Googlenet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CNN2.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "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": "3.9.4"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ewa5uOvRx-Tt"
      },
      "source": [
        "import os \n",
        "import zipfile \n",
        "\n",
        "import tensorflow as tf \n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator \n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.layers import Dense, Input, Dropout,Flatten, Conv2D, MaxPool2D\n",
        "from tensorflow.keras.layers import BatchNormalization, Activation, MaxPooling2D\n",
        "from tensorflow.keras.models import Model, Sequential\n",
        "from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "\n",
        "import glob\n",
        "from PIL import Image\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from keras.utils.np_utils import to_categorical # convert to one-hot-encoding\n",
        "\n",
        "from IPython.display import clear_output\n",
        "from tqdm import tqdm\n",
        "\n",
        "from sklearn.metrics import confusion_matrix\n",
        "import seaborn as sns\n",
        "\n",
        "from keras import backend as K\n",
        "\n",
        "from sklearn.utils import resample"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sOSKOnc3OkVX",
        "outputId": "8abc8d52-d1f5-4412-ddbe-58faf735cc78"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive/')"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive/; to attempt to forcibly remount, call drive.mount(\"/content/drive/\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5HyWq_mhW75S"
      },
      "source": [
        "# data_dir = \"images/\"\n",
        "# metadata_path = \"HAM10000_metadata.csv\"\n",
        "data_dir = \"/content/drive/MyDrive/Colab Notebooks/images/\"\n",
        "metadata_path = \"/content/drive/MyDrive/Colab Notebooks/HAM10000_metadata.csv\""
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Gf2eWAukQrVe"
      },
      "source": [
        "#Lesion Dictionary\n",
        "lesion_type_dict = {\n",
        "    'nv': 'Melanocytic nevi',\n",
        "    'mel': 'Melanoma',\n",
        "    'bkl': 'Benign keratosis-like lesions ',\n",
        "    'bcc': 'Basal cell carcinoma',\n",
        "    'akiec': 'Actinic keratoses',\n",
        "    'vasc': 'Vascular lesions',\n",
        "    'df': 'Dermatofibroma'\n",
        "}"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GEXRn0E25KCy"
      },
      "source": [
        "# Read metadata into dataframe\n",
        "metadata = pd.read_csv(metadata_path)\n",
        "skin_df = metadata"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZvxnWcZt5KCy"
      },
      "source": [
        "# Dictionary for Image Names\n",
        "imageid_path_dict = {os.path.splitext(os.path.basename(x))[0]: x for x in glob.glob(os.path.join(data_dir, '*.jpg'))}"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VNqP0ahB5KCy"
      },
      "source": [
        "#Create useful Columns - Images Path, Lesion Type and Lesion Categorical Code\n",
        "skin_df['path'] = skin_df['image_id'].map(imageid_path_dict.get)\n",
        "skin_df['cell_type'] = skin_df['dx'].map(lesion_type_dict.get) \n",
        "skin_df['cell_type_idx'] = pd.Categorical(skin_df['cell_type']).codes"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0POOFz2V5KCz",
        "outputId": "59d87f92-eccd-4c6c-fdb8-9b8a64d7674e"
      },
      "source": [
        "# Lesion code dictionary\n",
        "lesion_code_dict = pd.Series(skin_df.dx.values,index=skin_df.cell_type_idx).to_dict()\n",
        "\n",
        "lesion_code_dict"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{0: 'akiec', 1: 'bcc', 2: 'bkl', 3: 'df', 4: 'nv', 5: 'mel', 6: 'vasc'}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 589
        },
        "id": "hqAIozEF5KCz",
        "scrolled": true,
        "outputId": "0c86ac15-709f-48ab-a95d-606002a579b6"
      },
      "source": [
        "skin_df"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>lesion_id</th>\n",
              "      <th>image_id</th>\n",
              "      <th>dx</th>\n",
              "      <th>dx_type</th>\n",
              "      <th>age</th>\n",
              "      <th>sex</th>\n",
              "      <th>localization</th>\n",
              "      <th>dataset</th>\n",
              "      <th>path</th>\n",
              "      <th>cell_type</th>\n",
              "      <th>cell_type_idx</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>HAM_0000118</td>\n",
              "      <td>ISIC_0027419</td>\n",
              "      <td>bkl</td>\n",
              "      <td>histo</td>\n",
              "      <td>80.0</td>\n",
              "      <td>male</td>\n",
              "      <td>scalp</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Benign keratosis-like lesions</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>HAM_0000118</td>\n",
              "      <td>ISIC_0025030</td>\n",
              "      <td>bkl</td>\n",
              "      <td>histo</td>\n",
              "      <td>80.0</td>\n",
              "      <td>male</td>\n",
              "      <td>scalp</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Benign keratosis-like lesions</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>HAM_0002730</td>\n",
              "      <td>ISIC_0026769</td>\n",
              "      <td>bkl</td>\n",
              "      <td>histo</td>\n",
              "      <td>80.0</td>\n",
              "      <td>male</td>\n",
              "      <td>scalp</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Benign keratosis-like lesions</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>HAM_0002730</td>\n",
              "      <td>ISIC_0025661</td>\n",
              "      <td>bkl</td>\n",
              "      <td>histo</td>\n",
              "      <td>80.0</td>\n",
              "      <td>male</td>\n",
              "      <td>scalp</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Benign keratosis-like lesions</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>HAM_0001466</td>\n",
              "      <td>ISIC_0031633</td>\n",
              "      <td>bkl</td>\n",
              "      <td>histo</td>\n",
              "      <td>75.0</td>\n",
              "      <td>male</td>\n",
              "      <td>ear</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Benign keratosis-like lesions</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10010</th>\n",
              "      <td>HAM_0002867</td>\n",
              "      <td>ISIC_0033084</td>\n",
              "      <td>akiec</td>\n",
              "      <td>histo</td>\n",
              "      <td>40.0</td>\n",
              "      <td>male</td>\n",
              "      <td>abdomen</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Actinic keratoses</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10011</th>\n",
              "      <td>HAM_0002867</td>\n",
              "      <td>ISIC_0033550</td>\n",
              "      <td>akiec</td>\n",
              "      <td>histo</td>\n",
              "      <td>40.0</td>\n",
              "      <td>male</td>\n",
              "      <td>abdomen</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Actinic keratoses</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10012</th>\n",
              "      <td>HAM_0002867</td>\n",
              "      <td>ISIC_0033536</td>\n",
              "      <td>akiec</td>\n",
              "      <td>histo</td>\n",
              "      <td>40.0</td>\n",
              "      <td>male</td>\n",
              "      <td>abdomen</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Actinic keratoses</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10013</th>\n",
              "      <td>HAM_0000239</td>\n",
              "      <td>ISIC_0032854</td>\n",
              "      <td>akiec</td>\n",
              "      <td>histo</td>\n",
              "      <td>80.0</td>\n",
              "      <td>male</td>\n",
              "      <td>face</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Actinic keratoses</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10014</th>\n",
              "      <td>HAM_0003521</td>\n",
              "      <td>ISIC_0032258</td>\n",
              "      <td>mel</td>\n",
              "      <td>histo</td>\n",
              "      <td>70.0</td>\n",
              "      <td>female</td>\n",
              "      <td>back</td>\n",
              "      <td>vidir_modern</td>\n",
              "      <td>/content/drive/MyDrive/Colab Notebooks/images/...</td>\n",
              "      <td>Melanoma</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>10015 rows × 11 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "         lesion_id      image_id  ...                       cell_type cell_type_idx\n",
              "0      HAM_0000118  ISIC_0027419  ...  Benign keratosis-like lesions              2\n",
              "1      HAM_0000118  ISIC_0025030  ...  Benign keratosis-like lesions              2\n",
              "2      HAM_0002730  ISIC_0026769  ...  Benign keratosis-like lesions              2\n",
              "3      HAM_0002730  ISIC_0025661  ...  Benign keratosis-like lesions              2\n",
              "4      HAM_0001466  ISIC_0031633  ...  Benign keratosis-like lesions              2\n",
              "...            ...           ...  ...                             ...           ...\n",
              "10010  HAM_0002867  ISIC_0033084  ...               Actinic keratoses             0\n",
              "10011  HAM_0002867  ISIC_0033550  ...               Actinic keratoses             0\n",
              "10012  HAM_0002867  ISIC_0033536  ...               Actinic keratoses             0\n",
              "10013  HAM_0000239  ISIC_0032854  ...               Actinic keratoses             0\n",
              "10014  HAM_0003521  ISIC_0032258  ...                        Melanoma             5\n",
              "\n",
              "[10015 rows x 11 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8pFORH3joGg2",
        "outputId": "5d0bc67b-eff2-40b3-98f3-a2eb03b8f155"
      },
      "source": [
        "# Attach images to the skin_df Dataframe\n",
        "temp_arr, count = [], 0\n",
        "for path in tqdm(skin_df['path']):\n",
        "  temp_arr.append(np.asarray( Image.open(path).resize((100,75)) ))\n",
        "  count += 1\n",
        "  # print(f\"Processed images: {count}\")\n",
        "  # if count % 10 == 0: clear_output()\n",
        "\n",
        "skin_df['image'] = temp_arr\n",
        "\n",
        "\n",
        "# skin_df['image'] = skin_df['path'].map(lambda x: np.asarray(Image.open(x).resize((100,75))))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 10015/10015 [02:19<00:00, 71.85it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "raUm2J0PgXYL"
      },
      "source": [
        "# Separate the dataframe into Features and Targets data\n",
        "features = skin_df.drop(columns=['cell_type_idx','dx'],axis=1)\n",
        "target = skin_df['cell_type_idx']"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yx5DqaAK5KCz",
        "outputId": "2c195c86-1a20-4245-d5c9-a91cb08e9410"
      },
      "source": [
        "# Create First Train and Test sets\n",
        "x_train_o, x_test_o, y_train_o, y_test_o = train_test_split(features, target, test_size=0.20, random_state=1)\n",
        "print(x_train_o.shape)\n",
        "print(y_train_o.shape)"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(8012, 10)\n",
            "(8012,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sbX83-Yp5KCz",
        "outputId": "0e25dd80-98fb-49d7-c738-977594fdb1e0"
      },
      "source": [
        "# Create image arrays from the train and test sets\n",
        "x_train = np.asarray(x_train_o['image'].tolist())\n",
        "x_test = np.asarray(x_test_o['image'].tolist())\n",
        "\n",
        "print(x_train.shape)\n",
        "print(x_test.shape)"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(8012, 75, 100, 3)\n",
            "(2003, 75, 100, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UUJSBhG-5KC0",
        "scrolled": true,
        "outputId": "8517665a-fc66-4a07-a1e5-e83ee7ce2ffe"
      },
      "source": [
        "# Class distribution before data augmentation\n",
        "train_value_counts = y_train_o.value_counts()\n",
        "print(train_value_counts)\n",
        "print(f'Total images: {train_value_counts.sum()}')"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "4    5378\n",
            "5     891\n",
            "2     871\n",
            "1     418\n",
            "0     266\n",
            "6     110\n",
            "3      78\n",
            "Name: cell_type_idx, dtype: int64\n",
            "Total images: 8012\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zFaa1RBfgE_B"
      },
      "source": [
        "# Data Augmentation\n",
        "# Image data generator is used to generate the possible versions of data as given below.\n",
        "\n",
        "dataaugment = ImageDataGenerator(\n",
        "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
        "        samplewise_center=False,  # set each sample mean to 0\n",
        "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
        "        samplewise_std_normalization=False,  # divide each input by its std\n",
        "        zca_whitening=False,  # apply ZCA whitening\n",
        "        rotation_range=90,  # randomly rotate images in the range (degrees, 0 to 180)\n",
        "        zoom_range = 0.1, # Randomly zoom image \n",
        "        width_shift_range=0.1,  # randomly shift images horizontally (fraction of total width)\n",
        "        height_shift_range=0.1,  # randomly shift images vertically (fraction of total height)\n",
        "        horizontal_flip=True,  # randomly flip images\n",
        "        vertical_flip=True,  # randomly flip images\n",
        "        shear_range = 10\n",
        "    ) "
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1xOuqdLf5KC0",
        "outputId": "dee1a057-1b87-4c81-c5a3-e6176fa71c60"
      },
      "source": [
        "# For each lesion code != 4, get the indices in y_train_o.\n",
        "# For the same indices in x_train_o, augment the image present in the row.\n",
        "reset_y_train_o = y_train_o.reset_index(drop=True)\n",
        "reset_index = reset_y_train_o.index\n",
        "\n",
        "print(len(y_train_o))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8012\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VvfAsDIGhnEx"
      },
      "source": [
        "# The normalisation is done using the training set Mean and Std. Deviation as reference\n",
        "x_train_mean = np.mean(x_train)\n",
        "x_train_std = np.std(x_train)\n",
        "\n",
        "x_train = (x_train - x_train_mean) / x_train_std\n",
        "x_test = (x_test - x_train_mean) / x_train_std"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UQ0EFbN5joCo"
      },
      "source": [
        "# Perform one-hot encoding on the labels\n",
        "y_train = to_categorical(y_train_o, num_classes = 7)\n",
        "y_test = to_categorical(y_test_o, num_classes = 7)"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vRrRWy_Njsnp"
      },
      "source": [
        "# Splitting training into Train and Validatation sets\n",
        "x_train, x_validate, y_train, y_validate = train_test_split(x_train, y_train, test_size = 0.1, random_state=1)\n",
        "\n",
        "# Reshaping the Images into 3 channels (RGB)\n",
        "# height = 75, width = 100, canal = 3\n",
        "x_train = x_train.reshape(x_train.shape[0], *(75, 100, 3))\n",
        "x_test = x_test.reshape(x_test.shape[0], *(75, 100, 3))\n",
        "x_validate = x_validate.reshape(x_validate.shape[0], *(75, 100, 3))"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Z9XLCqXO5KC1",
        "outputId": "759144a6-ff89-4ee7-fc23-af9fbc49b50c"
      },
      "source": [
        "from keras.layers import Concatenate, AveragePooling2D\n",
        "\n",
        "input_shape = (75, 100, 3)\n",
        "num_classes = 7\n",
        "\n",
        "# GoogleNet\n",
        "\n",
        "def inception(x, filters):\n",
        "    # 1x1\n",
        "    path1 = Conv2D(filters=filters[0], kernel_size=(1,1), strides=1, padding='same', activation='relu')(x)\n",
        "\n",
        "    # 1x1->3x3\n",
        "    path2 = Conv2D(filters=filters[1][0], kernel_size=(1,1), strides=1, padding='same', activation='relu')(x)\n",
        "    path2 = Conv2D(filters=filters[1][1], kernel_size=(3,3), strides=1, padding='same', activation='relu')(path2)\n",
        "    \n",
        "    # 1x1->5x5\n",
        "    path3 = Conv2D(filters=filters[2][0], kernel_size=(1,1), strides=1, padding='same', activation='relu')(x)\n",
        "    path3 = Conv2D(filters=filters[2][1], kernel_size=(5,5), strides=1, padding='same', activation='relu')(path3)\n",
        "\n",
        "    # 3x3->1x1\n",
        "    path4 = MaxPooling2D(pool_size=(3,3), strides=1, padding='same')(x)\n",
        "    path4 = Conv2D(filters=filters[3], kernel_size=(1,1), strides=1, padding='same', activation='relu')(path4)\n",
        "\n",
        "    return Concatenate(axis=-1)([path1,path2,path3,path4])\n",
        "\n",
        "\n",
        "def auxiliary(x, name=None):\n",
        "    layer = AveragePooling2D(pool_size=(5,5), strides=3, padding='valid')(x)\n",
        "    layer = Conv2D(filters=128, kernel_size=(1,1), strides=1, padding='same', activation='relu')(layer)\n",
        "    layer = Flatten()(layer)\n",
        "    layer = Dense(units=256, activation='relu')(layer)\n",
        "    layer = Dropout(0.4)(layer)\n",
        "    layer = Dense(units=num_classes, activation='softmax', name=name)(layer)\n",
        "    return layer\n",
        "\n",
        "\n",
        "def googlenet():\n",
        "    layer_in = Input(shape=input_shape)\n",
        "    \n",
        "    # stage-1\n",
        "    layer = Conv2D(filters=64, kernel_size=(7,7), strides=2, padding='same', activation='relu')(layer_in)\n",
        "    layer = MaxPooling2D(pool_size=(3,3), strides=2, padding='same')(layer)\n",
        "    layer = BatchNormalization()(layer)\n",
        "\n",
        "    # stage-2\n",
        "    layer = Conv2D(filters=64, kernel_size=(1,1), strides=1, padding='same', activation='relu')(layer)\n",
        "    layer = Conv2D(filters=192, kernel_size=(3,3), strides=1, padding='same', activation='relu')(layer)\n",
        "    layer = BatchNormalization()(layer)\n",
        "    layer = MaxPooling2D(pool_size=(3,3), strides=2, padding='same')(layer)\n",
        "\n",
        "    # stage-3\n",
        "    layer = inception(layer, [ 64,  (96,128), (16,32), 32]) #3a\n",
        "    layer = inception(layer, [128, (128,192), (32,96), 64]) #3b\n",
        "    layer = MaxPooling2D(pool_size=(3,3), strides=2, padding='same')(layer)\n",
        "    \n",
        "    # stage-4\n",
        "    layer = inception(layer, [192,  (96,208),  (16,48),  64]) #4a\n",
        "    aux1  = auxiliary(layer, name='aux1')\n",
        "    layer = inception(layer, [160, (112,224),  (24,64),  64]) #4b\n",
        "    layer = inception(layer, [128, (128,256),  (24,64),  64]) #4c\n",
        "    layer = inception(layer, [112, (144,288),  (32,64),  64]) #4d\n",
        "    aux2  = auxiliary(layer, name='aux2')\n",
        "    layer = inception(layer, [256, (160,320), (32,128), 128]) #4e\n",
        "    layer = MaxPooling2D(pool_size=(3,3), strides=2, padding='same')(layer)\n",
        "    \n",
        "    # stage-5\n",
        "    layer = inception(layer, [256, (160,320), (32,128), 128]) #5a\n",
        "    layer = inception(layer, [384, (192,384), (48,128), 128]) #5b\n",
        "    layer = AveragePooling2D(pool_size=(7,7), strides=1, padding='same')(layer)\n",
        "    \n",
        "    # stage-6\n",
        "    layer = Flatten()(layer)\n",
        "    layer = Dropout(0.4)(layer)\n",
        "    layer = Dense(units=256, activation='linear')(layer)\n",
        "    main = Dense(units=num_classes, activation='softmax', name='main')(layer)\n",
        "    \n",
        "    model = Model(inputs=layer_in, outputs=[main, aux1, aux2])\n",
        "    \n",
        "    return model\n",
        "\n",
        "\n",
        "# train model\n",
        "model = googlenet()\n",
        "model.summary()\n",
        "'''\n",
        "optimizer = ['Adam', 'SGD', 'Adam', 'SGD']\n",
        "epochs = [20, 30, 20, 30]\n",
        "history_all = {}\n",
        "\n",
        "for i in range(len(optimizer)):\n",
        "    print('Usnig optimizer: ' + optimizer[i] + ', Epoch: ' + str(epochs[i]))\n",
        "    \n",
        "    model.compile(loss='categorical_crossentropy', \n",
        "                  loss_weights={'main': 1.0, 'aux1': 0.3, 'aux2': 0.3},\n",
        "                  optimizer=optimizer[i], metrics=['accuracy'])\n",
        "    \n",
        "    train_history = model.fit_generator(\n",
        "            train_generator,\n",
        "            steps_per_epoch=EPOCH_STEPS,\n",
        "            epochs=epochs[i],\n",
        "            #callbacks=[checkpoint]\n",
        "            shuffle=True\n",
        "            )\n",
        "    \n",
        "    # save history    \n",
        "    if len(history_all) == 0:\n",
        "        history_all = {key: [] for key in train_history.history}\n",
        "    \n",
        "    for key in history_all:\n",
        "        history_all[key].extend(train_history.history[key])\n",
        "\n",
        "model.save(MODEL_NAME)\n",
        "'''\n",
        "# show train history\n",
        "'''\n",
        "def show_train_history(history, xlabel, ylabel, train):\n",
        "    for item in train:\n",
        "        plt.plot(history[item])\n",
        "    plt.title('Train History')\n",
        "    plt.xlabel(xlabel)\n",
        "    plt.ylabel(ylabel)\n",
        "    plt.legend(train, loc='upper left')\n",
        "    plt.show()\n",
        "\n",
        "show_train_history(history_all, 'Epoch', 'Accuracy', ('main_acc', 'aux1_acc', 'aux2_acc'))\n",
        "show_train_history(history_all, 'Epoch', 'Loss', ('main_loss', 'aux1_loss', 'aux2_loss'))\n",
        "'''"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 75, 100, 3)] 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 38, 50, 64)   9472        input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D)    (None, 19, 25, 64)   0           conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization (BatchNorma (None, 19, 25, 64)   256         max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 19, 25, 64)   4160        batch_normalization[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 19, 25, 192)  110784      conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_1 (BatchNor (None, 19, 25, 192)  768         conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2D)  (None, 10, 13, 192)  0           batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_4 (Conv2D)               (None, 10, 13, 96)   18528       max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_6 (Conv2D)               (None, 10, 13, 16)   3088        max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2D)  (None, 10, 13, 192)  0           max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 10, 13, 64)   12352       max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_5 (Conv2D)               (None, 10, 13, 128)  110720      conv2d_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 10, 13, 32)   12832       conv2d_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 10, 13, 32)   6176        max_pooling2d_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate (Concatenate)       (None, 10, 13, 256)  0           conv2d_3[0][0]                   \n",
            "                                                                 conv2d_5[0][0]                   \n",
            "                                                                 conv2d_7[0][0]                   \n",
            "                                                                 conv2d_8[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 10, 13, 128)  32896       concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 10, 13, 32)   8224        concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2D)  (None, 10, 13, 256)  0           concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 10, 13, 128)  32896       concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 10, 13, 192)  221376      conv2d_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_13 (Conv2D)              (None, 10, 13, 96)   76896       conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_14 (Conv2D)              (None, 10, 13, 64)   16448       max_pooling2d_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_1 (Concatenate)     (None, 10, 13, 480)  0           conv2d_9[0][0]                   \n",
            "                                                                 conv2d_11[0][0]                  \n",
            "                                                                 conv2d_13[0][0]                  \n",
            "                                                                 conv2d_14[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2D)  (None, 5, 7, 480)    0           concatenate_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_16 (Conv2D)              (None, 5, 7, 96)     46176       max_pooling2d_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_18 (Conv2D)              (None, 5, 7, 16)     7696        max_pooling2d_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_5 (MaxPooling2D)  (None, 5, 7, 480)    0           max_pooling2d_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_15 (Conv2D)              (None, 5, 7, 192)    92352       max_pooling2d_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_17 (Conv2D)              (None, 5, 7, 208)    179920      conv2d_16[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_19 (Conv2D)              (None, 5, 7, 48)     19248       conv2d_18[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_20 (Conv2D)              (None, 5, 7, 64)     30784       max_pooling2d_5[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_2 (Concatenate)     (None, 5, 7, 512)    0           conv2d_15[0][0]                  \n",
            "                                                                 conv2d_17[0][0]                  \n",
            "                                                                 conv2d_19[0][0]                  \n",
            "                                                                 conv2d_20[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_23 (Conv2D)              (None, 5, 7, 112)    57456       concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_25 (Conv2D)              (None, 5, 7, 24)     12312       concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_6 (MaxPooling2D)  (None, 5, 7, 512)    0           concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_22 (Conv2D)              (None, 5, 7, 160)    82080       concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_24 (Conv2D)              (None, 5, 7, 224)    226016      conv2d_23[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_26 (Conv2D)              (None, 5, 7, 64)     38464       conv2d_25[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_27 (Conv2D)              (None, 5, 7, 64)     32832       max_pooling2d_6[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_3 (Concatenate)     (None, 5, 7, 512)    0           conv2d_22[0][0]                  \n",
            "                                                                 conv2d_24[0][0]                  \n",
            "                                                                 conv2d_26[0][0]                  \n",
            "                                                                 conv2d_27[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_29 (Conv2D)              (None, 5, 7, 128)    65664       concatenate_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_31 (Conv2D)              (None, 5, 7, 24)     12312       concatenate_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_7 (MaxPooling2D)  (None, 5, 7, 512)    0           concatenate_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_28 (Conv2D)              (None, 5, 7, 128)    65664       concatenate_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_30 (Conv2D)              (None, 5, 7, 256)    295168      conv2d_29[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_32 (Conv2D)              (None, 5, 7, 64)     38464       conv2d_31[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_33 (Conv2D)              (None, 5, 7, 64)     32832       max_pooling2d_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_4 (Concatenate)     (None, 5, 7, 512)    0           conv2d_28[0][0]                  \n",
            "                                                                 conv2d_30[0][0]                  \n",
            "                                                                 conv2d_32[0][0]                  \n",
            "                                                                 conv2d_33[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_35 (Conv2D)              (None, 5, 7, 144)    73872       concatenate_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_37 (Conv2D)              (None, 5, 7, 32)     16416       concatenate_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_8 (MaxPooling2D)  (None, 5, 7, 512)    0           concatenate_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_34 (Conv2D)              (None, 5, 7, 112)    57456       concatenate_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_36 (Conv2D)              (None, 5, 7, 288)    373536      conv2d_35[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_38 (Conv2D)              (None, 5, 7, 64)     51264       conv2d_37[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_39 (Conv2D)              (None, 5, 7, 64)     32832       max_pooling2d_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_5 (Concatenate)     (None, 5, 7, 528)    0           conv2d_34[0][0]                  \n",
            "                                                                 conv2d_36[0][0]                  \n",
            "                                                                 conv2d_38[0][0]                  \n",
            "                                                                 conv2d_39[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_42 (Conv2D)              (None, 5, 7, 160)    84640       concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_44 (Conv2D)              (None, 5, 7, 32)     16928       concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_9 (MaxPooling2D)  (None, 5, 7, 528)    0           concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_41 (Conv2D)              (None, 5, 7, 256)    135424      concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_43 (Conv2D)              (None, 5, 7, 320)    461120      conv2d_42[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_45 (Conv2D)              (None, 5, 7, 128)    102528      conv2d_44[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_46 (Conv2D)              (None, 5, 7, 128)    67712       max_pooling2d_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_6 (Concatenate)     (None, 5, 7, 832)    0           conv2d_41[0][0]                  \n",
            "                                                                 conv2d_43[0][0]                  \n",
            "                                                                 conv2d_45[0][0]                  \n",
            "                                                                 conv2d_46[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_10 (MaxPooling2D) (None, 3, 4, 832)    0           concatenate_6[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_48 (Conv2D)              (None, 3, 4, 160)    133280      max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_50 (Conv2D)              (None, 3, 4, 32)     26656       max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_11 (MaxPooling2D) (None, 3, 4, 832)    0           max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_47 (Conv2D)              (None, 3, 4, 256)    213248      max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_49 (Conv2D)              (None, 3, 4, 320)    461120      conv2d_48[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_51 (Conv2D)              (None, 3, 4, 128)    102528      conv2d_50[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_52 (Conv2D)              (None, 3, 4, 128)    106624      max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_7 (Concatenate)     (None, 3, 4, 832)    0           conv2d_47[0][0]                  \n",
            "                                                                 conv2d_49[0][0]                  \n",
            "                                                                 conv2d_51[0][0]                  \n",
            "                                                                 conv2d_52[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_54 (Conv2D)              (None, 3, 4, 192)    159936      concatenate_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_56 (Conv2D)              (None, 3, 4, 48)     39984       concatenate_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_12 (MaxPooling2D) (None, 3, 4, 832)    0           concatenate_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_53 (Conv2D)              (None, 3, 4, 384)    319872      concatenate_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_55 (Conv2D)              (None, 3, 4, 384)    663936      conv2d_54[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_57 (Conv2D)              (None, 3, 4, 128)    153728      conv2d_56[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_58 (Conv2D)              (None, 3, 4, 128)    106624      max_pooling2d_12[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_8 (Concatenate)     (None, 3, 4, 1024)   0           conv2d_53[0][0]                  \n",
            "                                                                 conv2d_55[0][0]                  \n",
            "                                                                 conv2d_57[0][0]                  \n",
            "                                                                 conv2d_58[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d (AveragePooli (None, 1, 1, 512)    0           concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_1 (AveragePoo (None, 1, 1, 528)    0           concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_2 (AveragePoo (None, 3, 4, 1024)   0           concatenate_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_21 (Conv2D)              (None, 1, 1, 128)    65664       average_pooling2d[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_40 (Conv2D)              (None, 1, 1, 128)    67712       average_pooling2d_1[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "flatten_2 (Flatten)             (None, 12288)        0           average_pooling2d_2[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 128)          0           conv2d_21[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "flatten_1 (Flatten)             (None, 128)          0           conv2d_40[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dropout_2 (Dropout)             (None, 12288)        0           flatten_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 256)          33024       flatten[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 256)          33024       flatten_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 256)          3145984     dropout_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dropout (Dropout)               (None, 256)          0           dense[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "dropout_1 (Dropout)             (None, 256)          0           dense_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "main (Dense)                    (None, 7)            1799        dense_2[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "aux1 (Dense)                    (None, 7)            1799        dropout[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "aux2 (Dense)                    (None, 7)            1799        dropout_1[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 9,325,381\n",
            "Trainable params: 9,324,869\n",
            "Non-trainable params: 512\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "\"\\ndef show_train_history(history, xlabel, ylabel, train):\\n    for item in train:\\n        plt.plot(history[item])\\n    plt.title('Train History')\\n    plt.xlabel(xlabel)\\n    plt.ylabel(ylabel)\\n    plt.legend(train, loc='upper left')\\n    plt.show()\\n\\nshow_train_history(history_all, 'Epoch', 'Accuracy', ('main_acc', 'aux1_acc', 'aux2_acc'))\\nshow_train_history(history_all, 'Epoch', 'Loss', ('main_loss', 'aux1_loss', 'aux2_loss'))\\n\""
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0oFIjDtr5KC2"
      },
      "source": [
        "# Define metrics \n",
        "\n",
        "def recall_m(y_true, y_pred):\n",
        "    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
        "    recall = true_positives / (possible_positives + K.epsilon())\n",
        "    return recall\n",
        "\n",
        "def precision_m(y_true, y_pred):\n",
        "    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "    predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
        "    precision = true_positives / (predicted_positives + K.epsilon())\n",
        "    return precision\n",
        "\n",
        "def f1_m(y_true, y_pred):\n",
        "    precision = precision_m(y_true, y_pred)\n",
        "    recall = recall_m(y_true, y_pred)\n",
        "    return 2*((precision*recall)/(precision+recall+K.epsilon()))"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t2QceAAB5KC2"
      },
      "source": [
        "# New baseline\n",
        "# Define the optimizer\n",
        "optimizer = Adam(learning_rate=0.001, \n",
        "                 beta_1=0.9, \n",
        "                 beta_2=0.999, \n",
        "                 # epsilon=None, \n",
        "                 decay=0.0, \n",
        "                 amsgrad=False)\n",
        "\n",
        "# Compile the model\n",
        "model.compile(optimizer = optimizer , loss = \"categorical_crossentropy\", metrics=[\"accuracy\", f1_m,precision_m, recall_m])\n",
        "\n",
        "# Set a learning rate annealer\n",
        "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', \n",
        "                                            patience=3, \n",
        "                                            verbose=1, \n",
        "                                            factor=0.5, \n",
        "                                            min_lr=0.00001)"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5WSfybEBiz-W"
      },
      "source": [
        "x_train = np.asarray(x_train)\n",
        "y_train = np.asarray(y_train)\n",
        "x_validate = np.asarray(x_validate)\n",
        "y_validate = np.asarray(y_validate)"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VTGRcxtb5KC2",
        "outputId": "f3ff0ab6-5429-4fcd-bd8c-b7c80f3bb499"
      },
      "source": [
        "# New baseline\n",
        "# Fit the model\n",
        "epochs = 25\n",
        "batch_size = 10\n",
        "history = model.fit_generator(dataaugment.flow(x_train,y_train, batch_size=batch_size),\n",
        "                              epochs = epochs, validation_data = (x_validate,y_validate),\n",
        "                              verbose = 1, steps_per_epoch=x_train.shape[0] // batch_size\n",
        "                              , callbacks=[learning_rate_reduction])\n",
        "\n",
        "# history = model.fit(x=x_train, y=y_train,\n",
        "#                               epochs = epochs, validation_data = (x_validate,y_validate),\n",
        "#                               verbose = 1, steps_per_epoch=x_train.shape[0] // batch_size\n",
        "#                               , callbacks=[learning_rate_reduction])"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:1844: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
            "  warnings.warn('`Model.fit_generator` is deprecated and '\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/25\n",
            "721/721 [==============================] - 356s 487ms/step - loss: 3.6134 - main_loss: 1.4470 - aux1_loss: 1.0677 - aux2_loss: 1.0987 - main_accuracy: 0.6465 - main_f1_m: 0.5841 - main_precision_m: 0.7218 - main_recall_m: 0.5196 - aux1_accuracy: 0.6555 - aux1_f1_m: 0.6048 - aux1_precision_m: 0.7715 - aux1_recall_m: 0.5140 - aux2_accuracy: 0.6657 - aux2_f1_m: 0.5829 - aux2_precision_m: 0.7428 - aux2_recall_m: 0.5025 - val_loss: 2.7094 - val_main_loss: 0.9215 - val_aux1_loss: 0.8808 - val_aux2_loss: 0.9072 - val_main_accuracy: 0.6808 - val_main_f1_m: 0.6994 - val_main_precision_m: 0.8205 - val_main_recall_m: 0.6142 - val_aux1_accuracy: 0.6958 - val_aux1_f1_m: 0.6782 - val_aux1_precision_m: 0.8749 - val_aux1_recall_m: 0.5577 - val_aux2_accuracy: 0.6808 - val_aux2_f1_m: 0.6989 - val_aux2_precision_m: 0.8214 - val_aux2_recall_m: 0.6130\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 2/25\n",
            "721/721 [==============================] - 353s 490ms/step - loss: 2.7720 - main_loss: 0.9445 - aux1_loss: 0.9006 - aux2_loss: 0.9270 - main_accuracy: 0.6773 - main_f1_m: 0.6705 - main_precision_m: 0.8009 - main_recall_m: 0.5899 - aux1_accuracy: 0.6879 - aux1_f1_m: 0.6638 - aux1_precision_m: 0.8072 - aux1_recall_m: 0.5752 - aux2_accuracy: 0.6834 - aux2_f1_m: 0.6696 - aux2_precision_m: 0.8059 - aux2_recall_m: 0.5836 - val_loss: 2.6124 - val_main_loss: 0.9153 - val_aux1_loss: 0.8248 - val_aux2_loss: 0.8723 - val_main_accuracy: 0.6833 - val_main_f1_m: 0.6896 - val_main_precision_m: 0.8008 - val_main_recall_m: 0.6094 - val_aux1_accuracy: 0.6970 - val_aux1_f1_m: 0.6838 - val_aux1_precision_m: 0.8336 - val_aux1_recall_m: 0.5841 - val_aux2_accuracy: 0.6908 - val_aux2_f1_m: 0.6852 - val_aux2_precision_m: 0.8208 - val_aux2_recall_m: 0.5925\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 3/25\n",
            "721/721 [==============================] - 351s 487ms/step - loss: 2.6296 - main_loss: 0.9072 - aux1_loss: 0.8492 - aux2_loss: 0.8732 - main_accuracy: 0.6861 - main_f1_m: 0.6844 - main_precision_m: 0.8071 - main_recall_m: 0.6027 - aux1_accuracy: 0.6914 - aux1_f1_m: 0.6824 - aux1_precision_m: 0.8210 - aux1_recall_m: 0.5938 - aux2_accuracy: 0.6898 - aux2_f1_m: 0.6842 - aux2_precision_m: 0.8116 - aux2_recall_m: 0.5999 - val_loss: 2.4511 - val_main_loss: 0.8575 - val_aux1_loss: 0.7770 - val_aux2_loss: 0.8166 - val_main_accuracy: 0.6833 - val_main_f1_m: 0.7106 - val_main_precision_m: 0.8175 - val_main_recall_m: 0.6334 - val_aux1_accuracy: 0.7282 - val_aux1_f1_m: 0.6956 - val_aux1_precision_m: 0.8557 - val_aux1_recall_m: 0.5901 - val_aux2_accuracy: 0.7132 - val_aux2_f1_m: 0.7140 - val_aux2_precision_m: 0.8410 - val_aux2_recall_m: 0.6250\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 4/25\n",
            "721/721 [==============================] - 351s 486ms/step - loss: 2.5861 - main_loss: 0.8890 - aux1_loss: 0.8405 - aux2_loss: 0.8566 - main_accuracy: 0.6849 - main_f1_m: 0.6806 - main_precision_m: 0.8155 - main_recall_m: 0.5947 - aux1_accuracy: 0.6980 - aux1_f1_m: 0.6793 - aux1_precision_m: 0.8173 - aux1_recall_m: 0.5913 - aux2_accuracy: 0.6943 - aux2_f1_m: 0.6841 - aux2_precision_m: 0.8201 - aux2_recall_m: 0.5965 - val_loss: 2.2705 - val_main_loss: 0.7785 - val_aux1_loss: 0.7305 - val_aux2_loss: 0.7616 - val_main_accuracy: 0.7269 - val_main_f1_m: 0.7094 - val_main_precision_m: 0.8645 - val_main_recall_m: 0.6058 - val_aux1_accuracy: 0.7382 - val_aux1_f1_m: 0.7168 - val_aux1_precision_m: 0.8667 - val_aux1_recall_m: 0.6154 - val_aux2_accuracy: 0.7244 - val_aux2_f1_m: 0.7125 - val_aux2_precision_m: 0.8644 - val_aux2_recall_m: 0.6106\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 5/25\n",
            "721/721 [==============================] - 352s 488ms/step - loss: 2.5027 - main_loss: 0.8602 - aux1_loss: 0.8079 - aux2_loss: 0.8346 - main_accuracy: 0.6926 - main_f1_m: 0.6789 - main_precision_m: 0.8163 - main_recall_m: 0.5922 - aux1_accuracy: 0.7008 - aux1_f1_m: 0.6920 - aux1_precision_m: 0.8237 - aux1_recall_m: 0.6071 - aux2_accuracy: 0.6964 - aux2_f1_m: 0.6857 - aux2_precision_m: 0.8258 - aux2_recall_m: 0.5963 - val_loss: 2.5095 - val_main_loss: 0.8890 - val_aux1_loss: 0.7972 - val_aux2_loss: 0.8233 - val_main_accuracy: 0.6696 - val_main_f1_m: 0.6287 - val_main_precision_m: 0.9182 - val_main_recall_m: 0.4820 - val_aux1_accuracy: 0.7195 - val_aux1_f1_m: 0.6821 - val_aux1_precision_m: 0.8604 - val_aux1_recall_m: 0.5673 - val_aux2_accuracy: 0.6908 - val_aux2_f1_m: 0.6672 - val_aux2_precision_m: 0.9006 - val_aux2_recall_m: 0.5325\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 6/25\n",
            "721/721 [==============================] - 351s 487ms/step - loss: 2.4296 - main_loss: 0.8329 - aux1_loss: 0.7877 - aux2_loss: 0.8091 - main_accuracy: 0.7035 - main_f1_m: 0.6893 - main_precision_m: 0.8204 - main_recall_m: 0.6057 - aux1_accuracy: 0.7117 - aux1_f1_m: 0.6997 - aux1_precision_m: 0.8224 - aux1_recall_m: 0.6194 - aux2_accuracy: 0.7108 - aux2_f1_m: 0.6983 - aux2_precision_m: 0.8303 - aux2_recall_m: 0.6145 - val_loss: 2.2446 - val_main_loss: 0.7848 - val_aux1_loss: 0.7114 - val_aux2_loss: 0.7484 - val_main_accuracy: 0.7170 - val_main_f1_m: 0.7204 - val_main_precision_m: 0.8325 - val_main_recall_m: 0.6394 - val_aux1_accuracy: 0.7494 - val_aux1_f1_m: 0.7226 - val_aux1_precision_m: 0.8517 - val_aux1_recall_m: 0.6322 - val_aux2_accuracy: 0.7269 - val_aux2_f1_m: 0.7271 - val_aux2_precision_m: 0.8504 - val_aux2_recall_m: 0.6394\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 7/25\n",
            "721/721 [==============================] - 347s 481ms/step - loss: 2.3089 - main_loss: 0.7906 - aux1_loss: 0.7495 - aux2_loss: 0.7688 - main_accuracy: 0.7131 - main_f1_m: 0.6999 - main_precision_m: 0.8319 - main_recall_m: 0.6147 - aux1_accuracy: 0.7223 - aux1_f1_m: 0.7110 - aux1_precision_m: 0.8337 - aux1_recall_m: 0.6297 - aux2_accuracy: 0.7193 - aux2_f1_m: 0.7095 - aux2_precision_m: 0.8367 - aux2_recall_m: 0.6264 - val_loss: 2.3156 - val_main_loss: 0.7975 - val_aux1_loss: 0.7560 - val_aux2_loss: 0.7621 - val_main_accuracy: 0.7170 - val_main_f1_m: 0.6896 - val_main_precision_m: 0.8510 - val_main_recall_m: 0.5841 - val_aux1_accuracy: 0.7369 - val_aux1_f1_m: 0.6995 - val_aux1_precision_m: 0.8600 - val_aux1_recall_m: 0.5938 - val_aux2_accuracy: 0.7195 - val_aux2_f1_m: 0.6865 - val_aux2_precision_m: 0.8622 - val_aux2_recall_m: 0.5745\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 8/25\n",
            "721/721 [==============================] - 352s 488ms/step - loss: 2.3003 - main_loss: 0.7912 - aux1_loss: 0.7474 - aux2_loss: 0.7617 - main_accuracy: 0.7113 - main_f1_m: 0.7008 - main_precision_m: 0.8273 - main_recall_m: 0.6164 - aux1_accuracy: 0.7292 - aux1_f1_m: 0.7156 - aux1_precision_m: 0.8278 - aux1_recall_m: 0.6369 - aux2_accuracy: 0.7237 - aux2_f1_m: 0.7093 - aux2_precision_m: 0.8334 - aux2_recall_m: 0.6257 - val_loss: 2.5192 - val_main_loss: 0.8829 - val_aux1_loss: 0.8065 - val_aux2_loss: 0.8299 - val_main_accuracy: 0.7107 - val_main_f1_m: 0.7163 - val_main_precision_m: 0.8278 - val_main_recall_m: 0.6358 - val_aux1_accuracy: 0.7182 - val_aux1_f1_m: 0.7178 - val_aux1_precision_m: 0.8229 - val_aux1_recall_m: 0.6406 - val_aux2_accuracy: 0.7219 - val_aux2_f1_m: 0.7247 - val_aux2_precision_m: 0.8313 - val_aux2_recall_m: 0.6466\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 9/25\n",
            "721/721 [==============================] - 351s 486ms/step - loss: 2.3284 - main_loss: 0.7976 - aux1_loss: 0.7558 - aux2_loss: 0.7750 - main_accuracy: 0.7085 - main_f1_m: 0.6970 - main_precision_m: 0.8330 - main_recall_m: 0.6105 - aux1_accuracy: 0.7201 - aux1_f1_m: 0.7086 - aux1_precision_m: 0.8292 - aux1_recall_m: 0.6286 - aux2_accuracy: 0.7186 - aux2_f1_m: 0.7053 - aux2_precision_m: 0.8364 - aux2_recall_m: 0.6197 - val_loss: 2.1297 - val_main_loss: 0.7368 - val_aux1_loss: 0.6834 - val_aux2_loss: 0.7095 - val_main_accuracy: 0.7382 - val_main_f1_m: 0.7324 - val_main_precision_m: 0.8333 - val_main_recall_m: 0.6575 - val_aux1_accuracy: 0.7544 - val_aux1_f1_m: 0.7507 - val_aux1_precision_m: 0.8463 - val_aux1_recall_m: 0.6791 - val_aux2_accuracy: 0.7469 - val_aux2_f1_m: 0.7455 - val_aux2_precision_m: 0.8438 - val_aux2_recall_m: 0.6719\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 10/25\n",
            "721/721 [==============================] - 351s 487ms/step - loss: 2.2867 - main_loss: 0.7790 - aux1_loss: 0.7428 - aux2_loss: 0.7649 - main_accuracy: 0.7088 - main_f1_m: 0.7012 - main_precision_m: 0.8434 - main_recall_m: 0.6105 - aux1_accuracy: 0.7242 - aux1_f1_m: 0.7122 - aux1_precision_m: 0.8299 - aux1_recall_m: 0.6329 - aux2_accuracy: 0.7183 - aux2_f1_m: 0.7034 - aux2_precision_m: 0.8367 - aux2_recall_m: 0.6164 - val_loss: 2.0943 - val_main_loss: 0.7375 - val_aux1_loss: 0.6637 - val_aux2_loss: 0.6931 - val_main_accuracy: 0.7294 - val_main_f1_m: 0.7346 - val_main_precision_m: 0.8290 - val_main_recall_m: 0.6623 - val_aux1_accuracy: 0.7519 - val_aux1_f1_m: 0.7535 - val_aux1_precision_m: 0.8429 - val_aux1_recall_m: 0.6851 - val_aux2_accuracy: 0.7394 - val_aux2_f1_m: 0.7384 - val_aux2_precision_m: 0.8424 - val_aux2_recall_m: 0.6611\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 11/25\n",
            "721/721 [==============================] - 349s 484ms/step - loss: 2.2340 - main_loss: 0.7627 - aux1_loss: 0.7252 - aux2_loss: 0.7461 - main_accuracy: 0.7197 - main_f1_m: 0.7035 - main_precision_m: 0.8312 - main_recall_m: 0.6195 - aux1_accuracy: 0.7270 - aux1_f1_m: 0.7174 - aux1_precision_m: 0.8280 - aux1_recall_m: 0.6410 - aux2_accuracy: 0.7208 - aux2_f1_m: 0.7121 - aux2_precision_m: 0.8381 - aux2_recall_m: 0.6279 - val_loss: 1.9967 - val_main_loss: 0.6823 - val_aux1_loss: 0.6482 - val_aux2_loss: 0.6662 - val_main_accuracy: 0.7444 - val_main_f1_m: 0.7349 - val_main_precision_m: 0.8822 - val_main_recall_m: 0.6334 - val_aux1_accuracy: 0.7581 - val_aux1_f1_m: 0.7488 - val_aux1_precision_m: 0.8822 - val_aux1_recall_m: 0.6538 - val_aux2_accuracy: 0.7544 - val_aux2_f1_m: 0.7487 - val_aux2_precision_m: 0.8960 - val_aux2_recall_m: 0.6466\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 12/25\n",
            "721/721 [==============================] - 350s 486ms/step - loss: 2.2139 - main_loss: 0.7548 - aux1_loss: 0.7237 - aux2_loss: 0.7354 - main_accuracy: 0.7196 - main_f1_m: 0.7108 - main_precision_m: 0.8385 - main_recall_m: 0.6264 - aux1_accuracy: 0.7330 - aux1_f1_m: 0.7251 - aux1_precision_m: 0.8416 - aux1_recall_m: 0.6452 - aux2_accuracy: 0.7309 - aux2_f1_m: 0.7147 - aux2_precision_m: 0.8440 - aux2_recall_m: 0.6287 - val_loss: 2.0260 - val_main_loss: 0.6921 - val_aux1_loss: 0.6632 - val_aux2_loss: 0.6707 - val_main_accuracy: 0.7431 - val_main_f1_m: 0.7255 - val_main_precision_m: 0.8746 - val_main_recall_m: 0.6238 - val_aux1_accuracy: 0.7544 - val_aux1_f1_m: 0.7477 - val_aux1_precision_m: 0.8607 - val_aux1_recall_m: 0.6647 - val_aux2_accuracy: 0.7456 - val_aux2_f1_m: 0.7397 - val_aux2_precision_m: 0.8643 - val_aux2_recall_m: 0.6502\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 13/25\n",
            "721/721 [==============================] - 350s 485ms/step - loss: 2.1377 - main_loss: 0.7276 - aux1_loss: 0.6961 - aux2_loss: 0.7140 - main_accuracy: 0.7303 - main_f1_m: 0.7129 - main_precision_m: 0.8298 - main_recall_m: 0.6325 - aux1_accuracy: 0.7363 - aux1_f1_m: 0.7270 - aux1_precision_m: 0.8299 - aux1_recall_m: 0.6541 - aux2_accuracy: 0.7343 - aux2_f1_m: 0.7245 - aux2_precision_m: 0.8385 - aux2_recall_m: 0.6448 - val_loss: 2.0606 - val_main_loss: 0.7298 - val_aux1_loss: 0.6616 - val_aux2_loss: 0.6692 - val_main_accuracy: 0.7406 - val_main_f1_m: 0.7173 - val_main_precision_m: 0.8252 - val_main_recall_m: 0.6394 - val_aux1_accuracy: 0.7569 - val_aux1_f1_m: 0.7424 - val_aux1_precision_m: 0.8469 - val_aux1_recall_m: 0.6659 - val_aux2_accuracy: 0.7631 - val_aux2_f1_m: 0.7365 - val_aux2_precision_m: 0.8537 - val_aux2_recall_m: 0.6526\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 14/25\n",
            "721/721 [==============================] - 349s 484ms/step - loss: 2.2406 - main_loss: 0.8064 - aux1_loss: 0.7083 - aux2_loss: 0.7259 - main_accuracy: 0.7261 - main_f1_m: 0.7216 - main_precision_m: 0.8321 - main_recall_m: 0.6448 - aux1_accuracy: 0.7424 - aux1_f1_m: 0.7305 - aux1_precision_m: 0.8357 - aux1_recall_m: 0.6559 - aux2_accuracy: 0.7371 - aux2_f1_m: 0.7279 - aux2_precision_m: 0.8366 - aux2_recall_m: 0.6516 - val_loss: 2.1878 - val_main_loss: 0.7505 - val_aux1_loss: 0.7164 - val_aux2_loss: 0.7208 - val_main_accuracy: 0.7257 - val_main_f1_m: 0.7252 - val_main_precision_m: 0.8419 - val_main_recall_m: 0.6406 - val_aux1_accuracy: 0.7506 - val_aux1_f1_m: 0.7418 - val_aux1_precision_m: 0.8505 - val_aux1_recall_m: 0.6623 - val_aux2_accuracy: 0.7406 - val_aux2_f1_m: 0.7256 - val_aux2_precision_m: 0.8452 - val_aux2_recall_m: 0.6394\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 15/25\n",
            "721/721 [==============================] - 353s 490ms/step - loss: 2.2949 - main_loss: 0.7845 - aux1_loss: 0.7504 - aux2_loss: 0.7599 - main_accuracy: 0.7165 - main_f1_m: 0.7070 - main_precision_m: 0.8205 - main_recall_m: 0.6287 - aux1_accuracy: 0.7313 - aux1_f1_m: 0.7148 - aux1_precision_m: 0.8243 - aux1_recall_m: 0.6386 - aux2_accuracy: 0.7308 - aux2_f1_m: 0.7106 - aux2_precision_m: 0.8244 - aux2_recall_m: 0.6321 - val_loss: 1.9536 - val_main_loss: 0.6687 - val_aux1_loss: 0.6353 - val_aux2_loss: 0.6496 - val_main_accuracy: 0.7494 - val_main_f1_m: 0.7283 - val_main_precision_m: 0.8885 - val_main_recall_m: 0.6214 - val_aux1_accuracy: 0.7706 - val_aux1_f1_m: 0.7422 - val_aux1_precision_m: 0.8742 - val_aux1_recall_m: 0.6502 - val_aux2_accuracy: 0.7581 - val_aux2_f1_m: 0.7369 - val_aux2_precision_m: 0.8789 - val_aux2_recall_m: 0.6382\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 16/25\n",
            "721/721 [==============================] - 354s 490ms/step - loss: 2.1051 - main_loss: 0.7228 - aux1_loss: 0.6790 - aux2_loss: 0.7033 - main_accuracy: 0.7324 - main_f1_m: 0.7255 - main_precision_m: 0.8390 - main_recall_m: 0.6487 - aux1_accuracy: 0.7491 - aux1_f1_m: 0.7389 - aux1_precision_m: 0.8402 - aux1_recall_m: 0.6676 - aux2_accuracy: 0.7472 - aux2_f1_m: 0.7280 - aux2_precision_m: 0.8435 - aux2_recall_m: 0.6497 - val_loss: 2.0544 - val_main_loss: 0.7025 - val_aux1_loss: 0.6662 - val_aux2_loss: 0.6857 - val_main_accuracy: 0.7431 - val_main_f1_m: 0.7468 - val_main_precision_m: 0.8402 - val_main_recall_m: 0.6767 - val_aux1_accuracy: 0.7544 - val_aux1_f1_m: 0.7461 - val_aux1_precision_m: 0.8416 - val_aux1_recall_m: 0.6743 - val_aux2_accuracy: 0.7519 - val_aux2_f1_m: 0.7452 - val_aux2_precision_m: 0.8375 - val_aux2_recall_m: 0.6755\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 17/25\n",
            "721/721 [==============================] - 355s 492ms/step - loss: 2.1313 - main_loss: 0.7289 - aux1_loss: 0.6940 - aux2_loss: 0.7084 - main_accuracy: 0.7326 - main_f1_m: 0.7256 - main_precision_m: 0.8362 - main_recall_m: 0.6495 - aux1_accuracy: 0.7427 - aux1_f1_m: 0.7374 - aux1_precision_m: 0.8333 - aux1_recall_m: 0.6693 - aux2_accuracy: 0.7450 - aux2_f1_m: 0.7320 - aux2_precision_m: 0.8423 - aux2_recall_m: 0.6552 - val_loss: 1.9921 - val_main_loss: 0.6877 - val_aux1_loss: 0.6493 - val_aux2_loss: 0.6551 - val_main_accuracy: 0.7444 - val_main_f1_m: 0.7425 - val_main_precision_m: 0.8291 - val_main_recall_m: 0.6767 - val_aux1_accuracy: 0.7718 - val_aux1_f1_m: 0.7489 - val_aux1_precision_m: 0.8360 - val_aux1_recall_m: 0.6839 - val_aux2_accuracy: 0.7643 - val_aux2_f1_m: 0.7522 - val_aux2_precision_m: 0.8471 - val_aux2_recall_m: 0.6815\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 18/25\n",
            "721/721 [==============================] - 354s 491ms/step - loss: 2.0281 - main_loss: 0.6940 - aux1_loss: 0.6582 - aux2_loss: 0.6759 - main_accuracy: 0.7502 - main_f1_m: 0.7408 - main_precision_m: 0.8492 - main_recall_m: 0.6669 - aux1_accuracy: 0.7547 - aux1_f1_m: 0.7462 - aux1_precision_m: 0.8401 - aux1_recall_m: 0.6788 - aux2_accuracy: 0.7503 - aux2_f1_m: 0.7406 - aux2_precision_m: 0.8432 - aux2_recall_m: 0.6686 - val_loss: 1.9267 - val_main_loss: 0.6649 - val_aux1_loss: 0.6189 - val_aux2_loss: 0.6429 - val_main_accuracy: 0.7643 - val_main_f1_m: 0.7524 - val_main_precision_m: 0.8614 - val_main_recall_m: 0.6719 - val_aux1_accuracy: 0.7668 - val_aux1_f1_m: 0.7617 - val_aux1_precision_m: 0.8642 - val_aux1_recall_m: 0.6863 - val_aux2_accuracy: 0.7656 - val_aux2_f1_m: 0.7579 - val_aux2_precision_m: 0.8699 - val_aux2_recall_m: 0.6755\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 19/25\n",
            "721/721 [==============================] - 353s 490ms/step - loss: 2.0081 - main_loss: 0.6852 - aux1_loss: 0.6518 - aux2_loss: 0.6711 - main_accuracy: 0.7570 - main_f1_m: 0.7402 - main_precision_m: 0.8376 - main_recall_m: 0.6711 - aux1_accuracy: 0.7601 - aux1_f1_m: 0.7492 - aux1_precision_m: 0.8416 - aux1_recall_m: 0.6821 - aux2_accuracy: 0.7562 - aux2_f1_m: 0.7420 - aux2_precision_m: 0.8486 - aux2_recall_m: 0.6677 - val_loss: 1.9933 - val_main_loss: 0.6829 - val_aux1_loss: 0.6503 - val_aux2_loss: 0.6601 - val_main_accuracy: 0.7594 - val_main_f1_m: 0.7605 - val_main_precision_m: 0.8355 - val_main_recall_m: 0.7031 - val_aux1_accuracy: 0.7668 - val_aux1_f1_m: 0.7550 - val_aux1_precision_m: 0.8398 - val_aux1_recall_m: 0.6911 - val_aux2_accuracy: 0.7556 - val_aux2_f1_m: 0.7587 - val_aux2_precision_m: 0.8460 - val_aux2_recall_m: 0.6935\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 20/25\n",
            "721/721 [==============================] - 354s 491ms/step - loss: 1.9944 - main_loss: 0.6843 - aux1_loss: 0.6462 - aux2_loss: 0.6639 - main_accuracy: 0.7522 - main_f1_m: 0.7402 - main_precision_m: 0.8425 - main_recall_m: 0.6682 - aux1_accuracy: 0.7591 - aux1_f1_m: 0.7433 - aux1_precision_m: 0.8430 - aux1_recall_m: 0.6717 - aux2_accuracy: 0.7569 - aux2_f1_m: 0.7413 - aux2_precision_m: 0.8473 - aux2_recall_m: 0.6674 - val_loss: 2.3601 - val_main_loss: 0.8038 - val_aux1_loss: 0.7643 - val_aux2_loss: 0.7920 - val_main_accuracy: 0.7207 - val_main_f1_m: 0.7322 - val_main_precision_m: 0.8104 - val_main_recall_m: 0.6719 - val_aux1_accuracy: 0.7444 - val_aux1_f1_m: 0.7345 - val_aux1_precision_m: 0.8096 - val_aux1_recall_m: 0.6767 - val_aux2_accuracy: 0.7344 - val_aux2_f1_m: 0.7330 - val_aux2_precision_m: 0.8218 - val_aux2_recall_m: 0.6659\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 21/25\n",
            "721/721 [==============================] - 352s 488ms/step - loss: 2.1363 - main_loss: 0.7261 - aux1_loss: 0.6990 - aux2_loss: 0.7113 - main_accuracy: 0.7408 - main_f1_m: 0.7306 - main_precision_m: 0.8398 - main_recall_m: 0.6549 - aux1_accuracy: 0.7514 - aux1_f1_m: 0.7331 - aux1_precision_m: 0.8335 - aux1_recall_m: 0.6623 - aux2_accuracy: 0.7455 - aux2_f1_m: 0.7351 - aux2_precision_m: 0.8376 - aux2_recall_m: 0.6636 - val_loss: 2.0403 - val_main_loss: 0.6984 - val_aux1_loss: 0.6744 - val_aux2_loss: 0.6675 - val_main_accuracy: 0.7531 - val_main_f1_m: 0.7552 - val_main_precision_m: 0.8380 - val_main_recall_m: 0.6911 - val_aux1_accuracy: 0.7506 - val_aux1_f1_m: 0.7481 - val_aux1_precision_m: 0.8319 - val_aux1_recall_m: 0.6839 - val_aux2_accuracy: 0.7569 - val_aux2_f1_m: 0.7572 - val_aux2_precision_m: 0.8432 - val_aux2_recall_m: 0.6911\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 22/25\n",
            "721/721 [==============================] - 353s 489ms/step - loss: 1.9977 - main_loss: 0.6802 - aux1_loss: 0.6507 - aux2_loss: 0.6667 - main_accuracy: 0.7494 - main_f1_m: 0.7443 - main_precision_m: 0.8378 - main_recall_m: 0.6771 - aux1_accuracy: 0.7596 - aux1_f1_m: 0.7486 - aux1_precision_m: 0.8395 - aux1_recall_m: 0.6828 - aux2_accuracy: 0.7627 - aux2_f1_m: 0.7495 - aux2_precision_m: 0.8455 - aux2_recall_m: 0.6800 - val_loss: 1.8601 - val_main_loss: 0.6445 - val_aux1_loss: 0.6002 - val_aux2_loss: 0.6154 - val_main_accuracy: 0.7668 - val_main_f1_m: 0.7569 - val_main_precision_m: 0.8392 - val_main_recall_m: 0.6935 - val_aux1_accuracy: 0.7918 - val_aux1_f1_m: 0.7628 - val_aux1_precision_m: 0.8501 - val_aux1_recall_m: 0.6959 - val_aux2_accuracy: 0.7781 - val_aux2_f1_m: 0.7540 - val_aux2_precision_m: 0.8391 - val_aux2_recall_m: 0.6887\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 23/25\n",
            "721/721 [==============================] - 352s 489ms/step - loss: 2.0245 - main_loss: 0.6947 - aux1_loss: 0.6587 - aux2_loss: 0.6711 - main_accuracy: 0.7467 - main_f1_m: 0.7360 - main_precision_m: 0.8320 - main_recall_m: 0.6670 - aux1_accuracy: 0.7485 - aux1_f1_m: 0.7380 - aux1_precision_m: 0.8313 - aux1_recall_m: 0.6707 - aux2_accuracy: 0.7457 - aux2_f1_m: 0.7377 - aux2_precision_m: 0.8318 - aux2_recall_m: 0.6697 - val_loss: 2.0860 - val_main_loss: 0.7256 - val_aux1_loss: 0.6784 - val_aux2_loss: 0.6821 - val_main_accuracy: 0.7269 - val_main_f1_m: 0.7212 - val_main_precision_m: 0.8004 - val_main_recall_m: 0.6611 - val_aux1_accuracy: 0.7494 - val_aux1_f1_m: 0.7436 - val_aux1_precision_m: 0.8130 - val_aux1_recall_m: 0.6899 - val_aux2_accuracy: 0.7344 - val_aux2_f1_m: 0.7382 - val_aux2_precision_m: 0.8200 - val_aux2_recall_m: 0.6755\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 24/25\n",
            "721/721 [==============================] - 353s 489ms/step - loss: 1.9929 - main_loss: 0.6797 - aux1_loss: 0.6515 - aux2_loss: 0.6617 - main_accuracy: 0.7520 - main_f1_m: 0.7396 - main_precision_m: 0.8285 - main_recall_m: 0.6749 - aux1_accuracy: 0.7638 - aux1_f1_m: 0.7517 - aux1_precision_m: 0.8380 - aux1_recall_m: 0.6876 - aux2_accuracy: 0.7635 - aux2_f1_m: 0.7452 - aux2_precision_m: 0.8336 - aux2_recall_m: 0.6804 - val_loss: 1.9287 - val_main_loss: 0.6716 - val_aux1_loss: 0.6210 - val_aux2_loss: 0.6362 - val_main_accuracy: 0.7506 - val_main_f1_m: 0.7620 - val_main_precision_m: 0.8602 - val_main_recall_m: 0.6887 - val_aux1_accuracy: 0.7768 - val_aux1_f1_m: 0.7631 - val_aux1_precision_m: 0.8562 - val_aux1_recall_m: 0.6923 - val_aux2_accuracy: 0.7643 - val_aux2_f1_m: 0.7679 - val_aux2_precision_m: 0.8694 - val_aux2_recall_m: 0.6923\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n",
            "Epoch 25/25\n",
            "721/721 [==============================] - 346s 480ms/step - loss: 2.0273 - main_loss: 0.6973 - aux1_loss: 0.6559 - aux2_loss: 0.6741 - main_accuracy: 0.7511 - main_f1_m: 0.7380 - main_precision_m: 0.8303 - main_recall_m: 0.6713 - aux1_accuracy: 0.7653 - aux1_f1_m: 0.7502 - aux1_precision_m: 0.8385 - aux1_recall_m: 0.6854 - aux2_accuracy: 0.7593 - aux2_f1_m: 0.7445 - aux2_precision_m: 0.8333 - aux2_recall_m: 0.6796 - val_loss: 1.9209 - val_main_loss: 0.6786 - val_aux1_loss: 0.6162 - val_aux2_loss: 0.6262 - val_main_accuracy: 0.7594 - val_main_f1_m: 0.7569 - val_main_precision_m: 0.8469 - val_main_recall_m: 0.6887 - val_aux1_accuracy: 0.7743 - val_aux1_f1_m: 0.7600 - val_aux1_precision_m: 0.8701 - val_aux1_recall_m: 0.6791 - val_aux2_accuracy: 0.7693 - val_aux2_f1_m: 0.7587 - val_aux2_precision_m: 0.8723 - val_aux2_recall_m: 0.6767\n",
            "WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,main_loss,aux1_loss,aux2_loss,main_accuracy,main_f1_m,main_precision_m,main_recall_m,aux1_accuracy,aux1_f1_m,aux1_precision_m,aux1_recall_m,aux2_accuracy,aux2_f1_m,aux2_precision_m,aux2_recall_m,val_loss,val_main_loss,val_aux1_loss,val_aux2_loss,val_main_accuracy,val_main_f1_m,val_main_precision_m,val_main_recall_m,val_aux1_accuracy,val_aux1_f1_m,val_aux1_precision_m,val_aux1_recall_m,val_aux2_accuracy,val_aux2_f1_m,val_aux2_precision_m,val_aux2_recall_m,lr\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "C8o3b0Pn5KC2",
        "scrolled": true,
        "outputId": "8792ec0f-1383-4cb9-da72-a94efbf0e41b"
      },
      "source": [
        "# plot graphs from history\n",
        "print(history.history.keys())\n",
        "print(f'Accuracy: {history.history[\"main_accuracy\"][-1]}')\n",
        "print(f'loss: {history.history[\"main_loss\"][-1]}')\n",
        "print(f'f1_m: {history.history[\"main_f1_m\"][-1]}')\n",
        "print(f'precision_m: {history.history[\"main_precision_m\"][-1]}')\n",
        "print(f'recall_m: {history.history[\"main_recall_m\"][-1]}')\n",
        "\n",
        "plt.plot(history.history[\"main_accuracy\"], label='accuracy')\n",
        "plt.plot(history.history[\"val_main_accuracy\"], label='Validaion accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'])\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history[\"main_f1_m\"], label='F1_score')\n",
        "plt.plot(history.history[\"val_main_f1_m\"], label='Validaion F1_score')\n",
        "plt.ylabel('F1_score')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'])\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history[\"main_precision_m\"], label='Precision')\n",
        "plt.plot(history.history[\"val_main_precision_m\"], label='Validaion Precision')\n",
        "plt.ylabel('Precision')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'])\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history[\"main_recall_m\"], label='Recall')\n",
        "plt.plot(history.history[\"val_main_recall_m\"], label='Validaion Recall')\n",
        "plt.ylabel('Recall')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'])\n",
        "plt.show()"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "dict_keys(['loss', 'main_loss', 'aux1_loss', 'aux2_loss', 'main_accuracy', 'main_f1_m', 'main_precision_m', 'main_recall_m', 'aux1_accuracy', 'aux1_f1_m', 'aux1_precision_m', 'aux1_recall_m', 'aux2_accuracy', 'aux2_f1_m', 'aux2_precision_m', 'aux2_recall_m', 'val_loss', 'val_main_loss', 'val_aux1_loss', 'val_aux2_loss', 'val_main_accuracy', 'val_main_f1_m', 'val_main_precision_m', 'val_main_recall_m', 'val_aux1_accuracy', 'val_aux1_f1_m', 'val_aux1_precision_m', 'val_aux1_recall_m', 'val_aux2_accuracy', 'val_aux2_f1_m', 'val_aux2_precision_m', 'val_aux2_recall_m', 'lr'])\n",
            "Accuracy: 0.7482662796974182\n",
            "loss: 0.6977332234382629\n",
            "f1_m: 0.737145185470581\n",
            "precision_m: 0.8304170966148376\n",
            "recall_m: 0.6696264743804932\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEGCAYAAAB2EqL0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3iUVfbA8e9NgYSQhFRKAiRAgNBL6GAvqKvYQFBZrNh1/a2ubnd117q7uiqu4tobKoriirhBASmhhCKdNAgkgVQgCZA69/fHncAQUmaSmcwkOZ/n4Rnyzvu+c8fInLn3nnuu0lojhBBCNMbL3Q0QQgjROkjAEEIIYRcJGEIIIewiAUMIIYRdJGAIIYSwi4+7G+As4eHhOiYmxt3NEEKIVmXTpk0FWusIe85tMwEjJiaG5ORkdzdDCCFaFaVUpr3nypCUEEIIu0jAEEIIYRcJGEIIIezSZuYw6lJZWUlWVhZlZWXuborL+fn5ER0dja+vr7ubIoRoo9p0wMjKyiIwMJCYmBiUUu5ujstorSksLCQrK4vY2Fh3N0cI0Ua16SGpsrIywsLC2nSwAFBKERYW1i56UkII92nTAQNo88GiRnt5n0II92nzAUMIIZwicy1kte+1XhIwXOzo0aO89tprDl93+eWXc/ToURe0SAjhsPIS+GQWLHnU3S1xKwkYLlZfwKiqqmrwuiVLltClSxdXNUsI4YgNb0LZUcjfAxaLu1vjNm06S8oTPP7446SnpzNixAh8fX3x8/MjJCSEPXv2kJKSwtVXX83BgwcpKyvjoYceYu7cucDpUielpaVcdtllTJ48mbVr1xIVFcXXX3+Nv7+/m9+ZEO1ExXFIehV8/KDyBBw7ACEx7m6VW7SbgPGXb3ayK6fYqfcc1COIP185uMFznn32WXbs2MHWrVtZsWIFV1xxBTt27DiV/vr2228TGhrKyZMnGTNmDNdddx1hYWFn3CM1NZVPPvmEN998kxkzZvDFF19w8803O/W9CCHqkfwOnCiEi5+CxD9C3p52GzBkSKqFjR079oy1Ei+//DLDhw9n/PjxHDx4kNTU1LOuiY2NZcSIEQCMHj2a/fv3t1RzhWie/avh81vgyH53t6RpKk/Cmn9B7Lkw6pfmWP5u97bJjdpND6OxnkBLCQgIOPX3FStWsGzZMpKSkujUqRPnnXdenWspOnbseOrv3t7enDx5skXaKkSzrfoHpP8IaT/CtFdh0FXubpFjNr8Px/Pg3HfAvwsEdjc9jHZKehguFhgYSElJSZ3PHTt2jJCQEDp16sSePXtYt25dC7dOCBc6XggZK2HYTAjrC5/NNllGla1kgWlVOax+CXpPgpjJ5ljEQOlhCNcJCwtj0qRJDBkyBH9/f7p27XrqualTp/L6668THx/PgAEDGD9+vBtbKoST7fkGdDVMuBci4uGHv5jJ4wPrYPq7Joh4si0fQkkOXG2T5RgZb+Y0LBbwan/ft5XW2t1tcIqEhARdewOl3bt3Ex8f76YWtbz29n6Fh3t/Ghw9AA9shppKBHuXwld3Q3UlXPkvGHq9e9tYn+pKeHkUBHaF2xNPt3/Te/DNg/DgFgjt4942OolSapPWOsGec9tfiBRCuN7xAtj3Ewy+9vSHLcCAqXD3aug6BL64HRY/ABUn3NfO+vy8wKTPnvvYme2PHGQe2+k8hksDhlJqqlJqr1IqTSn1eB3Pv6iU2mr9k6KUOmrzXC+l1P+UUruVUruUUjGubKsQwol2LwZtgcHXnP1ccDTc8i1M+TVs/gDevMCzPoCrq2DV36HHSOh30ZnPRQwwj+10HsNlAUMp5Q3MAy4DBgGzlFKDbM/RWj+stR6htR4BvAJ8afP0+8ALWut4YCyQ56q2CiGcbMeXEBYHXevJTvT2gQv/BDd/AScKYP55Zs7AE4bIdyw0acDnPHpm7wLALwiCoj0rwLUgV/YwxgJpWusMrXUFsACY1sD5s4BPAKyBxUdrnQigtS7VWntgv1UIcZaSXMhcY3oXjVVR7nehGaKKToCv74NFd5m6Te5iqYafXoCuQ2HA5XWfEzkQ8qSH4WxRwEGbn7Osx86ilOoNxAI/Wg/1B44qpb5USm1RSr1g7bHUvm6uUipZKZWcn5/v5OYLIZqkoeGougR2g19+Def9DrZ/bnob7voGv3MRFKbBOY/UH+wiBkJBigkunmDfKlNJtwV4yqT3TGCh1rrmN+ADTAEeAcYAfYBbal+ktZ6vtU7QWidERES0VFuFEA3Z+RWEDzApqPby8obzHoM538CJIvjf713XvvpYLPDT301AiG9ggWFkPFSXQ9G+lmtbfYpzzEr6Jb9pkaKIrgwY2UBPm5+jrcfqMhPrcJRVFrDVOpxVBXwFjHJJKz1M586d3d0EIZqu5LAZjhpybePDUXWJmQz9p8LhHc5vW2P2fGMms895tOE1FhHWQOjuie/qShMsqsrg+rdbZF2IK19hIxCnlIpVSnXABIXFtU9SSg0EQoCkWtd2UUrVdBsuAHa5sK1CCGfY9TWgYdDVTb9HZDyUHjY9jZaitZm7COvX+FBaTaaUuye+E/8EB9fDVa9ARP8WeUmXBQxrz+B+4HtgN/CZ1nqnUupJpZRtf28msEDbrCC0Dk09AvyglNoOKOBNV7XVlR5//HHmzZt36ucnnniCv/71r1x44YWMGjWKoUOH8vXXX7uxhUI40c5FZq1C5MCm36NrzVqHFvyOmLIUDm+HKY+Y4bGGdOwMwb3c28PYuQjWvQbj7ja9uRbi0tIgWuslwJJax/5U6+cn6rk2ERjmtMZ897j5H8KZug2Fy55t8JQbbriBX/3qV9x3330AfPbZZ3z//fc8+OCDBAUFUVBQwPjx47nqqqtkX27RuhXnwIEkOL+Z8w+nFsftPl3DyZW0hpXPmZLlQ6fbd03kQPf1MApS4ev7IXqsKbnegqSWlIuNHDmSvLw8cnJyyM/PJyQkhG7duvHwww/z008/4eXlRXZ2Nrm5uXTr1s3dzRWi6XZZe8r2ZkfVJ7A7+HWB3J3Nb5M90n6AnC1w5ctmfYg9IuMhY4VZ5GfvNc5QcRw+nQ0+HU09Lp8OLffatKeA0UhPwJWmT5/OwoULOXz4MDfccAMfffQR+fn5bNq0CV9fX2JiYuosay5Eq7JzkVm/EB7XvPsoZXoZLbHWoaZ3EdwThs+y/7qIeKiugKKMFps/QGv478Nmm9jZX0JwnasUXMpT0mrbtBtuuIEFCxawcOFCpk+fzrFjx4iMjMTX15fly5eTmZnp7iYK0TzHsswE7OBmTHbb6moNGK5e+b1vJWRtgMm/cuzbes0cTUvOYyS/Dds+hfN/B30vaLnXtSEBowUMHjyYkpISoqKi6N69OzfddBPJyckMHTqU999/n4EDmzFBKIQn2PmVeWzucFSNyHgoPwbF9WXiO8nKF8wQ2AgHtzwOHwCollvxnb0Zlj4O/S42E/Nu0n6GpNxs+/bTE+7h4eEkJSXVeV5paWlLNUkI59m5CLoNc94eF5HWGlS5u0yxQlfYvwYyV8PU58DXz7FrO3SCkN4tEzBOFMFnc6BzV7h2vlv34ZAehhCieY5kQnay83oXcHrIx5WptT89DwGRMHpO066PiDfzCa5kscCXc826lBnvQadQ175eIyRgCCGax1nZUbb8QyAoynUB4+BGk+U06UHw9W/aPSIHmrpTVRVObdoZVv0D0hJh6jMQNdp1r2OnNh8w2sqOgo1pL+9TeKCdi8zeEaGxzr1vZLzrAsbuxeDdAUbf2vR7RMSDpQqK0p3XLlvpy2H532DoDEi43TWv4aA2HTD8/PwoLCxs8x+mWmsKCwvx83NwHFaI5jqyH3I2O7d3USMyHvJTzFoHZ8vZYnb969iM2m2nhs1cMI9xLNvsSBgxEK58qWl1uVygTU96R0dHk5WVRXsofe7n50d0tIsmB4Woz85F5rE5taPqEznYWhXWyWsdLBbI2QrDb2jefcL7g/JqcB7jYNEJwjt3xL9DI+VGbFVVWIsKlsMNH0CHgOa104nadMDw9fUlNtbJ3WQhxGk7F5mx9ZDezr93TXn0vJ3ODRiFqVBRAj2aWQDb1x9CYuvtYfxv52Hu/nATPt5eTOgTxvkDIjh/YCS9wxoJAIl/NGtDpr/b/EWQTtamh6SEEC5UmA6HfnbNcBSYqrDKy/lDPtmbzWOUE3ZMiKw7U2pb1lEeWrCVIVHBzB7fm4NHTvDEN7s494UVXPD3FTz5zS5WpeZTXlVrE6b0H2H96zDuHtf9d22GNt3DEEK40C7rYj1XDEeB+QYf2sf5NaVyNoNvgBlSaq6IgbD3OzN85NMRgKwjJ7jt3WTCOnfgrTljiAjsyB9/MYjMwuOs2JvP8r15fLQ+k7fX7MPf15tJ/cI5f2AE5w2IJGr7F+AXDBc/2fy2uYAEDCHagqJ9ZpvO4bNabmHXzkWmYmqXno2f21SRg5wfMLI3m6yuxsqY2yMyHnS1Sa/tOphjJyu57d2NlFdV88md44gI7Hjq1N5hAcyZGMCciTGcrKhmXUYhy/fm8eOePJbtzgU0m/yXcLhLArt/ziMusjP9IjsT0NFzPqY9pyVCiKb78SnY8YXZ7c6RqqtNVZBmtgu49BnXvk7kINj9DVScMKurm6uqwrR73Nzm3wtMDwMgbzcVYfHc+9Em9hUc571bxxLXNbDey/w7eHP+wEjOHxjJX67SpOcfZ1vyKsI2FPF8fhyffv7zqXOjuvgT17Uz/bsG0i+y86lAEujn65z34AAJGEK0dpZqM/YdFAVbP4KyY3DdW46Xu3DEqeyoaa57DbBupqShYK/pFTRX3i6TedXcCe8a4XGgvNF5u/n9nu2sSSvk79OHM7FfuN23UErRL7Iz/QLNXM3fHnmIuZVBpOaWkpZXQkpuKal5paxNL6Si6vS+3T2C/ejXNZC4yM4Miw5m2gjXV6+VgCFEa5e9CU4egSv+AccL4btH4aPrYdYn0LH+b7nNsnMR9Bzv+hLbNZsp5e5yTsDIceKEN5h5i9A+ZOxK5vPsUTx4YRzXj25ienvaD9BtKD5detAX6BvRGTi9R061RXOg6ASpuSWk5pWeelyfUcjw6C4SMIQQdkhbZrKJ+pxvag35d4FFd8N7V8JNX0BAmHNfL3+vSXWd+pxz71uX0D7g3dF5K76zN4F/KHRxXhpwTocYVPZ2rh0ZxcMXNTENtqwYDq6DiQ/Ue4q3lyI2PIDY8AAuGXz6eLVFU1JW2bTXdZCk1QrR2qUmQlTC6cJ0w2bAzI9MOuo7U82qYWfa+RWgXDIcVVVt4XeLtvP0kt1UVVvMxHTEACcGjC2md+GkldPrMwr5MiuQGK88npkW1/RtlvetNGVG+l3s8KXeXoounVpm5z0JGEK0ZscLTJmLfhedeXzAZXDzl1ByGN6+1ExSO8vORdB7IgR1d949MSVufr9oBx+vP8D8nzK44/1kSsuroOtg56zFqDhuNjxy0vxFen4pcz/YRGFAX7yw0PFIM2pKpSZCxyDoOdYpbXMVCRhCtGbpPwIa4i46+7mYSTDnG6g8aYLGoZ/PPsdRebvNh64LFpW98P1ePk0+yIMXxvH0NUNZlVrAjNeTKA6Kg5JDZl+I5ji0DbTFKfMXhaXl3PrORny8FHOvvcIcbGqpc63NsGKfc8G75TOfHCEBQ4jWLG0ZdAqD7vVMCPcYAbd9bxbBvfsLs1ajOXYuAhTEX9W8+9Tyn1UZvLYinRvH9eLhi+K4cVwv3pqTQGbhcf68zpoZ1NxeRs2EdzN7GGWV1dzxfjK5xWX8Z04C3fsMBi+fprcvf4/ZWbAJw1EtTQKGaN9OFJlidK2RxWIya/pe2PBivfB+cNtSCOwGH1wDKd83/fV2LoKYyRDYtWn3qMOiLVn89dvdXDakG09NG3JqHuC8AZF8fvdE0ugFQOqODc17oezNJvW4GW23WDT/99lWth48yks3jGBkrxCzF3hYv6b3MFITzWPtYUUPJFlSov1K/9F8gHbobNI3uw2FbkPMVqOR8R5VJbROh7bAiQKIs+ObaXA03LoUPrwWFtwIV78Ow6bXf37lSfON+fB2yN1hHg/vMEX7xt/jtLewfE8ej36+jYl9w3hp5gi8vc6cNB7UI4j5911Jyb8C2LB+NZsir2Pm2F5Ne7Gczc1OzX1u6R6WbD/M7y+P57KhNnM4EQObPuSXtsz8/+fqFGUnkIAh2q/VL0HnbibbJ3cHbP8ckt+yPqnMt8ZuQ0wg6TrUPAZ285i9CUj7AVDQ9wL7zg8IM3MaC26EL++EsqMw9k4ozYPD20xAqAkQBSlmvB+gQ6CZeB4+01SmHdbMsuBWmzKPcM9HmxjYPZA3Zo+mo0/dpTq6d+lEVdRQEvIPc+mX2zl45AS/vngAXl4O/B5OHjFl0kfe3KS2Hiw6wYKNB3jjpwxuHt+LO6bUqoIdGW92HnR0RXp5KRxIgnF3NaldLU0ChmifDm0zqYwXPQGTHzbHtIajB878Vp296fSqZjDzBdFjTerqgMtdu5q6MamJ5htzgP2rivELgpsWwsJbYckjsOJZ00upEdzTbCwUf9XpYNklxun1qVJyS7jt3Y10C/Lj3VvHNlrmwqfbYPrnf86sMT2Ztzydg0UneWH6sHqDzFlytphHO+cvtNbsyC4mcddh/rcrlz2HSwC4bEg3nrhy8NnpsxEDadKK9P2roLqiVcxfgAQM0V4lzTMVS0ffcvqYUmZfh5DeEP+L08dPHjUF8HJ3mG/i6csh5TtTVXTwtTDiRoge07I9jxNFkJ0MUx5x/FpfP5jxAax81qzR6GbtPXUdfHothwtlHTnBL9/aQEcfLz64fRzhnTs2flFkPKq8mKcvCqV3eADPfreHw8fKeGP2aEIC7FiDUFPSvIEP84oqC+v3FZK4K5fEXbkcOlaGl4KE3qH84Yp4Lh7Utf69LE7t3bHHsYCRmmj+P+w1wf5r3EgChmh/inNgx0IYcwf4hzR+vn8Xk6IaM8n8bKk2vZOtn8DPC2DTO2b4avhMGDbTtdVba2QsN0NG9sxf1MXbBy74g3PbZIei4xX88u0NnKio4rO7J9Az1M7hm65mabPK283d515MVBd/fv35z1z377W8c+uYxjclytkCoX3N79JGSVklK/bmk7grl+V78ygpq8LP14spcRH838X9uWBgJGH2BLTQPuDla1KO7aU1pCWadFqflll411wSMET7s/4N82Hb1MlbL28zb9D3AlPSYdfX8PMn8ONf4ce/QewUGH4jDLrKdRPnqcvAr4uZU2gljpdXces7G8g+cpIP7xjHwG5B9l986hv8Loi7mCuH96BbsB93vp/MNa+t5c1fJjC6twn+ldUWSsqqKD5ZSXFZJcUnq0jI3EhuyGj+91OG9VglGQXHWZdRSGW1JjSgA1MHd+OSwd2Y3C/csS1VwayfCI8zPQx7FaaZIdBJv3LstdzIpQFDKTUV+BfgDfxHa/1sredfBM63/tgJiNRad7F5PgjYBXyltb7flW0V7UR5qekRxF8JITHNv59fEIyabf4U7YNtn5rg8dXd8O2vzYT6iFnQe7Lz5gEsFpNZ0/cC5+zp0AIqqizc/eEmduQU88bNoxkT4+DQl38IBPYwRQitxsSEsujeSdz6zgZmzV9HaEAHissqOVFx5i52ERxho18u7xWH8XbGbrwUBPr5EhnYkVsmxnDJ4G6M6hVyVoaWwyLjIWuj/ee3onTaGi4LGEopb2AecDGQBWxUSi3WWp/6jWutH7Y5/wGg9uDfU8BPrmqjaIe2fGjKf0+ov8hbk4XGwnmPw7mPmcyXrR+buks/fwwT7odL/+ac18ndDsfzmj4c1cJq1i6sSi3gheuHcdGgJq6DiIw/q6ZUbHgAX947iRcTU6ioshDo50OQvy9Bpx596ZW/ApbDPTdN5+E+Ewno4ONYhpW9IuLNniTlpdCxc+Pnpy0zu/65Yj90F3FlD2MskKa1zgBQSi0ApmF6DHWZBfy55gel1GigK7AUSHBhO0V7YamGda9Bz3HQc4zrXkcpU2up90S47Hn4+j7Y+B+Y/H/OqRybtsw89r2w+fdyMa01f/lmJ//ddojfXjaQ6QnNmN/pOgjWz4fqqjM2iAoN6MBTVw+p/7ofU0F5E9FvDHRwYemNSOtmSgV7Gx8qrDgB+1fDmNtd1x4XcOVK7yjgoM3PWdZjZ1FK9QZigR+tP3sB/wAaTAFRSs1VSiUrpZLz8/Od0mjRhu3+Bo5mmm/7LaVDJ9PjqCqzWePRTKnLzOJCJ662doWqagtP/ncX7yVlMvecPtx1bt/m3TBykNn86Mg+x67L3mxdiOmEHfsaEmGTKdWY/avNe2lFw1HgOaVBZgILtdY1g4/3Aku01lkNXaS1nq+1TtBaJ0RERLi8kaKVS3rVzFsMvKJlXzdyoMmz3zAfKsuad6+yY3Bwvcd/0Bw7Wclt7yXzzpr93DophsenDmz+TWsmvh3Z41trp6zwtktorNm7w55MqbRl4OMPvSe5vl1O5MqAkQ3Y9j+jrcfqMhP4xObnCcD9Sqn9wN+BXyqlnq3rQtFGFR8yC+ec5cB6MyE5/l73TBRPvB+O58P2z5p3n4wVoKs9ev5iX8FxrnltDWvTCnj6mqH8+crBzpkziBgIKMeK/B3Zb1Z5O2uHvYZ4eZs5CXt6GGmJJpvOnQs/m8CVAWMjEKeUilVKdcAEhcW1T1JKDQRCgKSaY1rrm7TWvbTWMZhhqfe11o+7sK3C0yx9DN680DkluQGSXjEL7Ubc5Jz7OSr2XFNeJGme+dbbVKmJ0DHYrDb3QKtS85n26mqOHK/gwzvGceO4JtZ9qouvv1nvkOdAD+PUlqwtlH4cObDxgFaYbsqUtJLV3bZcFjC01lXA/cD3wG7gM631TqXUk0op29rIM4EFWjfnX5FoUywW2LfKjPF+9kuz0ro5ijJg938h4Tb7sldcQSnTy8jfc3rS2lFam/pRfc49Y9LXE2iteXfNPm55ZyPdg/1ZfP9kxvdx8tawYCa+HelhZG8GH7/Te4O7WsRAKM4y63Pqk/aDeezn+UkLtbl0DkNrvURr3V9r3Vdr/TfrsT9prRfbnPNEQ70HrfW7sgajnSnYCyeLTNmOY1nw1T3N+1a+7nWzX8FYNxd4G3wtBHaHta807fq8XVCS43HDURVVZlvVJ77ZxfkDIvji3on2r+B2VOQg8wWg8qR95+dsMWVPWmpjopp5lvy99Z+Tlmh6SmHNTAJwA0+Z9BbitP2rzeOkX8Elf4W9S2DNv5p2r5NHzNqLodc7fUtRh/l0MFVJ961s2vyMk9NpN2UeYW1aAZXVTd8PpOh4BTe/tZ5PNhzk3vP6Mn92Ap07urD3EznIrNJv6AO5hqUacrY6bUtWu0RYJ/frm/iuLDO9Zw9PWqiPZ/VrhQDIXGM2ugmJgXF3m6ygH/4C0Qlm8x5HJL8Dlcdhwn0uaarDRt8CK18wcxnXvO7YtamJEDnYKfsmbDlwhJnzk6is1gT5+XDBwEguGdyNc/pH2P2Bv+dwMXe8l0xeSTn/mjmCaSNaYD+HmqGlvF1mN8GGFKSY331LTHjXCIkx2U/1TXxnroGqk61y/gKkhyE8jdZmG9HeE824v1Jw1SumcNznt0LJYfvvVVVhUln7nGeGJTyBf4jZk2H7QpMJZq/yEjiwru69ux1UWFrOvR9tpmuQH6/eOJKLB3VjZUo+9360mVFPJnLLOxv4eP0B8orrTwFO3JXLda+tpaLKwmd3TWiZYAFmKMe741krvuuU7ZwtWR3i5Q0R/evvYaT9YNrv6BcfDyEBQ3iWwnQozT0zP71jINzwAVSUwsLbzEpfe+z4AkoOuaYMSHOMv8ekxm54w/5r9v0ElspmD2VUVVt44JMtFB6v4PWbR/OLYT34x4zhbPz9RXw6dzyzJ/QmI/84v1u0nbFP/8DV89bw2oo00vJK0FqjtWbe8jTmfpBM38jOLL5/MiN6dmn8hZ3F28d8IOfaEzA2QccgU0m4JUXE19/DSEs0VY9dvYjQRWRISniWTOv8Re1vYJHxcOW/zE5xP/wFLnmq4ftobRbqRcR7XjZKaCwM/AUkv232s7Ancys10Wwl23N8s176H4kprE0v5PnrhzEkKvjUcR9vL8b1CWNcnzD+cEU8Kbml/G/nYRJ35/L80r08v3QvseEBdA/2Y216IVcO78EL1w/Dz9cNa1oiB5sA2piczdB9uNM3f2pU5EDYtsBk99mWUz+SaYbJRt/asu1xIulhCM+SuRYCIuv+VjhsBiTcDmtfNmmyDclYYTY8mnCf52ypamviA2bV9taPGj+3Jp02tnn7JizdcZh/r0hn1thezGigppNSigHdAnngwjgW3z+ZpN9ewFPTBhMd4s+uQ8U8ckl/Xp45wj3BAsyXh5Ick9BQn6pys+VsS85f1IioJ1OqJmmhlU54g/QwhCfRGvavOT1/UZepz5hUya/uMR8c9aUmJr1qAs+wGa5rb3P0HGsW3617zWzk1NDq84IUOHYApjxc/zmNyMgv5ZHPf2Z4dDBPXOXYmoTuwf7MnhDD7AkxTX59p7JupkTebvP/Sl1yd5ghvJacv6gRaZMp1Wvc6eNpy6BLL7NvRislPQzhOY5mmkVPDU0I+nSEGe+ZD9jP5tSdj5+32/zjHDvXnO+pJt5vSlfsaaS31MxvpsfLq7j7w034eiteu3m0/ftgeyp7akrVTHi7o4cR3At8O525wLCqAjJWmuwoT+zx2kkChvAc+9eYx8YKsnXpBdf+x3yL/LaOgsZJr5rURk8vHT3wFyYNc+2rDZ+XmgjhA8z7dpDWmse+2EZaXimvzBpFVBf/prXVkwRFmfIoDa34ztkCncIhuAW2y63NywsiBpzZvgNJJsW3FQ9HgQQM4Uky15q00wg7KpvGXQTn/ga2fgib3z99vCQXtn0GI26ETg7u6tbSvLxNMcSsDXBwQ93nVBw3uftNXN39zpr9/HfbIX59yQAmx4U3o7EeRKk6N1M6Q/Zm07tw17f5iHhTBqZG2jKz53fsOe5pj5NIwBCeI3O16V3Ym9Vy7mNmm9JvHzldpHDjm1BdaT6IW4MRN5miiEn19DL2r4bqiiZlem3YV8TTS3Zz8aCu3NPcvSg8TddBJgULtyUAACAASURBVGDUVTKmvNSUl3HH/EWNyIEmPfxEkfk5bRn0nuC+WmZOIgFDeIZj2WY835H9Aby8zdBUQLgpUlh8CDa+BQMug/AWzr1vqo6dTVHE3d+Y919baqIZD+9Vz+RuPfKKy7jv481Eh/jzjxnDXbMlqTtFDjJZZsU5Zz936GdTPsQd8xc1alak5+8x9dDydrXa1d22JGAIz5C51jzWl/VSn4AwmP6eCThvnm+KFrbkjnrOMPYuUN6w7t9nP5e2DGIc2zehstrCfR9vprSsitdnjybIr4UK77WkUyVC6pjHyHHDCu/aaoZV83bbVKdt3fMXIAFDeIrM1WYisyklPHqOgUv/ZlZ19xjpeNBxt6Dupjji5g/OXFtQmG62I3Vw/uLZ7/awcf8Rnr1uKAO7BTm5sR6iJlOqrr0xsjebye7ObtyFMzgaOgRay9knmon6mja3YhIwhGfYvwZ6jW/6bnhj58JlL8BVr3pU2mJVtYXPkg+yM+dYwydOuM9k0Wx69/Sx1ETz6MA3029+zuGt1fu4ZWJMy9V3codOoaZUfH09jJbYkrUhSplMqcM7rOm0F3nU/5dNJQFDuF9pHhSmNq9noBSMmwvdhjivXc20M+cY17y2lt8s3Mat72yksLS8/pO7DTVFEte/YXL2wQxHhfY1pUTskJJbwmNfbGN07xB+d3nr/zbbqMj4s9dinCgyc0HunL+oETkQDqyF8uI2MRwFEjCEJ8i0rr9opRU8ayurrOa5pXu46tU1HDpWxm8vG8jRE5U89sU2GtxYcsL9Zlht55dmQeL+VXYPR5WUVXL3B5vo1MGH124aRQefdvBPO3KQKb9hqT59zBPmL2rUlAjx8jG7JLYBUhrEk508aiq1NnWYprXYvwZ8A0yhuFZubXoBv/tyO/sLT3BDQk9+d3k8wZ188fX24sn/7uL9pEzmTIyp++J+F5nJ0rWvmkVnVWV2fTOttmge/nQrmUUn+OiOcXQNsn+CvFWLHGS28S3KOF1uI3uLeWxsr4yWUFMipOc4kzrdBrSDryGtVMUJeGkYbPnA3S1xvcy1puZOS22j6QLHTlTy+BfbuPHN9Wjg4zvG8dz1wwjuZN7TrZNiOH9ABH9bspvdh+rZ71kpM5eRux1+fMrsRW1Hr+v5pXtYtjuPP/1ikGv20fZUXW02U6qRsxnC4jzjA7rrEFBe0P9Sd7fEaSRgeKrCVCg/1nC9nLbgRJHJdGltmU1WWmuWbD/Ehf9cyeebsrjr3D4sfegcJvY7c1W1UooXpg8nyM+XBz/ZwsmK6rpvOHQGBETAoa1mTYpvw6U8Pk8+yBs/ZXDz+F78ckJvZ72t1iF8AKDOnPjO3gxRo93WpDMEdoM7l8O4e9zdEqeRgOGpClLN47Es97bD1U6tv2h98xeHj5Ux94NN3PvRZroFd+Tr+ybx28vi8e9Q9xBieOeO/HPGcFLzSvnrt/WUtfD1Mxlf0Oj8xcb9Rfxu0XYm9Qvjz1cORrWBLByHdOhkEgJqvlQV50DpYc+Y8K7RY0SzStJ7GpnD8FSFaebx2EH3tsPVMteYoRdP+kfeCItF8/GGAzz33R4qLRZ+d/lAbpsUi49349+/zukfwV3n9OGNnzKYEhfB1CHdzj5p7FyTOTa0/tLsB4tOcNcHm+gZ0onXbhyNrx2v3SZFDjrdw3DHlqztjF0BQyn1JfAW8J3W2uLaJgnA7IEA7aCHsQaix3h2GXIb+wuO8+jCn9m4/wiT+oXx9DVD6R0W4NA9fn3JAJIyCnn8y20M7xlM9+Baw07+XeCKv9d7fUlZJbe/t5Gqagv/mZNwap6kXYocBHuXmKyynM0mI8lT9m9vg+z9WvIacCOQqpR6Vik1wIVtEnB6SOrkEVNMrS0qOwaHtztWP8qN1mcUMm3eGlJyS3nh+mF8ePs4h4MFQAcfL/41cyQVVRZ+tWAr1ZYGUm1rqbZoHlqwlfT847x202j6RLTuYnbN1nWQqRtVkGJ6GJGDHCqjIhxjV8DQWi/TWt8EjAL2A8uUUmuVUrcqpdrx1xsXsVjMkFRApPm5ONu97XGVA+vMP/YYzw8YX2/NZvZbGwjr3IFv7p/M9ISezZoziA0P4MlpQ1i/r4jXlqfZfd2z3+3mxz15PHHV4LZTrrw5ampK5e40e2C0oqHN1sjugU+lVBhwC3AHsAX4FyaAJLqkZe1ZSQ5UnoC+55uf2+o8RuYas0dAVIK7W1IvrTXzlqfx0IKtjOzVhS/vmUivsE5Oufd1o6K4angPXvohlU2ZDexPbfXpxgO8uWofcyb0Zvb4dpYRVZ/QvuDdAfZ8C2VHZf7CxewKGEqpRcAqoBNwpdb6Kq31p1rrB4B23id2gZrhqD41AaONzmPsX2NSIDs45wPY2SqrLTz+xXZe+H4vV4/owfu3j6VLJ+dlvCil+Os1Q+ge7MdDC7ZQXFZZ77nrMwr5w1c7mBIXzh9/4die3G2at49Jr01Zan6WHoZL2dvDeFlrPUhr/YzW+pDtE1prz/162FrVBIzYKabsdVsMGOWlZgjBQ4ejSsoque3djXyafJAHL+jHizeMcMle2EF+vrw8aySHjpXx+0U76iwdcqDwBHd/uImeoZ14ddYou7Kx2pWug8BSZbbljWgHNbTcyN7/8wYppbrU/KCUClFKtZItzVqhwlRTGjkoylTkbIsBI2sD6GqPXLCXc/Qk019PIim9kOevH8b/XTLApWscRvUK4eGL4vjm5xwWbjrzd11szYiyaHhrzpj2nRFVn5qy4d2HmR6HcBl7A8adWuujNT9orY8AdzZ2kVJqqlJqr1IqTSn1eB3Pv6iU2mr9k6KUOmo9PkIplaSU2qmU2qaUusHeN9QmFKSYHeOUMnX122LA2L/G9J56jnN3S86wI/sYV89bQ/aRk7x761hmJPRskde957x+jO8Typ8X7yQj32TFVVVbeODjLewrOM6/bx5FbLjjGVntQuRg8yjzFy5nb8DwVjZfsZRS3kCDg7nWc+YBlwGDgFlKqTMGX7XWD2utR2itRwCvAF9anzoB/FJrPRiYCrxk28Np8wrSILy/+XtwdNuc9M5cY1bBdgx0d0tOWb4njxlvJOHjpVh4z8QWzULy9lK8eMMIOvh48dCCrVRUWXh6yR5WpuTzl2mDmdhXMqLq1WOkKV7ZhH3PhWPsDRhLgU+VUhcqpS4EPrEea8hYIE1rnaG1rgAWANMaOH+W9b5orVO01qnWv+cAeYAbt89qQRXHoTjLFFADa8DINqm2bUXlScje5FHrLz5cl8nt722kT0QAi+6bxIBuLR/Iugf789x1w9iefYyZ85N4e43ZCOmmcZIR1aDOEfD4AYd3JhSOszdgPAYsB+6x/vkB+E0j10QBtl+Ns6zHzqKU6g3EAj/W8dxYTG8mvY7n5iqlkpVSyfn5+Xa8jVagpiRIuE3AsFTC8Tz3tcnZspKhusIjAobFonlmyW7+8NUOzhsQyadzJ7i1PPilg7tx07hebD5wlHP6R/CHK2QS1y4yd9Ei7PqvbC0H8m/rH1eYCSzUWp9RwlMp1R34AJhTV0kSrfV8YD5AQkKC/ctlPVlNhtSpgGEdQz+WZapftgWZawBltmR1o7ySMv6yeBffbj/E7PG9+fOVgzwiA+mPvxjEkKhgfjGsu0e0R4ga9taSigOewcxFnPr6pbXu08Bl2YDtjGG09VhdZgL31XrNIOBb4Pda63X2tLNNKEgFFIRa/9MGR5vHYwchuo1kMGeuMfV+/Ft+Wkprzfp9RXy4LpOlOw5TrTV/uCKe2yfHeky1Vz9fb2aN7eXuZghxFnv7ce8AfwZeBM4HbqXx4ayNQJxSKhYTKGZi6lGdQSk1EAgBkmyOdQAWAe9rrRfa2ca2oTAVuvQ6vQ/CqYDRRjKlqirg4EYYfUuLvmxJWSWLtmTzQVImqXmlBPn5MGdiDDeN6yX1mISwk70Bw19r/YNSSmmtM4EnlFKbgD/Vd4HWukopdT/wPeANvK213qmUehJI1lovtp46E1igz1yxNAM4BwhTSt1iPXaL1nqr/W+tlSpIOT0cBWbnsA6BZuLbFYpz4EBS4+fZUl7Q5zzwD3H89XI2Q9XJFluwt/tQMR+uy2TRlmxOVFQzNCqY568bxpXDe9S7b4UQom72BoxypZQXplrt/ZgeQ6Nfy7TWS4AltY79qdbPT9Rx3YfAh3a2re2wWKAwHWKmnD52ai2Gi1Jrv/sN7P7G8evC4mD2Iuji4DqF/avNYy/XLdgrr6pm6Y7DfJCUSXLmETr6eHHl8B7MHt+b4T3bT3a2EM5mb8B4CFNH6kHgKcyw1BxXNardqik6GNbvzOOuXLyXnwJ9L4Cpz9p/zZH98MWd8PalMPsriOhv/7WZa035hgDn7z19sOgEH284wGcbD1J4vIKYsE784Yp4rh8d7dQaUEK0V40GDOsCvBu01o8ApZj5C+EKNZsm2Q5JgQkYOVuc/3qWajiyDwZMhQgHtjiJGAC3/Bc+vBbemQo3f2EWTzWmugoOrofhM5ve5loKS8tZsv0QX2/NITnzCF4KLorvyuwJvZnUNxwvL8+YyBaiLWg0YGitq5VSrW/D5daooGYNRq1v7MHRcKLALHjz9T/7uqYqzjbrIUIbSnarR/dhcNv38P7V8O6VMOsTUyyxIYd+horSZq+/KC2vInHXYb7emsOq1AKqLZq4yM48eukArhkZRY8uTvxvJIQ4xd4hqS1KqcXA58DxmoNa6y/rv0Q4rCDFTHB37nrm8VNrMbJNjSlnKcowj6F9m3Z9WF+43Ro0PrwOpr8LAy+v//zMNeaxCQGjvKqalXvzWfxzDst251JWaSGqiz9zz+nDVcN7MLBboMekxQrRVtkbMPyAQuACm2Oa07WfhDMUpprhqNoffLZrMVwSMJrQw6gR1ANu/Q4+uh4+vRmufq3+IafMNWZ+JrBr3c/XUm3RrN9XyOKtOSzZfojisipCAzowfXRPpo3owaheITLkJEQLsnelt8xbtISCtLrTTV21FqMwHXz8TAn15ggIgzmLYcGNsOgusw/5+HvOPMdSDZlJMPhqu2756cYD/DMxhdzicgI6eHPp4G5cOaIHk/uF4yurn4VwC3tXer+D6VGcQWt9m9Nb1F7VFB2sPeEN5ls8yvkBo2gfhMSClxM+gDsGwo2fwxe3w9LHTdA477ene0u5O6H8GMQ0Ph329up9PPnfXYyJCeGPvxjEhQO7ypoJITyAvUNS/7X5ux9wDZDj/Oa0YzVFB8PqCBjevq7ZSKkow8xDOIuvH0x/D755CFY+Z4LG1OdMQDo1f9Hw+ov/rMrgr9/u5tLBXXll1ig6+EhvQghPYe+Q1Be2PyulPgFWu6RF7VXtooO1OXvxnsViUmrjLnLePcFUDZ32qqkTlfQqnDxq5jX2r4YuvU8Pr9WhJlhcNqQbL88aKUNPQniYptYEjgMindmQdu9U0cF6vvEHR5u0VGcpyYGqsuZNeNdHKbjkr6Z0yI9PQXkxHNwAAy6r95L5P6Xz9JI9XD60G/+aKcFCCE9k7xxGCWfOYRzG7JEhnOVU0cF69mIIjoY934LWZ2dRNUVzU2oboxSc84jpaXz7CKDrHY7694p0nlu6hyuGdeelG0ZIsBDCQ9k7JOU5+2i2VQUpZy/YsxXcE6rL4XiB2WGsuZyRUmuPMXeAXxdImgf9zt4Rbd7yNF74fi9XDu/BizOGy/4PQngwu/51KqWuUUoF2/zcRSllX36kaFxN0cH65i/gzLUYzlCYDt4dIajOTRDrtfnAEa7/91oe+GQL27OO2XfR0Oth7vKz1l+8+mMqL3y/l2kjJFgI0RrY+y/0z1rrU58OWuujmP0xhDMUZ9dddNCWs9diFGVASIzdKbVV1RZeTExh+utJZB05yYo9eVz56mpmzk/ixz25WCyObXj48g+p/P1/KVwzMop/zhghwUKIVsDeSe+6/jXLJrrOUliTIdXQkJSzA8Y+u1Nq9xcc51efbmXrwaNcOzKKJ6YNBuDTDQd5e80+bns3mbjIztw5pQ/TRvago0/DayZeWpbCS8tSuXZUFC9cPxxvWa0tRKtg79e6ZKXUP5VSfa1//glscmXD2pVTRQcbGJLyDwHfAOcEDIvF9DAamb/QWrNgwwEuf3kVGfmlvDJrJP+8YQRBfr4E+fly5zl9+Ok35/PiDWY46TdfbGPyc8uZtzyNYycq67zfPxNNsLh+dLQECyFaGXt7CQ8AfwQ+xWRLJVJrD27RDAUp0DHo7KKDtpy5kVLpYbPrXWhsvacUlpbz2BfbWbY7l0n9wvj79OF0Dz67CqyvtxfXjIzm6hFRrEkrZP6qDF74fi/zlqcxI6Ent0+OpWdop1PB4pUf05iREM2z1w6TOlBCtDL2ZkkdBx53cVvar8JUM3/RWLqsszZSaiSldvmePB5duI3ik5X84Yp4bpsU2+iHu1KKyXHhTI4LZ/ehYt5clcGH6zJ5P2k/lw/tTmhAB95PymTmmJ48fc1QCRZCtEL2rsNIBKZbJ7tRSoVg9uG+1JWNazcKUu2qsURwNBze3vzXqyel9mRFNX9bsosP1x1gYLdAPrxjLAO7BTl8+/juQfxzxggevXQA767dz8frDlBSXsWssb3429VDJFgI0UrZOyQVXhMsALTWR5RSstLbGcpLTZZUQ/MXNYJ7wvE8qCyrf4GfPQrTwcv3jDId27KO8qtPt5KRf5w7p8Ty60sG4OfbvIJ/3YP9+e1l8dx/fj9+PniMiX3DJFgI0YrZGzAsSqleWusDAEqpGOqoXiuaoKGig7XVfMAXZzevaOCplFpvqi2af69I46VlqYR37shHd4xjUr/wpt+7DoF+vkyOc+49hRAtz96A8XtgtVJqJaCAKcBcl7WqPSmsZ1vWutim1jYrYJiU2mqL5u4PN5G4K5crhnXn6auHEtzJt+n3FUK0afZOei9VSiVggsQW4CvgpCsb1m6cKjpoR4kOZ6zF0Nr0MGKn8PzSPSTuyuUPV8Rz++RY2eJUCNEgeye97wAeAqKBrcB4IIkzt2wVTVGQ0nDRQVvO2EipNBcqj7O5NIQ3kjOYPb43d0xxcT0pIUSbYO/CvYeAMUCm1vp8YCRwtOFLhF0KU+0bjgLw6WjWajRnLYY1Q+qVrRYm9QvjT1cOavq9hBDtir0Bo0xrXQaglOqotd4DDHBds9oJi8Ws8rYnQ6pGM9diFB3cA0BFUAyv3ThaSokLIexm76dFllKqC2buIlEp9TWQ6bpmtRPF2WbFdQsFjJKySr5ftYZKvHlqzlSZ4BZCOMTeSe9rrH99Qim1HAgGlrqsVe1FTdFBe1JqawRHQ8r3Dm+kVG3RPPjJFq4/cZCqLr3o07WLg40VQrR3Do9HaK1Xaq0Xa60rGjtXKTVVKbVXKZWmlDqrtIhS6kWl1FbrnxSl1FGb5+YopVKtf+Y42s5WobF9vOsS3NP0Sk4UOfRSzyzZzfK9+YwPOYZ/NwdeTwghrFxWolwp5Q3MAy4GsoCNSqnFWutdNedorR+2Of8BzGQ6SqlQzH4bCZgFgpus1x5xVXvdoiC18aKDtdlupBQQZtcln248wH9W72PO+F6E7cqC0PMcb6sQot1z5YznWCBNa51h7Y0sAKY1cP4s4BPr3y8FErXWRdYgkQhMdWFb3aMgxb6ig7YcXIuxLqOQP3y1gylx4fzx/AioKHX9tqxCiDbJlQEjCrDN/8yyHjuLUqo3EAv86Mi1Sqm5SqlkpVRyfn6+UxrdogrT7E+prRHc0zzaETAOFJ7gng830TO0E6/eOAqfY/vNE/VUqRVCiIZ4Sk7lTGCh1rrakYu01vO11gla64SIiAgXNc1FThUdbGBb1rp0CgUf/0bXYhSXVXL7exuxaHh7zhiC/X1N0UFocB8MIYSojysDRjbQ0+bnaOuxuszk9HCUo9e2To4UHbR1aiOl+nsYVdUWHvh4C/sKjvPvm0cREx5gnijKAOVtVpYLIYSDXBkwNgJxSqlYpVQHTFBYXPskpdRAIARTaqTG98AlSqkQ694bl1iPtR2OFB2srZGA8fSSPaxMyecv0wYzsa9NldiiDAjpDd6y/kII4TiXBQytdRVwP+aDfjfwmdZ6p1LqSaXUVTanzsRsxqRtri0CnsIEnY3Ak9ZjbUdBCnYXHaytgYDxyYYDvL1mH7dOiuGmcb3PfNKOfbyFEKI+LkurBdBaLwGW1Dr2p1o/P1HPtW8Db7usce5WkGq+7TdlI6TgntZ9uctNfSmrRVuy+ONXOzi3fwS/vzz+zGtqqtT2HNvMhgsh2iuXBgzRgMJUx+cvapzaSCkHQmMpLa/iT1/t4Mst2YyJCeGVG0fiU7tG1IlCKC+WHoYQoskkYLhDTdHBmClNu95mLcbPx0N4cMEWDhad4KEL43jggn5nBwuw2cdbUmqFEE0jAcMdmlJ00JY1YPywfjN3bSshMrAjC+ZOYGxsaP3XnEqplR6GEKJpJGC4Q0GKeWzikFSuCqMrsGX7di4ZPJlnrhnWeOXZogxQXpJSK4RoMgkY7tCMlNplu3J5dOHP/E8Hc00fC31uHGXf1qpFGSZY+HRw+DWFEAIkYLhHQYq16GCk3ZeUVVbzzJLdvJeUyaDuQQT6xhLR4aj9daiK0mU4SgjRLJ5SGqR9KUh1qOhgSm4JV89bw3tJmdw+OZZF903EL6yX/RspaQ2FsgZDCNE80sNwh0L7MqS01ny0/gBP/XcXnTv68M6tYzh/gLVXEtwT0n6wbyOlk0eg/JgEDCFEs0jAaGl2Fh2s2SHv2+2HmBIXzj9mDCcy0GaRX3A0VB6HsqPgH9Lwa0pKrRDCCSRgtDQ7J7y/2pLNt9sP8fBF/Xnggn54edXqRdjui9FYwJCUWiGEE8gcRksraHwf77LKav6ZmMLQqOC6gwU4tpFSTUptSO/GzxVCiHpIwGhphak0VnTw/aT9ZB89yW8vH1h3sADHA0Zw9Bl1p4QQwlESMFpaI0UHj56o4NUf0zhvQMSZpclr6xQO3h0b3UgJkJRaIYRTSMBoaQUNFx18bUU6JeVVPDZ1YMP38fKC4Cj7exgSMIQQzSQBoyVZLA3u45199CTvrt3PtSOjie8e1Pj9GtlICYATRSatVgKGEKKZJGC0pOIsa9HBulNq//G/vQD8+hI7S4YE92w8YBTtM4+SUiuEaCYJGC2pgQypXTnFLNqSza2TYujRxd+++wVHQ8khqK6s/5xTazCkhyGEaB4JGC2pgTUYzy7dQ5CfL/ee2/CCvjMER4O2mKBRn6IMQEFIjENNFUKI2iRgtKR6ig6uTi3gp5R87j+/X+Nlym3Zk1pbk1LblK1ghRDChgSMllSQajZNsqn9ZLFonvluN1Fd/Jk9wcGFdcE9zWODASMdQmOb0FghhDiTBIyWVEdK7TfbctiZU8wjl/bHz9fbsfsFRZnHhtZiSEqtEMJJJGC0lPJSKMk5I0OqvKqaF77fy6DuQUwbHuX4PTt0gk5h9fcwTh6FE4WSISWEcAoJGC2ljgnvD5IyyTrSSAmQxjS0FkMypIQQTiQBo6XUSqk9drKSV5enMSUunClxEU2/b0NrMSRgCCGcSAJGSylMNRVjrR/er69M5+iJysZLgDSmwR6GddGepNQKIZxA9sNwVGE6bP0Y9q8GXW3/dUX7oEsv8PUj5+hJ3l69j6tH9GBIVHDz2hMcDeXFUHYM/GrdqyjdTIx36NS81xBCCCRg2OfkUdj5JWz9BLI2mJ5C9FjoGGj/PboPgwGXA/BiYgpaw68vGdD8ttmuxTgrYEiGlBDCeSRg1Ke6CtJ/hJ8/hj1LoLocIuLh4idh6AwI6t6k2+45XMwXm7O4bVIsPUOd8M3fdi1G18FnPleUAQMua/5rCCEEEjDOlrvTDDlt/xxKc8E/FEbPgeGzoMfIMxbdNcVz3+0hoKMP953vQAmQhpzqYdRai1FWDMfzJaVWCOE0Lg0YSqmpwL8Ab+A/Wutn6zhnBvAEoIGftdY3Wo8/D1yBmZhPBB7SWmuXNPR4gQkQWz+Gw9vAywfiLoURs8yjTwenvExSeiHL9+bz2NSBhAQ4554ERIKX79kT35IhJYRwMpcFDKWUNzAPuBjIAjYqpRZrrXfZnBMH/BaYpLU+opSKtB6fCEwChllPXQ2cC6xwekOLMuDVMWCpgu7DYepzMPR6CGhgt7smqCkB0j3Yj1snxTjvxvVtpCQBQwjhZK7sYYwF0rTWGQBKqQXANGCXzTl3AvO01kcAtNZ51uMa8AM6AArwBXJd0sqQWLjgD6Yn0XWQS14C4Kut2WzLOsYL1w9zvARIY+pai3EqYEgdKSGEc7gyYEQBtgPrWcC4Wuf0B1BKrcEMWz2htV6qtU5SSi0HDmECxqta6921X0ApNReYC9CrV6+mtVIpmPxw0661w8b9RcxbnsaKvfkM6h7EtaOinf8iwdEmzddWUQYEdocOAc5/PSFEu+TuSW8fIA44D4gGflJKDQXCgXjrMYBEpdQUrfUq24u11vOB+QAJCQmumd9oAq01K1Ly+ffydDbsLyI0oAOPXNKf2RNi8G5qCZCGBEdDcY7J7PK2/kolpVYI4WSuDBjZQE+bn6Otx2xlAeu11pXAPqVUCqcDyDqtdSmAUuo7YAKwCg9WbdF8t+MQry1PZ9ehYnoE+/HElYO4YUwv/Ds4eRjKVnC0WURYevh01lRRBsRd7LrXFEK0O64MGBuBOKVULCZQzARurHXOV8As4B2lVDhmiCoD6APcqZR6BjMkdS7wkgvb2iwVVRYWbcni9ZUZ7Cs4Tp+IAJ6/fhhXj4iig08LVF+xXbwXHA3lJSYlWFJqhRBO5LKAobWuUkrdD3yPmZ94W2u9Uyn1JJCstV5sfe4SpdQuoBp4VGtdqJRaCFwAbMdMgC/VWn/jqrY21YmKKhZsOMibqzI4dKyMIVFBvHbTKC4d3M01Q0/1qb2RUk0NKRmSEkI4kUvnMLTWS4AltY79Pk3aSgAACchJREFUyebvGvg/6x/bc6qBu1zZtuYor6pm/soM3l6zjyMnKhkbG8qz1w3jnLhwVDMX9jVJ7Y2UJKVWCOEC7p70bpVe+SGNV5enccHASO49ry8JMaHubVDHzuAfYtPDkJRaIYTzScBw0PHyKt5P2s/Uwd14ffZodzfnNNsy50Xp0LmrY8URhRCiEbIfhoMWbDxIcVkVd53rYcM9tov3ivbJcJQQwukkYDigstrCW6syGBsbysheIe5uzpmCo8+cw5CAIYRwMgkYDvjvthxyjpVxt6f1LsAEjLJjUJILJYckYAghnE4Chp201ryxMoP+XTtzXv9IdzfnbDVrMfZb1zZKwBBCOJkEDDutTMlnz+ES5p7TF6+WXGNhr5q1GPtWmkcJGEIIJ5OAYafXV6bTLciPq4b3cHdT6lbTw8iQgCGEcA0JGHbYevAo6zKKuH1ybMuU+miKzl3Nxk9HMyEgAvyC3N0iIUQb46Gffp5l/k/pBPr5MHNsz8ZPdhcvbwiy9n6kdyGEcAEJGI3YX3Cc73YcZvb43gT6+bq7OQ2rmceQgCGEcAEJGI14c1UGvl5e3OLMbVVdpWYeQ6rUCiFcQAJGA/JLyvl8UxbXjY4iMtDP3c1p3KmAITWkhBDOJwGjAe8n7aey2sIdU1rJEM+pgNFK2iuEaFWk+GA9TJHBTC4Z1JW+EZ3d3Rz7xE+D4wXQfbi7WyKEaIMkYNTj040HOXaykrvObUXzAQFhcO5v3N0KIUQbJUNSdaistvDW6n2MjQlllKcVGRRCCDeRgFGHb7cdIvvoSc8rYS6EEG4kAaMWrTWvr0wnLrIz5w/wwCKDQgjhJhIwavkptcBaZLCPZxYZFEIIN5GAUcsbK9PpGtSRaSOi3N0UIYTwKBIwbGzLOsra9ELPLjIohBBuIp+KNt5YmUGgnw+zxvZyd1OEEMLjSMCwMkUGD3FzaygyKIQQbiABw+o/qzPw8fLi1okx7m6KEEJ4JAkYQEFpOZ8nZ3HtqCgig1pBkUEhhHADCRjA+2v3U1Ft4c5zZKGeEELUp90HjOPlVbyXlMnF8a2oyKAQQrhBuy8+WFpexeR+4dw2WfaQEEKIhri0h6GUmqqU2quUSlNKPV7POTOUUruUUjuVUh/bHO+llPqfUmq39fkYV7Sxa5Af824axejeUmRQCCEa4rIehlLKG5gHXAxkARuVUou11rtszokDfgtM0lofUUrZFm96H/ib1jpRKdUZsLiqrUIIIRrnyh7GWCBNa52hta4AFgDTap1zJzBPa30EQGudB6CUGgT4aK0TrcdLtdYnXNhWIYQQjXBlwIgCDtr8nGU9Zqs/0F8ptUYptU4pNdXm+FGl1JdKqS1KqResPZYzKKXmKqWSlVLJ+fn5LnkTQgghDHdnSfkAccB5wCzgTaVUF+vxKcAjwBigD3BL7Yu11vO11gla64SIiIiWarMQQrRLrgwY2UBPm5+jrcdsZQGLtdaVWut9QAomgGQBW63DWVXAV8AoF7ZVCCFEI1wZMDYCcUqpWKVUB2AmsLjWOV9hehcopcIxQ1EZ1mu7KKVqug0XALsQQgjhNi4LGNaewf3A98Bu4DOt9U6l1JNKqausp30PFCqldgHLgUe11oVa62rMcNQPSqntgALedFVbhRBCNE5prd3dBqdISEjQycnJ7m6GEEK0KkqpTVrrBLvObSsBQymVD2Q24xbhQIGTmtPayHtvv9rz+2/P7x1Ov//eWmu7sobaTMBoLqVUsr1Rtq2R994+3zu07/ffnt87NO39uzutVgghRCshAUMIIYRdJGCcNt/dDXAjee/tV3t+/+35vUMT3r/MYQghhLCL9DCEEELYRQKGEEIIu7T7gGHPJk9tmVJqv1Jqu1Jqq1KqTa98VEq9rZTKU0rtsDkWqpRKVEqlWh/b7E5a9bz/J5RS2dbf/1al1OXubKOrKKV6KqWW22zW9pD1eJv//Tfw3h3+3bfrOQxryfQUbDZ5AmbZbvLU1iml9gMJWus2v4BJKXUOUAq8r7UeYj32PFCktX7W+oUhRGv9mDvb6Sr1vP8ngFKt9d/d2TZXU0p1B7prrTcrpQKBTcDVmCrYbfr338B7n4GDv/v23sOwZ5Mn0UZorX8Cimodnga8Z/37e5h/SG1SPe+/XdBaH9Jab7b+vQRT3y6KdvD7b+C9O6y9Bwx7Nnlq6zTwP6XUJqXUXHc3xg26aq0PWf9+GOjqzsa4yf1KqW3WIas2NyRTm1IqBhgJrKed/f5rvXdw8Hff3gOGgMla61HAZcB91mGLdkmb8dn2Nkb7b6AvMAI4BPzDvc1xLaVUZ+AL4Fda6+L/b+9+QmyMwjiOf3/+JEaxwEYhbFBMKQtDKWVN+ZN/TVYWNnYiUsoWG8WCIkP+ZCJLFpNZaEhEWMliLGY2UhRp5rE452poTGeu5t7x3t9nc+8973vfztPp3ue+573veUZuq/r4jxL7uMe+1RNGSZGnSouIj/lxEOgmTdO1koE8x1ub6x1scn8aKiIGImIoIoZJJQQqO/6SppO+MLsi4m5ubonxHy32esa+1RNGSZGnypLUli+CIakN2AK8HvtdlXMf6MzPO4F7TexLw9W+LLNtVHT8JQm4BLyNiDMjNlV+/P8Wez1j39L/kgLIfyU7B0wFLkfE6SZ3qWEkLSWdVUCqo369yvFLukGq8DgPGABOkqo+3gIWkZbH3xkRlbww/Jf4N5GmJAL4ABwcMadfGZI2AI+BV8Bwbj5Gmsuv9PiPEftuxjn2LZ8wzMysTKtPSZmZWSEnDDMzK+KEYWZmRZwwzMysiBOGmZkVccIwmwQkbZL0oNn9MBuLE4aZmRVxwjAbB0n7JPXl+gEXJU2V9EXS2Vxr4JGk+XnfdklP8uJu3bXF3SQtl/RQ0ktJzyUty4efLemOpHeSuvIdumaThhOGWSFJK4BdQEdEtANDwF6gDXgWEauAHtId1ABXgSMRsZp0l22tvQs4HxFrgPWkhd8grSJ6GFgJLAU6Jjwos3GY1uwOmP1HNgNrgaf5x/9M0mJ1w8DNvM814K6kOcDciOjJ7VeA23ntroUR0Q0QEd8A8vH6IqI/v34BLAF6Jz4sszJOGGblBFyJiKO/NUon/tiv3vV2vo94PoQ/nzbJeErKrNwjYLukBfCrHvRi0udoe95nD9AbEZ+BT5I25vb9QE+ueNYvaWs+xgxJsxoahVmd/AvGrFBEvJF0nFShcArwAzgEfAXW5W2DpOsckJbLvpATwnvgQG7fD1yUdCofY0cDwzCrm1erNftHkr5ExOxm98NsonlKyszMivgMw8zMivgMw8zMijhhmJlZEScMMzMr4oRhZmZFnDDMzKzIT6RtvgRFCNTAAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEGCAYAAAB2EqL0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXxU1dnA8d+THbJBSICQAAl72HeQXRHEDcTdautSl7pUrW/drVqtrW9rF62oxYpV66sioqDixo4CSlCQBBCBgAQICRBCFrKf948zgRBCMklmMpmZ5/v5zGdm7tx759wMzDNne44YY1BKKaXqE+DpAiillPIOGjCUUko5RQOGUkopp2jAUEop5RQNGEoppZwS5OkCuEpsbKxJSkrydDGUUsqrrF+//qAxJs6ZfX0mYCQlJZGamurpYiillFcRkd3O7qtNUkoppZyiAUMppZRTNGAopZRyis/0YdSmrKyMzMxMiouLPV0UtwsLCyMxMZHg4GBPF0Up5aN8OmBkZmYSGRlJUlISIuLp4riNMYZDhw6RmZlJcnKyp4ujlPJRPt0kVVxcTLt27Xw6WACICO3atfOLmpRSynN8OmAAPh8sqvjLdSqlPMfnA4ZSysvsXA5713u6FKoWbg0YIjJNRH4Qke0i8kAtr/9dRDY4bttE5Ei1164VkR8dt2vdWU53OnLkCC+88EKDjzvvvPM4cuRI/Tsq5UuK8+Dta+Dd66Gi3NOlUTW4LWCISCAwCzgX6AtcJSJ9q+9jjPmNMWawMWYw8E9gvuPYGOAxYBQwEnhMRNq6q6zudLqAUV5e93+GRYsW0aZNG3cVS6mWaf1/oDQfjuyGLQs9XRpVgztrGCOB7caYncaYUuBtYEYd+18FvOV4fA7whTHmsDEmF/gCmObGsrrNAw88wI4dOxg8eDAjRoxg/PjxTJ8+nb59bey86KKLGDZsGP369WP27NnHj0tKSuLgwYPs2rWLlJQUbrrpJvr168fUqVM5duyYpy5HKfcpL4W1L0HXcRDTHVb/E3RF0BbFncNqE4A91Z5nYmsMpxCRrkAysLSOYxNqOe5m4GaALl261FmY33+YzuZ9R50sunP6dorisQv71bnP008/TVpaGhs2bGD58uWcf/75pKWlHR/+OmfOHGJiYjh27BgjRozgkksuoV27died48cff+Stt97i5Zdf5vLLL+e9997jmmuucem1KOVx6fMhfx9Mfw6O/AQf3wO7V0PSWE+XTDm0lE7vK4F5xpiKhhxkjJltjBlujBkeF+dUskWPGzly5ElzJZ577jkGDRrE6NGj2bNnDz/++OMpxyQnJzN48GAAhg0bxq5du5qruEo1D2NsjSKuD/Q4GwZdBa3b2W2qxXBnDWMv0Lna80THttpcCdxe49hJNY5d3pTC1FcTaC7h4eHHHy9fvpzFixezZs0aWrduzaRJk2qdSxEaGnr8cWBgoDZJKd+zYykcSIMZs0AEQlrDiJtgxdOQ8wPE9fZ0CRXurWGsA3qKSLKIhGCDwim9WCLSB2gLrKm2+TNgqoi0dXR2T3Vs8zqRkZHk5+fX+lpeXh5t27aldevWbN26lbVr1zZz6ZRqIVb/EyI6woDLTmwbeRMEhcGa5z1XLnUStwUMY0w5cAf2i34LMNcYky4iT4jI9Gq7Xgm8bcyJ3i1jzGHgSWzQWQc84djmddq1a8fYsWPp378/995770mvTZs2jfLyclJSUnjggQcYPXq0h0qplAdlbYKdy2DULRB0ojZNeCwM/hlsfBvyD3iufOo4MT4yCmH48OGm5gJKW7ZsISUlxUMlan7+dr3KR8y/BbZ8CPekQ6sao+cPbofnh8OE38JZjzRfmSrK4NB2yN5ibzlbbNNYz6lwzlPNV45mICLrjTHDndnXp5MPKqVauLxMSJtn+ytqBguA2B7Q53xY928Y9xsICT91n6aoKIfcDEdQ2ArZmyF7qw0WlWV2HwmAtskQGAJrX4TRt0H0KYM2/YIGDKWU53z9kh0hNfrW0+8z5k7Y+hF89yaMutk177vrS/j0QVtrqCg5sb1NV2jfF3pPg7gUaJ8CsT0huJUd6vvsIFj3Mpz9uGvK4WU0YCilPKM4D1L/A/0ugrZdT79fl1GQONJ2fo/4JQQENu19j+6Hudfa2sqomx2BoQ/E9obQiNMf16YLpFwIqa/ChPvsSC4/owFDKeUZ375u04CccUf9+475Ncz9ue3r6HdR49+zsgLm3wRlRXD9ooYP1x19G2xeAN+/DcNvaHw5TudwBnz1D6hsYB6ttkkw4d56d2sqDRhKqeZXUWb7A5LGQ8LQ+vfvcz7EdIPVz0HfGXauRmOs+hvsWmXnezRmbkfnURA/2JZ96HUQ4OKBpp8/Ats+g4j2DTuu40DXluM0NGAopZpf2nw4uhcu+Idz+wcEwhm3w8f/Az+tha5nNPw9d6+G5X+0cz0GX93w48EGqtG3wfs3w86ldla6q+z7zvbVTHoIJt3vuvO6UEtJDaIcIiLqaENVyhfUTAPirEE/g1YxtpbRUEWH4b0bbdPNBX9vfA0FoN9MiOhgaxmutOyPdqRYXQMAPEwDhlKqee1cDgc22b6LhjTphLS2s79/WAQHT825dlrGwAe3QmEOXPoqhEY2uMgnCQqxw4C3L7ajrFxhzzfw4+cw9i4Ii3LNOd1AA4abPfDAA8yaNev488cff5w//OEPTJ48maFDhzJgwAAWLFjgwRIq1cxWP2d/oQ+8vOHHjmhEupC1L8K2T2HKk9BpcMPfszbDr4fAUDss2BWW/gHC42Cki4YNu4n/9GF88oBNQeBKHQfAuU/XucsVV1zB3Xffze2329yKc+fO5bPPPuPOO+8kKiqKgwcPMnr0aKZPn67rcivfl5VmEw1OfvTkNCDOioizmWw3/B+c+Yh9Xpe938IXj0Lv823qEVcJj7UBb8NbcNbvoHVM48+VsQoyVsA5f3T9xEQX0xqGmw0ZMoTs7Gz27dvHxo0badu2LR07duShhx5i4MCBnH322ezdu5cDBzRXjvIDa56H4HAYdn3jz3HG7VBRCt/Mrnu/4qMw7wZbm5nxfNP6LWoz+lYoPwbfvtb4cxgDy56CyHj3DNN1Mf+pYdRTE3Cnyy67jHnz5pGVlcUVV1zBm2++SU5ODuvXryc4OJikpKRa05or5VPy9sKmd2HEjU37RR7bE3qfZ2dcj/tN7RPojIGP7razs6/7uGnvdzod+kHyRPjmZdsfExjc8HPsWAo/rYHznrGzyVs4rWE0gyuuuIK3336befPmcdlll5GXl0f79u0JDg5m2bJl7N6929NFVMr9vn4JTKVrRgGN+TUcy4UNb9b++ndvQNp7cOaDjRuC66zRt9nhwVs+bPixxti+i+guMPQXri+bG2jAaAb9+vUjPz+fhIQE4uPjufrqq0lNTWXAgAG8/vrr9OnTx9NFVMq9io/C+v9A34vs0Nam6jIaEkfAmll29nZ12Vtg0X321/+4e5r+XnXpOdVOKGzMENttn8K+b2HivY3rz/EA/2mS8rBNm050uMfGxrJmzZpa9ysoKGiuIil/t/DXdohrQwQEQdI46HexnaUd6ORXyLevQ8lRWzNwBRFHupBfwNaPoa9jiZ3SInj3epsT6uKXm553qj4BATDqVvjkXshMhUSnsoRDZSUsfcpmwR10lXvL6EIaMJTyR8V58N1/bZqLhqTIKMm3s7S/fd0OA02ZDv0vhi5nnP7LuSoNSNdxzqUBcVafC2xtZfVzNimgCHz6gF274pr5ENnBde9Vl8FXwdIn7TVe+opzx2xZaOeizJzduL4PD9GAoZQ/2r3G9idMeQKSxzfs2LJjdpJZ+vt2eGvqK3Z51b4zbPBIHHnyhLz09+FoJpz/V9deQ0Cg7Wxe9FvY87XtS/j2NdsR3mOya9+rLqGRtg/i65cg74n618qorIDlf7LZcQdc2jxldBGfDxjGGL+Y3+ArKyeqZpKx0k48SxzR8GODW9ng0HcGlBbatvi0+baP4pt/QVSC7avofzF0GmprALG9bXu/qw2+2qbU+OIxOJBug9WZD7v+feoz8iZY+4Jd6OnsxzDGsHRrNq+vsQNa2keG0j4qlPaRYQzO/ZxBOVs5eO5sIiogzIt6kn06YISFhXHo0CHatWvn00HDGMOhQ4cICwvzdFGUt8hYadeZCG7iv5mQcOh/ib2V5MMPn9jgse5lWDvL1jwKsmD6P12f2RXskNoRN8LKP0NYtG0S8kQTT9sk6HM+Zv2rLO9wLX9bnsmmvXkktGlFTHgIW7OOcrCgFCrLWRzyDFvownnvt8a8/ylRYUG0jwqzQSUylI7RregWF073uAh6xEUQ3brlNFn5dMBITEwkMzOTnJwcTxfF7cLCwkhMTPR0MZQ3KDxk289dvUZ2aKSd/Tzwcjh2xOZ8SpsPxV1gQCPSgDhr5M12pvS4e+wiRx5gjGF9xysZvuVDPn/7OfKiL+TPlw5k5pAEggNtoKyoNBSt/Q+Rnx9g0/gX+d+owWTnF5OdX0L20RKy84tJ3Z3LgaP7Kas40WIQGxFCt7gIusdF0D0unO7tbSDp1KYVgQHN+0NYfKUpY/jw4SY1NdXTxVCq5du8wI4uuuFzW8tQjWaMYfGWbJ5dso20vXl81up3xIdDq7tTCQ6qMQigvBT+OQzC28FNy04787yi0rDncBE7cgrsLbuQHTkFbM8p4EhR2fH9QoMCSI61AWRI5zbcOL5bo65BRNYbY5wa3uXTNQylVC0yVtr0HK4cseRnjDF8sfkAzy75kfR9R+narjV/uXQQ3QPvJWjBrbBr+akd79+9AXk/1ZtePTBASIoNJyk2nMkpJ4/0OlxY6ggijmCSU0ja3jzyisoaHTAaQgOGUv4mYyV0HeNVwzlbCmMMn28+wHPVAsUzlw3iosGdCAoMgPJLYMnjdoht9YBRVgwrn7Er9jVhBFdMeAgx4TGMSDo51UllZfO0FGnAUMqf5GfBwW0w5BpPl8TrLPshm798+gOb9x8lqV1r/nrZIGZUBYoqQaG2E37ZU5CzDeJ62e3rX4X8fXDxv1yfBBEIaKa+DC8a0KWUarKMVfY+eYJny+FFjhSVcs87G7j+1XUUlZbzt8sHsfieiVwyLPHkYFFlWI21MkoL7VriSeO9/u+uNQyl/EnGCjv8tONAT5fEK3yx+QAPvb+J3MJS7pzckzvO7EFIUD2/syPiYOBlsPEtOxLt29ehMBuueKN5Cu1GGjCU8ie7VtkUHe7OseTljhSV8vjCdD7YsI8+HSN59boR9E+Idv4Eo261qVfWzILUOXbt8i6j3VfgZqIBQ3m3ygq7nkDafDvjePStrktw52tyd0PuLvtl1kIZY/gkLYvIsCDG96xnNT03+Tw9i4feT+NIUSl3Te7J7c7UKmrq2N82Qa16xj4/8yHXF9QDNGAo71NZCZnf2CCxeYGdSRzUyi6bueIvMOTn0KqNp0vZ8uxq2f0XP2Tl87sFaXyTcRiAc/t35NEL+xIf3TwLC+UWlvL4h+ks2LCPlPgoXrthBP06NaBWUdPo2+zfvPf5kDDMdQX1IA0YyjsYA3vX20R26e/bRHOBodBzis1Z1PMcOLwT/jXeLt058T5Pl7jlyVgFrWOhfYqnS3KSgpJynl28jTlf7SIyLIinZvbnSFEZzy35kZXbcvjNlF5cNyap9g5mF/k0LYtHPrC1irvP7sltkxpRq6ip1zSY/JjXJRisiwYM1XIZA/s3Qvp8GySO/AQBwXYc++THoPe5EBZ1Yv/4gXbpzjWzYNSvTn7N3xlj518kj3fLsM7GMMawaFMWT360mayjxVwxvDP3n9uHmPAQAC4c2IlHF6bxh4+38N63e/njzP4M6dLWpWU4XFjKYwvT+XDjPvrGR/H6DSPp28lF/24CAmC8mxdwamYaMFTLtPVj+PwRW2sICIJuk2DiA9Dn/LqbmybcCy+faZPfjf+f5ipty3doh50H0EKao3bmFPDYwnRW/XiQvvFRzLp6KMO6nhwMurRrzavXjeDTtCwe/zCdi19czc9GduG+c/q4JCHfp2n7eeSDNPKOlXHPlF7cOqn78bxPqnYaMFTLc2QPzL8F2nSGC5+1i/S0jqn/OLDpLnpOhdXPw8hb7Mpryg6nBUjybMA4VlrBrGXbmb1yJ6FBATx+YV+uGd31tM1NIsK5A+IZ3yuOv32+jf+szuCz9CwePj+FiwYnNCgL9ZGiUtbsOMSX2w+yeschMg4W0j8hijd+OYqUeK2NOkMDhmpZjIGP7raL+1z1VuPWf55wH7xytl3YZ+xdLi+iV9q1CiI7QbvuHivC4s0HePzDdDJzjzFzSAIPnteH9pHOpVePCA3i0Qv7cvHQBB75II3fvLORuesyefKi/vRoX/uPguKyCtbtOmwDxPZDpO3LwxgIDwlkVLd23DAumStHdNZaRQNowFAty8a3YPtiOPfPjQsWAJ1HQPez4KvnYMRNds0Ef1ZZaTu8e5ztkf6LPYeL+P2H6Szekk3P9hG8ffNoRndr16hz9U+IZv6tY3hr3U/87ydbOffZlfxqYnduP7MHQQHCpr15rN5xiC9/PMj6n3IpLa8kKEAY2qUtd03uybgesQzq3EaDRCNpwFAtR36WXZO582j7Rd8UE++HOefYHD5n3O6a8nmrnC1QdNBt/ReVlYZDhaVk5RWTdbSYrLxj7D/+uJhvf8olQISHzuvD9WOTm/xlHRAgXD2qK1P7duRPi7bwz6XbmZu6h6LSCvKLywFIiY/iF6O7MrZnLCOTYggP1a86V3DrX1FEpgHPAoHAv40xT9eyz+XA44ABNhpjfubYXgFscuz2kzFmujvLqjzMGPjoHigvgRmzmr46W5fR9gvyq2dh+A12WVF/dTx/VN1rd1dUGgpLyykqqaCgpJyi0nJ7X1JBYWk5hSUVFJSUceBoyfFgkJVXTHZ+8UkL/gAEBQgdosLoEBXKzCEJ3Dm5p8vnU8RFhvK3KwZz6fBEXly+g4Q2rRjTI5Yx3dsRGxHq0vdSltsChogEArOAKUAmsE5EFhpjNlfbpyfwIDDWGJMrIu2rneKYMWawu8qnWpj0+fDDxzDlCYjt4ZpzTrwf/nO+zeUz6hbXnNMbZay0zXvVVqM7VlrBQ+9vYv3u3OOBobis0qnTtQoOJD46jA5RYYxKjqFDdNjx5/HRYXSMDiM2PLTZMqiO6R7LmO6xzfJe/s6dNYyRwHZjzE4AEXkbmAFsrrbPTcAsY0wugDEm243lUS1V4UFYdC90GgqjXdh8lDQOuo6FL/8Bw66zqaf9TWUF7PoS+s04vim/uIxfvpbKul2HOa9/PNGtg4kIDaJ1SKDjPojw0EDCQ4IID3U8Dg1yPLf7NGR0kvId7gwYCcCeas8zgZrrQfYCEJGvsM1WjxtjPnW8FiYiqUA58LQx5oOabyAiNwM3A3Tp4pm1fJULfHIfFB+1TVGBLv4nOeFeeOMimwhuxC9de25vkPU9lOQdH06bW1jKda9+Q/q+ozx35RAuHNTJwwVU3sTTQwWCgJ7AJOAq4GURqZqV1dWxzuzPgH+IyCnjAY0xs40xw40xw+PiPJOoTDXRlo8g7T2byqNDX9efv9skSBwJX/7drqnsbzJW2vvk8WTnF3Pl7LVsycrnpWuGabBQDebOgLEX6FzteaJjW3WZwEJjTJkxJgPYhg0gGGP2Ou53AsuBIW4sq/KEY7nw8T3QYQCM+4173kPE9mXk7bFDdv1NxkqI7U1meRSXv7SGPblFvHrdCM7u26H+Y5WqwZ0BYx3QU0SSRSQEuBJYWGOfD7C1C0QkFttEtVNE2opIaLXtYzm570P5gs8esf0XM5537/rSPSbb/pFVf4WKMve9T0tTUQa713A0fjSXv7SGQ4WlvPHLUYztoR3EqnHcFjCMMeXAHcBnwBZgrjEmXUSeEJGqIbKfAYdEZDOwDLjXGHMISAFSRWSjY/vT1UdXKR+wfTFs+C+Muxs6uXkwXFUt48hu+H6ue9+rJdn7LZQV8of0OErKK3n75tGn5GtSqiHEGFP/Xl5g+PDhJjU11dPFUM4oPgovnGFnYN+yCoKdSw/RJMbAvyZAaQHcvs71nest0P6FTxD/7V+ZGvwaL9x09mlTaCj/JiLrHf3F9fJ0p7fyR4sft+tZzJjVPMECTtQyDu+0new+bu3OQ/y0/lN+lCReuXWqBgvlEr7/M0u1LBmrbFLA0bdD55HN+969z4MO/WHlX+yiNp5c19oYKMm3KTsKD0Jhzsn3ptKmZ49seOf0sh+yufONNawP2kbZkBsIj/HzXFrKZTRg+LvMVPsF1Xua+9+rtBAW3gFtk+GsR9z/fjUFBNh5Ge9eaxdkaq6V0Na/ZtcdL8xx3A7Z+4qS2vcPiYTyY5C9GX6xoEGBbdGm/dz19ndcGpNJSH4ZIb3PctFFKKUBQ634s10f+96dTc/fVJ+lT0HuLrj2I89lkE2ZDnF9YOUz0O9i91/zV8/BF7+DyHiI6ADh7aF9P7v+eHgshMc5brF2+dTwWJv36rv/woLbbTkn3e/UW81bn8l98zYypEtbHut2CNYGQNcz3Ht9yq9owPB3BQfsfIgDaXaJU3fZ8w2sfQGG/7LeJHhuVVXLeO+XsPVD6Duj/mMaa8NbNlj0mwmXvNKwJrDBV8POFbDiaZviJGnsaXctKCnn+aXbeWnFDsb2aMfLvxhO2Bu/h05DICzaBReilKWd3v6uMMfeV80IdoeyYvtrOToRpvzefe/jrH4zoV1PW7uqdC7hXoNt+8xec/JEmPmvhveXiMAFf7PNd+/daJuxaiivqOS/a3cz6S/LeGnFDi4Zmsgr146gtSmGvaktZjlW5Ts0YPgzY5onYGxeAAe3wfl/hdBI972PswICYcJvba1q2yeuP/+eb2DutdCxP1z5ZuOTHoZGwmWv2o7xBbfZzwswxrB06wGmPbuKRz5Io1tsBAtuH8tfLx9EWHAg/LQWKsshyYM1OeWTNGA0l/IS2+nbkhQfgYpSCAyB3auhotw977NjiW2f7zHFPedvjP6X2l/vK/73+BexS2RvhTcvg6h4uPo9SgJb89rqXby+Zhd7jxxr+PniB8HUP8C2T2HtC2zed5RrXvmaG/6TSnlFJS9dM4x3bhnNoM5tThyzayUEBNs1QZRyIe3DaA6VFfDquTb53S0rPDucs7oCR+2i51TY+hHs3wCJTs3fcV5lJexYapdMdXcHc0MEBtm+jAW32VFT5//Ndjg3RV4m/PdiW6O4Zj5rDgTw8Pur2HnQ/lB4dEE6KfFRTElpz+SUDgxIiHZuzYiRN1O8bSlBnz/KAyWV/BTWm8cu7MvVo7oSElTL3zRjJSSOgJDwpl2PUjVowGgO374Ge9fbx5vmwaArPFueKoWO5Uf6zbQBI2OF6wPGgTTb7NW9BQ7vHHSV7fRf9kdbw7rgH5ByQePOVXQY/nsJlOSTd8UHPLnkKPPWb6ZLTGtev2EkCW1bsXjzAZZsyeb5Zdt5bul22keGMjmlPWendGBsj1jbnFRDYUk5/1q5k7nbLmN+wDpei3yRgF+tIrrtadbEPnYE9m+ECfc17jqUqoMGDHcrOgxLnoSu46A4D5b/Cfpf7N5ke84qcASM9n3tUM+MlXaymCvtWGrvu5/p2vO6QkAAjL8Hep0D7/8K3rkaBl4J5z4NrRqQc6m0CP7vCszhDFaOfIm7/3uI/OJybpvUnTsn9zweCLpPjOCWid3JLSxl2Q/ZLN5ygA837uetb/YQFhzAuB5xTOnbnjP7tKddeCjvpu7hr19sIye/hPMHdiNgwBzazr8YFv8WLp1jO8Zr2r3aTvrz5Eg05bM0YLjbsqdsoDjvz5C7G96+yqbZHvoLT5fsRId3RHv7BbP+NdvX4sqV6XYsscEosqPrzulqHfrBjUtg1TN23kPGCpj+PPQ8u/5jK8rg3eswmev4e9tHeG5ZKMO6RvDHmQPo3bH2Dv624SFcPDSRi4cmUlJewdc7D7NkywEWb7FBBCA2IpSDBSUM7dKGl64ZdiJpYO5DsPRJu87HsGtPPXnGSggKs01SSrmYBgx3ytoEqXNgxE32S6l9X5tme8WfYeAVnl8ytOAASCC0irFDML9+yc78rmPMf4OUFtoRO96wnnZQCJz5EPSaBh/cCm9eAkOvhXOeOv3ILmOoWPBrAn/8jEcrbuSD3AH8cWYKV47o7PR61qFBgUzoFceEXnE8Pt2wNSufxZsPsDEzj5lDEjhvQMeTl0Mddw/sWgWf3G9Tq7RPOfmEGSttZ7en/20pn9SCeiF9jDGw6D7btHHmg3abCJz1sF3M59vXPVs+sE1S4bG2aabrWJAA1w6v3b3ajsJqif0Xp5MwFG5eAWPvsp/Ri2Ns/qta7Jt3H4Hfv8Xfyi4lr+81LPmfifxsVBeng0VNIkJKfBS/ntyTf187nPMHxp+6dnZAAMycDaER8O51tjmsSuFByE7X4bTKbTRguMumefDTapj82Mnt4d0nQ5cz7GI+ZY0YZulKhTm2OQqgVRs7hHNX7V+OjbJjqW0e6eJl6SmCw2DKE3DDZxAQBK9dYH/RO76ccwtL+fDFh+iUPpv5QdMY9os/8dxVQ2gf2UyZdyM7wMWzIecH+LRa2pCqzy55YvOUQ/kdDRjuUFJgU0J0GgJDfn7yayJw5sOQv982V3lSQbbNbVQlabyddFb9V2tTbF9iay7BrVxzvubWZRT86ksYeQt8/RLlL4zl/YXzeeaZJ7nwwCy2xpzFub99g4m929d/LlfrfpZd1vbb1+2PE7C1w5BI++9OKTfQgOEOK/9iA8J5z9Q+9yB5vO0z+PLvnp3MV72GAfaXaWUZ7Fnb9HPnZcLBH7yrOaoWuWXBvNnudn4f8zRZuUeZvv4Gfm9mUdhpDH1ue5tWYSGeK9yZD0PnUfDh3XBohw0YXc/wi8WhlGdowHC1g9thzSybPK6uOQ1nPmK/sL+Z3Xxlq84YRw0j7sS2LqNtE4wr+jF2LLP3XhgwikrLWbBhL7/8zzpGPLWYh99PY2VZCh+NmUfRgJ8T1H0i4b94x/Mdy4FBjqSGAfDWVXBou+aPUm6lP0VcyRjbphzcCs5+vO59u4yyqTK+etZmcA2Lao4SnlBy1K7HUL2GERoBCYuQ+wkAAB5gSURBVMNO28nbIDuW2JTeNUfxtFBlFZWs+jGHBRv28Xn6AY6VVRAfHcYvxyUzfXAn+sZHOTqg/+npop6sTWeY8YKdQwIaMJRbacBwpR8+ge2L4Zw/nvxFfDpnPQyzJ8HaF51e88BlqtKChNcoZ/IE2yFfnNf41NiVFbBzuV3hrrbJZW6Qd6yMd9b9xKGCUkKDAggJCiA0KNBxf7rnARSVVrBo034WbdpPblEZbVoHM3NoAjMGdWJEUkyjRzw1q5QLYMyvYevH0GGAp0ujfJgGDFcpK4bPHrSL84y82bljOg2BPhfAmudh5E3QOsa9ZayuKi1IRNzJ25Mn2D6Y3Wsavwrf/g12jY1maI7KO1bGnC8zmPNVBvnF5YQGBVBaUdmgfIJhwQFM7duRGYM7Mb5nXO35mVq6qX+As59oWfm6lM/RgOEqq/9pV5P7xYKGpf2Y9KDN47TmeZj8qNuKd4qqtCA1axiJIyEw1A7RbGzA2LEUEDsb2U1qBopz+nXgzsk96dcpGmMMZRWG0opKSsoqHPeV1e4rKCmrpKTCroUxMimG8FAf+K+gwUK5mQ/8L2kBjuyxzTgp0xv+Jdmxv10qdO1LMPq2pmdMdVb1tCDVBYfZGcQZKxp/7u1L7ZwON1xLXlEZr3yVwatfZpBfcnKgqCIihAQJIUEBRPhCIFCqhdD/Ta7w+SP2/pynGnf8pAdh8wd2mG1jz9FQBdl2ZnfrWrKeJk+EZX+wiRMb2kxWfNSuET7mTteU0yGvqIxXvtzJq1/tIr+knGn9OnLn5J707dTMgwWU8mMaMKBxX4xVdq6wX/aTHoI2XRp3jrheNrfUun/bzsvmSNRXmG2DRW1rcySPh2XAri+h7/SGnXfXl3a1Nxf1XxwpKmXOlxnHA8W5/W2gSInXQKFUc9OAcSwX/trbdkD3uxj6zrCrpTmjosymjGjTFcY28Rf1xPvg+7m2aeu8vzTtXM6oOcu7uk5DITjczsdoaMDYsdQe23lUo4t2rLSCzNwiFm7cx6tf7aKgpJzzBnTk12dpoFDKkzRgAEy8H9Lft3MoPn0Auo6xiwr1nVH38Nh1/4acLXDl/zU9/UVMNxhyNaz/j23OadO5aeerT0H2qSOkqgSF2BnDjZnAt2OJraEEnX4GdEl5BXtzj5GZe4w9uUX2/rC9z8wt4mBB6fF9zx8Qz68n96BPRw0USnmaBoxWbWHCb+0tZ5sNHOnzYdFv4ZP7IGmcrXmkTIfwau39BTmw7E82mWDv81xTlgn3wca37boMFz7rmnOeTmE2xHRj75FjrN5+8JSXewcNZODBxSz88jtKwk7uvDZARaWhvKKSsgpDeaW9b124h+sP7+SLyJl8/dFmyisNZRWVlFcYih1BYk9uEdn5JScNew0KEBLatiKxbSvOTulAYttWdI5pTf+EaLrHRbj376CUcpoGjOrietkJdJPuhwObTwSPj+6Gj/8Huk10BI8LYMnjUFYE5/6v6yantels12BY/yqMvRtikl1z3pqMgYIc8oNjmPH8VxwsKDlllwESw4ehsHjRPBZWjnHqtD8LXALB8Jftndiz8yeCAoXgwACCAoTQ4AA6RbdiXI84Ose0IrFtazq3bUViTGs6RoUR6A0T5JTyc2IaMsOpBRs+fLhJTU11/YmNsQshpc+HtPlwZDcEBNskfWPuhKlPuvb9ju6H5wbbwDTzRdeeu0pJPvwpkdmh1/F8yXnMuW4EHaJqpOaurCBhdl+Kel5A7uS/nnKKoEAhKCCA4EAhyBEUQub9AsnaiNy9qdlmeCulmkZE1htj6kh8d4LWMOojAvED7W3yY7DvOxs8DmfYjmpXi4qHETfC2hds+uq4Xi5/i9K8A4QAPxa24qXrhzE86TQjxJLHEbFvNRExres/aUU57Fpp+340WCjlk3RqaEOI2BXZpv4Brnzz9Et3NtXYuyGoFax42uWnNsbwr0U2ffmMcYMZ072OyXXJE+zs9SM/1X/ivettQkMvzE6rlHKOBoyWKCLOroOdNh8OpLv01M8t2U7ath8BGDeob907V2U+dSZ77Y6ldiJgN13tTSlfpQGjpRrza3u/5SOXnfL97zL5++JtTOni+Njry6gblwKtY50bXrtjiU2NXn05WqWUT6kzYIhIvogcreWWLyJHm6uQfql1jP1Cz9vjktOt2XGI++Z9z5ju7ZjZMxgQGwzqEhBghxXvWkWd6V+P5domKW2OUsqn1RkwjDGRxpioWm6Rxph6Z1KJyDQR+UFEtovIA6fZ53IR2Swi6SLyf9W2XysiPzpu1zb80nxAVAIc3dvk02zPzueWN1JJahfOi9cMI/BYjg1IzizlmTzBluHwztPvk7ESTKUGDKV8XJ3fGCJSZ4IlY8zhOo4NBGYBU4BMYJ2ILDTGbK62T0/gQWCsMSZXRNpXe9/HgOHYeWLrHcfmOndZPiI6EXJ+aNIpcvJLuO7VdYQGB/Lq9SOIbhVcd1qQmpIdfRIZK6Bd99r32bEUQqMgwamReUopL1XfT8z12C/s2sZJGqBbHceOBLYbY3YCiMjbwAxgc7V9bgJmVQUCY4xjkQbOAb6oCkgi8gUwDXirnvL6luhE2L7ENgc1YqjqsdIKbnxtHYcKSnnnltEktnUMjy3MOX1akJradbdLrWasguE3nPq6MTadefIE52osSimvVV+TVLIxppvjvuatrmABkABUb4DPdGyrrhfQS0S+EpG1IjKtAcciIjeLSKqIpObk5NRTHO+RX1zGi8t38Pz6Y1BWyDtfppGTf+ps7LpUVBruevs7Nu3N459XDWFgYpsTLzakhiFig0HGytr7MQ7tgLyftDlKKT/g9E9CEWkL9ASOTwk2xjQiO90p798TmAQkAitFxOlFiY0xs4HZYGd6N7EsHpdbWMqrq3fxn68yOFpczl3xnSAXXl30JQ8u+olRye04b2A80/p1JC4ytM5zPfXxFj7ffIDHL+zL2X07nPxiQbZza45XSZ4A378D2VugQ42huDuW2vsek50/n1LKKzkVMETkRuAu7Jf6BmA0sAao62flXqB6ytVEx7bqMoGvjTFlQIaIbMMGkL3YIFL92OXOlNUbHThazL9X7eTNr3+iqLSCc/p14LZJPRgkHeDfT/HyRR14N68HH2/az+8+SOOxBWmMTI7h/AHxnNO/I+0jT07r8epXdunSG8Ymc93YGvmoSguhrBDCnWySAkgab+8zVtYSMJbYTLttkxp+4Uopr+JsDeMuYASw1hhzpoj0Af5YzzHrgJ4ikowNAFcCP6uxzwfAVcCrIhKLbaLaCewA/uio1QBMxXaO+5Q9h4t4acUO3k3NpLyykumDOnHbmT3o1cExg/yobYXrHHCYe6b25jdTerHtQAEfb9rPok37+d2CdB5dmM7IpBjOHxjPtP4d2fDTEZ74aDNT+3bg4fNTTn3TqrW8Izqc+trptO1q1/zYtQpG/+rE9vJS27cx+KpG/gWUUt7E2YBRbIwpFhFEJNQYs1VEetd1gDGmXETuAD4DAoE5xph0EXkCSDXGLHS8NlVENgMVwL3GmEMAIvIkNugAPFHXiCxvsz07nxeW72DBhn0EinDJsER+NbEbXduFn7xjRAeb6DDPVsxEhN4dI+ndMZJ7pvRi24F8PvreBo9HF6Tz2MJ0ggKEgQnRPHvlkNozwJ5uLe/6JE+ALQuhsuLEKn2Z39jaSndtjlLKHzgbMDJFpA22RvCFiOQCu+s7yBizCFhUY9uj1R4b4B7Hreaxc4A5TpbPK6TtzWPWsu18mp5FWFAg141J4qbx3egYHVb7AQEBNhlhXmatL/fqEMk9U04Ej4+/38+2A/k8MaM/rUJqWXoVTtQwGtIkBTZgfPcGZH1vVycE238REGQn9ymlfJ5TAcMYM9Px8HERWQZEA5+6rVQ+ZN+RY3yWnsUnm7L4ZtdhIsOCuH1SD64fm0S7iLo7rgGISnRq8l6vDpH0muJEMsTCqiapBtYwjvdjrDoRMLYvgcSREKar4SnlD5zt9B4NpBtj8o0xK0QkChgCfO3W0nmpXQcL+SQti0/Ts9i45wgAvTtEcv+0Plw9ugtRYcHOnyw6Efa48M9c4GiSamgNIyoeYnvZju+xd0LhQdi/Ec582HVlU0q1aM42Sb0IDK32vKCWbX7LGMO2AwV8krafT9Oy2JqVD8CgxGjum9abaf060q2xS41GJ0D6PqistE1UTVWYbRMEBjYgaFVJGm+XkK0og53LAQM9dP6FUv7C2YAhptrSfMaYShHx62m9xhi+z8zj0/QsPk3LIuNgISIwomsMj17Ql3P6dyShTaumv1FUgl3drzAbIjs2/XwNmbRXU/IESH3FLiK1Y5kNPPGDm14mpZRXcPZLf6eI3ImtVQDchh3+6pfS9+Vx8+vr2XvkGEEBwhnd23Hj+GSm9q1/Ql2DRTumsuTtdU3AKMxpeP9Flap+jJ0r7PyLbpNOjJhSSvk8ZwPGr4DngEewOaSWADe7q1At3axl2ykoKeeZywZxdkp72rQOcd+bRTsyouTtgcRhTT9fQTbED2rcseHtoMMA+PY1yN+vw2mV8jPOjpLKxk6883uHCkr4YvMBfj46iUuHJbr/DaMcAcMFac6BptUwAJLH2/XGAbqf6ZoyKaW8glO9qCLSS0SWiEia4/lAEXnEvUVrmd7/bi9lFYYrRnSuf2dXaNUWgsOPT95rkrJjdt3tho6Qqq5q2dbY3nYEl1LKbzg77OZlbGqOMgBjzPf4YY3DGMPc1D0M7tyG3h2dmPPgCiK2WcoVK+8VNHIORnVdx0BgCPSc0vTyKKW8irN9GK2NMd/IyWsylLuhPC3ad3uOsO1AAX+62OmEuq7hopX3jqcFaewoKYCwaLhxMbRNrn9fpZRPcbaGcVBEumM7vBGRS4H9bitVCzV33R5aBQdywcD45n3j6ETXNEkdr2E0oUkKbKe5zu5Wyu84W8O4HbvuRB8R2QtkAFe7rVQtUGFJOR9u3Mf5A+OJbMhMbVeIToSCAzY7bFATRmQVNiJTrVJKOThVwzDG7DTGnA3EAX2AiYBfZZz7eNN+CksruLK5Oruri0oADOTva9p5GpsWRCmlqCdgiEiUiDwoIs+LyBSgCLgW2A5c3hwFbCnmrttDt7hwhnVtW//OrlY1GqmpzVKF2bYPIsjFkwuVUn6hviapN4Bc7Op6NwEPAwLMNMZscHPZWozt2fmk7s7lwXP7UKPjv3kcDxi1pzl3WlPSgiil/F59AaObMWYAgIj8G9vR3cUYU+z2krUgc1MzCQoQLh7qoXkHxyfvNTFgNHXSnlLKr9XXh1FW9cAYUwFk+luwKKuoZP63mZzVp73r80Q5K6Q1tIpxUQ1D+y+UUo1TXw1jkIgcdTwWoJXjuWAXzPP5sZVLtmRzsKCUK0d6oLO7uugE1/RhRGg6D6VU49QZMIwxfp+KdG7qHjpEhTKhp4d/mUclNm22d3kJFOdpH4ZSqtFcsCKP78rKK2b5D9lcOiyRoEAP/6mimxgwqmZ5N3XSnlLKb2nAqMO89XuoNHD5cA83R4FtkirOg5KCxh1fNctbaxhKqUbSgHEalZWGuamZjO4WQ9d24Z4ujm2SgsbnlHJF4kGllF/TgHEaazMO8dPhIq4c0cXTRbGOz8VoZLNUVVoQHSWllGokDRinMXfdHiLDgpjW3wXLorrC8ZX3tIahlPIMDRi1yCsqY1FaFhcNTiAsuIUMFIuMB6TxTVKFORASCcGtXFospZT/0IBRiwUb91JaXtl8q+o5IzDYBo3GTt4ryNbahVKqSTRg1OKddXvoGx9F/4RoTxflZNEJjQ8YmhZEKdVEGjBqSNubR/q+oy2rdlGlKSvvaVoQpVQTacCoYW7qHkKCArhocIKni3Kq6ERbwzCm4ccWapOUUqppNGBUU1xWwfvf7eXc/h2Jbt3Mq+o5IzoRyouh6HDDjqsog2O5OmlPKdUkGjCq+TQti/zicq5oCTO7a9PYNOeaFkQp5QIaMKp5Z90eOse0YnS3dp4uSu0au5CSpgVRSrmABgyH3YcKWbPzEJcP60xAgAdW1XNGY5dqPV7D0IChlGo8DRgO76ZmEiBw6XAPrarnjNaxEBjS8CapAk0LopRqOg0YQHlFJfPWZzKxVxzx0S14JnRAgO3HaGiTVKGmBVFKNZ0GDGDljzlkHS1umXMvaopObHiTVEE2BIdDSAvIuquU8lpuDRgiMk1EfhCR7SLyQC2vXyciOSKywXG7sdprFdW2L3RnOd9Zt4d24SGc1aeDO9/GNRozea8gW0dIKaWarL41vRtNRAKBWcAUIBNYJyILjTGba+z6jjHmjlpOccwYM9hd5auSk1/Cki3ZXD82iZAgL6hwRSfC0X1QWQEBTiZGLMzWEVJKqSZz5zfkSGC7MWanMaYUeBuY4cb3a5RWIYH87oK+XDWyhax7UZ/oBDAVkJ/l/DEFmkdKKdV07gwYCUD11X4yHdtqukREvheReSJSvRMhTERSRWStiFxU2xuIyM2OfVJzcnIaVciI0CCuHZNEt7iIRh3f7Bqz8p6mBVFKuYCn22A+BJKMMQOBL4DXqr3W1RgzHPgZ8A8R6V7zYGPMbGPMcGPM8Lg4P2mjb+jKexXlNpWINkkppZrInQFjL1C9xpDo2HacMeaQMabE8fTfwLBqr+113O8ElgND3FhW79HQlfeKDgJGO72VUk3mzoCxDugpIskiEgJcCZw02klE4qs9nQ5scWxvKyKhjsexwFigZme5fwqLtivnOdskpWlBlFIu4rZRUsaYchG5A/gMCATmGGPSReQJINUYsxC4U0SmA+XAYeA6x+EpwL9EpBIb1J6uZXSV/6pKc+4MnbSnlHIRtwUMAGPMImBRjW2PVnv8IPBgLcetBga4s2xerSEr7xU4BgNoWhClVBN5utNbNUZDJu9pDUMp5SIaMLxRdGebgbasuP59C7IhqBWEeMmwYaVUi6UBwxtVjZRyppZRmGNHSEkLTdmulPIaGjC8UVQDAkbBAR0hpZRyCQ0Y3qghK+9pWhCllItowPBGUQ2YvFeYrSOklFIuoQHDGwWH2dX36lt5r7ICig5pDUMp5RIaMLyVM5P3ig6BqdQ+DKWUS2jA8FbOrLxXoHMwlFKuowHDWzkzeU8n7SmlXEgDhreKToSSo1Ccd/p9jqcF0YChlGo6DRjeypk058drGDpKSinVdBowvJUzK+8VZENgKIRGNU+ZlFI+TQOGt3Jm5b1Cx6Q9TQuilHIBDRjeKrIjSGDdTVIFOmlPKeU6GjC8VUAgRMbX3SRVmK0jpJRSLqMBw5vVN3mvIEdrGEopl9GA4c3qWnmvsvJEH4ZSSrmABgxvFpUAR/fZ4FDTscNgKnQOhlLKZTRgeLPozlBRAkUHT32tQOdgKKVcSwOGNzs+ea+WZqmqSXtaw1BKuYgGDG8WVUfAqEoLon0YSikX0YDhzaI72/vahtZq4kGllItpwPBmrWMgKOw0NYxsCAyBsDbNXy6llE/SgOHNRGyzVK19GI45GJoWRCnlIhowvF10Yu1NUpoWRCnlYhowvN3pVt7TtCBKKRfTgOHtohIgfz9UlJ28vSBHh9QqpVxKA4a3i04EjA0aVYxxpAXRJimllOtowPB2ta28dywXKsu0hqGUcikNGN6uauW96iOlCnQOhlLK9TRgeLuqGsbRagHjeFoQbZJSSrmOBgxvFxoJYdEnN0lpDUMp5QYaMHxBVI2FlAodeaS0D0Mp5UIaMHxBdOLJTVIF2Xa971ZtPVcmpZTP0YDhC6ITTm6SKnTM8g7Qj1cp5Tpu/UYRkWki8oOIbBeRB2p5/ToRyRGRDY7bjdVeu1ZEfnTcrnVnOb1eVIJdYa+0yD4v0KVZlVKuF+SuE4tIIDALmAJkAutEZKExZnONXd8xxtxR49gY4DFgOGCA9Y5jc91VXq9WPc15bE9NC6KUcgt31jBGAtuNMTuNMaXA28AMJ489B/jCGHPYESS+AKa5qZzer+bKe5oWRCnlBu4MGAnAnmrPMx3barpERL4XkXki0rkhx4rIzSKSKiKpOTk5riq396m+8p4xjhqGzsFQSrmWp3tFPwSSjDEDsbWI1xpysDFmtjFmuDFmeFycH39BRnUCxDZJFedBRanWMJRSLufOgLEX6FzteaJj23HGmEPGmBLH038Dw5w9VlUTFGr7LPIyT8zB0D4MpZSLuTNgrAN6ikiyiIQAVwILq+8gIvHVnk4HtjgefwZMFZG2ItIWmOrYpk6nauW9Ak0LopRyD7eNkjLGlIvIHdgv+kBgjjEmXUSeAFKNMQuBO0VkOlAOHAaucxx7WESexAYdgCeMMYfdVVafEJ0IOVtP5JHSGoZSysXcFjAAjDGLgEU1tj1a7fGDwIOnOXYOMMed5fMp0YmwfUm1GoYGDKWUa3m601u5SlQClBXCwW0gAdA6xtMlUkr5GA0YviLasS7Gvu+gdSwEBHq2PEopn6MBw1dUBYysTdp/oZRyCw0YvqJq8l5FqY6QUkq5hQYMXxHRAQKCHY+1hqGUcj0NGL4iIACiHNNaNGAopdxAA4YviXL0Y+iQWqWUG2jA8CVVHd9aw1BKuYEGDF9SleZcO72VUm6gAcOXaA1DKeVGbk0NoppZynQ4sgfiUjxdEqWUD9KA4Usi2sOU33u6FEopH6VNUkoppZyiAUMppZRTNGAopZRyigYMpZRSTtGAoZRSyikaMJRSSjlFA4ZSSimnaMBQSinlFDHGeLoMLiEiOcDuJpwiFjjoouJ4G712/+XP1+/P1w4nrr+rMcapBHQ+EzCaSkRSjTHDPV0OT9Br989rB/++fn++dmjc9WuTlFJKKadowFBKKeUUDRgnzPZ0ATxIr91/+fP1+/O1QyOuX/swlFJKOUVrGEoppZyiAUMppZRT/D5giMg0EflBRLaLyAOeLk9zE5FdIrJJRDaISKqny+NOIjJHRLJFJK3athgR+UJEfnTct/VkGd3pNNf/uIjsdXz+G0TkPE+W0V1EpLOILBORzSKSLiJ3Obb7/Odfx7U3+LP36z4MEQkEtgFTgExgHXCVMWazRwvWjERkFzDcGOPzE5hEZAJQALxujOnv2PZn4LAx5mnHD4a2xpj7PVlOdznN9T8OFBhjnvFk2dxNROKBeGPMtyISCawHLgKuw8c//zqu/XIa+Nn7ew1jJLDdGLPTGFMKvA3M8HCZlJsYY1YCh2tsngG85nj8GvY/kk86zfX7BWPMfmPMt47H+cAWIAE/+PzruPYG8/eAkQDsqfY8k0b+Ib2YAT4XkfUicrOnC+MBHYwx+x2Ps4AOniyMh9whIt87mqx8rkmmJhFJAoYAX+Nnn3+Na4cGfvb+HjAUjDPGDAXOBW53NFv4JWPbZ/2tjfZFoDswGNgP/NWzxXEvEYkA3gPuNsYcrf6ar3/+tVx7gz97fw8Ye4HO1Z4nOrb5DWPMXsd9NvA+tpnOnxxwtPFWtfVme7g8zcoYc8AYU2GMqQRexoc/fxEJxn5hvmmMme/Y7Beff23X3pjP3t8Dxjqgp4gki0gIcCWw0MNlajYiEu7oBENEwoGpQFrdR/mchcC1jsfXAgs8WJZmV/Vl6TATH/38RUSAV4Atxpi/VXvJ5z//0117Yz57vx4lBeAYSvYPIBCYY4x5ysNFajYi0g1bqwAIAv7Pl69fRN4CJmHTOh8AHgM+AOYCXbDp8S83xvhkx/Bprn8StknCALuAW6q16fsMERkHrAI2AZWOzQ9h2/J9+vOv49qvooGfvd8HDKWUUs7x9yYppZRSTtKAoZRSyikaMJRSSjlFA4ZSSimnaMBQSinlFA0YSrUAIjJJRD7ydDmUqosGDKWUUk7RgKFUA4jINSLyjWP9gH+JSKCIFIjI3x1rDSwRkTjHvoNFZK0judv7VcndRKSHiCwWkY0i8q2IdHecPkJE5onIVhF50zFDV6kWQwOGUk4SkRTgCmCsMWYwUAFcDYQDqcaYfsAK7AxqgNeB+40xA7GzbKu2vwnMMsYMAsZgE7+BzSJ6N9AX6AaMdftFKdUAQZ4ugFJeZDIwDFjn+PHfCpusrhJ4x7HPf4H5IhINtDHGrHBsfw1415G7K8EY8z6AMaYYwHG+b4wxmY7nG4Ak4Ev3X5ZSztGAoZTzBHjNGPPgSRtFfldjv8bm2ymp9rgC/f+pWhhtklLKeUuAS0WkPRxfD7or9v/RpY59fgZ8aYzJA3JFZLxj+8+BFY4VzzJF5CLHOUJFpHWzXoVSjaS/YJRykjFms4g8gl2hMAAoA24HCoGRjteysf0cYNNlv+QICDuB6x3bfw78S0SecJzjsma8DKUaTbPVKtVEIlJgjInwdDmUcjdtklJKKeUUrWEopZRyitYwlFJKOUUDhlJKKadowFBKKeUUDRhKKaWcogFDKaWUU/4fhp6NfJZYM3gAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "52HsGwKn5KC2",
        "outputId": "86f5d4dc-f2cd-4873-838f-f1774d1964a6"
      },
      "source": [
        "# Evaluate model\n",
        "test_eval = model.evaluate(x_test, y_test, verbose=1, return_dict=True)\n",
        "validation_eval = model.evaluate(x_validate, y_validate, verbose=1, return_dict=True)\n",
        "# model.save(\"model.h5\")"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "63/63 [==============================] - 14s 217ms/step - loss: 2.0556 - main_loss: 0.7274 - aux1_loss: 0.6599 - aux2_loss: 0.6682 - main_accuracy: 0.7449 - main_f1_m: 0.7496 - main_precision_m: 0.8330 - main_recall_m: 0.6838 - aux1_accuracy: 0.7718 - aux1_f1_m: 0.7643 - aux1_precision_m: 0.8693 - aux1_recall_m: 0.6856 - aux2_accuracy: 0.7708 - aux2_f1_m: 0.7597 - aux2_precision_m: 0.8629 - aux2_recall_m: 0.6812\n",
            "26/26 [==============================] - 6s 212ms/step - loss: 1.9209 - main_loss: 0.6786 - aux1_loss: 0.6162 - aux2_loss: 0.6262 - main_accuracy: 0.7594 - main_f1_m: 0.7569 - main_precision_m: 0.8469 - main_recall_m: 0.6887 - aux1_accuracy: 0.7743 - aux1_f1_m: 0.7600 - aux1_precision_m: 0.8701 - aux1_recall_m: 0.6791 - aux2_accuracy: 0.7693 - aux2_f1_m: 0.7587 - aux2_precision_m: 0.8723 - aux2_recall_m: 0.6767\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0tC4GrGl5KC2",
        "scrolled": true,
        "outputId": "d54048f6-9746-43f4-d8ab-2a74409f0f8a"
      },
      "source": [
        "for i, j in validation_eval.items():\n",
        "    print(f'{i}: {j}')"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "loss: 1.9209141731262207\n",
            "main_loss: 0.6785610914230347\n",
            "aux1_loss: 0.6161662936210632\n",
            "aux2_loss: 0.6261867880821228\n",
            "main_accuracy: 0.7593516111373901\n",
            "main_f1_m: 0.7569131255149841\n",
            "main_precision_m: 0.8468733429908752\n",
            "main_recall_m: 0.6887019276618958\n",
            "aux1_accuracy: 0.7743142247200012\n",
            "aux1_f1_m: 0.7600494027137756\n",
            "aux1_precision_m: 0.8701063990592957\n",
            "aux1_recall_m: 0.6790865659713745\n",
            "aux2_accuracy: 0.7693266868591309\n",
            "aux2_f1_m: 0.7587065100669861\n",
            "aux2_precision_m: 0.8722543716430664\n",
            "aux2_recall_m: 0.676682710647583\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BPNXGVMO5KC3",
        "scrolled": false,
        "outputId": "5c56f87d-a6eb-4ecd-f070-96ca8570a338"
      },
      "source": [
        "for i, j in test_eval.items():\n",
        "    print(f'{i}: {j}')"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "loss: 2.0555758476257324\n",
            "main_loss: 0.7274333834648132\n",
            "aux1_loss: 0.6599315404891968\n",
            "aux2_loss: 0.668210506439209\n",
            "main_accuracy: 0.7448827028274536\n",
            "main_f1_m: 0.7495604753494263\n",
            "main_precision_m: 0.833039402961731\n",
            "main_recall_m: 0.6838189363479614\n",
            "aux1_accuracy: 0.7718422412872314\n",
            "aux1_f1_m: 0.7642954587936401\n",
            "aux1_precision_m: 0.8693410158157349\n",
            "aux1_recall_m: 0.6856464147567749\n",
            "aux2_accuracy: 0.7708437442779541\n",
            "aux2_f1_m: 0.759667694568634\n",
            "aux2_precision_m: 0.8628720641136169\n",
            "aux2_recall_m: 0.6811821460723877\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3DqyE7-W5KC3",
        "outputId": "da8dbe6e-4a15-45e2-b876-46f301b7ff37"
      },
      "source": [
        "# Get predictions for plotting the confusion matrix\n",
        "predictions = model.predict(x_test, verbose=1)"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "63/63 [==============================] - 14s 217ms/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 358
        },
        "id": "pFXzuzzh5KC3",
        "scrolled": true,
        "outputId": "6843f5a9-a5d4-4284-94a2-f368a750a00e"
      },
      "source": [
        "# Plot the confusion matrix\n",
        "predictions = np.argmax(predictions, axis=0)\n",
        "y_test = np.argmax(y_test, axis=0)\n",
        "\n",
        "sns.heatmap(confusion_matrix(y_test, predictions), annot=True, cmap = 'coolwarm', fmt='.0f')\n",
        "plt.show()"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "error",
          "ename": "TypeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-49-b89fef9ebc07>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0my_test\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_test\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0msns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mheatmap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfusion_matrix\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_test\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpredictions\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mannot\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcmap\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m'coolwarm'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfmt\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'.0f'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py\u001b[0m in \u001b[0;36mconfusion_matrix\u001b[0;34m(y_true, y_pred, labels, sample_weight, normalize)\u001b[0m\n\u001b[1;32m    266\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    267\u001b[0m     \"\"\"\n\u001b[0;32m--> 268\u001b[0;31m     \u001b[0my_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_true\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_check_targets\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_true\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_pred\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    269\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0my_type\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32min\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0;34m\"binary\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"multiclass\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    270\u001b[0m         \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"%s is not supported\"\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0my_type\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py\u001b[0m in \u001b[0;36m_check_targets\u001b[0;34m(y_true, y_pred)\u001b[0m\n\u001b[1;32m     78\u001b[0m     \u001b[0my_pred\u001b[0m \u001b[0;34m:\u001b[0m \u001b[0marray\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mindicator\u001b[0m \u001b[0mmatrix\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     79\u001b[0m     \"\"\"\n\u001b[0;32m---> 80\u001b[0;31m     \u001b[0mcheck_consistent_length\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_true\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_pred\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     81\u001b[0m     \u001b[0mtype_true\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtype_of_target\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_true\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     82\u001b[0m     \u001b[0mtype_pred\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtype_of_target\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_pred\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/sklearn/utils/validation.py\u001b[0m in \u001b[0;36mcheck_consistent_length\u001b[0;34m(*arrays)\u001b[0m\n\u001b[1;32m    206\u001b[0m     \"\"\"\n\u001b[1;32m    207\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 208\u001b[0;31m     \u001b[0mlengths\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0m_num_samples\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mX\u001b[0m \u001b[0;32min\u001b[0m \u001b[0marrays\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mX\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    209\u001b[0m     \u001b[0muniques\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munique\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlengths\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    210\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0muniques\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/sklearn/utils/validation.py\u001b[0m in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m    206\u001b[0m     \"\"\"\n\u001b[1;32m    207\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 208\u001b[0;31m     \u001b[0mlengths\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0m_num_samples\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mX\u001b[0m \u001b[0;32min\u001b[0m \u001b[0marrays\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mX\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    209\u001b[0m     \u001b[0muniques\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munique\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlengths\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    210\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0muniques\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/sklearn/utils/validation.py\u001b[0m in \u001b[0;36m_num_samples\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m    150\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    151\u001b[0m             raise TypeError(\"Singleton array %r cannot be considered\"\n\u001b[0;32m--> 152\u001b[0;31m                             \" a valid collection.\" % x)\n\u001b[0m\u001b[1;32m    153\u001b[0m         \u001b[0;31m# Check that shape is returning an integer or default to len\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    154\u001b[0m         \u001b[0;31m# Dask dataframes may not return numeric shape[0] value\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mTypeError\u001b[0m: Singleton array 0 cannot be considered a valid collection."
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-6ys5M0w8SSm"
      },
      "source": [
        "# Model Parameters\n",
        "input_shape = (75, 100, 3)\n",
        "\n",
        "# Number of classes\n",
        "num_classes = 7\n",
        "\n",
        "# Optimizer\n",
        "optimizer = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)\n",
        "\n",
        "# Epoch is number of times a node is trained\n",
        "epochs = 9\n",
        "batch_size = 20\n",
        "\n",
        "# Callback functions (To follow the path of gradient descent)\n",
        "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', patience=5, verbose=0, factor=0.5, min_lr=0.00001)\n",
        "early_stopping_monitor = EarlyStopping(patience=20, monitor='val_accuracy')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B3COod77Db5c",
        "outputId": "e99a24fe-b6a1-428d-9c8c-64e007fa7e2d"
      },
      "source": [
        "# dataaugment.flow(x_train,y_train, batch_size=batch_size),\n",
        "# callbacks=[learning_rate_reduction,early_stopping_monitor]\n",
        "history = model.fit(\n",
        "    x = x_train,\n",
        "    y = y_train, \n",
        "    batch_size=batch_size,\n",
        "    verbose = 0, \n",
        "    steps_per_epoch=x_train.shape[0] // batch_size,\n",
        "    epochs=10,\n",
        "    validation_data = (x_validate, y_validate)\n",
        ")\n",
        "\n",
        "loss, accuracy = model.evaluate(x_test, y_test, verbose=0)\n",
        "predictions = model.predict(x_test)\n",
        "loss_v, accuracy_v = model.evaluate(x_validate, y_validate, verbose=0)\n",
        "loss_t, accuracy_t = model.evaluate(x_train, y_train, verbose=0)\n",
        "\n",
        "print(\"Training: accuracy = %f\" % (accuracy_t))\n",
        "print(\"Validation: accuracy = %f\" % (accuracy_v))\n",
        "print(\"Test: accuracy = %f\" % (accuracy))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7l6D3elvzeOc",
        "outputId": "aa5f62ba-4636-4e80-c7ac-f7d814fbbb25"
      },
      "source": [
        "predictions = np.argmax(predictions, axis=1)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NVVDXdOHxYcY"
      },
      "source": [
        "y_test = np.argmax(y_test, axis=1)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IAGYMa0HxUsy",
        "outputId": "f3c2f649-b11b-4c68-efdc-1b9c627d14c6"
      },
      "source": [
        "\n",
        "sns.heatmap(confusion_matrix(y_test, predictions), annot=True, cmap = 'coolwarm', fmt='.0f')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jL1zPJ27xXmE",
        "outputId": "18204e51-7a3b-4c0f-df3f-bd99c34f6d58"
      },
      "source": [
        "len(predictions)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qj27aH_-z1f4",
        "outputId": "db34107b-edb9-436e-aa12-b978a68a1074"
      },
      "source": [
        "target.value_counts()\n",
        "len(x_train)\n",
        "len(x_validate)"
      ]
    }
  ]
}