdY0T089Q6kG18iTT+cm82TMiFYRjCleGFXrYvk4bWPgPi3SyepNSFYU7iTPqwKO3sD0lMF4GKDpe3YjIIo8IfWwtFwEAtpW/kWT0wPt1tqRL7PwCkLf5qZvNuP+5HbqebSCUO9DASYDcs9YePqQ9YdOfs8Y8nFmGLk/PCPHzoK2meu1WGkUNvMMhqc7I/A+SRAf8WUjp/NTQam1o/cENNznOTksQ7QCtePR3Mvaw3m9MKwH3FbOm0PqFEQOkBf/rgK2bw4AACAASURBVID3TgC3WlDPCnreJntF3oBHE9ny2k1J/YQSe9wRSHvye8ZG80UGQeOkOqmjfmNjYTzsBUatHf9qyYy37dWhqwruU+Ovd/LVo9C+4mseaHv9EI255vF+OrmUPxuBSkjJ6vUMKXvRCvQeeTrWez+m557nnfVtlTxQio5VZ/hZAtY97g9ri+6n576KHDgjr72efnrHWtabi4wtfK0C3ArakZ4C4/cNZbQTwA3kyxYAx0Daro/i2ZESmOdpr2HVr3pEMjH/aBJqmIXb4w2mAURUT9QfnmdpZfg9wiNFy4BbQwYa82M5WR4vVumBMcukbfT6TtsdKb72m4I1f3qM6+YnUNnLimLi3D4dV53Ile2TyjeqS4+53VlfrvXMVXetTd4XprJx5u9Ksm5anurTtR5YezKonlh+rTvSfSvL+XQrreWN5hLXm43NjKHdGeBmyhSOOyTzAHkS6qBpfg9MtZwnozfYkTJpfg94dVCjyRr1kQJQxKOiZFxWb9RkP3v1ZwTGOgm4b1hOz9hGiq3bBnXcGSS8UIm3IuB+8MpwOzLQzgyRto/JG+/IiHv8PB7KfwQUOg7aRuD463EN2PiaEjtHxjuaJ1w+eriJwy885ub9G19P700eDwOYp8axlaK55FEFoBUblHYSuIGjVkrTNQYVdYR6zqrUFy5cwDXXXHPEC+OB8RQ1mjAKPN7SLSsTpXsT2gNC+9nDKV6oR70LvWmTKZz2oU0KS9MnKRW0RkBWaasX+so8NB7TCLS9MlaX7r2OQiZ6zkZOgS/q3xmKwDSqz84rFOVTw2FpPB+jFUk07qyD3lxRfVU+pnd6L0s/+quA7ul9NOd0vmR67LVXDZDnUK2hnQVunaBMOuAeMHsW2/NyL1y4gIsXLx6GTbTTR0CjCs1pURs8z0yVVhU0MiIe7+gmTuTVaFq2ijEywOTHzaOxYv4e36iMl679qjclvTp10kWeIIO219fcBg+INb8H/l4bK6Dq5Z+Z/JGueJTNu9Gc9F5ylskU6URmIKO5pWPkbVrw4srG0/69Mcva4umb1w47r6x2RrQTwM2dZed6vTI5LK/egNOtcXysL4viAYoGl+uP2qN5RpPT8+ayiTWSweMVgaj2bXQzUY2gxoPVu1ceCnx83bumbfCepPMmKbfDA2/uk6jvrD5PPpWN+0R5Knhr2yNZIt3LAF91OxoPlcVzEoy8uHxkvExvPK87W3laGU7LwNV7CZNd824MGj/dNaa8vXZGXnZkWLx/T9c9qsxro50AbsC3jvbvTVZ+ACQj7nytg/c26+4Lr+OiyZ6BZ2RdIwAayTAiD7AzkKwAkl7TJWxUV9S2CMC1fgWhqG2RMVAAydoaxS2jskxab8WDrgK15qnoewTYFUeAz9mh4rGwvvIA1NMLlS8CuqiMysWGlUMlHulKyxvLykpN5RoBt/LzDKR3XKWdAG71uK2BuiNCr1k5G0QNCVh+/qoKdyADN09+tt7Kr0IzQF0tbzRaZkXKUZkwUd0cqrE8vDrxvLwIeLRez7CM2qZhETYkWV0R+Gbpozwe6K2ZiFnd2+CnPI2vjpmGILx30Xg3JIHjKzVP1yr/dpwZGttS6AE34wLfKI3eUaL6pCDP6dE8ztqi/eG1a9b47wRwA+PlIysZT3pWotHmf93a5r3MSb24NcS8Ij6Rh6UDH/HQ/hjJYsdVRVMeOsmZvPsNqqTe5NB2Rx5K5m3bL7sZOxqHKnBrmsc3M1xR/R6t8eJnyRweqy/6ReDlbefjeRcBstUXOQwVAxk5HqoP0W4iTz+9tnF9Wm40t1imqM5ROyPaGeCOBjdbUnp3pIHnPrKb1cM7SbK35Y3krYJmlZ93nHmkM0o+kivjpeCdGZiKZ+HJ6AFoxTB5E5fryW6QZTTqtwqgqjwV8rz4ar3R2Gs/ah/pu0ii8eFzXfVaehR2rOhdZsh4Vc5GI5PdA+IIZC3d0xnWz+y6HY9WkZ5RnDH6OwPcQL6csM7yPEW28spHeQDPvaJVl07acV4HVpSsel1lG3lzLKMeVynyLCOQ8jy/CFyjyc7XR32qaZFCe8bN897tPNvKV+0TbbOee2mjN0VGuhZN3oqx8Mp7Y2Y/3VKn5XSfM3uQ0dzz5mkkS5Sm7fBuRqpuRfqlczwzLJ6uKC+WIWqvzgvtby8P58vGeaeAm8nrHD73ljg8gLq/144NrKN3k4y8LK5zJHtUnvlE+SpgnE2AkYwRPy3HMWy+toa/Lhm1DTOGUGWJ8o34ZgCdGX8+92Ks9kTh6PW+szQC7ci4MmXhEUtnUM4MWwTQI6M4GrOobUYK6Ap0ihORPFE7MtmNb1VfI+fDo+rKbCeAW606cHQp4Vkj79h48TVetvEDNvriJVX4bGAypazmi/hngBSRTr5Zb9FT3Mh73YS8STOaWJ7MMwBv5xHoZzw0jzceHuhFux2q4xlRZSUYrX4yuaveqpfmGYoICFXOCnnzMuI147WqQ5fJ6snr6W5EKrNhUNYHI547AdxA/H7qTPGj12bav9149G5Q6DXLH9XlAVpEFS9S2zk698gDC253xmME7J6y8XUPKDJDFF2vTmyvT9W4sNwjz9MDJE8er5zqJMd4Nd7KT7HO8J0BOO0bbXfGowIe3FejMMJItixvJJuOrScbOywRaFd0aqQvzKsK2JHsXv6qMdgZ4I68vqgxOqiWph387LPPHm7347vdo8efRzc3R23x0iKr7X3Tbsbb9h7/t37zHhOvejFW3lPqDCA9AzQydN44RHm1/RHoeW3i+rh8VEc2DrYlTT+aYDzX3hQd0Ug/vPaqMdHVgNcH3k+va9nKGFbJG6ORYVcAz96HXjXaXrp6zNFqyFuReUDOOlMJse0EcFcnrTeAGS/2qkfbe6oezhrF5DqiycRUiYtmy9oonwdI6plHsqih4zhp5ImoMYkAJwPuCkBpf2WrNCP+XNbIqHiT0UD74sWLxyYlh0pGbcioUs5zVjg9AhJuX3QcgXU0d7YJ2h6NDLTn4FVlmsGBTJ+z/jad83hZXu8DFR7tBHADxz29mXKZJ2BW7KS8n0yuKJ0BOxv4SDEVKLT9UR+qkmi+qN8ZvLP2KVh4ckVe/oy3w/Xxu521HvVkuIylVb189ZoMtPmjzN7ujG2B9pp5weOh/aJ5uc5o5ap9FoH3SPe89q1p16jNJnelXpV/jRHSuaTz1Muj5dXoR1T5WPDzAPw2gGuX/O/vvf9Sa+37ALwXwPMBfBrAT/ben26tXQvgPQD+HICHAfz13vtXR/VEAz4qE51nyqmdt8lgzcil9eojwyPrzWnRUtGONV+kmB74R+3gvuOwkyeLV29EDABVL5MV3ANufc2BGu+Kpx3JwwaDt3VF7/VYC1JR/bPEfWL8PGDwVkqeI5R522sM8DZJnZNZj9uOMwdqROoUecCtRpLLVN7bXtmr9BSA1/TefxDArQBe11p7JYC3Abij9/4SAN8EcNuS/zYA31zS71jylRprFHkFEbiwJ6BKxR3hWcToutbB/179WreWjdpbMTZapkLeKsTzlDIZtZz2kdd2rsN7bFjrZJ4ZL52Q0dgxMNk1846jlz1Vvf8ob2b8NqHMcLMMej7qQ54zXjgx+rRdVhfLMpoTlTyZwaj+vEfcvTZ77YxeRhW1I6JoDFl32QEZvXvFaAjc/YAeW06vXn4dwGsAvH9JfzeAn1iOX7+cY7n+F9sE4kSK4OXJgAM4+qVvb4J7x5lcmZyZklodo3ZoPRVZo7pHyldRSq8N+lOeI94joxGNOcvg8fPyq0HIqKKiDPjcv1F/bIsip6MibwTgI4BX/amUGfGfqTuSf6YOT7aMr9d/2yBv3kTgrbuSIirFuFtrF3AQDnkJgF8B8BUA3+q9f3vJcj+AFy3HLwJwHwD03r/dWnsEB+GUh6oN5Q7O9jt6g+M1mm9EefmpnaXByoCn0iZeHkV8VbZqfQr8IyAcgZ6RdwMwaoO3mlC+Xv+P2sM8ZsplE5fbUiUPSE4KtLdJLHO09TXqqwjEo3LbljkboxE+VOuo8PRIX3DnhTy8694cqjoBJeDuvX8HwK2ttRsBfADAD5RbFVBr7XYAtwPANddcc6zzWGGsExweIXgwv6uuuurIJ7U43jk7eSvKugaM1tKsMq+VI1Iw7YNRPyqfGRkiLz97i6M3RpFxq1DVWztr8sbDjkdjyeUrHi3/a/mTpMhgMi5U5NjE6I6AebTK1zRu18Yet1Hv/VuttY8DeBWAG1trFxev+2YADyzZHgBwC4D7W2sXAXw3Dm5SKq87AdwJADfccEP37v7a2/ysMXrd/jPvla1f70df+j/rOW/bw4gs/VlMgjUeLANBtpKo8NT0iJcqtOf5qlc/AzjcLpVJHxVnx0InndfOWYDIeM3wsLq91VDEP1qdjPKupZEDMuMMeOlqrKO6K96u5dM+jEA7u+5tq630Y2VXyQsAPLOA9nUAXouDG44fB/AGHOwseTOA31yKfHA5/93l+sf6QBL2mqjeY28d0wlbXT5Fk9PyjG7CePIq32iw13rfs55glffIQ9Gymafq1ecpdCYXT6wIqPmYlZ3LMB/vHdIReOt4e94Pn+vNI3YutkkKOpE+zxjJik7ZdW/r56wBXCNnxsMDSiXVl7VUmc8KzJYWfc3LA23+N8cSOP6iK6WKx30TgHe3gzj3VQDe13u/q7X2RQDvba39KwCfBfCuJf+7APyn1tq9AP4fgDcW6jgWc1OljSY7/2eemEc6aU/Ds/XqipR+E8XbhKL+VfKAb1Zuzh9NSJ0gCiTKR+WZMcqR4VCjwWlcZ8VwzVC1PzOAszRte7SyiOaFl57p8UnNJ89BiPAhcjCYKnm4npHB0PNIv1trhztJOH/l+YIhcPfePwfg5U76/wHwCif9MoC/NuIrZdy30GWgzWW9YzvPQKeS76SoaliUKisJuzZaPo6uVYBbjyPlzlYWUTnvOBo3Lr/GGKsB8Za4dqw7ASLvfBPjOzPO2o6oXDaOHshH4D3KU6lzE4rG2uv30RzyaDR3MvBWD9zyRi8cUx22jRTb8LhPhTj2rI0H8sGIACYbNDYOGc16aRVvqzK5RqsMr7wHWlVvjflH9xWY+OZuVCe3pbJM9zxuBsjIa450JTL4mRysf/ZUJMuvk3BkBCODlnm/GThEdWXno3S+xvOiAtzKd03dEY1WLwqk1f6L5kzEay2pHqsOeg+xea/n8GhngJsnRfWxT28Se0qnEw84DjxrvO4Z7z+SP8qbKVdURyVPJouVMyOqpF7CrDdT9fQtryq7B3qZJ6N5R0bP/vlhCJVDjVHkaGi7PGMStV3TRkYioopTEqXN/q+Ro+L5VjxqzyiOQDtaJUX8PT7Rk7J83yN6qIZfIaD6VenTnQFuAMcaEg0K5/fOPQ8wmlzKJwLykaLOgv6aCeB51KM8s9eZd9VbN75cLlN+D6SVnyeH56VGqzPNP+oTj6fyz9rg1b+JF2hpEaDYeWSMZgx2lr4WuDcBdCD3sKPjCJSjunVco7EaORZe3a01F7Q9Gdc4XDsF3IAPnN4yQ/N6g+C9WGo04aIf1+uVy65n9a25tpanXa8A8gjgIwPJxxXQGhlUz4OeAYUqsKnHrV5Q5PUq4I7AZdRO5lnlE7WxSl7+kwTsGZnUEfB0wtOvqL8zcI4oqt+u2b/9og9pWD3MK9LDEe0McHMHGOhmgG3/3pfcAf91pdFAKgCt8WSq3t2obCV9ExpNesvjeY3eufLKPKXR9Ug+nWA6qa1cpC967k2WyMuOeEWyjvpBy44magZYUf4sT9Q/3nh7/yPDdxKUgTdf13bM8OSyHmn9lt+O9aVnnpfN2MJf5lJdqPTlTgG3WSs712vA+GvNXppOep1AbP08EPdo20A940VmVPHGIs9XFdMD6axPKvVE3mTmkXK93te1rTy/qY/l98CG5WDdU09b+ydrYwSKm4Kajou2bRMDrzwiYB7VUdHp0SpuxN8r7wF4ZEBH45CtiPg65+OndlmHtJz2MePYyOHwaKeAW28GZQ3Sxut1Jg+QvTzR9cwDyeqdpYqRYJod7DXyzNbhgf4ImD2D4fGsgsfI21UA9F5e7xkYLpOBUgWoR4ZqlCcC3Eq9yoPlz/jN6JjnPEUO1SzPzOvO5srs/PLyqVOgDqXnEHhl7ZN2VedQaaeAe7RUZbCueBt8jS0j89JvUmp9Ec9NQXvWoxnxqJbPQMcDnwrARLJ5Y6gAoXkikB8puLeU5nHWOpTYY/J0a7Q0rxopT+5NaRtGe61jU51/22qr8hode/JtotuqY3ot01PLz09KWrlZ2nng5gdzdCJGChZZdg2L6MeDK0Adpc3Qpt5MlWfkeWaAOpoQszJ5fEe0pow3UVUXvGUz65q3BM7k8+RVzyqT1ZOrShXHpcJD/z2w3RS0K+lVWaMwRKS3SpG37uWL5NCxZh2yR9aNh3eDMjPu1f7ZCeCOPGz754HwQFuvR96SGoBMYb08kXzVNs54NWsp83KiMADnzSbELHmefHbd85ZmVgEjQz6i3p+Lk3s8s8k8yjMqP7oW6Z8ez1LmmETXszpnwFzP16xSMv1inlWHwZOvmm43HPnrS/bvgXUFEyLaCeAGjt+AZBo1SIE2mqzRF8szzyMD6Kryesui6kTYlGY8Og+0Z+SKFLx6neVlWbzJNwK7qM7IY8vyRLxYPs854PPRGHB7qv3k/Ue8R9dm9H1tyGOT8hVDvamTYf/sVWtdqi+GVfo1eXYweYec6rNiQ7Vvdga4R1Y+amAEul5HZ/9VQPfknWmb1lcts5ZG3q4da93Zcs6TbwSAs0bA4xl55JuQjil7TFZn5OlnoBz1aXSu7arKXdWjNeDt5anwrOjGNmlWv2bl8RwavW4hEgVwrsfbtVQdR6WdAW5gDKR8HllXBm62nlE9UX0akpn1lGc9oGraWtp2GGRmklj+ClBlYQH1ikbGZ7Yt7B3xe0rYg9J2eeC9pp2jfopkrrQ10t3ZEExFZzJZvfZuQirPTPjMZOj9eHhsVgYjqz+6+Zg9m6I8ItmNdgK4FVwZfEeg63XEyDvw6qqUrXokWTs3OV9LEWiPlCYCxkr7KyGJ7Hp07v1vQp6BZ6+J94bzjSevrdmx53BEHvg2DXtVZ6uGYFs6GRktpsrYVoxJRXfVC/bCdZF+qj5670Ey8j4BqLJUaCeAGzj+4vYITDmPHlcAdsSjUi7LNzvBZjyXNRQBRuTZRUo+miCZAmYgq5PEky+SaQ1oV8BedS8Kl3C6dz3qY8ufgfdItmp6hTLHxTNEa+o6yTIjnRg5FTzOSt6qPXJoDLSfeeaZY143yxKF2GYckZ0EbuC4tZpRWA+Es87SAdTyngyRd1zxAKKyGc0o/lornnnYHrhqHZHSafnouidvFg5RHiOKJibzyLYBjtrB8o6MXGRA13q72zL0m/KrGNmqUV9Da50KS4/kqHrtnneuPDifynDuQiURcNvxDK/snNMioPbyePyyciO513j4VaoAS5Y+8hS886zeDOyylcAmxsqjCCztGm8l1RAc51N+OhFHcqocI+A6KV2p8N0mqGb8t9Wuisft1Z+tkjJSfc30IZpDa9q+E8AN+A88KGiOqAraGciO/qP8lTKRvCNQq/Dyyli5GeUYeQvZZN80dFFZtWwKIlF/qK615j8Cz3kz7yxaJUSgXRmfk/K4q3q6TVInYZNV4qiOURlvPEdG1ZN/pg1Rm7cWKmmtPQ/AbwO4dsn//t77L7XWfhXAnwfwyJL1b/be724HNb8DwI8DeGJJ/8yoHn3LnzWkqtAjYNEJMgPaWld2XFW4Sv41E6YyCbWvqsv/TevOeGee90h+TR8p/4xhMB30gN0r77UtujaSjdOyc16eR+3JDEmVtgngJxUayZyOKM/sCq+68vTq8fQ004WIKh73UwBe03t/rLV2NYDfaa399+XaP+y9v1/y/xiAly6/HwbwzuW/TCMvNyuTlY+uZ2CeyVFNP2tSwKnIN/L2Z0MTM3VoucrxjFyZrLps9vQga1PkfUVldN+vxy+6NiKVI5tPmb5Xrnt1z9CozjW0JuRW4TmaT54Rrcy7mZVx5WPBHcBjy+nVyy/j/noA71nK/V5r7cbW2k29968ldaQ3hSyPd8xWS/OMgFuP+TxLz0CmSrMTYS1VlmMKSJkyrp0MUfhA00b1R3xHXpDSKI96Q9Ux98IgWX0erwpYe0akAgwZL+84kk2PN/HkRzJtyqOiC0rR+EaGuWLseGWk4zaLIflHHZ9rxIXW2t0AHgTwkd77J5ZLb22tfa61dkdr7dol7UUA7qPi9y9pKfGb+vStffyL3g6oaVFHRnmqPKNrWR2jek+Sqv1Qlbkqb6XNlTpmx3mmPzNdAY7e/VcvKpIpauOofzTd0/NRPZlcmRyenNFYeMeVvhy1fyTnJjRbl6dzM7hTHfNN+qB0c7L3/h0At7bWbgTwgdbanwbwFgBfB3ANgDsB/GMA/6LCDwBaa7cDuB0ALl68eEwpMotmx56VqgJMNuGrCjyaQDO0TWVd4/1EnqLymY3bahnPg428Zu84WkFE/DeJpeoEjfJwnfw2y0jmTP4MHD3KPEM9H4VlvPpnPPoZuStl13jK2yQvLOd5zNsIyYwwT2lqV0nv/VuttY8DeF3v/ZeX5Kdaa/8RwD9Yzh8AcAsVu3lJU1534gDwce211/Y1ALwpSGZgW/U0gPmXYlXzrKUR7xHAjoA04jkCg8woRHVo3LlCWR0ef+9f+WWGJAJOldurQ/s4mgfVPvNCG/pq0ar+Vzy/bdKmdc04FF65SljRKDLQI8Pj9fUa56Kyq+QFAJ5ZQPs6AK8F8DaLW7eDWn8CwD1LkQ8C+JnW2ntxcFPykZ7Et43OQkGqXvToeBve9bbav6mHWYlzr5UlAuYqmEd51evdFmjPtpfl9WLuKhvLvC2dqPJZ6/RsIsdJlpvhmemad68lun8SOTvR/PFoNr9RxeO+CcC7W2sXcBATf1/v/a7W2scWUG8A7gbw95b8H8LBVsB7cbAd8G9VBFnjJc7yyzznNYBdqWdWxk1prWfsKWOUP/MkRvWsyZPVVQ2NjNKrXv2Mp5R56VG+LK1yjfN4fRPp7bZ0cRM+I5k27TfjUdF/Po5WWLOgXZk3VWqn6elGdO211/abbropzbPGk5z1NGZAemZibctzGtGsp1lJn4nXrs0bHXOahky8MESWJ2tbli9bUWXlPDk8GSNSnfPi05ExqC7XvfOqjm87TyW/9tvaeVbRf0+3snyZns3Uy3T33Xfj0qVL7sWdeXJyRCdpyTVNJ8asMp8VZQpfLZ+FIbSOGY/TKzPy0quevZdndpJo2dZa6WtKWZ1Z383I5XlwVdDKZNglXT7LVepMvV74xMtnNNJ3Lz3Kz3RugHstVcE3mgiZh17lddakYYQqoHrLR4+v5a2A6lnQGgMwGnfNpxPaC1VUZfPqqBqASvlqmTW0BoRmeG/SD1G50byohkZmnYg142K0k8C9TeCbienNhErWyHBa5ZRGAMzXtN0zYD+KQY5igKdFVWPL4x7FPbN075hpBEKZs3ESxnctVY3bpnnOQm+isZ+Ja2cr4bXt2Ung3gaNANubVGuWkSelSNs2XmvAe7Qs9MozRaGDTTyNbVIFcCJArfSn1mFt7/3onuoRoG6iC5nxPQ+6O+K7DT3KHAy7XjGAazzutbQzwD1S1DXWaeRB23EG3BGwR/XsKlW95xFQaXpEnlda4RnJFVGFT2U14F2LdGa0fNY6tWxVjkw2pizM4xnKs/Baz4L3Np2DbCWl43waTslOAHfk2azhk6VnHod3XPF6trUMXMNjbdxy1oOeMZoznsla8J5p91rQWgOoGXgzzyyksi3ywNu7PkpbQ5s6WSPep0Gb6Jh3fdu0E8BdoZHXq/lm0jLQ3taS9SRomyGHkVc4A/ZRTHANv02pNf+FQCqD95KztTLOhlKiOPlMfXqeGYYZ/rum8zMUyb7NOTMTTtkmnRvgBmrgnYG0Km4llJKljWRZk4+puuMgKrfWS58NLSj/DIxOU7m1PqW1QBlRtISO8nl514KqhvSqDs6mnmJ1tbzG4VLahfsiRjPhsZOgnQfukYJlS8AsPj0KjUT819BaHidRruqFaL41YZlNvJG19TF/87ZHY78JIHj6Ga00ZvnMyhBdy8a2ovOzxmRkLI1mbuqeBs2uCivhsZOinQNu7oAquEbpWdhDz8/zknCWPFA+i4miMmRpXshDKQJMKzt65/tJkSdPllfzZMY2AmbP+/f0fZux2co8W0OZMdglL/w0aeeAu+otjkC9avWj8pvQaXvYGa314Cqhkk1uGlZpBOhA7OUYgOnK6ywmeyV8tIanx9vLozRa0s/OiTWe+7YB/jzQmhvlHu0McK/xpPm4ohSR1zFb/yb5Tqp8xrfiqTJ54YWTAruT8PS9ycFe6GmCt3q+UR4N70R8ovSZ1ekorr8JaM/Ms1kQO0s9PS2qyl/6As5p0kz8bVaxZkG7yvtK8xTOqj0nUa+Nz66MURS+WyPfKDzoGSsDu6hPtiHPScyzrL7zSjYe/KvSznjcTJXBjuKVm3rOI9oGn/OucEajpb7nbVa8ycgTU0+ZvdSo3DZ3jXgyachhFIKIbpbOeJ+z7fG84IjXJo7RqK5R3ZW8UQjP0vUDBzP6puO6dmW2Vte53hHtDHBXlGomfUYpTmsLz57WEwNCFveOJtpZL6EzQFtjXCr5K/cptkmVEM1JyqH1bwK+zPOs7olk/XQugXt0fSb9NAblJL30NdvmtsFnmzQC3W3xPw2qeNqz2/NmYt6cf1bf18zBk6IZAze6+Tzqqyh2fhJb/DI+2QpSaaeB+zTS93S2dFKgvSZueFoUrR52TUd3UZ7IiHngXQHcCOi34a0rVXZI2cvHRn2/M8DNlMXdRumbLjnPWwx7VNfaGG8EKiOPbVNFr4Q6TmpinSR58dOM1gJ6tqPjJEMnm/Ladnkv9M6ebQAABcZJREFUNDWjK5VYPevgyEOPYtnM3z7qXNGPnQRuYC5Mss0bIWcVTqnIsZbProDbCJS9LXGjG00z9UR5105uYF2s2Su/jTGq8Ng1LzqjNX2yTV33wDhL8zBpJtzSez+8uXougFstz4wXfBpeQ0WeXQHH06ZRu6O4neeBZACuZTSPhkYqYZKzGrMopsrnezofFHn2HmDrTVNO0+t6TWkngBvYzIPexJuepbVhB49Gy+FdNxAMkh4Yefn0P1pSRkAe8ffAe6Yd2xzXEe8opjrLO+N53ug8yR8BbBRaHHnf7LiyB3+ub05uq9x5WiIC50feNSDppXne8sjb5rxZPbO7CrZJa0JVuxTe2lNMHgBXjDJf4593LaK2CwDRWrsE4MtnLccJ0fcCeOishTgBulLbBVy5bdu363zRn+i9v8C7sCse95d77z901kKcBLXWPnUltu1KbRdw5bZt364rh3buXSV72tOe9rSnnPbAvac97WlP54x2BbjvPGsBTpCu1LZdqe0Crty27dt1hdBO3Jzc0572tKc91WlXPO497WlPe9pTkc4cuFtrr2utfbm1dm9r7RfOWp5Zaq39h9bag621eyjte1prH2mt/cHy/0eX9NZa+7dLWz/XWvuzZyd5Tq21W1prH2+tfbG19oXW2s8u6ee6ba2157XWPtla+/2lXf98Sf++1tonFvn/a2vtmiX92uX83uX6i89S/hG11i601j7bWrtrOb9S2vXV1trnW2t3t9Y+taSda13chM4UuFtrFwD8CoAfA/AyAG9qrb3sLGVaQb8K4HWS9gsAPtp7fymAjy7nwEE7X7r8bgfwzlOScQ19G8DP995fBuCVAH56GZvz3ranALym9/6DAG4F8LrW2isBvA3AHb33lwD4JoDblvy3Afjmkn7Hkm+X6WcBfInOr5R2AcBf6L3fSlv/zrsurid9euc0fwBeBeDDdP4WAG85S5lWtuPFAO6h8y8DuGk5vgkH+9QB4N8BeJOXb9d/AH4TwGuvpLYBuB7AZwD8MA4e4Li4pB/qJYAPA3jVcnxxydfOWvagPTfjAMBeA+AuAO1KaNci41cBfK+kXTG6OPs761DJiwDcR+f3L2nnnV7Ye//acvx1AC9cjs9le5dl9MsBfAJXQNuWcMLdAB4E8BEAXwHwrd77t5csLPthu5brjwB4/ulKXKZ/A+AfAbDv+j0fV0a7AKAD+B+ttU+31m5f0s69Lq6lXXly8oql3ntvrZ3brTuttRsA/DqAn+u9PyrvZjiXbeu9fwfAra21GwF8AMAPnLFIG1Nr7S8DeLD3/unW2o+etTwnQD/Se3+gtfbHAHyktfa/+OJ51cW1dNYe9wMAbqHzm5e0805/2Fq7CQCW/weX9HPV3tba1TgA7V/rvf/GknxFtA0Aeu/fAvBxHIQQbmytmSPDsh+2a7n+3QAePmVRK/RqAH+ltfZVAO/FQbjkHTj/7QIA9N4fWP4fxIGxfQWuIF2cpbMG7v8J4KXLne9rALwRwAfPWKZt0AcBvHk5fjMO4sOW/lPLXe9XAniElno7Re3AtX4XgC/13t9Ol85121prL1g8bbTWrsNB3P5LOADwNyzZtF3W3jcA+FhfAqe7RL33t/Teb+69vxgH8+hjvfe/gXPeLgBorf2R1tp32TGAvwTgHpxzXdyIzjrIDuDHAfxvHMQZ/8lZy7NC/v8C4GsAnsFBLO02HMQKPwrgDwD8FoDvWfI2HOyi+QqAzwP4obOWP2nXj+Agrvg5AHcvvx8/720D8GcAfHZp1z0A/umS/v0APgngXgD/DcC1S/rzlvN7l+vff9ZtKLTxRwHcdaW0a2nD7y+/LxhOnHdd3OS3f3JyT3va057OGZ11qGRPe9rTnvY0SXvg3tOe9rSnc0Z74N7Tnva0p3NGe+De0572tKdzRnvg3tOe9rSnc0Z74N7Tnva0p3NGe+De0572tKdzRnvg3tOe9rSnc0b/H27GURSuaLcOAAAAAElFTkSuQmCC\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1, 400, 600, 3)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 9 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "from __future__ import print_function\n",
        "from skimage.io import imread\n",
        "from skimage import exposure, color\n",
        "from skimage.transform import resize\n",
        "from keras_preprocessing.image import ImageDataGenerator\n",
        "\n",
        "# Pre-process\n",
        "def imgGen(img, zca=False, rotation=0., w_shift=0., h_shift=0., shear=0., zoom=0., h_flip=False, v_flip=False,  preprocess_fcn=None, batch_size=9):\n",
        "    datagen = ImageDataGenerator(\n",
        "            zca_whitening=zca,\n",
        "            rotation_range=rotation,\n",
        "            width_shift_range=w_shift,\n",
        "            height_shift_range=h_shift,\n",
        "            shear_range=shear,\n",
        "            zoom_range=zoom,\n",
        "            fill_mode='nearest',\n",
        "            horizontal_flip=h_flip,\n",
        "            vertical_flip=v_flip,\n",
        "            preprocessing_function=preprocess_fcn,\n",
        "          )\n",
        "    \n",
        "    datagen.fit(img)\n",
        "\n",
        "    i=0\n",
        "    for img_batch in datagen.flow(img, batch_size=9, shuffle=False):\n",
        "        for img in img_batch:\n",
        "            plt.subplot(330 + 1 + i)\n",
        "            plt.imshow(img)\n",
        "            i=i+1    \n",
        "        if i >= batch_size:\n",
        "            break\n",
        "    plt.show()\n",
        "\n",
        "# Visualize the original image\n",
        "img = imread(\"./Multi-class Weather Dataset/Cloudy/cloudy1.jpg\")\n",
        "plt.imshow(img)\n",
        "plt.show()\n",
        "\n",
        "# Reshape raw images, ready for data enhancement\n",
        "img = img.astype('float32')\n",
        "img /= 255\n",
        "h_dim = np.shape(img)[0]\n",
        "w_dim = np.shape(img)[1]\n",
        "num_channel = np.shape(img)[2]\n",
        "img = img.reshape(1, h_dim, w_dim, num_channel)\n",
        "print(img.shape)\n",
        "\n",
        "# Data enhancement\n",
        "imgGen(img, rotation=30, h_shift=0.5)\n",
        "\n"
      ],
      "id": "CqwExPnF4A4E"
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "E6GirxLZH9Zn"
      },
      "outputs": [],
      "source": [
        "img_height,img_width = 90,90 # Setup the image size as 90,90"
      ],
      "id": "E6GirxLZH9Zn"
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "78313df2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "62cbfd3b-3c08-4b96-b334-2e409ee7e11e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 901 images belonging to 4 classes.\n",
            "Found 224 images belonging to 4 classes.\n",
            "Found 224 images belonging to 4 classes.\n"
          ]
        }
      ],
      "source": [
        "from keras_preprocessing.image import ImageDataGenerator\n",
        "datagen = ImageDataGenerator( # Pre-process the image, including width,height,size...\n",
        "    rescale = 1.0/255.,# Scale the dimensions\n",
        "    rotation_range=90,# The image is rotated randomly by a certain angle, and the maximum rotation angle is the set value\n",
        "    width_shift_range=0.2,# The image pans horizontally randomly, and the maximum pan value is the set value. If the value is less than 1 float value, it can be considered to be proportional translation, if greater than 1, the translation is pixels; If the value is integer, the translation is also pixels; Assuming a pixel of 2.0, the movement range is between [-1,1].\n",
        "    height_shift_range=0.2,# Image randomly pans vertically, same as above\n",
        "    shear_range=0.2,# Images are randomly trimmed\n",
        "    zoom_range=0.3,# Images are randomly zoomed\n",
        "    validation_split=0.2, # Optional float between 0 and 1, fraction of data to reserve for validation.\n",
        "    vertical_flip=True)# The image is randomly flipped vertically\n",
        "train_generator = datagen.flow_from_directory( # Divide the dataset into tarin and valid\n",
        "    dataset,\n",
        "    batch_size=16, # Size of the batches of data. Default: 32. If None, the data will not be batched (the dataset will yield individual samples).\n",
        "    shuffle=True, # Whether to shuffle the data. Default: True.\n",
        "    class_mode='categorical', # one of \"binary\", \"categorical\", \"input\", \"multi_output\", \"raw\", sparse\" or None. Default: \"categorical\". \"categorical\": 2D numpy array of one-hot encoded labels. Supports multi-label output.\n",
        "    subset='training', # Subset of the data to return. One of \"training\", \"validation\" or \"both\". Only used if validation_split is set. When subset=\"both\", the utility returns a tuple of two datasets (the training and validation datasets respectively).\n",
        "    target_size=(img_height,img_width))\n",
        "valid_generator = datagen.flow_from_directory(\n",
        "    dataset, \n",
        "    batch_size=16,\n",
        "    shuffle=True,\n",
        "    class_mode='categorical',\n",
        "    subset='validation',\n",
        "    target_size=(img_height,img_width))\n",
        "test_generator = datagen.flow_from_directory(\n",
        "    dataset, \n",
        "    batch_size=16,\n",
        "    shuffle=True,\n",
        "    class_mode='categorical',\n",
        "    subset='validation',\n",
        "    target_size=(img_height,img_width))"
      ],
      "id": "78313df2"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "npt82eiQNVTl"
      },
      "source": [
        "# 3.ResNet152V2 Model"
      ],
      "id": "npt82eiQNVTl"
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "uOmizxCJOf4u"
      },
      "outputs": [],
      "source": [
        "# Matrix contents required when compile\n",
        "METRICS = [\n",
        "        'accuracy',\n",
        "        tf.keras.metrics.Precision(),\n",
        "        tf.keras.metrics.Recall(),\n",
        "        tf.keras.metrics.AUC()\n",
        "    ]\n",
        "\n",
        "# Create a function that can generate the images\n",
        "def acc_loss(history): \n",
        "  train_acc = history.history['accuracy'] # import the argument\n",
        "  train_loss = history.history['loss']\n",
        "  train_precision = history.history['precision']\n",
        "  train_recall = history.history['recall']\n",
        "  train_auc = history.history['auc']\n",
        "\n",
        "  val_acc = history.history['val_accuracy']\n",
        "  val_loss = history.history['val_loss']\n",
        "  val_precision = history.history['val_precision']\n",
        "  val_recall = history.history['val_recall']\n",
        "  val_auc = history.history['val_auc']\n",
        "\n",
        "  # 1.Loss and accuacy\n",
        "  fig1 = plt.figure(figsize=(10, 4)) \n",
        "  plt.subplot(1, 2, 1) # row,column,number\n",
        "  plt.plot(train_acc, label='Training')\n",
        "  plt.plot(val_acc, label='Validation')\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.ylabel('Accuracy')\n",
        "  plt.legend()\n",
        "  \n",
        "  plt.subplot(1, 2, 2) \n",
        "  plt.plot(train_loss, label='Training')\n",
        "  plt.plot(val_loss, label='Validation')\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.ylabel('Loss')\n",
        "  plt.legend()\n",
        "  plt.show() # Show that a graph contains 2 subgraphs\n",
        "\n",
        "  # 2.Precision, recall and AUC\n",
        "  fig2 = plt.figure(figsize=(15, 4)) \n",
        "  plt.subplot(1, 3, 1) \n",
        "  plt.plot(train_precision, label='Training')\n",
        "  plt.plot(val_precision, label='Validation')\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.ylabel('Precision')\n",
        "  plt.legend()\n",
        "\n",
        "  plt.subplot(1, 3, 2) \n",
        "  plt.plot(train_recall, label='Training')\n",
        "  plt.plot(val_recall, label='Validation')\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.ylabel('Recall')\n",
        "  plt.legend()\n",
        "\n",
        "  plt.subplot(1, 3, 3) \n",
        "  plt.plot(train_auc, label='Training')\n",
        "  plt.plot(val_auc, label='Validation')\n",
        "  plt.xlabel('Epochs')\n",
        "  plt.ylabel('AUC')\n",
        "  plt.legend()\n",
        "  plt.show() # Show that a graph contains 3 subgraphs\n",
        "\n",
        "# Create a function that can generate the confusion matrix and classification report\n",
        "from sklearn.metrics import confusion_matrix, classification_report\n",
        "from sklearn.metrics._plot.confusion_matrix import ConfusionMatrixDisplay\n",
        "def prediction(model): \n",
        "  probabilities = model.predict(valid_generator)\n",
        "  predictions = []\n",
        "  for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "  labels = valid_generator.classes\n",
        "  # tn,fp,fn,tp = confusion_matrix(labels, predictions).ravel()\n",
        "  cm = confusion_matrix(labels, predictions) #[[tp, fp],[fn, tn]]\n",
        "  cr = classification_report(labels, predictions)\n",
        "  print(cr)\n",
        "  cm_dis = ConfusionMatrixDisplay(confusion_matrix=cm,display_labels=class_names).plot()"
      ],
      "id": "uOmizxCJOf4u"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dF3u2LarVMQC"
      },
      "source": [
        "## 3.1Original Model"
      ],
      "id": "dF3u2LarVMQC"
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "U1qxzFvwpGqv",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "17b2016d-e5fa-474a-c6d0-d5f8bd83eb07"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet152v2_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "234545216/234545216 [==============================] - 8s 0us/step\n"
          ]
        }
      ],
      "source": [
        "from tensorflow.keras.applications import ResNet152V2\n",
        "base_model = ResNet152V2(include_top=False, input_shape=(img_height,img_width,3)) # Choose the CNN structure ResNet152V2, and decide the input size of images\n",
        "base_model.trainable = False"
      ],
      "id": "U1qxzFvwpGqv"
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "UO4pRtBXNUt2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "65497038-c4c5-44b7-ef05-85780017d904"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " resnet152v2 (Functional)    (None, 3, 3, 2048)        58331648  \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 3, 3, 2048)        0         \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 18432)             0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 64)                1179712   \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 4)                 260       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 59,511,620\n",
            "Trainable params: 1,179,972\n",
            "Non-trainable params: 58,331,648\n",
            "_________________________________________________________________\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/optimizers/optimizer_v2/adam.py:110: UserWarning:\n",
            "\n",
            "The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "\n"
          ]
        }
      ],
      "source": [
        "model1=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3), # Base the ResNet152V2, we add some layers to make the result more precise\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model1.summary() # Check the model structue\n",
        "\n",
        "from keras.optimizers import Adam\n",
        "optimizer=Adam(lr=0.01, beta_1=0.5)\n",
        "model1.compile(\n",
        "    optimizer = optimizer, #'Adam'\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "UO4pRtBXNUt2"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GtvBiCxqNwxu",
        "outputId": "571b3f67-ed9f-497f-ecab-d7362a652eb6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "57/57 [==============================] - 407s 7s/step - loss: 2.2874 - accuracy: 0.7503 - precision: 0.7683 - recall: 0.7358 - auc: 0.8874 - val_loss: 0.9174 - val_accuracy: 0.7188 - val_precision: 0.7586 - val_recall: 0.6875 - val_auc: 0.9193\n",
            "Epoch 2/100\n",
            "57/57 [==============================] - 52s 909ms/step - loss: 0.7661 - accuracy: 0.8135 - precision: 0.8390 - recall: 0.7925 - auc: 0.9444 - val_loss: 1.3392 - val_accuracy: 0.7500 - val_precision: 0.7512 - val_recall: 0.7277 - val_auc: 0.9016\n",
            "Epoch 3/100\n",
            "57/57 [==============================] - 51s 900ms/step - loss: 0.6795 - accuracy: 0.8579 - precision: 0.8647 - recall: 0.8368 - auc: 0.9599 - val_loss: 0.6958 - val_accuracy: 0.7812 - val_precision: 0.8086 - val_recall: 0.7545 - val_auc: 0.9435\n",
            "Epoch 4/100\n",
            "57/57 [==============================] - 55s 956ms/step - loss: 0.8239 - accuracy: 0.8257 - precision: 0.8499 - recall: 0.7669 - auc: 0.9416 - val_loss: 0.9931 - val_accuracy: 0.7321 - val_precision: 0.7979 - val_recall: 0.6875 - val_auc: 0.9108\n",
            "Epoch 5/100\n",
            "57/57 [==============================] - 52s 910ms/step - loss: 0.5032 - accuracy: 0.8657 - precision: 0.8949 - recall: 0.8413 - auc: 0.9662 - val_loss: 1.0476 - val_accuracy: 0.7902 - val_precision: 0.7936 - val_recall: 0.7723 - val_auc: 0.9280\n",
            "Epoch 6/100\n",
            "57/57 [==============================] - 53s 925ms/step - loss: 0.4573 - accuracy: 0.8790 - precision: 0.9023 - recall: 0.8613 - auc: 0.9728 - val_loss: 0.4369 - val_accuracy: 0.8170 - val_precision: 0.8551 - val_recall: 0.7902 - val_auc: 0.9684\n",
            "Epoch 7/100\n",
            "57/57 [==============================] - 50s 880ms/step - loss: 0.4039 - accuracy: 0.9023 - precision: 0.9197 - recall: 0.8768 - auc: 0.9762 - val_loss: 0.8626 - val_accuracy: 0.8259 - val_precision: 0.8465 - val_recall: 0.8125 - val_auc: 0.9453\n",
            "Epoch 8/100\n",
            "57/57 [==============================] - 50s 885ms/step - loss: 0.4361 - accuracy: 0.8923 - precision: 0.9097 - recall: 0.8835 - auc: 0.9725 - val_loss: 0.7277 - val_accuracy: 0.7768 - val_precision: 0.8333 - val_recall: 0.7589 - val_auc: 0.9355\n",
            "Epoch 9/100\n",
            "57/57 [==============================] - 53s 937ms/step - loss: 0.4547 - accuracy: 0.8890 - precision: 0.9138 - recall: 0.8701 - auc: 0.9731 - val_loss: 0.7990 - val_accuracy: 0.8036 - val_precision: 0.8610 - val_recall: 0.7188 - val_auc: 0.9364\n",
            "Epoch 10/100\n",
            "57/57 [==============================] - 54s 952ms/step - loss: 0.3001 - accuracy: 0.9134 - precision: 0.9372 - recall: 0.8779 - auc: 0.9842 - val_loss: 0.5047 - val_accuracy: 0.8705 - val_precision: 0.8942 - val_recall: 0.8304 - val_auc: 0.9661\n",
            "Epoch 11/100\n",
            "57/57 [==============================] - 54s 958ms/step - loss: 0.4658 - accuracy: 0.8846 - precision: 0.9262 - recall: 0.8635 - auc: 0.9737 - val_loss: 0.8549 - val_accuracy: 0.8036 - val_precision: 0.8660 - val_recall: 0.7500 - val_auc: 0.9405\n",
            "Epoch 12/100\n",
            "57/57 [==============================] - 54s 943ms/step - loss: 0.3684 - accuracy: 0.9068 - precision: 0.9266 - recall: 0.8824 - auc: 0.9787 - val_loss: 0.6356 - val_accuracy: 0.8125 - val_precision: 0.8424 - val_recall: 0.7634 - val_auc: 0.9485\n",
            "Epoch 13/100\n",
            "57/57 [==============================] - 55s 965ms/step - loss: 0.4400 - accuracy: 0.9179 - precision: 0.9359 - recall: 0.8912 - auc: 0.9804 - val_loss: 0.5823 - val_accuracy: 0.8036 - val_precision: 0.8485 - val_recall: 0.7500 - val_auc: 0.9481\n",
            "Epoch 14/100\n",
            "57/57 [==============================] - 55s 968ms/step - loss: 0.4571 - accuracy: 0.8923 - precision: 0.9236 - recall: 0.8724 - auc: 0.9745 - val_loss: 0.5739 - val_accuracy: 0.8080 - val_precision: 0.8607 - val_recall: 0.7723 - val_auc: 0.9507\n",
            "Epoch 15/100\n",
            "57/57 [==============================] - 54s 952ms/step - loss: 0.3840 - accuracy: 0.9001 - precision: 0.9290 - recall: 0.8713 - auc: 0.9774 - val_loss: 0.5700 - val_accuracy: 0.7991 - val_precision: 0.8914 - val_recall: 0.6964 - val_auc: 0.9461\n",
            "Epoch 16/100\n",
            "57/57 [==============================] - 52s 914ms/step - loss: 0.3891 - accuracy: 0.8846 - precision: 0.9361 - recall: 0.8291 - auc: 0.9765 - val_loss: 0.5425 - val_accuracy: 0.8482 - val_precision: 0.9053 - val_recall: 0.7679 - val_auc: 0.9551\n",
            "Epoch 17/100\n",
            "57/57 [==============================] - ETA: 0s - loss: 0.3345 - accuracy: 0.9057 - precision: 0.9506 - recall: 0.8757 - auc: 0.9821"
          ]
        }
      ],
      "source": [
        "history1 = model1.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=100) "
      ],
      "id": "GtvBiCxqNwxu"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "sXgp822lfoVN"
      },
      "outputs": [],
      "source": [
        "acc_loss(history1) # Plot the image\n",
        "prediction(model1)"
      ],
      "id": "sXgp822lfoVN"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "r33AhsAYnVP7"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "import pandas as pd\n",
        "\n",
        "probabilities = model1.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "r33AhsAYnVP7"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3I2uHB2FoxUv"
      },
      "source": [
        "## 3.2Turning the number of epochs during training"
      ],
      "id": "3I2uHB2FoxUv"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Ic06TfvpoxUx"
      },
      "outputs": [],
      "source": [
        "model2=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model2.summary()\n",
        "model2.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "Ic06TfvpoxUx"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "YH2VriUcoxUx"
      },
      "outputs": [],
      "source": [
        "history2 = model2.fit(\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=60) # Change the number of epochs, exploring the influence"
      ],
      "id": "YH2VriUcoxUx"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "2eJbOnamoxUx"
      },
      "outputs": [],
      "source": [
        "acc_loss(history2)\n",
        "prediction(model2)"
      ],
      "id": "2eJbOnamoxUx"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "HzKvVEVFoxUx"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "probabilities = model2.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "HzKvVEVFoxUx"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uj9RxW6CpB6z"
      },
      "source": [
        "##3.3Turning the flatten layers in ResNet"
      ],
      "id": "uj9RxW6CpB6z"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "w3-ZpscApB60"
      },
      "outputs": [],
      "source": [
        "model3=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Flatten(), # Add anothe fatten layer, exporing the influence of flatten\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model3.summary()\n",
        "model3.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "w3-ZpscApB60"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ORXJ5bmHpB60"
      },
      "outputs": [],
      "source": [
        "history3 = model3.fit(\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=30) "
      ],
      "id": "ORXJ5bmHpB60"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "O_j_DPSwpB60"
      },
      "outputs": [],
      "source": [
        "acc_loss(history3)\n",
        "prediction(model3)"
      ],
      "id": "O_j_DPSwpB60"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "p3ih3jvipB60"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "from sklearn.metrics import confusion_matrix\n",
        "import seaborn as sns\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "probabilities = model3.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "p3ih3jvipB60"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8sBh7B3AqGaE"
      },
      "source": [
        "## 3.4Turning the activation function of the dense layer"
      ],
      "id": "8sBh7B3AqGaE"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Mq5av7jwqGaS"
      },
      "outputs": [],
      "source": [
        "model4=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(64,activation='sigmoid'), # Change the activation function to explore the influence\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model4.summary()\n",
        "model4.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "Mq5av7jwqGaS"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "No4dpLqUqGaT"
      },
      "outputs": [],
      "source": [
        "history4 = model4.fit(\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=30) "
      ],
      "id": "No4dpLqUqGaT"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "fmXQmmMeqGaT"
      },
      "outputs": [],
      "source": [
        "acc_loss(history4)\n",
        "prediction(model4)"
      ],
      "id": "fmXQmmMeqGaT"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "BI3ZCRcrqGaT"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "probabilities = model4.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "BI3ZCRcrqGaT"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "a6Jz6SieVAwV"
      },
      "source": [
        "## 3.5Turning the optimiser of model"
      ],
      "id": "a6Jz6SieVAwV"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "JcM0NSKaVAwW"
      },
      "outputs": [],
      "source": [
        "model5=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3), # Base the ResNet152V2, we add some layers to make the result more precise\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model5.summary() # Check the model structue\n",
        "model5.compile(\n",
        "    optimizer = 'sgd', # Change the optimiser\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "JcM0NSKaVAwW"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "lwhHiUHPVAwX"
      },
      "outputs": [],
      "source": [
        "history5 = model5.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=30) "
      ],
      "id": "lwhHiUHPVAwX"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-bP850RgVAwX"
      },
      "outputs": [],
      "source": [
        "acc_loss(history5) # Plot the image\n",
        "prediction(model5)"
      ],
      "id": "-bP850RgVAwX"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "dcCZVLTMVAwX"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "probabilities = model5.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "dcCZVLTMVAwX"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zq-YecN7ezCX"
      },
      "source": [
        "## 3.6Turning the Batch Normalisation layer of model"
      ],
      "id": "zq-YecN7ezCX"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "O0Q82j10ezCj"
      },
      "outputs": [],
      "source": [
        "model6=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3), # Base the ResNet152V2, we add some layers to make the result more precise\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.BatchNormalization(),\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "model6.summary() # Check the model structue\n",
        "model6.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "id": "O0Q82j10ezCj"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "fhDcfSeOezCl"
      },
      "outputs": [],
      "source": [
        "history6 = model6.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=30) "
      ],
      "id": "fhDcfSeOezCl"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "GqMfwhUVezCm"
      },
      "outputs": [],
      "source": [
        "acc_loss(history6) # Plot the image\n",
        "prediction(model6)"
      ],
      "id": "GqMfwhUVezCm"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "tAYutOeeezCm"
      },
      "outputs": [],
      "source": [
        "# Evaluate the CNN Model\n",
        "\n",
        "probabilities = model6.predict(valid_generator)\n",
        "predictions = []\n",
        "for prob in probabilities:\n",
        "    best_index = np.argmax(prob)\n",
        "    predictions.append(best_index)\n",
        "#confusion_matrix(valid_generator.classes, predictions)\n",
        "\n",
        "# Kappa coefficient is the model evaluation parameter obtained based on the calculation of confusion matrix (the better to close 1).\n",
        "def kappa_cal(matrix):\n",
        "    n = np.sum(matrix)\n",
        "    sum_po = 0\n",
        "    sum_pe = 0\n",
        "    for i in range(len(matrix[0])):\n",
        "        sum_po += matrix[i][i]\n",
        "        row = np.sum(matrix[i, :])\n",
        "        col = np.sum(matrix[:, i])\n",
        "        sum_pe += row * col\n",
        "    po = sum_po / n\n",
        "    pe = sum_pe / (n * n)\n",
        "    # print(po, pe)\n",
        "    return (po - pe) / (1 - pe)\n",
        "\n",
        "Kappa_confficient = kappa_cal(confusion_matrix(valid_generator.classes, predictions))\n",
        "print(\"Kappa Confficient: \", Kappa_confficient)\n",
        "\n",
        "# Hamming distance （the better to close 0)\n",
        "from sklearn.metrics import hamming_loss\n",
        "ham_distance = hamming_loss(valid_generator.classes, predictions)\n",
        "print(\"Hamming Distance:  \", ham_distance)\n",
        "\n",
        "from sklearn.metrics import jaccard_score\n",
        "# Jaccard coefficient (the better to close 1)\n",
        "# Because the dataset is multiclassified, it needs to change the average parameter to micro instead of binary\n",
        "# The original function is -> jaccard_similarity_score, but it is modified to jaccard_score in 2021\n",
        "jaccrd_score = jaccard_score(valid_generator.classes, predictions,pos_label = \"PAIDOFF\",average='micro')\n",
        "print(\"Jaccard_Index:     \", jaccrd_score)"
      ],
      "id": "tAYutOeeezCm"
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 4.Cross-folding verification"
      ],
      "metadata": {
        "id": "Zv3zAh-u6Zxp"
      },
      "id": "Zv3zAh-u6Zxp"
    },
    {
      "cell_type": "code",
      "source": [
        "from keras_preprocessing.image import ImageDataGenerator\n",
        "datagen = ImageDataGenerator( # Pre-process the image, including width,height,size...\n",
        "    rescale = 1.0/255.,# Scale the dimensions\n",
        "    rotation_range=90,# The image is rotated randomly by a certain angle, and the maximum rotation angle is the set value\n",
        "    width_shift_range=0.2,# The image pans horizontally randomly, and the maximum pan value is the set value. If the value is less than 1 float value, it can be considered to be proportional translation, if greater than 1, the translation is pixels; If the value is integer, the translation is also pixels; Assuming a pixel of 2.0, the movement range is between [-1,1].\n",
        "    height_shift_range=0.2,# Image randomly pans vertically, same as above\n",
        "    shear_range=0.2,# Images are randomly trimmed\n",
        "    zoom_range=0.3,# Images are randomly zoomed\n",
        "    validation_split=0.5, # Optional float between 0 and 1, fraction of data to reserve for validation.\n",
        "    vertical_flip=True)# The image is randomly flipped vertically\n",
        "train_generator = datagen.flow_from_directory( # Divide the dataset into tarin and valid\n",
        "    dataset,\n",
        "    batch_size=16, # Size of the batches of data. Default: 32. If None, the data will not be batched (the dataset will yield individual samples).\n",
        "    shuffle=True,\n",
        "    class_mode='categorical',\n",
        "    subset='training', # Subset of the data to return. One of \"training\", \"validation\" or \"both\". Only used if validation_split is set. When subset=\"both\", the utility returns a tuple of two datasets (the training and validation datasets respectively).\n",
        "    target_size=(img_height,img_width))\n",
        "valid_generator = datagen.flow_from_directory(\n",
        "    dataset, \n",
        "    batch_size=16,\n",
        "    shuffle=True,\n",
        "    class_mode='categorical',\n",
        "    subset='validation',\n",
        "    target_size=(img_height,img_width))\n",
        "test_generator = datagen.flow_from_directory(\n",
        "    dataset, \n",
        "    batch_size=16,\n",
        "    shuffle=True,\n",
        "    class_mode='categorical',\n",
        "    subset='validation',\n",
        "    target_size=(img_height,img_width))\n",
        "\n",
        "model1.summary()\n",
        "\n",
        "model1.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)\n",
        "\n",
        "history7_1 = model1.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=30) \n",
        "\n",
        "acc_loss(history7_1) # Plot the image\n",
        "\n",
        "model1.compile(\n",
        "    optimizer = 'Adam',\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)\n",
        "\n",
        "history7_2 = model1.fit( # Import the dataset to train the model\n",
        "    valid_generator,\n",
        "    validation_data = train_generator,\n",
        "    epochs=30) \n",
        "\n",
        "acc_loss(history7_2) # Plot the image\n"
      ],
      "metadata": {
        "id": "ygZ-YFb76hkz"
      },
      "id": "ygZ-YFb76hkz",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 5.Modify the overfit"
      ],
      "metadata": {
        "id": "LqOmnHDrMf7Z"
      },
      "id": "LqOmnHDrMf7Z"
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 5.1 Learning rate "
      ],
      "metadata": {
        "id": "q6QW4e6qNSIO"
      },
      "id": "q6QW4e6qNSIO"
    },
    {
      "cell_type": "code",
      "source": [
        "model1.summary() # Check the model structue\n",
        "\n",
        "from keras.optimizers import Adam\n",
        "optimizer=Adam(lr=0.01, beta_1=0.5)\n",
        "model1.compile(\n",
        "    optimizer = optimizer, #'Adam'\n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "metadata": {
        "id": "yaM40hhJM0kH"
      },
      "id": "yaM40hhJM0kH",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history1 = model1.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=100) \n",
        "acc_loss(history1) # Plot the image\n",
        "prediction(model1)"
      ],
      "metadata": {
        "id": "cVS40oOBNGJ2"
      },
      "id": "cVS40oOBNGJ2",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 5.2 Regularization\n"
      ],
      "metadata": {
        "id": "TUkvBRYaNxDZ"
      },
      "id": "TUkvBRYaNxDZ"
    },
    {
      "cell_type": "code",
      "source": [
        "model1=tf.keras.models.Sequential([\n",
        "     base_model,\n",
        "    tf.keras.layers.Dropout(0.3), # Base the ResNet152V2, we add some layers to make the result more precise\n",
        "    tf.keras.layers.Flatten(),\n",
        "    Dense(256, kernel_regularizer=keras.regularizers.l1(0.01)),\n",
        "    '''\n",
        "    Dense(256, kernel_regularizer=keras.regularizers.l1(0.01))\n",
        "    Dense(256, kernel_regularizer=keras.regularizers.l2(0.01))\n",
        "    Dense(256, kernel_regularizer=keras.regularizers.l1_l2(0.01))\n",
        "\n",
        "    '''\n",
        "    tf.keras.layers.Dense(64,activation='relu'),\n",
        "    tf.keras.layers.Dense(4,activation='softmax') # num_classes\n",
        "])\n",
        "\n",
        "model1.summary() # Check the model structue\n",
        "\n",
        "from keras.optimizers import Adam\n",
        "\n",
        "model1.compile(\n",
        "    optimizer = 'Adam', \n",
        "    loss = tf.keras.losses.CategoricalCrossentropy(),\n",
        "    metrics = METRICS)"
      ],
      "metadata": {
        "id": "c3rYJ02vN6jl"
      },
      "id": "c3rYJ02vN6jl",
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history1 = model1.fit( # Import the dataset to train the model\n",
        "    train_generator,\n",
        "    validation_data = valid_generator,\n",
        "    epochs=100) \n",
        "acc_loss(history1) # Plot the image\n",
        "prediction(model1)"
      ],
      "metadata": {
        "id": "MhFffNTPN97I"
      },
      "id": "MhFffNTPN97I",
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [
        "a6Jz6SieVAwV"
      ],
      "machine_shape": "hm",
      "provenance": []
    },
    "gpuClass": "premium",
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.12"
    },
    "papermill": {
      "default_parameters": {},
      "duration": 223.887359,
      "end_time": "2022-04-05T17:04:52.841765",
      "environment_variables": {},
      "exception": null,
      "input_path": "__notebook__.ipynb",
      "output_path": "__notebook__.ipynb",
      "parameters": {},
      "start_time": "2022-04-05T17:01:08.954406",
      "version": "2.3.3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}