nc-severstal02-classification-basecnn-1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:35.467634Z",
     "iopub.status.busy": "2021-03-19T04:18:35.467005Z",
     "iopub.status.idle": "2021-03-19T04:18:41.411691Z",
     "shell.execute_reply": "2021-03-19T04:18:41.412980Z"
    },
    "papermill": {
     "duration": 5.973049,
     "end_time": "2021-03-19T04:18:41.413368",
     "exception": false,
     "start_time": "2021-03-19T04:18:35.440319",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import cv2\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras.callbacks import Callback, ModelCheckpoint, ReduceLROnPlateau,EarlyStopping\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import Sequential, Model, load_model\n",
    "from tensorflow.keras.layers import Input, Dense, Conv2D, MaxPooling2D\n",
    "from tensorflow.keras.layers import GlobalAveragePooling2D, AveragePooling2D\n",
    "from tensorflow.keras.layers import ZeroPadding2D, Dropout, Flatten, add, Reshape\n",
    "from tensorflow.keras.layers import Activation, BatchNormalization\n",
    "from tensorflow.keras.optimizers import Adam, SGD, Nadam\n",
    "from tensorflow.keras.utils import plot_model,to_categorical\n",
    "from tensorflow.keras import backend as K\n",
    "from tensorflow.keras import initializers\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, classification_report\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 0.046757,
     "end_time": "2021-03-19T04:18:41.495630",
     "exception": false,
     "start_time": "2021-03-19T04:18:41.448873",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Read train and test files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:41.576071Z",
     "iopub.status.busy": "2021-03-19T04:18:41.575328Z",
     "iopub.status.idle": "2021-03-19T04:18:42.086578Z",
     "shell.execute_reply": "2021-03-19T04:18:42.087029Z"
    },
    "papermill": {
     "duration": 0.551266,
     "end_time": "2021-03-19T04:18:42.087184",
     "exception": false,
     "start_time": "2021-03-19T04:18:41.535918",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7095, 3)\n"
     ]
    },
    {
     "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>ImageId</th>\n",
       "      <th>ClassId</th>\n",
       "      <th>EncodedPixels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0002cc93b.jpg</td>\n",
       "      <td>1</td>\n",
       "      <td>29102 12 29346 24 29602 24 29858 24 30114 24 3...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0007a71bf.jpg</td>\n",
       "      <td>3</td>\n",
       "      <td>18661 28 18863 82 19091 110 19347 110 19603 11...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>000a4bcdd.jpg</td>\n",
       "      <td>1</td>\n",
       "      <td>37607 3 37858 8 38108 14 38359 20 38610 25 388...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>000f6bf48.jpg</td>\n",
       "      <td>4</td>\n",
       "      <td>131973 1 132228 4 132483 6 132738 8 132993 11 ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0014fce06.jpg</td>\n",
       "      <td>3</td>\n",
       "      <td>229501 11 229741 33 229981 55 230221 77 230468...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId  ClassId                                      EncodedPixels\n",
       "0  0002cc93b.jpg        1  29102 12 29346 24 29602 24 29858 24 30114 24 3...\n",
       "1  0007a71bf.jpg        3  18661 28 18863 82 19091 110 19347 110 19603 11...\n",
       "2  000a4bcdd.jpg        1  37607 3 37858 8 38108 14 38359 20 38610 25 388...\n",
       "3  000f6bf48.jpg        4  131973 1 132228 4 132483 6 132738 8 132993 11 ...\n",
       "4  0014fce06.jpg        3  229501 11 229741 33 229981 55 230221 77 230468..."
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = pd.read_csv('../input/severstal-steel-defect-detection/train.csv')\n",
    "\n",
    "print(train_df.shape)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:42.132179Z",
     "iopub.status.busy": "2021-03-19T04:18:42.131657Z",
     "iopub.status.idle": "2021-03-19T04:18:42.155111Z",
     "shell.execute_reply": "2021-03-19T04:18:42.154487Z"
    },
    "papermill": {
     "duration": 0.047811,
     "end_time": "2021-03-19T04:18:42.155226",
     "exception": false,
     "start_time": "2021-03-19T04:18:42.107415",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5506, 3)\n"
     ]
    },
    {
     "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>ImageId</th>\n",
       "      <th>EncodedPixels</th>\n",
       "      <th>ClassId</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000f269f.jpg</td>\n",
       "      <td>1 409600</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000ccc2ac.jpg</td>\n",
       "      <td>1 409600</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>002451917.jpg</td>\n",
       "      <td>1 409600</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>003c5da97.jpg</td>\n",
       "      <td>1 409600</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0042e163f.jpg</td>\n",
       "      <td>1 409600</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId EncodedPixels  ClassId\n",
       "0  0000f269f.jpg      1 409600        0\n",
       "1  000ccc2ac.jpg      1 409600        0\n",
       "2  002451917.jpg      1 409600        0\n",
       "3  003c5da97.jpg      1 409600        0\n",
       "4  0042e163f.jpg      1 409600        0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission_df = pd.read_csv('../input/severstal-steel-defect-detection/sample_submission.csv')\n",
    "print(submission_df.shape)\n",
    "submission_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:42.203580Z",
     "iopub.status.busy": "2021-03-19T04:18:42.202757Z",
     "iopub.status.idle": "2021-03-19T04:18:42.208025Z",
     "shell.execute_reply": "2021-03-19T04:18:42.207630Z"
    },
    "papermill": {
     "duration": 0.032308,
     "end_time": "2021-03-19T04:18:42.208132",
     "exception": false,
     "start_time": "2021-03-19T04:18:42.175824",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5506"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_submission_images = submission_df['ImageId'].unique()\n",
    "len(unique_submission_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:42.256916Z",
     "iopub.status.busy": "2021-03-19T04:18:42.256073Z",
     "iopub.status.idle": "2021-03-19T04:18:42.262565Z",
     "shell.execute_reply": "2021-03-19T04:18:42.263024Z"
    },
    "papermill": {
     "duration": 0.034624,
     "end_time": "2021-03-19T04:18:42.263169",
     "exception": false,
     "start_time": "2021-03-19T04:18:42.228545",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5506, 1)\n"
     ]
    },
    {
     "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>ImageId</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000f269f.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000ccc2ac.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>002451917.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>003c5da97.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0042e163f.jpg</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId\n",
       "0  0000f269f.jpg\n",
       "1  000ccc2ac.jpg\n",
       "2  002451917.jpg\n",
       "3  003c5da97.jpg\n",
       "4  0042e163f.jpg"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission_df = pd.DataFrame(unique_submission_images, columns=['ImageId'])\n",
    "print(submission_df.shape)\n",
    "submission_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "papermill": {
     "duration": 0.021173,
     "end_time": "2021-03-19T04:18:42.306427",
     "exception": false,
     "start_time": "2021-03-19T04:18:42.285254",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Combine defected and no-defect images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:42.356164Z",
     "iopub.status.busy": "2021-03-19T04:18:42.355479Z",
     "iopub.status.idle": "2021-03-19T04:18:44.140827Z",
     "shell.execute_reply": "2021-03-19T04:18:44.140271Z"
    },
    "papermill": {
     "duration": 1.813065,
     "end_time": "2021-03-19T04:18:44.140960",
     "exception": false,
     "start_time": "2021-03-19T04:18:42.327895",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(12568, 2)\n",
      "1    6666\n",
      "0    5902\n",
      "Name: defect_label, dtype: int64\n"
     ]
    },
    {
     "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>ImageId</th>\n",
       "      <th>defect_label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0002cc93b.jpg</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId  defect_label\n",
       "0  0002cc93b.jpg             1"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TRAIN_PATH = '../input/severstal-steel-defect-detection/train_images/'\n",
    "TEST_PATH = '../input/severstal-steel-defect-detection/test_images/'\n",
    "\n",
    "train_flist = os.listdir(TRAIN_PATH) \n",
    "\n",
    "image_defects = train_df['ImageId'].unique().tolist()\n",
    "image_no_defects=[item for item in train_flist if item not in image_defects]\n",
    "\n",
    "#df_bin is the dataframe with label defect based on train_df \n",
    "df_bin = train_df.drop_duplicates(subset=['ImageId'], keep='last').copy()\n",
    "df_bin.loc[:, 'defect_label'] = 1\n",
    "\n",
    "df_noDefect_bin = pd.DataFrame(image_no_defects,columns=['ImageId'])\n",
    "df_noDefect_bin.loc[:, 'defect_label'] = 0\n",
    "\n",
    "frames = [df_bin[['ImageId', \"defect_label\"]], df_noDefect_bin]\n",
    "df = pd.concat(frames).reset_index(drop=True)\n",
    "print(df.shape)\n",
    "print(df['defect_label'].value_counts())\n",
    "df.head(1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 0.022106,
     "end_time": "2021-03-19T04:18:44.185973",
     "exception": false,
     "start_time": "2021-03-19T04:18:44.163867",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Split train data, load and resave images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:44.235354Z",
     "iopub.status.busy": "2021-03-19T04:18:44.234857Z",
     "iopub.status.idle": "2021-03-19T04:18:44.258280Z",
     "shell.execute_reply": "2021-03-19T04:18:44.257783Z"
    },
    "papermill": {
     "duration": 0.050467,
     "end_time": "2021-03-19T04:18:44.258404",
     "exception": false,
     "start_time": "2021-03-19T04:18:44.207937",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "nb_images = df.shape[0]\n",
    "df_train, df_test = train_test_split(df, test_size=0.2, \n",
    "                                                    stratify = df['defect_label'], \n",
    "                                                    random_state=2021)\n",
    "\n",
    "df_train, df_val = train_test_split(df_train, test_size=0.25, \n",
    "                                                  stratify = df_train['defect_label'],\n",
    "                                                  random_state=2021) \n",
    "\n",
    "# 0.25 x 0.8 = 0.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:18:44.313372Z",
     "iopub.status.busy": "2021-03-19T04:18:44.312406Z",
     "iopub.status.idle": "2021-03-19T04:22:01.711213Z",
     "shell.execute_reply": "2021-03-19T04:22:01.710584Z"
    },
    "papermill": {
     "duration": 197.429111,
     "end_time": "2021-03-19T04:22:01.711387",
     "exception": false,
     "start_time": "2021-03-19T04:18:44.282276",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 7540/7540 [01:57<00:00, 63.94it/s]\n",
      "100%|██████████| 2514/2514 [00:40<00:00, 62.35it/s]\n",
      "100%|██████████| 2514/2514 [00:39<00:00, 64.27it/s]\n",
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:32: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    }
   ],
   "source": [
    "IMAGE_SIZE = 64 #256\n",
    "\n",
    "def load_img(code, base, resize=True):\n",
    "    path = f'{base}/{code}'\n",
    "    img = cv2.imread(path)\n",
    "    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
    "    if resize:\n",
    "        img = cv2.resize(img, (IMAGE_SIZE, IMAGE_SIZE))\n",
    "    return img\n",
    "\n",
    "def validate_path(path):\n",
    "    if not os.path.exists(path):\n",
    "        os.makedirs(path)\n",
    "\n",
    "train_path = '../tmp/train'\n",
    "validate_path(train_path)\n",
    "val_path = '../tmp/val'\n",
    "validate_path(val_path)\n",
    "test_path = '../tmp/test'\n",
    "validate_path(test_path)\n",
    "\n",
    "def process_img(df, train_path):\n",
    "    #read image, resize and save\n",
    "    for code in tqdm(df['ImageId']):\n",
    "        img = load_img(\n",
    "            code,\n",
    "            base='../input/severstal-steel-defect-detection/train_images'\n",
    "        )\n",
    "        path = code.replace('.jpg', '')\n",
    "        cv2.imwrite(f'{train_path}/{path}.png', img)\n",
    "\n",
    "    df['ImageId'] = df['ImageId'].apply(lambda x: x.replace('.jpg', '.png'))\n",
    "    return df\n",
    "df_train = process_img(df_train,train_path)\n",
    "df_val = process_img(df_val,val_path)\n",
    "df_test = process_img(df_test,test_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:02.622759Z",
     "iopub.status.busy": "2021-03-19T04:22:02.622183Z",
     "iopub.status.idle": "2021-03-19T04:22:02.631276Z",
     "shell.execute_reply": "2021-03-19T04:22:02.630769Z"
    },
    "papermill": {
     "duration": 0.464699,
     "end_time": "2021-03-19T04:22:02.631394",
     "exception": false,
     "start_time": "2021-03-19T04:22:02.166695",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7540, 2) (2514, 2) (2514, 2)\n",
      "1    3999\n",
      "0    3541\n",
      "Name: defect_label, dtype: int64\n",
      "1    1334\n",
      "0    1180\n",
      "Name: defect_label, dtype: int64\n",
      "1    1333\n",
      "0    1181\n",
      "Name: defect_label, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(df_train.shape,df_val.shape,df_test.shape)\n",
    "print(df_train['defect_label'].value_counts())\n",
    "print(df_val['defect_label'].value_counts())\n",
    "print(df_test['defect_label'].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 0.455417,
     "end_time": "2021-03-19T04:22:03.535002",
     "exception": false,
     "start_time": "2021-03-19T04:22:03.079585",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define ImageDataGenerator and data augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:04.467657Z",
     "iopub.status.busy": "2021-03-19T04:22:04.466842Z",
     "iopub.status.idle": "2021-03-19T04:22:04.557953Z",
     "shell.execute_reply": "2021-03-19T04:22:04.558777Z"
    },
    "papermill": {
     "duration": 0.565116,
     "end_time": "2021-03-19T04:22:04.558998",
     "exception": false,
     "start_time": "2021-03-19T04:22:03.993882",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 7540 validated image filenames.\n",
      "Found 2514 validated image filenames.\n",
      "Found 2514 validated image filenames.\n"
     ]
    }
   ],
   "source": [
    "BATCH_SIZE = 32\n",
    "\n",
    "# Using original generator\n",
    "data_generator = ImageDataGenerator(\n",
    "        zoom_range=0.1,  # set range for random zoom\n",
    "        # set mode for filling points outside the input boundaries\n",
    "        fill_mode='constant',\n",
    "        cval=0.,\n",
    "        rotation_range=10,\n",
    "        height_shift_range=0.05,\n",
    "        width_shift_range=0.05,\n",
    "        horizontal_flip=True,\n",
    "        vertical_flip=True,\n",
    "        rescale=1/255.,\n",
    "    )\n",
    "val_generator = ImageDataGenerator(\n",
    "        rescale=1/255.\n",
    "    )\n",
    "\n",
    "train_gen = data_generator.flow_from_dataframe(\n",
    "        df_train, \n",
    "        directory='../tmp/train',\n",
    "        x_col='ImageId', \n",
    "        y_col='defect_label', \n",
    "        class_mode='raw',\n",
    "        target_size=(64, 64),\n",
    "        batch_size=BATCH_SIZE,\n",
    "        )\n",
    "\n",
    "val_gen =  val_generator.flow_from_dataframe(\n",
    "        df_val, \n",
    "        directory='../tmp/val',\n",
    "        x_col='ImageId', \n",
    "        y_col='defect_label', \n",
    "        class_mode='raw',\n",
    "        target_size=(64, 64),\n",
    "        batch_size=BATCH_SIZE,\n",
    "        )\n",
    "test_gen =  val_generator.flow_from_dataframe(\n",
    "        df_test, \n",
    "        directory='../tmp/test',\n",
    "        x_col='ImageId', \n",
    "        y_col='defect_label', \n",
    "        class_mode='raw',\n",
    "        target_size=(64, 64),\n",
    "        batch_size=BATCH_SIZE,\n",
    "        shuffle=False\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:05.518383Z",
     "iopub.status.busy": "2021-03-19T04:22:05.517801Z",
     "iopub.status.idle": "2021-03-19T04:22:09.238387Z",
     "shell.execute_reply": "2021-03-19T04:22:09.237746Z"
    },
    "papermill": {
     "duration": 4.223489,
     "end_time": "2021-03-19T04:22:09.238589",
     "exception": false,
     "start_time": "2021-03-19T04:22:05.015100",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 5506 validated image filenames.\n"
     ]
    }
   ],
   "source": [
    "submission_gen = ImageDataGenerator(rescale=1/255.).flow_from_dataframe(\n",
    "        submission_df,\n",
    "        directory='../input/severstal-steel-defect-detection/test_images/',\n",
    "        x_col='ImageId',\n",
    "        class_mode=None,\n",
    "        target_size=(64, 64),\n",
    "        batch_size=BATCH_SIZE,\n",
    "        shuffle=False\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 0.553079,
     "end_time": "2021-03-19T04:22:10.246856",
     "exception": false,
     "start_time": "2021-03-19T04:22:09.693777",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Define Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:11.585181Z",
     "iopub.status.busy": "2021-03-19T04:22:11.584629Z",
     "iopub.status.idle": "2021-03-19T04:22:14.821512Z",
     "shell.execute_reply": "2021-03-19T04:22:14.820763Z"
    },
    "papermill": {
     "duration": 3.850366,
     "end_time": "2021-03-19T04:22:14.821713",
     "exception": false,
     "start_time": "2021-03-19T04:22:10.971347",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 64, 64, 64)        1792      \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 64, 64, 64)        36928     \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 64, 64, 64)        256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 32, 32, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 32, 32, 64)        36928     \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 32, 32, 64)        36928     \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 32, 32, 64)        256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 16, 16, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 16, 16, 128)       73856     \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 16, 16, 128)       147584    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 16, 16, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 8, 8, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 8, 8, 128)         147584    \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 8, 8, 128)         147584    \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 8, 8, 128)         512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 4, 4, 128)         0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 2048)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 128)               262272    \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 901,313\n",
      "Trainable params: 900,545\n",
      "Non-trainable params: 768\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input_shape = (IMAGE_SIZE,IMAGE_SIZE,3)\n",
    "prob_drop_hidden = 0.5\n",
    "\n",
    "def build_model():\n",
    "    model = Sequential()\n",
    "\n",
    "    # conv layer\n",
    "    model.add(Conv2D(filters=64, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu', \n",
    "                   input_shape=input_shape))    \n",
    "    model.add(Conv2D(filters=64, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu', ))    \n",
    "    model.add(BatchNormalization())\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'))\n",
    "    \n",
    "    model.add(Conv2D(filters=64, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(Conv2D(filters=64, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding='same'))\n",
    "\n",
    "    model.add(Conv2D(filters=128, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(Conv2D(filters=128, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2,2), padding='same'))\n",
    "    \n",
    "    model.add(Conv2D(filters=128, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(Conv2D(filters=128, kernel_size=(3, 3), strides=(1,1), padding='same', activation='relu' ))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2,2), padding='same'))\n",
    "\n",
    "    model.add(Flatten())\n",
    "\n",
    "    # fc1 layer\n",
    "    model.add(Dense(128, activation='relu'))\n",
    "    model.add(Dropout(rate=prob_drop_hidden))\n",
    "\n",
    "    # fc2 layer\n",
    "    model.add(Dense(64, activation='relu'))\n",
    "    model.add(Dropout(rate=prob_drop_hidden))\n",
    "\n",
    "    # The output of the model \n",
    "    model.add(layers.Dense(1, activation='sigmoid'))\n",
    "    \n",
    "    model.compile(\n",
    "        loss='binary_crossentropy',\n",
    "        optimizer= Adam(learning_rate = 0.001),\n",
    "        metrics=['accuracy']\n",
    "    )\n",
    "    \n",
    "    return model\n",
    "model = build_model()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 0.458079,
     "end_time": "2021-03-19T04:22:15.734379",
     "exception": false,
     "start_time": "2021-03-19T04:22:15.276300",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:16.649847Z",
     "iopub.status.busy": "2021-03-19T04:22:16.648163Z",
     "iopub.status.idle": "2021-03-19T04:22:16.650411Z",
     "shell.execute_reply": "2021-03-19T04:22:16.650852Z"
    },
    "papermill": {
     "duration": 0.467761,
     "end_time": "2021-03-19T04:22:16.650990",
     "exception": false,
     "start_time": "2021-03-19T04:22:16.183229",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "epochs = 200\n",
    "\n",
    "import math\n",
    "# We take the ceiling because we do not drop the remainder of the batch\n",
    "compute_steps_per_epoch = lambda x: int(math.ceil(1. * x / BATCH_SIZE))\n",
    "\n",
    "steps_per_epoch = compute_steps_per_epoch(df_train.shape[0])\n",
    "validation_steps = compute_steps_per_epoch(df_val.shape[0])\n",
    "\n",
    "es = EarlyStopping(monitor='val_loss', mode='min', verbose=2, patience=50)\n",
    "\n",
    "mc = ModelCheckpoint(\n",
    "    'model.h5', \n",
    "    monitor='val_accuracy', \n",
    "    verbose=1, \n",
    "    save_best_only=True, \n",
    "    save_weights_only=False,\n",
    "    mode='auto'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:22:17.572384Z",
     "iopub.status.busy": "2021-03-19T04:22:17.571861Z",
     "iopub.status.idle": "2021-03-19T04:59:01.100544Z",
     "shell.execute_reply": "2021-03-19T04:59:01.101759Z"
    },
    "papermill": {
     "duration": 2203.98465,
     "end_time": "2021-03-19T04:59:01.102136",
     "exception": false,
     "start_time": "2021-03-19T04:22:17.117486",
     "status": "completed"
    },
    "scrolled": false,
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "236/236 [==============================] - 18s 61ms/step - loss: 0.9127 - accuracy: 0.5349 - val_loss: 0.6780 - val_accuracy: 0.5672\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.56722, saving model to model.h5\n",
      "Epoch 2/200\n",
      "236/236 [==============================] - 15s 62ms/step - loss: 0.6622 - accuracy: 0.5926 - val_loss: 0.6623 - val_accuracy: 0.5863\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.56722 to 0.58632, saving model to model.h5\n",
      "Epoch 3/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.6008 - accuracy: 0.6545 - val_loss: 0.5642 - val_accuracy: 0.6810\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.58632 to 0.68099, saving model to model.h5\n",
      "Epoch 4/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.5567 - accuracy: 0.6932 - val_loss: 0.6568 - val_accuracy: 0.5963\n",
      "\n",
      "Epoch 00004: val_accuracy did not improve from 0.68099\n",
      "Epoch 5/200\n",
      "236/236 [==============================] - 15s 61ms/step - loss: 0.5372 - accuracy: 0.7214 - val_loss: 0.5251 - val_accuracy: 0.7156\n",
      "\n",
      "Epoch 00005: val_accuracy improved from 0.68099 to 0.71559, saving model to model.h5\n",
      "Epoch 6/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.5311 - accuracy: 0.7220 - val_loss: 2.1139 - val_accuracy: 0.4714\n",
      "\n",
      "Epoch 00006: val_accuracy did not improve from 0.71559\n",
      "Epoch 7/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.5030 - accuracy: 0.7447 - val_loss: 0.5162 - val_accuracy: 0.7267\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.71559 to 0.72673, saving model to model.h5\n",
      "Epoch 8/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.5225 - accuracy: 0.7372 - val_loss: 0.5685 - val_accuracy: 0.6989\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.72673\n",
      "Epoch 9/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.4903 - accuracy: 0.7565 - val_loss: 0.4849 - val_accuracy: 0.7530\n",
      "\n",
      "Epoch 00009: val_accuracy improved from 0.72673 to 0.75298, saving model to model.h5\n",
      "Epoch 10/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.4627 - accuracy: 0.7747 - val_loss: 0.6796 - val_accuracy: 0.6169\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.75298\n",
      "Epoch 11/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.4556 - accuracy: 0.7749 - val_loss: 0.7493 - val_accuracy: 0.5692\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.75298\n",
      "Epoch 12/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.4592 - accuracy: 0.7690 - val_loss: 0.6002 - val_accuracy: 0.7224\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.75298\n",
      "Epoch 13/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.4468 - accuracy: 0.7840 - val_loss: 0.5473 - val_accuracy: 0.7239\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.75298\n",
      "Epoch 14/200\n",
      "236/236 [==============================] - 15s 64ms/step - loss: 0.4319 - accuracy: 0.7929 - val_loss: 0.5686 - val_accuracy: 0.6877\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.75298\n",
      "Epoch 15/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.4263 - accuracy: 0.7924 - val_loss: 0.4287 - val_accuracy: 0.7808\n",
      "\n",
      "Epoch 00015: val_accuracy improved from 0.75298 to 0.78083, saving model to model.h5\n",
      "Epoch 16/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.4122 - accuracy: 0.8033 - val_loss: 1.3433 - val_accuracy: 0.5680\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.78083\n",
      "Epoch 17/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.4433 - accuracy: 0.7794 - val_loss: 0.5410 - val_accuracy: 0.7391\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.78083\n",
      "Epoch 18/200\n",
      "236/236 [==============================] - 15s 64ms/step - loss: 0.4203 - accuracy: 0.8000 - val_loss: 0.3953 - val_accuracy: 0.8055\n",
      "\n",
      "Epoch 00018: val_accuracy improved from 0.78083 to 0.80549, saving model to model.h5\n",
      "Epoch 19/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.4276 - accuracy: 0.7935 - val_loss: 0.8550 - val_accuracy: 0.5963\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.80549\n",
      "Epoch 20/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.3906 - accuracy: 0.8076 - val_loss: 0.9414 - val_accuracy: 0.5986\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.80549\n",
      "Epoch 21/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.3839 - accuracy: 0.8185 - val_loss: 0.4998 - val_accuracy: 0.7784\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.80549\n",
      "Epoch 22/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.3909 - accuracy: 0.8107 - val_loss: 0.3609 - val_accuracy: 0.8222\n",
      "\n",
      "Epoch 00022: val_accuracy improved from 0.80549 to 0.82220, saving model to model.h5\n",
      "Epoch 23/200\n",
      "236/236 [==============================] - 15s 64ms/step - loss: 0.3742 - accuracy: 0.8313 - val_loss: 0.4735 - val_accuracy: 0.7856\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.82220\n",
      "Epoch 24/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.3563 - accuracy: 0.8302 - val_loss: 0.4847 - val_accuracy: 0.7717\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.82220\n",
      "Epoch 25/200\n",
      "236/236 [==============================] - 15s 64ms/step - loss: 0.3575 - accuracy: 0.8284 - val_loss: 0.3964 - val_accuracy: 0.7932\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.82220\n",
      "Epoch 26/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.3595 - accuracy: 0.8350 - val_loss: 0.3793 - val_accuracy: 0.8286\n",
      "\n",
      "Epoch 00026: val_accuracy improved from 0.82220 to 0.82856, saving model to model.h5\n",
      "Epoch 27/200\n",
      "236/236 [==============================] - 16s 67ms/step - loss: 0.3510 - accuracy: 0.8349 - val_loss: 0.4041 - val_accuracy: 0.8126\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.82856\n",
      "Epoch 28/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.3620 - accuracy: 0.8294 - val_loss: 0.3790 - val_accuracy: 0.8329\n",
      "\n",
      "Epoch 00028: val_accuracy improved from 0.82856 to 0.83294, saving model to model.h5\n",
      "Epoch 29/200\n",
      "236/236 [==============================] - 15s 64ms/step - loss: 0.3448 - accuracy: 0.8336 - val_loss: 0.4019 - val_accuracy: 0.8079\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.83294\n",
      "Epoch 30/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.3456 - accuracy: 0.8404 - val_loss: 0.4486 - val_accuracy: 0.7745\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.83294\n",
      "Epoch 31/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.3244 - accuracy: 0.8487 - val_loss: 0.5253 - val_accuracy: 0.7478\n",
      "\n",
      "Epoch 00031: val_accuracy did not improve from 0.83294\n",
      "Epoch 32/200\n",
      "236/236 [==============================] - 16s 66ms/step - loss: 0.3405 - accuracy: 0.8457 - val_loss: 0.3680 - val_accuracy: 0.8278\n",
      "\n",
      "Epoch 00032: val_accuracy did not improve from 0.83294\n",
      "Epoch 33/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.3293 - accuracy: 0.8457 - val_loss: 0.8413 - val_accuracy: 0.6905\n",
      "\n",
      "Epoch 00033: val_accuracy did not improve from 0.83294\n",
      "Epoch 34/200\n",
      "236/236 [==============================] - 16s 66ms/step - loss: 0.3211 - accuracy: 0.8505 - val_loss: 0.8345 - val_accuracy: 0.6647\n",
      "\n",
      "Epoch 00034: val_accuracy did not improve from 0.83294\n",
      "Epoch 35/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.3278 - accuracy: 0.8485 - val_loss: 0.3268 - val_accuracy: 0.8564\n",
      "\n",
      "Epoch 00035: val_accuracy improved from 0.83294 to 0.85640, saving model to model.h5\n",
      "Epoch 36/200\n",
      "236/236 [==============================] - 16s 68ms/step - loss: 0.3185 - accuracy: 0.8576 - val_loss: 0.3951 - val_accuracy: 0.8214\n",
      "\n",
      "Epoch 00036: val_accuracy did not improve from 0.85640\n",
      "Epoch 37/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.3435 - accuracy: 0.8451 - val_loss: 0.4360 - val_accuracy: 0.8043\n",
      "\n",
      "Epoch 00037: val_accuracy did not improve from 0.85640\n",
      "Epoch 38/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.3114 - accuracy: 0.8555 - val_loss: 0.4537 - val_accuracy: 0.8111\n",
      "\n",
      "Epoch 00038: val_accuracy did not improve from 0.85640\n",
      "Epoch 39/200\n",
      "236/236 [==============================] - 16s 70ms/step - loss: 0.3100 - accuracy: 0.8579 - val_loss: 0.4221 - val_accuracy: 0.7820\n",
      "\n",
      "Epoch 00039: val_accuracy did not improve from 0.85640\n",
      "Epoch 40/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.3006 - accuracy: 0.8673 - val_loss: 0.3635 - val_accuracy: 0.8218\n",
      "\n",
      "Epoch 00040: val_accuracy did not improve from 0.85640\n",
      "Epoch 41/200\n",
      "236/236 [==============================] - 16s 68ms/step - loss: 0.3120 - accuracy: 0.8537 - val_loss: 0.4512 - val_accuracy: 0.7904\n",
      "\n",
      "Epoch 00041: val_accuracy did not improve from 0.85640\n",
      "Epoch 42/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.3012 - accuracy: 0.8674 - val_loss: 0.3889 - val_accuracy: 0.8329\n",
      "\n",
      "Epoch 00042: val_accuracy did not improve from 0.85640\n",
      "Epoch 43/200\n",
      "236/236 [==============================] - 16s 68ms/step - loss: 0.3093 - accuracy: 0.8489 - val_loss: 0.3284 - val_accuracy: 0.8433\n",
      "\n",
      "Epoch 00043: val_accuracy did not improve from 0.85640\n",
      "Epoch 44/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.2850 - accuracy: 0.8697 - val_loss: 0.3190 - val_accuracy: 0.8552\n",
      "\n",
      "Epoch 00044: val_accuracy did not improve from 0.85640\n",
      "Epoch 45/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.2828 - accuracy: 0.8642 - val_loss: 0.3793 - val_accuracy: 0.8103\n",
      "\n",
      "Epoch 00045: val_accuracy did not improve from 0.85640\n",
      "Epoch 46/200\n",
      "236/236 [==============================] - 16s 66ms/step - loss: 0.2795 - accuracy: 0.8794 - val_loss: 0.3439 - val_accuracy: 0.8711\n",
      "\n",
      "Epoch 00046: val_accuracy improved from 0.85640 to 0.87112, saving model to model.h5\n",
      "Epoch 47/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2683 - accuracy: 0.8807 - val_loss: 0.3245 - val_accuracy: 0.8492\n",
      "\n",
      "Epoch 00047: val_accuracy did not improve from 0.87112\n",
      "Epoch 48/200\n",
      "236/236 [==============================] - 16s 69ms/step - loss: 0.2728 - accuracy: 0.8777 - val_loss: 0.4349 - val_accuracy: 0.8023\n",
      "\n",
      "Epoch 00048: val_accuracy did not improve from 0.87112\n",
      "Epoch 49/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2747 - accuracy: 0.8799 - val_loss: 0.3393 - val_accuracy: 0.8560\n",
      "\n",
      "Epoch 00049: val_accuracy did not improve from 0.87112\n",
      "Epoch 50/200\n",
      "236/236 [==============================] - 16s 69ms/step - loss: 0.2705 - accuracy: 0.8807 - val_loss: 0.3329 - val_accuracy: 0.8512\n",
      "\n",
      "Epoch 00050: val_accuracy did not improve from 0.87112\n",
      "Epoch 51/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.2630 - accuracy: 0.8871 - val_loss: 0.5111 - val_accuracy: 0.7932\n",
      "\n",
      "Epoch 00051: val_accuracy did not improve from 0.87112\n",
      "Epoch 52/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2664 - accuracy: 0.8770 - val_loss: 0.2645 - val_accuracy: 0.8910\n",
      "\n",
      "Epoch 00052: val_accuracy improved from 0.87112 to 0.89101, saving model to model.h5\n",
      "Epoch 53/200\n",
      "236/236 [==============================] - 16s 69ms/step - loss: 0.2660 - accuracy: 0.8894 - val_loss: 0.7855 - val_accuracy: 0.7259\n",
      "\n",
      "Epoch 00053: val_accuracy did not improve from 0.89101\n",
      "Epoch 54/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2598 - accuracy: 0.8837 - val_loss: 0.3143 - val_accuracy: 0.8648\n",
      "\n",
      "Epoch 00054: val_accuracy did not improve from 0.89101\n",
      "Epoch 55/200\n",
      "236/236 [==============================] - 16s 67ms/step - loss: 0.2588 - accuracy: 0.8894 - val_loss: 0.3339 - val_accuracy: 0.8528\n",
      "\n",
      "Epoch 00055: val_accuracy did not improve from 0.89101\n",
      "Epoch 56/200\n",
      "236/236 [==============================] - 15s 62ms/step - loss: 0.2583 - accuracy: 0.8914 - val_loss: 0.3974 - val_accuracy: 0.8397\n",
      "\n",
      "Epoch 00056: val_accuracy did not improve from 0.89101\n",
      "Epoch 57/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.2484 - accuracy: 0.8910 - val_loss: 0.3094 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00057: val_accuracy did not improve from 0.89101\n",
      "Epoch 58/200\n",
      "236/236 [==============================] - 16s 70ms/step - loss: 0.2456 - accuracy: 0.8863 - val_loss: 0.4211 - val_accuracy: 0.8214\n",
      "\n",
      "Epoch 00058: val_accuracy did not improve from 0.89101\n",
      "Epoch 59/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2477 - accuracy: 0.8937 - val_loss: 0.3242 - val_accuracy: 0.8612\n",
      "\n",
      "Epoch 00059: val_accuracy did not improve from 0.89101\n",
      "Epoch 60/200\n",
      "236/236 [==============================] - 18s 76ms/step - loss: 0.2586 - accuracy: 0.8886 - val_loss: 0.3016 - val_accuracy: 0.8695\n",
      "\n",
      "Epoch 00060: val_accuracy did not improve from 0.89101\n",
      "Epoch 61/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2378 - accuracy: 0.8963 - val_loss: 0.4213 - val_accuracy: 0.8238\n",
      "\n",
      "Epoch 00061: val_accuracy did not improve from 0.89101\n",
      "Epoch 62/200\n",
      "236/236 [==============================] - 16s 69ms/step - loss: 0.2237 - accuracy: 0.9023 - val_loss: 0.4026 - val_accuracy: 0.8719\n",
      "\n",
      "Epoch 00062: val_accuracy did not improve from 0.89101\n",
      "Epoch 63/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.2616 - accuracy: 0.8930 - val_loss: 0.3816 - val_accuracy: 0.8508\n",
      "\n",
      "Epoch 00063: val_accuracy did not improve from 0.89101\n",
      "Epoch 64/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2329 - accuracy: 0.8948 - val_loss: 0.2625 - val_accuracy: 0.8846\n",
      "\n",
      "Epoch 00064: val_accuracy did not improve from 0.89101\n",
      "Epoch 65/200\n",
      "236/236 [==============================] - 17s 71ms/step - loss: 0.2282 - accuracy: 0.9028 - val_loss: 0.2970 - val_accuracy: 0.8803\n",
      "\n",
      "Epoch 00065: val_accuracy did not improve from 0.89101\n",
      "Epoch 66/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.2409 - accuracy: 0.8984 - val_loss: 0.3270 - val_accuracy: 0.8663\n",
      "\n",
      "Epoch 00066: val_accuracy did not improve from 0.89101\n",
      "Epoch 67/200\n",
      "236/236 [==============================] - 17s 73ms/step - loss: 0.2314 - accuracy: 0.9048 - val_loss: 0.2787 - val_accuracy: 0.8914\n",
      "\n",
      "Epoch 00067: val_accuracy improved from 0.89101 to 0.89141, saving model to model.h5\n",
      "Epoch 68/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.2344 - accuracy: 0.9000 - val_loss: 0.3197 - val_accuracy: 0.8568\n",
      "\n",
      "Epoch 00068: val_accuracy did not improve from 0.89141\n",
      "Epoch 69/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.2205 - accuracy: 0.9096 - val_loss: 0.2884 - val_accuracy: 0.8898\n",
      "\n",
      "Epoch 00069: val_accuracy did not improve from 0.89141\n",
      "Epoch 70/200\n",
      "236/236 [==============================] - 17s 70ms/step - loss: 0.2262 - accuracy: 0.9055 - val_loss: 0.3849 - val_accuracy: 0.8528\n",
      "\n",
      "Epoch 00070: val_accuracy did not improve from 0.89141\n",
      "Epoch 71/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2244 - accuracy: 0.9055 - val_loss: 0.3574 - val_accuracy: 0.8512\n",
      "\n",
      "Epoch 00071: val_accuracy did not improve from 0.89141\n",
      "Epoch 72/200\n",
      "236/236 [==============================] - 17s 72ms/step - loss: 0.2128 - accuracy: 0.9034 - val_loss: 0.3219 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00072: val_accuracy did not improve from 0.89141\n",
      "Epoch 73/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2299 - accuracy: 0.9008 - val_loss: 0.3741 - val_accuracy: 0.8687\n",
      "\n",
      "Epoch 00073: val_accuracy did not improve from 0.89141\n",
      "Epoch 74/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.2226 - accuracy: 0.9062 - val_loss: 0.4790 - val_accuracy: 0.8373\n",
      "\n",
      "Epoch 00074: val_accuracy did not improve from 0.89141\n",
      "Epoch 75/200\n",
      "236/236 [==============================] - 18s 75ms/step - loss: 0.2148 - accuracy: 0.9068 - val_loss: 0.5035 - val_accuracy: 0.7904\n",
      "\n",
      "Epoch 00075: val_accuracy did not improve from 0.89141\n",
      "Epoch 76/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2175 - accuracy: 0.9084 - val_loss: 0.3691 - val_accuracy: 0.8588\n",
      "\n",
      "Epoch 00076: val_accuracy did not improve from 0.89141\n",
      "Epoch 77/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2213 - accuracy: 0.9031 - val_loss: 0.5171 - val_accuracy: 0.8075\n",
      "\n",
      "Epoch 00077: val_accuracy did not improve from 0.89141\n",
      "Epoch 78/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.2077 - accuracy: 0.9101 - val_loss: 0.3680 - val_accuracy: 0.8671\n",
      "\n",
      "Epoch 00078: val_accuracy did not improve from 0.89141\n",
      "Epoch 79/200\n",
      "236/236 [==============================] - 16s 70ms/step - loss: 0.2095 - accuracy: 0.9146 - val_loss: 0.3781 - val_accuracy: 0.8576\n",
      "\n",
      "Epoch 00079: val_accuracy did not improve from 0.89141\n",
      "Epoch 80/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.2042 - accuracy: 0.9137 - val_loss: 0.5487 - val_accuracy: 0.7570\n",
      "\n",
      "Epoch 00080: val_accuracy did not improve from 0.89141\n",
      "Epoch 81/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1998 - accuracy: 0.9165 - val_loss: 0.3565 - val_accuracy: 0.8715\n",
      "\n",
      "Epoch 00081: val_accuracy did not improve from 0.89141\n",
      "Epoch 82/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.2028 - accuracy: 0.9185 - val_loss: 0.3289 - val_accuracy: 0.8922\n",
      "\n",
      "Epoch 00082: val_accuracy improved from 0.89141 to 0.89220, saving model to model.h5\n",
      "Epoch 83/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.2033 - accuracy: 0.9142 - val_loss: 0.2876 - val_accuracy: 0.8946\n",
      "\n",
      "Epoch 00083: val_accuracy improved from 0.89220 to 0.89459, saving model to model.h5\n",
      "Epoch 84/200\n",
      "236/236 [==============================] - 18s 78ms/step - loss: 0.2026 - accuracy: 0.9181 - val_loss: 0.3073 - val_accuracy: 0.8735\n",
      "\n",
      "Epoch 00084: val_accuracy did not improve from 0.89459\n",
      "Epoch 85/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.2082 - accuracy: 0.9186 - val_loss: 0.5053 - val_accuracy: 0.7991\n",
      "\n",
      "Epoch 00085: val_accuracy did not improve from 0.89459\n",
      "Epoch 86/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.1937 - accuracy: 0.9206 - val_loss: 0.3091 - val_accuracy: 0.8815\n",
      "\n",
      "Epoch 00086: val_accuracy did not improve from 0.89459\n",
      "Epoch 87/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1999 - accuracy: 0.9192 - val_loss: 0.3046 - val_accuracy: 0.8839\n",
      "\n",
      "Epoch 00087: val_accuracy did not improve from 0.89459\n",
      "Epoch 88/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1912 - accuracy: 0.9161 - val_loss: 0.3299 - val_accuracy: 0.8687\n",
      "\n",
      "Epoch 00088: val_accuracy did not improve from 0.89459\n",
      "Epoch 89/200\n",
      "236/236 [==============================] - 18s 77ms/step - loss: 0.1980 - accuracy: 0.9145 - val_loss: 0.3410 - val_accuracy: 0.8663\n",
      "\n",
      "Epoch 00089: val_accuracy did not improve from 0.89459\n",
      "Epoch 90/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1914 - accuracy: 0.9177 - val_loss: 0.3595 - val_accuracy: 0.8691\n",
      "\n",
      "Epoch 00090: val_accuracy did not improve from 0.89459\n",
      "Epoch 91/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1845 - accuracy: 0.9232 - val_loss: 0.5485 - val_accuracy: 0.8134\n",
      "\n",
      "Epoch 00091: val_accuracy did not improve from 0.89459\n",
      "Epoch 92/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1878 - accuracy: 0.9204 - val_loss: 0.3945 - val_accuracy: 0.8429\n",
      "\n",
      "Epoch 00092: val_accuracy did not improve from 0.89459\n",
      "Epoch 93/200\n",
      "236/236 [==============================] - 17s 74ms/step - loss: 0.1784 - accuracy: 0.9250 - val_loss: 0.4055 - val_accuracy: 0.8640\n",
      "\n",
      "Epoch 00093: val_accuracy did not improve from 0.89459\n",
      "Epoch 94/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.1927 - accuracy: 0.9200 - val_loss: 0.3167 - val_accuracy: 0.8866\n",
      "\n",
      "Epoch 00094: val_accuracy did not improve from 0.89459\n",
      "Epoch 95/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.1777 - accuracy: 0.9281 - val_loss: 0.3377 - val_accuracy: 0.8675\n",
      "\n",
      "Epoch 00095: val_accuracy did not improve from 0.89459\n",
      "Epoch 96/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1842 - accuracy: 0.9275 - val_loss: 0.3444 - val_accuracy: 0.8831\n",
      "\n",
      "Epoch 00096: val_accuracy did not improve from 0.89459\n",
      "Epoch 97/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1947 - accuracy: 0.9230 - val_loss: 0.3758 - val_accuracy: 0.8667\n",
      "\n",
      "Epoch 00097: val_accuracy did not improve from 0.89459\n",
      "Epoch 98/200\n",
      "236/236 [==============================] - 18s 75ms/step - loss: 0.1857 - accuracy: 0.9242 - val_loss: 0.2837 - val_accuracy: 0.8978\n",
      "\n",
      "Epoch 00098: val_accuracy improved from 0.89459 to 0.89777, saving model to model.h5\n",
      "Epoch 99/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1824 - accuracy: 0.9275 - val_loss: 0.2603 - val_accuracy: 0.9077\n",
      "\n",
      "Epoch 00099: val_accuracy improved from 0.89777 to 0.90772, saving model to model.h5\n",
      "Epoch 100/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1872 - accuracy: 0.9227 - val_loss: 0.5047 - val_accuracy: 0.8254\n",
      "\n",
      "Epoch 00100: val_accuracy did not improve from 0.90772\n",
      "Epoch 101/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1686 - accuracy: 0.9304 - val_loss: 0.3597 - val_accuracy: 0.8755\n",
      "\n",
      "Epoch 00101: val_accuracy did not improve from 0.90772\n",
      "Epoch 102/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1751 - accuracy: 0.9260 - val_loss: 0.3311 - val_accuracy: 0.8759\n",
      "\n",
      "Epoch 00102: val_accuracy did not improve from 0.90772\n",
      "Epoch 103/200\n",
      "236/236 [==============================] - 18s 76ms/step - loss: 0.1787 - accuracy: 0.9286 - val_loss: 0.4779 - val_accuracy: 0.8341\n",
      "\n",
      "Epoch 00103: val_accuracy did not improve from 0.90772\n",
      "Epoch 104/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1842 - accuracy: 0.9163 - val_loss: 0.5058 - val_accuracy: 0.8413\n",
      "\n",
      "Epoch 00104: val_accuracy did not improve from 0.90772\n",
      "Epoch 105/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.1722 - accuracy: 0.9303 - val_loss: 0.3035 - val_accuracy: 0.9025\n",
      "\n",
      "Epoch 00105: val_accuracy did not improve from 0.90772\n",
      "Epoch 106/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1624 - accuracy: 0.9298 - val_loss: 0.3825 - val_accuracy: 0.8560\n",
      "\n",
      "Epoch 00106: val_accuracy did not improve from 0.90772\n",
      "Epoch 107/200\n",
      "236/236 [==============================] - 15s 63ms/step - loss: 0.1626 - accuracy: 0.9329 - val_loss: 0.4755 - val_accuracy: 0.8107\n",
      "\n",
      "Epoch 00107: val_accuracy did not improve from 0.90772\n",
      "Epoch 108/200\n",
      "236/236 [==============================] - 17s 73ms/step - loss: 0.1673 - accuracy: 0.9368 - val_loss: 0.3502 - val_accuracy: 0.8644\n",
      "\n",
      "Epoch 00108: val_accuracy did not improve from 0.90772\n",
      "Epoch 109/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1645 - accuracy: 0.9274 - val_loss: 0.4235 - val_accuracy: 0.8437\n",
      "\n",
      "Epoch 00109: val_accuracy did not improve from 0.90772\n",
      "Epoch 110/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1670 - accuracy: 0.9269 - val_loss: 0.4053 - val_accuracy: 0.8775\n",
      "\n",
      "Epoch 00110: val_accuracy did not improve from 0.90772\n",
      "Epoch 111/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1595 - accuracy: 0.9298 - val_loss: 0.9505 - val_accuracy: 0.7383\n",
      "\n",
      "Epoch 00111: val_accuracy did not improve from 0.90772\n",
      "Epoch 112/200\n",
      "236/236 [==============================] - 16s 69ms/step - loss: 0.1806 - accuracy: 0.9275 - val_loss: 0.3286 - val_accuracy: 0.8886\n",
      "\n",
      "Epoch 00112: val_accuracy did not improve from 0.90772\n",
      "Epoch 113/200\n",
      "236/236 [==============================] - 16s 67ms/step - loss: 0.1443 - accuracy: 0.9392 - val_loss: 0.6400 - val_accuracy: 0.8290\n",
      "\n",
      "Epoch 00113: val_accuracy did not improve from 0.90772\n",
      "Epoch 114/200\n",
      "236/236 [==============================] - 13s 56ms/step - loss: 0.1662 - accuracy: 0.9321 - val_loss: 0.5721 - val_accuracy: 0.8481\n",
      "\n",
      "Epoch 00114: val_accuracy did not improve from 0.90772\n",
      "Epoch 115/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1592 - accuracy: 0.9334 - val_loss: 0.5774 - val_accuracy: 0.8031\n",
      "\n",
      "Epoch 00115: val_accuracy did not improve from 0.90772\n",
      "Epoch 116/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1616 - accuracy: 0.9360 - val_loss: 0.3810 - val_accuracy: 0.8687\n",
      "\n",
      "Epoch 00116: val_accuracy did not improve from 0.90772\n",
      "Epoch 117/200\n",
      "236/236 [==============================] - 18s 76ms/step - loss: 0.1627 - accuracy: 0.9399 - val_loss: 0.7232 - val_accuracy: 0.8130\n",
      "\n",
      "Epoch 00117: val_accuracy did not improve from 0.90772\n",
      "Epoch 118/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1590 - accuracy: 0.9350 - val_loss: 0.3454 - val_accuracy: 0.8986\n",
      "\n",
      "Epoch 00118: val_accuracy did not improve from 0.90772\n",
      "Epoch 119/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1487 - accuracy: 0.9400 - val_loss: 0.5041 - val_accuracy: 0.8437\n",
      "\n",
      "Epoch 00119: val_accuracy did not improve from 0.90772\n",
      "Epoch 120/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.1554 - accuracy: 0.9409 - val_loss: 0.4259 - val_accuracy: 0.8683\n",
      "\n",
      "Epoch 00120: val_accuracy did not improve from 0.90772\n",
      "Epoch 121/200\n",
      "236/236 [==============================] - 13s 56ms/step - loss: 0.1687 - accuracy: 0.9351 - val_loss: 0.4427 - val_accuracy: 0.8636\n",
      "\n",
      "Epoch 00121: val_accuracy did not improve from 0.90772\n",
      "Epoch 122/200\n",
      "236/236 [==============================] - 19s 80ms/step - loss: 0.1480 - accuracy: 0.9384 - val_loss: 0.4694 - val_accuracy: 0.8636\n",
      "\n",
      "Epoch 00122: val_accuracy did not improve from 0.90772\n",
      "Epoch 123/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1647 - accuracy: 0.9366 - val_loss: 0.5300 - val_accuracy: 0.8345\n",
      "\n",
      "Epoch 00123: val_accuracy did not improve from 0.90772\n",
      "Epoch 124/200\n",
      "236/236 [==============================] - 14s 61ms/step - loss: 0.1481 - accuracy: 0.9423 - val_loss: 0.3757 - val_accuracy: 0.8846\n",
      "\n",
      "Epoch 00124: val_accuracy did not improve from 0.90772\n",
      "Epoch 125/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1390 - accuracy: 0.9445 - val_loss: 0.3132 - val_accuracy: 0.9033\n",
      "\n",
      "Epoch 00125: val_accuracy did not improve from 0.90772\n",
      "Epoch 126/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.1440 - accuracy: 0.9393 - val_loss: 0.3869 - val_accuracy: 0.8632\n",
      "\n",
      "Epoch 00126: val_accuracy did not improve from 0.90772\n",
      "Epoch 127/200\n",
      "236/236 [==============================] - 19s 80ms/step - loss: 0.1433 - accuracy: 0.9432 - val_loss: 0.4720 - val_accuracy: 0.8349\n",
      "\n",
      "Epoch 00127: val_accuracy did not improve from 0.90772\n",
      "Epoch 128/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1374 - accuracy: 0.9459 - val_loss: 0.4489 - val_accuracy: 0.8333\n",
      "\n",
      "Epoch 00128: val_accuracy did not improve from 0.90772\n",
      "Epoch 129/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.1535 - accuracy: 0.9338 - val_loss: 0.5644 - val_accuracy: 0.8477\n",
      "\n",
      "Epoch 00129: val_accuracy did not improve from 0.90772\n",
      "Epoch 130/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1546 - accuracy: 0.9371 - val_loss: 0.3306 - val_accuracy: 0.8926\n",
      "\n",
      "Epoch 00130: val_accuracy did not improve from 0.90772\n",
      "Epoch 131/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1511 - accuracy: 0.9419 - val_loss: 0.4128 - val_accuracy: 0.8878\n",
      "\n",
      "Epoch 00131: val_accuracy did not improve from 0.90772\n",
      "Epoch 132/200\n",
      "236/236 [==============================] - 19s 79ms/step - loss: 0.1387 - accuracy: 0.9408 - val_loss: 0.3475 - val_accuracy: 0.8608\n",
      "\n",
      "Epoch 00132: val_accuracy did not improve from 0.90772\n",
      "Epoch 133/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1573 - accuracy: 0.9366 - val_loss: 0.3479 - val_accuracy: 0.8771\n",
      "\n",
      "Epoch 00133: val_accuracy did not improve from 0.90772\n",
      "Epoch 134/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1539 - accuracy: 0.9344 - val_loss: 0.4775 - val_accuracy: 0.8369\n",
      "\n",
      "Epoch 00134: val_accuracy did not improve from 0.90772\n",
      "Epoch 135/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1376 - accuracy: 0.9455 - val_loss: 0.8126 - val_accuracy: 0.7693\n",
      "\n",
      "Epoch 00135: val_accuracy did not improve from 0.90772\n",
      "Epoch 136/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1511 - accuracy: 0.9364 - val_loss: 0.3212 - val_accuracy: 0.9037\n",
      "\n",
      "Epoch 00136: val_accuracy did not improve from 0.90772\n",
      "Epoch 137/200\n",
      "236/236 [==============================] - 19s 82ms/step - loss: 0.1368 - accuracy: 0.9480 - val_loss: 0.6748 - val_accuracy: 0.7757\n",
      "\n",
      "Epoch 00137: val_accuracy did not improve from 0.90772\n",
      "Epoch 138/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1467 - accuracy: 0.9442 - val_loss: 0.3665 - val_accuracy: 0.8783\n",
      "\n",
      "Epoch 00138: val_accuracy did not improve from 0.90772\n",
      "Epoch 139/200\n",
      "236/236 [==============================] - 14s 60ms/step - loss: 0.1462 - accuracy: 0.9411 - val_loss: 0.4620 - val_accuracy: 0.8469\n",
      "\n",
      "Epoch 00139: val_accuracy did not improve from 0.90772\n",
      "Epoch 140/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1405 - accuracy: 0.9407 - val_loss: 0.5363 - val_accuracy: 0.8496\n",
      "\n",
      "Epoch 00140: val_accuracy did not improve from 0.90772\n",
      "Epoch 141/200\n",
      "236/236 [==============================] - 14s 57ms/step - loss: 0.1345 - accuracy: 0.9457 - val_loss: 0.3157 - val_accuracy: 0.9065\n",
      "\n",
      "Epoch 00141: val_accuracy did not improve from 0.90772\n",
      "Epoch 142/200\n",
      "236/236 [==============================] - 20s 85ms/step - loss: 0.1395 - accuracy: 0.9431 - val_loss: 0.3273 - val_accuracy: 0.8906\n",
      "\n",
      "Epoch 00142: val_accuracy did not improve from 0.90772\n",
      "Epoch 143/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1424 - accuracy: 0.9428 - val_loss: 0.3514 - val_accuracy: 0.9089\n",
      "\n",
      "Epoch 00143: val_accuracy improved from 0.90772 to 0.90891, saving model to model.h5\n",
      "Epoch 144/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1335 - accuracy: 0.9458 - val_loss: 0.3695 - val_accuracy: 0.8890\n",
      "\n",
      "Epoch 00144: val_accuracy did not improve from 0.90891\n",
      "Epoch 145/200\n",
      "236/236 [==============================] - 13s 57ms/step - loss: 0.1309 - accuracy: 0.9490 - val_loss: 0.4046 - val_accuracy: 0.8536\n",
      "\n",
      "Epoch 00145: val_accuracy did not improve from 0.90891\n",
      "Epoch 146/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1256 - accuracy: 0.9511 - val_loss: 0.5138 - val_accuracy: 0.8488\n",
      "\n",
      "Epoch 00146: val_accuracy did not improve from 0.90891\n",
      "Epoch 147/200\n",
      "236/236 [==============================] - 19s 82ms/step - loss: 0.1365 - accuracy: 0.9471 - val_loss: 0.4329 - val_accuracy: 0.8492\n",
      "\n",
      "Epoch 00147: val_accuracy did not improve from 0.90891\n",
      "Epoch 148/200\n",
      "236/236 [==============================] - 14s 59ms/step - loss: 0.1403 - accuracy: 0.9382 - val_loss: 0.3777 - val_accuracy: 0.8866\n",
      "\n",
      "Epoch 00148: val_accuracy did not improve from 0.90891\n",
      "Epoch 149/200\n",
      "236/236 [==============================] - 14s 58ms/step - loss: 0.1275 - accuracy: 0.9464 - val_loss: 0.3779 - val_accuracy: 0.8982\n",
      "\n",
      "Epoch 00149: val_accuracy did not improve from 0.90891\n",
      "Epoch 00149: early stopping\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    train_gen,\n",
    "    steps_per_epoch= steps_per_epoch,\n",
    "    validation_data=val_gen,\n",
    "    validation_steps= validation_steps,\n",
    "    epochs=epochs,\n",
    "    callbacks=[es,mc]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:59:17.646775Z",
     "iopub.status.busy": "2021-03-19T04:59:17.644992Z",
     "iopub.status.idle": "2021-03-19T04:59:17.647344Z",
     "shell.execute_reply": "2021-03-19T04:59:17.647751Z"
    },
    "papermill": {
     "duration": 8.032845,
     "end_time": "2021-03-19T04:59:17.647890",
     "exception": false,
     "start_time": "2021-03-19T04:59:09.615045",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def history_plot(model_id, history1):\n",
    "\tplt.figure(figsize=(14,16)) \n",
    "\t# plot loss\n",
    "\tplt.subplot(2, 1, 1)\n",
    "\tplt.title('Cross Entropy Loss - ' + model_id, fontsize = 12)\n",
    "\tplt.plot(history1.history['loss'], color='blue', label='train base')\n",
    "\tplt.plot(history1.history['val_loss'], color='cyan', label='val base')\n",
    "\tplt.xlabel(\"Epochs\",fontsize = 12)\n",
    "\tplt.ylabel(\"Loss\", fontsize = 12)\n",
    "\tplt.legend(loc='upper right')\n",
    "\n",
    "\t# plot accuracy\n",
    "\tplt.subplot(2, 1, 2)\n",
    "\tplt.title('Classification Accuracy ' + model_id, fontsize = 10)\n",
    "\tplt.plot(history1.history['accuracy'], color='blue', label='train base')\n",
    "\tplt.plot(history1.history['val_accuracy'], color='cyan', label='val base')\n",
    "\tplt.xlabel(\"Epochs\", fontsize = 12)\n",
    "\tplt.ylabel(\"Accuracy\", fontsize = 12)\n",
    "\tplt.legend(loc='lower right')\n",
    "\n",
    "\tplt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:59:33.564721Z",
     "iopub.status.busy": "2021-03-19T04:59:33.563850Z",
     "iopub.status.idle": "2021-03-19T04:59:33.566666Z",
     "shell.execute_reply": "2021-03-19T04:59:33.567070Z"
    },
    "papermill": {
     "duration": 7.834441,
     "end_time": "2021-03-19T04:59:33.567205",
     "exception": false,
     "start_time": "2021-03-19T04:59:25.732764",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.history.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T04:59:50.166579Z",
     "iopub.status.busy": "2021-03-19T04:59:50.165698Z",
     "iopub.status.idle": "2021-03-19T04:59:50.496298Z",
     "shell.execute_reply": "2021-03-19T04:59:50.496718Z"
    },
    "papermill": {
     "duration": 8.386147,
     "end_time": "2021-03-19T04:59:50.496876",
     "exception": false,
     "start_time": "2021-03-19T04:59:42.110729",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x1152 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "history_plot('CNN Model', history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:00:06.534546Z",
     "iopub.status.busy": "2021-03-19T05:00:06.533624Z",
     "iopub.status.idle": "2021-03-19T05:00:06.550011Z",
     "shell.execute_reply": "2021-03-19T05:00:06.549126Z"
    },
    "papermill": {
     "duration": 7.951277,
     "end_time": "2021-03-19T05:00:06.550134",
     "exception": false,
     "start_time": "2021-03-19T04:59:58.598857",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "model.load_weights('model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:00:23.331770Z",
     "iopub.status.busy": "2021-03-19T05:00:23.330042Z",
     "iopub.status.idle": "2021-03-19T05:00:23.332340Z",
     "shell.execute_reply": "2021-03-19T05:00:23.332733Z"
    },
    "papermill": {
     "duration": 8.22497,
     "end_time": "2021-03-19T05:00:23.332874",
     "exception": false,
     "start_time": "2021-03-19T05:00:15.107904",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "category_names = [\"No Defect\", \"Defect\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:00:39.269147Z",
     "iopub.status.busy": "2021-03-19T05:00:39.268267Z",
     "iopub.status.idle": "2021-03-19T05:00:41.038787Z",
     "shell.execute_reply": "2021-03-19T05:00:41.039185Z"
    },
    "papermill": {
     "duration": 9.583871,
     "end_time": "2021-03-19T05:00:41.039341",
     "exception": false,
     "start_time": "2021-03-19T05:00:31.455470",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicting...\n",
      "79/79 [==============================] - 2s 20ms/step\n"
     ]
    }
   ],
   "source": [
    "y_test = df_test['defect_label']\n",
    "\n",
    "STEP_SIZE_TEST = compute_steps_per_epoch(y_test.shape[0])\n",
    "print('predicting...')\n",
    "\n",
    "predictions = model.predict(\n",
    "    test_gen,\n",
    "    steps = STEP_SIZE_TEST,\n",
    "    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:00:57.416316Z",
     "iopub.status.busy": "2021-03-19T05:00:57.414541Z",
     "iopub.status.idle": "2021-03-19T05:00:57.416891Z",
     "shell.execute_reply": "2021-03-19T05:00:57.417282Z"
    },
    "papermill": {
     "duration": 8.052417,
     "end_time": "2021-03-19T05:00:57.417415",
     "exception": false,
     "start_time": "2021-03-19T05:00:49.364998",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "y_pred = (predictions > 0.5).astype(\"int32\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:01:13.694374Z",
     "iopub.status.busy": "2021-03-19T05:01:13.693513Z",
     "iopub.status.idle": "2021-03-19T05:01:13.704169Z",
     "shell.execute_reply": "2021-03-19T05:01:13.704674Z"
    },
    "papermill": {
     "duration": 8.114658,
     "end_time": "2021-03-19T05:01:13.704848",
     "exception": false,
     "start_time": "2021-03-19T05:01:05.590190",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "   No Defect       0.92      0.92      0.92      1181\n",
      "      Defect       0.93      0.93      0.93      1333\n",
      "\n",
      "    accuracy                           0.92      2514\n",
      "   macro avg       0.92      0.92      0.92      2514\n",
      "weighted avg       0.92      0.92      0.92      2514\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# classification report\n",
    "report_string = classification_report(y_test, y_pred, target_names=category_names)\n",
    "print(report_string)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:01:30.235595Z",
     "iopub.status.busy": "2021-03-19T05:01:30.234793Z",
     "iopub.status.idle": "2021-03-19T05:01:30.404304Z",
     "shell.execute_reply": "2021-03-19T05:01:30.404725Z"
    },
    "papermill": {
     "duration": 8.276056,
     "end_time": "2021-03-19T05:01:30.404890",
     "exception": false,
     "start_time": "2021-03-19T05:01:22.128834",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_cm(y_true, y_pred, figsize=(6,5)):\n",
    "    cm = confusion_matrix(y_true, y_pred, labels=np.unique(y_true))\n",
    "    cm_sum = np.sum(cm, axis=1, keepdims=True)\n",
    "    cm_perc = cm / cm_sum.astype(float) * 100\n",
    "    annot = np.empty_like(cm).astype(str)\n",
    "    nrows, ncols = cm.shape\n",
    "    for i in range(nrows):\n",
    "        for j in range(ncols):\n",
    "            c = cm[i, j]\n",
    "            p = cm_perc[i, j]\n",
    "            if i == j:\n",
    "                s = cm_sum[i]\n",
    "                annot[i, j] = '%.1f%%\\n%d/%d' % (p, c, s)\n",
    "            elif c == 0:\n",
    "                annot[i, j] = ''\n",
    "            elif i<j:\n",
    "                annot[i, j] = '%.1f%%\\n%d\\nFP' % (p, c)\n",
    "            else:\n",
    "                annot[i, j] = '%.1f%%\\n%d\\nFN' % (p, c)\n",
    "    #cm = pd.DataFrame(cm, index=np.unique(y_true), columns=np.unique(y_true))\n",
    "    cm = pd.DataFrame(cm, index=category_names, columns=category_names)\n",
    "    cm.index.name = 'Actual'\n",
    "    cm.columns.name = 'Predicted'\n",
    "    fig, ax = plt.subplots(figsize=figsize)\n",
    "    sns.set(font_scale=1.4)\n",
    "    sns.heatmap(cm, cmap= \"YlGnBu\", annot=annot, fmt='', ax=ax, annot_kws={\"fontsize\":12})\n",
    "\n",
    "plot_cm(y_test, y_pred, figsize=(6,5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:01:46.778982Z",
     "iopub.status.busy": "2021-03-19T05:01:46.778054Z",
     "iopub.status.idle": "2021-03-19T05:01:46.967399Z",
     "shell.execute_reply": "2021-03-19T05:01:46.966900Z"
    },
    "papermill": {
     "duration": 8.419396,
     "end_time": "2021-03-19T05:01:46.967530",
     "exception": false,
     "start_time": "2021-03-19T05:01:38.548134",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROC AUC=0.974\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAAEWCAYAAACufwpNAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Il7ecAAAACXBIWXMAAAsTAAALEwEAmpwYAABjhElEQVR4nO3dd1hT1xvA8W/YIgREqCCCA8WJA0Vwb1FwW6vWWVdbR911trV1T6pVa12tWnet1AHULVrr3ltR3IMlYYdxf3/wIzUyDBgI43yep8/T3HPHexLMm3vPkkmSJCEIgiAIWqCn6wAEQRCEwkMkFUEQBEFrRFIRBEEQtEYkFUEQBEFrRFIRBEEQtEYkFUEQBEFrRFIRBEEQtMZA1wHoWkREDCkp2R+qU7KkGWFh0bkQUf4l6lw0iDoXDTmts56ejBIlimdaXuSTSkqKlKOkknZsUSPqXDSIOhcNuVFn8fhLEARB0BqdJpVHjx7x7bff0rlzZ6pVq0aHDh00PtbX15d27drh4uKCt7c3fn5+uRipIAiCoAmdPv66d+8ex48fp1atWqSkpKDpNGQBAQFMmjSJYcOG0ahRIw4dOsS4ceMoXrw4zZo1y+WoBUEQhMzoNKm0bNmS1q1bAzB58mSuX7+u0XFLly6lXbt2jB8/HgAPDw8ePnzITz/9JJKKIAiCDun08ZeeXvYv/+TJEx48eIC3t7fadm9vb65du0Z4eLi2whMEQRCyqcA11D948AAAJycnte0VK1ZUKxcEQRDSkySJ5OSUXDt/getSHBkZCYBcLlfbbmFhoVYuCLnp2OVnnLnxKt32Ri52NK5pR1SskpW70z/ObeFqT/2qpQhXxLNm78105Z71HaldyZoXYTFsDLiTrrxDo3JUL2fF41dRbD10L11592ZOVCxjwf2nkew6HpSuvHfrSjiWMudGcDj7/glOV96/XWVsbMy5fC+Uv88+Tlc+tGM1rOQmnL31iqMXn6UrH961BuamRpy8+oJ/rr1IVz7mk1oYG+pz5OJTzt16na58Uh9XAALOPObK/VC1MkNDPcZ9UhuAPf885FZwhFq5WTFDRnRzAeCPY0EEPVP/LighN2ZYx+oAbDl0lyev/hujYWikj5WZMQPbVwHgN//bvAqPVTveoZQZn7Z2BmD13htEKBLUyp3sLfi4eeqP3RV/XiM6LlGtvGq5EnRqVB6AJTsuk5io/sVeq6I17dwdAZi/+WK698at6ke0dC1DQmIyP+64kq5ck7+9MhamjBl7gAQjOBUwMN0+2lDgkkoamUym9jqtkf/d7e9TsqRZjmOwsTHP8bEFlS7rHPBvMMcvPaVCaQuGdkn98li8+QKhkXFq+1Upa8UA72oAzPntLFGxSrXyWpVs6NWmMgDfrfkXZWKyWrlbVVu6tUi9852y8mS6OBrXsqdjs4okJEvceBCmVmZuboKNjTlG0QkYGumnO1ZuXiz1PTQwyLDcwiK1PEEiw3LL/5dHKVMyLrc0xcbGnNDoxAzLS5Qojo2NOZYRcRmWW1kVV8WRcbkZNiWKIX+qyLC8ZEkzLMyMMTePyLDc2toMEyMDzMxMMixP+/sqXtw4XbmRoX7W5cYGqnJTU6N05SbGhv+VF0tfXqzYf+XFihmmKzctZqQqNzE2xNAoSb3c9L9yI2MDDN+5Gyhe3Pi/ciMDkCVnWp7Re2Nmlvq3Fa9MyrD8fX975mYmfPGFP6dPP8XUzJDoaGWu/HuW5ZeVH9Ma6vft25flfsePH2fYsGH4+fmpPQK7evUqPXr0YPPmzdSrV0/j64aFRedoAJCNjTkhIVHZPq4gy6zO2v7VCFDKyjTdr8Y7T94A0LpemTz7xbhk5xUSle8mndRfjIWV+NsuvM6de86KFeeZN68lLi52Oaqznp4syx/jBe5OpUKFCkBq28nbSSUoKEitXNCuY5efcfFeqNoX7NuPI/JCZQdL3KuXonlte9W2tMSUmbTElpn3xT93eOMi8WUjFD5JSSmsWnWBx48VLFjQCgA3t9L89lunXL1ugUsqDg4OVKhQAT8/P9q0aaPavm/fPlxcXLCystJhdAVbVs+54xOSefQiitIlTdOVd2pUXvXLPyNpdw2ZSbvryEzaHYsgCJq5cSOEsWMPcPlyarvfgAE1qV7dJk+urdOkEhcXx/HjxwF49uwZ0dHRBAQEAODi4oK9vT1Tp07F19eXmzf/a9T86quvGDt2LI6OjjRs2JDDhw/zzz//8Msvv+ikHgXNu43Mzg6WeDUom+Ux7dwd6dehuvjVLgj5WEJCEj4+Z1i27BxJSSnY25uzaFHrPEsooOOkEhYWxujRo9W2pb2eO3cu3bp1IyUlheRk9Wfa7du3Jz4+nlWrVrFu3TocHR1ZvHixGPiooQNnn/AyPJbKDpYA3H3yhhJyY1q6linUbQWCUJidP/+csWMPcudOaueRzz6rxfTpjTE3N87TOPJNQ72uFIWG+rQ7k3J25nh5lOXK/TCSUlLU2iY0UZDqrC2izkVDYajz+PEH2bTpGhUqWOLj05YGDbL+gZjTOhe6hnohc2nJ492xBGm9pgAu3A3JdjIRBCF/io5WYmZmBMB33zXFzs6MESPqUayYoc5iEkmlEEhLJm8nj7dl1GtKEISCKzIynhkzAjl9+hlHjvSlWDFD5HJjJkxooOvQRFIpqI5dfkZlB0vsShbn1LWX3H8WqUoeVnITAOpXLUX9qqV0HKkgCNrk73+fr78+zKtXMRgZ6XP+/AuaNHHUdVgqIqkUEO/22Lrz5A2NXGwZ7F0NL4+yvIlJEHciglCIvX4dw7RpR/nrr7sA1Ktnx48/tsXZuaSOI1MnkkoBYWNZDD09mapTQWUHS5zsU+c7q13JWpehCYKQy/buvcuECYeIiIjH1NSQ6dMb89lntdDXz39zAoukko+l3Z1YmBnh5VGWib3r6DokQRB0ICkphYiIeJo1K8vixa1xdLTQdUiZEkklnzp2+ZlqltrKDpY8eKHAsVTRm8BSEIqilBSJq1dfUbu2LQBdulTGwsKEFi3KZnvS3Lwmkko+dfV+6gCm/u0qi7YSQShC7t8PZ9y4g1y48ILDh/tSpYo1MpmMli3L6To0jYikkk95eZSlZsWSIqEIQhGRlJTCypXnWbjwXxISkrGxMeX161iqFLCp70RSyWeOXX7G1ftheHmUFQlFEIqIa9deM3bsAa5eTZ3QtWfPavzwQzNKlCim48iyTySVfCZtEGPNiiWpWCb/NsYJgqAd27ffZOzYAyQlpVCmjDmLFrUpMI+6MiKSSj6R1tPr8etoKjtYirsUQSgi3N1LY2Skz4ABNZk2rbFq2pWCSiSVfKKCnZzbjyJwxAz36mIUvCAUVtHRSrZuvc6QIXWQyWSUK2fJuXODsbFJv1ZRQSSSio4du/yMc7de49WgLF90rqHrcARByEVHjwYzYcIhnjxRYGJiQL9+NQEKTUIBkVR06u2xKG5VP9JxNIIg5JaIiDi++y6QbdtuAODi8pFqDEphI5KKDqXN5SXGoghC4bV3710mTz5CSEgsxsb6TJzYgC+/rIuhob6uQ8sVIqnomGiUF4TCa8+euwwZsg8Ad3d7fHzaULGilY6jyl0iqejAscvPsCxuTP92lXUdiiAIuah9eyfc3e3p0qUyn31WCz29/D3FijbkaIrLFy9ecP36dWJiYrQdT5Fw5sYr/E4/wq5kcexKFtd1OIIgaMnjx5EMG7afkJBYAAwN9dmz5xMGD65dJBIKZDOpHDp0iLZt29KyZUt69OjB1atXAQgPD6djx44cPHgwV4IsTI5dTl3eV7+I/IEJQlGQkiKxdu0lmjbdiK/vHebMOakqy+8TQGqbxknl2LFjjBo1CisrK0aMGIEkSaoyKysr7Ozs+PPPP3MlyMLi7d5eYiyKIBQOd++G0bHjdqZOPUpsbCKdOjkzZUojXYelMxq3qaxYsQJXV1c2b95MREQEy5cvVyuvXbs2O3fu1HqAhUnNCiVp4WqPw0dmonFeEAq4xMRkVqw4z6JFp1Eqk/noo+LMn98Sb+9Kug5NpzROKnfv3uXrr7/OtNzGxoawsDCtBFVYWclN6NdWNM4LQmFw/XoIc+f+gyTBp59WZ8aMZlhamug6LJ3TOKkYGRmhVCozLX/+/Dnm5mIRqaycvZU6LqV+VfHoSxAKoqSkFAwMUlsN6tSxZcqURtSubUvz5mV1HFn+oXGbiqurK35+fhmWRUdHs2vXLtzd3bUWWGF09OIzjl58puswBEHIgdOnn9G06QaOH3+k2jZmjLtIKO/QOKmMHDmSO3fu8Nlnn3H06FEAbt68ydatW+natStRUVGMGDEi1wIt6NJ6fQmCULBERyuZPPkwnTpt5/79CH755aKuQ8rXNH785eLiwtq1a/n222+ZOnUqAAsXLgSgbNmyrF27Ficnp9yJshBIm5JF9PoShILjyJGHTJhwiKdPozAw0OOrr9wYO1Y8kclKtkbU169fn4CAAG7fvs3Dhw+RJAkHBwdq1KiRo77YwcHBzJw5k4sXL2JsbIy3tzcTJkygWLGsVzuLjY1l5cqVBAQEEBISQqlSpejUqRPDhg3DyCj/rkUgpmQRhILhzZt4pk8/xo4dNwGoVasUPj5tqVHDRseR5X8aJxVfX1/q1atHmTJlqFKlClXeWTj56dOnnD9/ni5dumh0PoVCQf/+/SldujRLly4lPDycuXPnEh4ejo+PT5bHzpgxg0OHDjF27FgqVarE1atXWbZsGQqFQnUXld8M7yqmtReEguTYsUeYmOjz9dcN+eKLuqoGeiFrGieVKVOmsGDBAsqUKZNh+dWrV5kyZYrGSWXbtm0oFAp8fX2xskqdYE1fX58JEyYwfPhwKlXKuK93UlISAQEBDBkyhH79+gHg4eHB8+fP2b9/f75MKscuP8NAT4/GNe10HYogCJl49SoaCwsTTEwMsLQ0YfVqL0qVMsPJqYSuQytQNE69b4+gz0h8fDz6+ppP5RwYGIiHh4cqoQB4enpiZGREYGBglnEkJyen674sl8tJSUnR+Pp5KW2uL0EQ8h9Jkli//hKNG29gyZLTqu0NGzqIhJIDWd6pPH/+nGfP/usC++DBA86dO5duv8jISLZt24a9vebtBUFBQXTv3l1tm5GREY6Ojjx48CDT4wwNDenWrRubNm3C1dWVihUrcu3aNXbs2EHfvn01vn5eSev1VdnBUtehCILwjkePIhk//iCBgY8BuH07jJQUqchM/pgbskwqf/75J8uXL0cmkyGTyVi1ahWrVq1Kt58kSejp6TFz5kyNL6xQKJDL5em2y+VyIiMjszx2xowZfPfdd3zyySeqbQMHDmTkyJEaXz9NyZJm2T4mjY3N+wd7XrwXCkBr97Ia7Z/fFYY6ZJeoc+GTnJzC8uVnmTr1CLGxiZQsWYylS9vx6acuRWoCyNz4nLNMKp6enlSoUAFJkhg/fjx9+vShbt266fYzNTWlWrVqfPTRhy+JK0nSez/URYsWcezYMWbNmkW5cuW4fPkyK1aswNramqFDh2bremFh0aSkZP1oLyM2NuaEhES9d79EZTKVHSypW7GkRvvnZ5rWuTARdS58FIoEevb8kwsXXgDQtWtlVq3qiEyWQmhotI6jyzs5/Zz19GRZ/hjPMqlUqlRJ1WCuVCqpV68eDg4O2Q4iI3K5HIVCkW57VFRUluNd7t69y/r161m5ciWtWrUCwM3NjaSkJJYtW0bv3r0xM8v53Ye2jfmklq5DEAThLebmRpQoYYKtbXEWLGhNu3ZO2NgUL9SJNC9p3Pura9euWr2wk5MTQUFBatuUSiWPHz+mW7dumR53//59AKpWraq2vVq1aiiVSl69epVvksqxy89ISZFo6ZpxjzlBEPLGlSuvMDMzwsmpBDKZDB+ftpiY6GNhISaA1LZsDX5UKpUcOHCAGzduoFAo0vW2kslkzJkzR6NzNW3alJ9//pmIiAhKlEjtYXHw4EGUSiXNmjXL9Li0zgA3btygdOnSqu3Xr19HJpOpbdO1MzdeERIZJ5KKIOhIXFwiCxf+y8qVF6hb1449ez5BX1+PUqXEiqu5ReOk8vLlSwYMGMCjR4+Qy+VERUVhYWGhSi4lSpTA1NRU4wv36tWL33//neHDhzN8+HDCwsKYN28eXl5eVKxYUbXf1KlT8fX15ebN1JGtNWrUoGbNmnz33XeEhYVRtmxZrl69yurVq+nevft7R+PnNRuL/BWPIBQVp049YezYgzx8+AY9PRl169qRlJSCvr4YxJibNE4qCxcuJCwsjC1btlCuXDkaNmyIj48PdevW5ddff2X79u389ttvGl9YLpezYcMGZs2axahRo1TTtEycOFFtv5SUFJKTk1Wv9fX1WbVqFUuXLmX16tWEhoZiZ2fHoEGD+PzzzzW+viAIhVNUVAI//HCCDRtSlzuvUqUkPj5tqVtXDD7OCxonlVOnTtG7d29cXV158+aNaruRkRGff/45QUFBzJkzh59//lnji5cvX55169Zluc+8efOYN2+e2raSJUvyww8/aHwdQRCKhsTEZNq23UJQUASGhnqMHl2fMWPcMTLSfGC28GE0vg+MiYnB0dERQDVpY0xMjKq8bt26GQ6MFARByCuGhvr07etCnTqlOHiwD19/3VAklDym8Z1KqVKlePUqdfp2U1NTLCwsuHXrFq1btwZSR98bGGSr3b/Qm9THVdchCEKhJkkSf/11l5QUiW7dUie5/eILV774wlW0neiIxlmgXr16nDx5UjVq3dPTk3Xr1qGvr09KSgobN26kefPmuRVngXPs8jPiE5Jp5+6o61AEoVB68SKKSZOOEBAQhFxuTJMmjtjYmIpkomMaJ5WBAwdy6tQpEhISMDY2ZsKECTx79oxly5YBqWut5McZgnXlzI1XPAuNEUlFELRMkiR+//0aM2YEEhWlxNzciO++a4K1tehpmR9onFQqV65M5cqVVa/lcjnr1q0jKioKmUyWbwYc5gdiEklByB0PH75h/PiDnDz5BABPzwosWNAKO7vCPVdZQfLB94nm5uaYmZlptLhWUSGWDhYE7ZMkieHD/Th58gklSxbjl1+82Lixs0go+YxGSUWSJEJDQ1EqlenKXrx4waxZs2jVqhWrV6/WeoAFkaGhHjUqWImlgwVBC9LWckqdsaMlH39clZMnB9K1a5UiNaNwQZHl4y9Jkvjxxx/ZvHkzMTExyGQyWrVqxZw5czA0NGTJkiVs3bqVpKQkmjdvzuDBg/Mq7nxt3Ce1dR2CIBR4SmUyS5ee5elTBUuXegJQp44tK1e213FkQlayTCobN27kl19+oXTp0jRq1IinT59y8OBB9PT0eP36NdeuXaNLly4MHjyY8uXL51XM+dqefx4C0KmReD8EIacuXnzB2LEHuHUrDIAvvqhL1arWOo5K0ESWSWXXrl3UrFmT33//XTXgccGCBaxfvx5bW1v+/PNPnJ2d8yTQguJWcAQgkoog5ERsbCLz55/il18ukpIiUb68JT4+bURCKUCybFN59OgRHTp0UCUUgB49egDw+eefi4QiCILWnDz5mGbNNvLzzxcAGDGiHkeP9qNhQ+2s4STkjSzvVBISElTT0qextLQEoGzZsrkWlCAIRc/evfd49CiSqlWtWbq0LbVr2+o6JCEH3jtOJbPeFXp6YtSqIAgf5s2beCwtUxfKmj69MY6OFgwdWkfM11WAvTepLFiwgJUrV6pep01DP3XqVExM1FdNk8lk7N+/X8shFixmxQx1HYIg5HuhobFMn36UCxdecvx4f0xNDTE3N2bEiHq6Dk34QFkmFTc3twy3f/TRR7kSTGEwopuLrkMQhHxLkiT+/PM206YdJTw8HlNTA65ceUWDBmJ11MIiy6SyadOmvIqjUPjjWBAAHzd30nEkgpD/PHsWxddfH+LgwdRu902aOLJ4cWvKlbPUbWCCVom56rUo6FmkrkMQhHxp586bTJp0hOhoJXK5MT/80IzevauLEfGFkEgqgiDkOhMTA6KjlbRr58SCBa2wtRUT0BZWIqkIgqB1SUkpXLjwAnf31PnvOnSoxJ49n+Dubi/uTgo50S9YEAStunEjBC+vrXTtupMbN0KA1J6hHh5lREIpAsSdihaVkBvrOgRB0JmEhCR8fM6wbNk5kpJSsLc3JyoqQddhCXlMJBUtGtaxuq5DEASdOHfuOWPHHuDu3XAABg2qxfTpTTAzM3rPkUJhk+2kEh8fz/Xr1wkLC8PNzQ0rK6vciEsQhALi99+vMX78QSQJKlSw5Mcf2+LhIcadFFXZalPZuHEjjRs3pl+/fowZM4Y7d+4AEB4eTr169di5c2euBFkQHLv8jEHzjrDl0F1dhyIIeapZs7LI5cZ89ZUbR4/2EwmliNM4qfz555/MmTOHJk2aMHv2bNVqbABWVlY0bNgQf3//XAmyIEhbQri0dXEdRyIIuSsyMp5ly86SkpL6HeDgIOf8+cFMn96EYmKaoiJP48dfv/76K82bN8fHx4eIiIh05TVq1OD333/XanAFTWUHS7GEsFCo+fndZ9Kkw7x6FYOZmRGDBtUGwMLCJOsDhSJD4zuV4OBgmjdvnml5iRIlMkw27zvn4MGDqVOnDh4eHsycOZO4uDiNjo2OjmbOnDk0bdqUGjVq0LJlS5YuXZqt6wuCoJnXr2MYMmQfAwfu4dWrGNzcStOkiaOuwxLyIY3vVMzMzFAoFJmWBwcHZ6vRXqFQ0L9/f0qXLs3SpUsJDw9n7ty5hIeH4+Pjk+Wx8fHx9O/fn5iYGMaOHYu9vT1Pnz7l+fPnGl9f20pZmers2oKQWyRJYufOW3zzzTEiIuIxNTXkm28a89lntdHTE2NOhPQ0TioeHh78+eefDBw4MF3Zy5cv2blzJ56enhpfeNu2bSgUCnx9fVXJSF9fnwkTJjB8+HAqVaqU6bGrV6/myZMn+Pv7Y22dusxo/fr1Nb52bhjYvopOry8IueGvv+4ycmQAAM2bl2XRotY4OlroOCohP9P48deYMWOIiIigW7dubN68GZlMRmBgIIsWLaJTp04YGBgwYsQIjS8cGBiIh4eH2t2Np6cnRkZGBAYGZnnsjh07aN++vSqh6Nqxy8/4zf+2rsMQBK3r0KESzZqVZdkyT7Zv7yYSivBeGieVsmXLsmXLFj766CNWrFiBJEn8+uuvrF27lmrVqrFlyxZsbTVf/jMoKIiKFSuqbTMyMsLR0ZEHDx5ketzTp08JCQnB1taWiRMnUqtWLerUqcO4ceMIDw/X+PradObGKwKv6O7RmyBoS1BQBJ06beXVqxgADAz02LGjG716iRmFBc1ka/Cjk5MT69evR6FQEBwcjCRJODg45GgApEKhQC6Xp9sul8uJjMx8CvnQ0FAA1qxZg7u7OytWrCAkJIQFCxYwZswYNm7cmO1YtKGyg6VOrisI2pCUlMLKledZuPBfEhKSsbAwYsmStkDmS4oLQkY0TipPnjzBwcEBSP3ir1mzZq4EJElSln/EacsZy+Vyli1bhpFR6jQQxYsXZ9SoUVy5coVatWppfL2SJXM+BbeNjTkAhv9fTzvtdWFWFOr4rsJe58uXXzJ48B4uXnwBwMCBtVm8uC1WVsV0HFneKuyfc0Zyo84aJ5U2bdpQs2ZNvLy8aN++PaVKlfqgC8vl8gx7k0VFReHklPnKiZaWlgC4urqqEgqkdiQAuH//fraSSlhYtGoQV3bY2JgTEhIFQKIyNdGlvS6s3q5zUVGY6xwfn8SSJaf56adzJCdLODjIWbSoNZ984kJISFShrXdGCvPnnJmc1llPT5blj3GN21QmTpxIcnIy8+bNo0WLFvTt25etW7fmuB3DycmJoKAgtW1KpZLHjx9ToUKFTI9zcHBQSybvSkjI+1lRHUqZ4VBKLDokFCy3b4eydGnqyPghQ2pz/Hh/WrQop+uwhAJO4zuVwYMHM3jwYJ48ecK+ffvw8/Pj+++/Z/bs2bi7u+Pl5UXbtm0xN9fsdqpp06b8/PPPREREUKJECQAOHjyIUqmkWbNmmR5nZGREo0aNuHDhAkqlUpVgTp06BaSO7M9rn7Z2zvNrCkJOJCQkYWyc+s++dm1bZsxohqurrWoxLUH4UDLp7Um8sun+/fvs27cPf39/Hj9+jKGhIVevXtXoWIVCQYcOHbC3t2f48OGEhYUxb948GjRooDb4cerUqfj6+nLz5k3VtuvXr9OrVy8aNGhA3759ef36NYsXL6Z69eqsW7cuW3X40Mdfxy4/4+6TN0Vi2nvxiKBgO3o0mIkTDzF/fitatSqf6X6Fqc6aEnXWnNYef2WkYsWKeHp60qZNG0xNTUlMTNT4WLlczoYNGzA1NWXUqFHMnTsXLy8v5syZo7ZfSkqKqnE+TY0aNVi7di1v3rxh5MiRLFmyhPbt27Ns2bIPqU6OnLnxitP/n0xSEPKjN2/i+eqrv+nZ808eP1awYYNmP/wEISdydKcSFBSEn58ffn5+BAcHY2BgQJMmTejQoQNeXl65EWeu+dA7lfmbLwIwqY+rtkPLd8SvuYJn3757TJ58hNevYzA21mfixAYMH14PA4PMf08W9DrnhKiz5t53p5KtLsV+fn7s37+fe/fuoaenh4eHB0OGDMlWW4ogCLkvLCyOiRMPsW/fPQDc3e3x8WlDxYpiUT0hd2WrS7FMJsPV1ZXp06fTvn17seqjIORThoZ6XLjwguLFDfnmmyYMHFhLTAAp5AmNk8rXX3+Nl5dXtqZiKQqc7MVcSEL+8OSJAmvrYhQrZohcbsyaNR2wszPDwSH9zBWCkFs0bqgfNGiQSCgZ+Li5Ex83z3ywpiDktpQUibVrL9GkyQYWLvxXtb1+/dIioQh5LtM7lXPnzgHg5uam9vp90vYvClb8eQ2AEd1cdByJUFTduxfO2LEHOHs2dULT58+j3jvVkSDkpkyTSr9+/ZDJZFy5cgUjIyPV68yk/SHfunUrVwLNj6LjNO9CLQjalJiYzIoV51m06DRKZTIffVScBQta4eVV8f0HC0IuyjSppM32mzZiXVez/wqCoC4yMp6uXXdy/XoIAH361OC775piaSnWiRd0L9Ok8u5KirpeWVEQhFRyuTFlyshRKJQsXtyaZs3K6jokQVDRuKG+f//+/Pvvv5mWnz59mv79+2slKEEQ1J0+/ZQ7d8KA1PVNlixpw/Hj/UVCEfIdjZPK2bNnVQtkZSQ8PFzjxvzComq5ElQtV0LXYQiFWHS0ksmTD9Op0w5Gj/6b5OQUAKytTSle3FDH0QlCetla+TErL1++pFixorWoT6dGmU/KJwgf6siRh0yYcIinT6MwMNCjefOypKRI6OvrOjJByFyWSeXQoUMcPnxY9XrHjh2qKebfplAoOHXqVLYWxyrovlvzL0plEuM+qa3rUIRCJjw8jm+/Pc6OHakzc9euXQofn7ZUr26j48gE4f2yTCr37t1j//79QOpz3EuXLnHlyhW1fWQyGcWKFaNu3bpMmzYt9yLNZ5SJySQmpug6DKGQSUxMxtNzC48eRWJios+kSY34/HPXLCeAFIT8JMuk8uWXX/Lll18CUKVKFebOnUvHjh3zJDBBKIoMDfUZOrQOfn73WbKkDRUqiDY7oWDRuE3l9u3buRmHIBRJkiSxdesN9PRk9OqVutDbkCF1GDKkjpgAUiiQtNZQLwhC9jx6FMn48QcJDHyMmZkRrVqVx8bGVCQToUDLNKm0bNkSPT09/P39MTQ0pGXLlu+dT0gmk3Ho0CGtB5kfuVW1JSYmQddhCAVQcnIK69ZdZs6ck8TGJmFlZcLs2S2wti5avSeFwinLEfUymQw9PT2110Kqbi0qFrmV4oQPd+dOGGPHHuD8+RcAdO1a+f8JxVTHkQmCduRoOeHC5EOXEy5KRJ0/jCRJeHlt5cKFl9jaFmfBgta0a5f/lk0Qn3PRoPPlhAV1U1aeJFGZXCTWphc+TNoM3jKZjAULWvPbb1f47rumyOXGug5NELRO487v586dSzdT8d69e/H09KRBgwbMmjWLlBQxbkMQ0sTFJfL994F8+aWfapuLy0csXtxGJBSh0NI4qSxfvpyLFy+qXgcFBTFlyhT09PSoUaMGmzdvFtPjC8L/nTr1hObNN7FixXl8fe9y+3bm8+YJQmGicVK5f/++2jQse/fuxcTEhJ07d7JmzRo6d+7Mrl27ciVIQSgooqISmDjxEF267OThwzdUrVoSP79eVKlirevQBCFPaNymEhUVhVz+33rXJ06coGHDhpiZpTbY1K1bl7///lv7EQpCAXHw4AMmTjzE8+fRGBrqMWaMO6NH18fISMwAKRQdGicVGxsb7t+/D8CrV6+4desWPXv2VJVHR0djYFB02v0b17InOjpe12EI+cjRo8E8fx6Nq6stPj5tqVpV3J0IRY/GWaBt27Zs3ryZxMRErl69ipGRES1btlSV3759GwcHh1wJMj/yblS+yHVBFNRJkkRISCwffVQcgKlTG1OpUkn693dBX19MACkUTRr/5Y8aNQpPT0/27NlDaGgoc+bMwdo69ZdYdHQ0Bw4coGHDhrkWaH4Tr0wiITFZ12EIOvLiRRQDBuyhffutREcrATAzM+Kzz2qJhCIUaRrfqZiamrJw4cJMywIDAzExMcnWxYODg5k5cyYXL17E2NgYb29vJkyYkK3Fvq5fv06PHj0wMTHh0qVL2br+h/h+7WkxTqUIkiSJ33+/xowZgURFKTE3N+LGjRDc3e11HZog5As5bgSJjo4GwMzMDD09PczNzbN1vEKhoH///pQuXZqlS5cSHh7O3LlzCQ8Px8fHR6NzpKSkMGPGDKysrIiNjc12HQQhOx4+fMP48Qc5efIJAJ6eFViwoBV2dtn72xeEwixbSeXZs2csXbqU48ePo1AoAJDL5TRv3pyvvvoKe3vNf61t27YNhUKBr68vVlZWAOjr6zNhwgSGDx9OpUqV3nuOHTt2EBUVRffu3dm0aVN2qiII2bJ58zWmTj1KXFwS1tbFmDOnJZ07O4v58AThHRo//H3w4AHdunVj37591KhRg/79+9OvXz9cXFzYu3cv3bt35+HDhxpfODAwEA8PD1VCAfD09MTIyIjAwMD3Hp92RzN16lQMDQ01vq4g5ESJEsWIi0uie/cqnDgxkC5dKouEIggZ0PhOZcmSJQDs2rWLqlWrqpXdvn2bAQMGsGTJEn766SeNzhcUFET37t3VthkZGeHo6MiDBw/ee/yiRYtwdXWlWbNmXL16VcNaCIJmlMpkTp9+RvfuqQtneXlV5ODBPtSqVUrHkQlC/qZxUjl79iz9+/dPl1Agdanhvn37ZusRlEKhUBtMmUYulxMZGZnlsZcuXWL//v3s27dP4+tlJqvZNrPSqp4jkDrTZ1FSFOp79uwzBg36i9u3Qzl71gpXVzsAWrcu/HVPUxQ+53eJOmuHxklFqVSqRs9nxNzcHKVS+cEBpc3ompnk5GS+//57Bg4cqJVxMTmd+r51fUdCQqKK1FiVwj49eGxsIvPmnWL16oukpEiUL29JYmJyoa5zRgr755wRUWfNvW/qe43bVJydndmzZw/x8elHkSuVSvbs2YOzs7PGgcnlclVj/9venQ7mXTt27CAkJIRPP/0UhUKBQqEgISF1BUaFQpFhfLkhMjqBqNgPT6JC/nDy5GOaNdvIqlUXABg5sh7HjvXD3b2MjiMThIJF4zuVzz//nJEjR9K9e3c+/fRTypcvD6Q24G/dupWHDx+yfPlyjS/s5OREUFCQ2jalUsnjx4/p1q1bpsc9ePCA0NBQmjZtmq7Mzc2N/v37M23aNI3jyKl5G8+JcSqFxK+/XmHSpMMAVKtmzY8/tqV2bVsdRyUIBZPGSaVVq1YsXLiQefPmMXPmTNUjKkmSsLa2ZuHChWrTtrxP06ZN+fnnn4mIiKBEiRIAHDx4EKVSSbNmzTI9rm/fvrRu3Vpt2+7du/Hz82PNmjXY2oovAyF72ratwIIFpxg6tA6jRrlhaCgmgBSEnMrWOJUOHTrQrl07bty4wbNnzwCwt7enevXq2Z5MslevXvz+++8MHz6c4cOHExYWxrx58/Dy8qJixYqq/aZOnYqvry83b94EoGzZspQtW1btXGfPnkVfXx93d/dsxSAUTSEhsaxff5kJEzzQ19fD3t6cc+cGY2ZmpOvQBKHAe28muHv3Llu2bOHJkydYWlri5eVFq1at1NZWyQm5XM6GDRuYNWsWo0aNUk3TMnHiRLX9UlJSSE4Wc2wJH06SJHbtus306UcJD4/HysqEoUNTH1+KhCII2iGTJCnTrk9Xr16lf//+ao3fMpmM7777jl69euVJgLktp72/luy8UuTaVApyD5lnz6L4+utDHDyYOkC3aVNHFi9uQ9myFlkeV5DrnFOizkWDTnp/rVixAgMDA37++WcuXbqEr68vVapUYdmyZSQlJWU7mMLEq0F5WriKSQTzu5QUid9+u0KTJhs4ePAhcrkxP/7Ylp07u783oQiCkH1ZJpXLly/Tp08fWrRoQbFixahSpQpTpkwhIiJC1aZSVDWpY0/9qmJ0dX7n63uHr78+THS0kvbtnTh5cgCfflpDTLEiCLkkyzYVhUKh6jqcxsnJCUmSeP36dboG86IkJCKOcEU8VvLsTfcv5K3OnZ3Zvfs2PXpUo2PHSiKZCEIuy/JORZIk9PXVu1fq6aUeUtQffy3ZeoE1e2/qOgzhHTduhNC9+x+8eJH6rFhfX49Nm7rQqZOYUVgQ8sJ7e38dO3aMly9fql7Hx8cjk8nw8/Pj+vXravvKZDKGDBmi/SgF4T0SEpLw8TnDsmXnSEpKYdGi0yxe3EbXYQlCkfPepLJ//37279+fbvvOnTvTbRNJRdCFc+eeM3bsAe7eDQdg0KBaTJ/eRMdRCULRlGVSOXz4cF7FIQjZFhOTyNy5J1mz5hKSBE5OJfDxaYOHh5ivSxB0Jcukkp2VHIua5OTsj20RtCsoKJy1ay+jpydjxIh6TJjQABOTHK+QLQiCFoh/gTn0cctKPHr+RtdhFDmxsYmYmqau9FmzZilmz26Bm5sdNWuK7t2CkB9oPPW98J8XYTGUtilO89riTi4v+fndx919PX///d/s1oMH1xYJRRDyEZFUcmBjwB1W/HFF12EUGa9fxzBkyD4GDtzDq1cx7Nx5S9chCYKQCfH4S8i3JElix45bfPPNUd68ScDU1JBvvmnMZ5/V1nVogiBkQiQVIV96/TqGr776myNHggFo0aIsixa1wcEh81VBBUHQPZFUhHzJ1NSQO3fCsLQ05ocfmtOzZzUxIl4QCoBstak8e/aM6dOn07p1a1xdXTl79iwA4eHhzJgxI90Ie0HIjqCgCGJiEoHU9U1+/bUjJ04MpFev6iKhCEIBoXFSCQoKolu3bvz999+ULVuWuLg41eJZVlZWXLlyhS1btuRaoPlJh0bl6NnaWddhFBqJicksW3aW5s03Mm/eP6rttWvbUqpUcR1GJghCdmn8+GvhwoUUL16cHTt2oKenR8OGDdXKmzZtSkBAgNYDzI+ql7Mqkov65IZr114zZswBrl17DUB0tBJJksSdiSAUUBonlfPnz/PFF19gbW1NREREunJ7e3tev36t1eDyq8evoohSpmBuJHpk51R8fBJLlpzmp5/OkZws4egoZ9GiNjRvXnSXUxCEwkDjpJKUlISpqWmm5W/evEk3TX5htfXQPQyN9BnXo5auQymQIiLi6NBhO/fuhSOTwdChdZgypZFYJ14QCgGNf2o7Oztz5syZDMskSeLgwYNUr15da4EJhVeJEsVwdraiUiUr9u7txezZLURCEYRCQuOkMmDAAP7++29WrFjBmzdvAEhJSSEoKIixY8dy/fp1Bg0alFtxCgXckSPB3LgRonrt49OWw4f7Ur9+aR1GJQiCtmn8+Mvb25tnz56xbNkyli9fDqBaO0VfX59JkybRrFmz3IlSKLAiIuL49tvjbN9+k1q1SuHv3xsDAz0sLcUyzIJQGGVr8OOwYcPo2LEjf//9N48ePSIlJQVHR0c8PT0pU0asYSGo27v3LpMnHyEkJBYTE306dRLdsAWhsMv2iHo7OzsGDhyYC6EUHN2bOWFpmXmnhaLu1atoJk8+wv799wHw8LDHx6ctTk4ldByZIAi5TUzTkgMVy1iIcSqZSExMpn37rTx9GkXx4oZ8800TBg6shZ6eGHciCEWBxkmlSpUqGg1Iu3Wr8E9Lfv9pJKHRiVibGeo6lHzH0FCfkSPdOHDgAYsWtaZMGTEBpCAUJRonlREjRqRLKsnJyTx9+pTDhw9Tvnx5WrRoka2LBwcHM3PmTC5evIixsTHe3t5MmDCBYsWKZXpMdHQ0v/76K4GBgTx8+BBDQ0NcXFwYN24cVapUydb1c2rX8SAxTuX/UlIk1q+/jLGxPv361QTgs89q8dlntcSoeEEogjROKqNGjcq07NWrV/To0YMKFSpofGGFQkH//v0pXbo0S5cuJTw8nLlz5xIeHo6Pj0+mxz1//pzt27fTvXt3Ro8eTUJCAuvXr6dnz5788ccfVKpUSeMYhA9z924YY8ce5Ny555iaGtK+fUWsrU1FMhGEIkwrbSqlSpWiV69erFixAi8vL42O2bZtGwqFAl9fX6ysrIDUrskTJkxg+PDhmSaHMmXKcPDgQbW7GQ8PD1q2bMm2bdv45ptvPrxCQpYSE5NZseI8ixadRqlMplSp4syf3wpra9F5QRCKOq1NXmVhYcHjx4813j8wMBAPDw9VQgHw9PTEyMiIwMDATI8zNTVN93isePHiODo6EhYWlv3AhWy5evUVbdtuYc6cf1Aqk+nTpwYnTw7Ay6uirkMTBCEf0EpSCQ8PZ8eOHdjb22t8TFBQEBUrqn8RGRkZ4ejoyIMHD7J1fYVCwb1799KdT9AuSZKYNOkIN26E4OhowR9/fIyPT1ssLMRARkEQUmn8+Kt///4Zbo+KiiIoKIikpCSWLFmi8YUVCgVyefqeQXK5nMjISI3PA6nT8gN07do1W8flVO/WlShRouis85GcnIK+vh4ymYxFi1qzbdsNJk9uRPHiovebIAjqNE4qkiSl2yaTyShTpgwNGzakR48elCtX7oMDyu5aGrt27WLHjh3Mnj07W3dKaUqWNMv2MTY25tk+piBSKBKYMuUQr17FsHNnDwCaN69A8+aad8goDIrK5/02UeeiITfqrHFS2bRpk1YvLJfLUSgU6bZHRUXh5OSk0TmOHz/Ot99+y/Dhw/n4449zFEdYWDQpKekTZlZuBIdjaVEM+xKZd30u6A4ffsiECYd49iwKAwM9Tp16RKNG5YrcgM+iOMhV1LloyGmd9fRkWf4Y16hNJT4+nv79+7Nr165sB5AZJycngoKC1LYplUoeP36sUdfky5cvM3r0aDp37szo0aO1Fpcm9v0TzPZDd/P0mnklPDyOESP86d17N8+eRVG7dikOHuyDs3NJXYcmCEIBoFFSMTEx4ebNm6o16bWhadOmnD59Wm0VyYMHD6JUKt872/H9+/f5/PPP8fDw4IcfftBaTEXdnj13adz4N3buvIWJiT7ffdcUP7/eVK9uo+vQBEEoIDTu/VW/fn3Onz+vtQv36tULc3Nzhg8fzokTJ/D19WXmzJl4eXmp9eKaOnUq1apVU70OCwtj8ODBGBsbM2DAAK5fv87ly5e5fPkyN2/e1Fp8RdH58y8IDY2jYcMyHDvWnxEj6mFgIJZMFgRBcxq3qUyfPp1BgwYxf/58Pv30U+zt7dHTy/kXjlwuZ8OGDcyaNYtRo0appmmZOHGi2n4pKSlqd0j379/n5cuXAOlmS7a3t+fIkSM5jqmokSSJ58+jsbdPbaybPLkh1apZ88kn1cQEkIIg5IhMyqhb1//5+vpSr149ypQpQ82aNZEkiaSkJAD09PQwMFDPSTKZjMuXL+dqwNqWk4b6+ZsvFvi5v4KD3zB+/EEePHhDYGB/zM2N33uMaMwsGkSdi4bcaqjP8k5lypQpLFiwgDJlyuDl5SXmdPq//u0qY2VVMMepJCensHbtZebOPUlsbBIlSxbj7t1w6ta103VogiAUAlkmlbdvYubNm5frwRQUdiWLF8hfNrdvhzJ27AEuXEh9fNitWxVmzWou5uwSBEFrxCJdOXD5XigWr2Mo/1HBuVtZt+4y3357jMTEFOzszFiwoBWenpqNBxIEQdDUe1vaxSOv9P4++5jdx+/rOoxscXSUk5iYQv/+NTlxYoBIKIIg5Ir33qlMmTKFadOmaXSygthQX1jFxiZy8uQT2rZNHUjapk0FTpwYQOXKYhCjIAi5571JpVatWjg4OORFLIKW/PPPE8aNO8ijR5H4+/emTh1bAJFQBEHIde9NKj179qRjx455EYvwgRSKBH744QQbN14FoGpVazF4URCEPCUa6guouLgYoqPfkJyc9P/XSURExNOzZyl69WqDubkxcrkxMlkcL18+0so1X7/WIyUlRSvnKihEnYsGUedU+voGmJlZUqxYzjshiaSSA0M7VsPKygz+PxA0r8XFxRAVFYGlpQ2GhkaEhsYSGRmFnl5xzMwMcXCQU6yY9tc6MTDQIympaP3DE3UuGkSdU4eQJCYqefMmBCDHiUU8G8kBK7kJNjqc9j46+g2WljYYGRkjk8mwtDTB0FCf0qXNqVTJKlcSiiAIhZtMJsPIyBhLSxuio9/k+DxZ3qncvn07xycuzM7eeoX8qYIqZdKvXJkXEhISCQlJoHRpI2QyGYaG+lStai3m6xIE4YMZGhqpHqvnhLhTyYGjF5/h9+/DPL9uSorExo1XefUqmpCQWEJCYlVlIqEIgqANHzo2USSVAuLBgwi6d9/JhAmHSEmRsLAwwdLSRNdhCYIgqBEN9flccnIKv/xykfnzTxEXl4S1dTFKlixGuXIWYrYDQRDyHZFU8rndu+8wY0YgAB9/XJWZM5uTmPi60CWUL74YxPXrV5k3bwmNGzdVK3vx4jk9enRi5sx5tGjROt2xjRvXY/jw0Xz6aT/VtqSkJPbs2U1AwH6Cgx+SlJSInV1pWrZsQ/fuPbG0tMztKhEWFsrSpYs5c+YUMpmMhg2b8NVX4zW69rZtv/PnnzsJCXmNra0dPXt+SpcuH6vKL148z1dffZHhsYaGhhw9+m+GZT/+uJA//thOt249GDdukmr78eNH2LZtM48fBxMXF4e1tQ1Nm7Zg4MAhmJn9N8357Nkz8Pffl+687342H3/ckZcvX6Tbb9++Q6r6JyYmsmbNSm7cuM6dO7eIj49XK89IcnIyQ4f25+7dOxn+PSiVSjZv3oC//z5CQl5jYWFJgwaNmDRpOpC99+3+/XusXr2CmzdvkJiopEKFigwaNBQ3Nw+14wIDj7Fp0688ehSMiYkJLi41+eKLUTg4OKr2+e23tRw9epiXL5+TnJyMvX0ZOnXqRrduPdT+LcfGxrBy5TKOHTtCbGwslSo58+WXo6hd21XtmrGxMaxYsZRjxw6jVCqpU6ceY8dOxM6utGofP7+9zJnzfbp6vvvZa5tIKvlct25VOHDgAZ98UpXWrVOnXPn/GmWFxvPnz7h+PXXA5sGD/umSSnYlJiYyadJYLl26QOfO3fjssyEYG5vw8GEQvr67uH//HnPnLtJG6JlKSkpi/PivSEpKZPr070lKSmLlyp+YPHkcP/+8LssfBb/+uobfflvLgAGDcXGpyT//nGDRonmAjC5dugNQuXIVVq36Ve245ORkxo8fSf36DTI87717d9m/fy/Fi6fvKqpQKKhd25VevfpgaWnJvXt3+fXXNQQF3cPHZ4XavqVL2/Ptt7PUtjk6OvKu5s1b0atXX7Vtbyeo+Ph49uzxpWrVatSsWYezZzNOhG/z9f2D0NDQDMskSWLatIncvXuHgQOHUK5ceUJCQrhz578VYTV93yIiwhkz5kvs7OyZNGkaRkbG7N69k4kTx7Bq1XqqVEldjfb8+bNMmzaRtm3bM2zYl0RFRbN+/WrGjBnOxo3bKF48tb5RUVF4eranXLkKGBoacv78WX78cSExMdH07z9Idd3Jk8fz4EEQX3wxEiurkuzcuZVx40bxyy/rqVSpsmq/GTOmcefObcaMmUjx4masXbuK0aO/ZOPG7ZiYqD8WX7z4J1UcACVL5u7MGiKp5MDwrjUoWdIMZZxS6+e+cOEF33xzjNWrvSlTRo6enozVq721fp385MABfwDq1avPyZOBxMbGYGqa88FX69ev5ty5MyxcuBQPj4aq7a6u9ejS5WONvrw+1PHjR7h//y4bN26nQoXUyTutrW348svBnD79Dw0aNM7wuISEeLZs2cjHH/di0KBhALi5efDq1UvWrv2ZDh06Y2BgQPHiZtSo4aJ27L//niQ+Ph5Pz/bpzitJEkuWzKN37774+e1NV96xYxfV/xsY6FGrlitGRsYsWDCb0NAQrK1tVOXGxsbprp0RKyurLPczNzfH3/8IMpkMP7+97/1cwsPDWLNmFV99NS7DX+Cp5zjNb79tpXz5Cqrtbdu2U/2/pu/b2bOnefPmDatXb6B0aXsg9e+nY8e2HDlySJVUDh36G1tbO6ZP/171Q8HW1pZhwwZy9eoVGjRoBMCoUWPVrunm5s7r16/w89unSiqXL1/k4sXzanfrdevW45NPOrN+/RrVD6EbN65z6tRJFi78UfV3VKFCRXr27Iyf3166deuhdq3KlavmyZ15GtFQnwPmpkZYmL1/pcTsiIlJ5JtvjuHltZXz51/g43NGq+fPzw4eDKBGjZoMGDCYhIQEjh8/muNzJSQksGvXDho3bqaWUNLo6+tn+oWuTf/++w9OThVVCQXAxaUWtrZ2/PvvP5ke9/DhA+Li4nB3V7/bcHdvwJs3b7hx41qmxx44EICZmXmG9du/fw+hoaH06dNf4zpYWFgAqFZ7zQ3ZeYy7YsVS3N09qFOnboble/bspm5dN7WEookDBwIwN1d/39Lq/PadlaGhIcbGxoCktp+pqalaPczMzP//f1mvKGthYaHWdffmzRvIZDLq1//v8ZqxsQm1a7ty5swpVUynT/+DmZkZ7u7//X3b2tri4lKL06cz/9vKKyKp5MDJqy84dPax1s534sRjmjffyC+/XEQmkzFyZD1mzWqutfPnZ7dv3+LRo2DatGlH7dqufPRRKdWdS07PFxsbo/qFmBNJSUnv/e99U3o8evSQcuXKp9tevnwFHj3KfNocPb3Uf5LvLtVtaGgEpCadjMTFxXHy5HFatGiNkZGRWplCEcmqVT8xcuRYjI2z7jGYnJxMQkICt2/f4tdf19KoURNsbdVXBX327Bnt2jWnWTN3PvvsUw4fPpDhuQ4cCKBly4a0bt2YceNGcudOzse9Xb58kcDAowwfPjrD8qSkJO7cuYWDgyM//rgIT89mtGzZiIkTR/Ps2dNMz5v2vrVsqf6+NWnSDCurkixbtoTQ0BAiI9+wbt0vxMbG0r79f3MhdujQmeDgh/zxxzaioqJ48eI5K1b8SLly5albt36GccbGxvDPPycICNjPxx/3VJXp6+uhp6en+htIY2hohFKp5Pnz1HoEBz/E0bFcuv3Kl3fi0aPgdNfs378nTZvWp0ePTqxfvzpXfySAePyVI/9ce4GhkT61ypf4oPNERsbz/feB/P77dQCqVbPmxx/bUru2bY7ON3/zxXTb3Kp+REvXMiQkJvPjjivpyhu52NG4ph1RsUpW7r6erryFqz31q5YiXBHP2n03kd758eVZ35HalaxzFC/AgQN+6Ovr07JlG2QyGa1bt2X79i2EhYVSsmT2zxsa+hqAUqVy9h4CNG/u8d592rfvwLRpMzItj4qKeusX63/MzeUEB2ecGADKlHFAT0+PW7du4OpaT7U97Q4lKkqR4XHHjx8hLi4uw0dfv/yygooVnWnWrEWm103j7d2K6OhoANzdGzJjxhy18kqVKlOlSjXKl69ATEw0e/f+xXffTSUhIQEvr/++bBs3bkq1ajUoVcqWly9fsGnTb4wYMYQ1azZm+04iKSmJJUvm07fvQEqVsuXFi+fp9omMfENSUhJ+fvuoUMGJGTNmExcXzy+/LGfChK/YtGlHukQN/71v7dp5qW2Xyy1YsWINX389hi5dUt9TMzNzFizwUYu/dm1X5sxZxPffT+fHH1MfT1Wo4ISPz4p0yf3evTt89lkf1esBAwbTs+d/rx0cypKcnMzdu7epVq0GkPrY8tatG0Bquxek/g1k/LdlrtoHoGRJawYP/pyqVaujr6/H6dOn2LBhHS9ePM/yb/dDiaSiQ48fK9i69QZGRvqMH+/ByJH1MDTU13VYeSY5OZnDhw/g5uZOiRKpCbpNm/Zs2bKJQ4f+VvsHp6m0JbA/pHfc2rUbVf+vr69HcnL6uxILC8v3niejGFLjyzw2U9PieHp6sXnzRipUqEj16i78++9J1d1bZvU6cCAAW1s7atWqo7b91q0b+PvvY/36ze+NF+Cnn34hMVHJvXv32LBhHZMmjcXHZwX6+ql/l5980ltt/yZNmvPVV1+wdu0qtaQyZsxE1f/XqlUHD4+G9OnzMb///hvffPODRrGk2blzGwkJCfTu3S/TfdLuHCUphfnzfVRtCPb2ZRg8uC/Hjx+hVau26Y5Le99q13YlOfm/X0wREeFMmTIBO7vSfPXVOPT1Ddiz50+mT/+aFSvWqu5Cr127wsyZ39ChQycaN25GZOQbfvttHZMmjePnn9eq3Rk6OpZl7dqNxMbGcvnyRTZv3oCenh6DB38OQP36HpQp48iiRXOZNu17rKys2LJlE0+fPgFAJvvvziSzv623t7u7N1B7jOrm5kHx4masX7+agQOHULZs+s4V2iCSSh5TKBKQy1PbY1xcPmLRotbUq1daK2udTOrjmmmZsaF+luXmpkZZllvJTZjav55WJ927ePEcYWFhfPZZU6KiogCwtbXDwcGRAwcCVEkl7Qsto0dOycnJwH+Pi2xsSgHw6lXOu8hVrOis+v/MJhp899HDu8zNzTO8q4iOjsLcPOvpfUaNGktYWBgTJ6Y+6rG0LMGQIV+wfPmPGd69hYeHceHCWT79tH+6Lxsfn4W0a+dNyZLWqvc4JSWFxMQkoqKiMDU1Vb2/kHonYmCgR7VqLlSuXJUhQ/oRGHg0w67caVq0aM3ixfOIiIhQ/Th4l4WFJa6ubtl+BPbmzRvWr1/N+PGTSEhIICEhgZiYGCC191h0dDRmZmaYm8uRyWRUqFBRrVG6cuUqmJmZERycfgaM9O/bf0ll8+aNREUpWLdu0//bUaBuXTf69fuE335by4wZswH48cdFuLq68dVX41XHVq/uQvfuHQgI8KNz526q7cbGJqoGflfXehQrZsqqVT/RpUt3Spa0xsDAgJkz5/Ltt1MYMKAXAE5OlejRozfbtv2u+uzNzeUZ/n1HR0djbp7+DuZtLVu2Yf361dy5c1sklYJOkiT++OMW06cfY8mSNnh7VwKgT5/396IprP7+O/XX96JF8/7fZVbdo0fBlC1bDgsLC/T09AgLS9+VNDQ0dUbVtC+zKlWqYmpanNOn/1Hr0ZQd2nj8VbZsee7du5tue3Dww/e298jlFixZ8hOhoSEoFJE4OJTlxInjQOoX1rsOHfqb5ORk2rZN/+jr8eNgbt68zp49u9W27927m717d/Prr5vVuqq+rVIlZ/T09Hj6NPM2Cfjv7vB9NN3vbSEhr4iLi2XWrO/Slc2ePQMzMzMCAo5hYmKCrW3pDM6QKiEhId22rN634OAHlC1bTpVQIPXHjZNTJbU2muDgB+m6wH/0USksLCyzbMuB1ISXnJzMixcvVAmjUqXKbNmyi6dPnyBJEo6OZVm8eD7W1jbY2qY+0i1Xrjznzp1Jd2eSFnNWcvIZZJdIKnng2bMoJk48xKFDqb+W9u+/r0oqRVVCQjyBgcdo2LBxurEMsbExTJkygQMH/Bk69EvVL7zAwGN88smnavueOHEMPT09atasDaR2d+3e/RM2b97AuXOn0w1US0lJ4cyZf7P8YtfG468GDRrx999+BAf/12B//fo1Xrx4rnEnAmtrG6ytbUhOTsbX9w/q1KmLo2PZdPsdOBCAs3PlDNsq5s//Md3kgDNmTKN6dRd69OiFvX3mq7peu3aFlJQUVZfajEiSxNGjh7C1tcv0LgVS7zguXjxHw4ZNMt0nI/b2DixbtkptW3h4GDNmTGPQoGFqjeGNGjVm715ftTum27dvEh0dTeXKVdOdO6v3zdbWjhMnjpGQEK96hJWcnMz9+3fVkrCtrR137txSO/blyxdERr5RG4iYkStXLiGTyShdWn0/mUymGjgZERHBkSMH6N79vwZ9D49G/PrrGs6c+VfVw/HVq5dcvXqZ0aMnZHnNI0cOIpPJqFy5Spb7fQiRVHJgzCe1sLY2IyoyLsv9UlIkNmy4ysyZJ4iOVmJhYczMmc3p2bNaHkWaf6WNR+nevadag3QaV9d6HDwYwNChXwIwePDnTJw4msmTx9GunTempsW5cuUSW7ZspHPn7moN84MGDeP27ZtMmjSOzp27U7++ByYmJgQHP8TXdxelS9tn+cWe9ogCcr7ORrNmLXFyqsT06ZP44osRJCcns2LFUmrUqImHx3/XThvouH27r6qX1YED/sTHx1OmjAOhoaHs2fMnQUH3WbVqfbrrPH4czO3bNxk5ckyGcdSqVTvdNiMjI2xsbNTe93HjRv6/O64TJibG3L59iy1bNuHkVImmTZsDqV+Ws2Z9R+vWnpQp40B0dBR79/7FpUsX1NpJDh4M4NSpk3h4NMTG5iNevHjO5s0bUCoT6dt3oFos//77D/Hxcdy+nTpA8Z9/AjE1NcXOrjRVqlTD1NQ03d9HWkN9+fIV1OrXu3c//v7bn6+/HsOAAYOIj4/nl19WUqGCU7pOCu973zp16srevb58/fU4evTohb6+Hnv27Obp0yeq0fmQOjrdx2chPj4LaNy4GQpFJBs2rMPSsgQtW7ZWvW+zZ8+gdWtP7O3LoFQquXDhLLt27aBz525YWf336Pu339bi4OBIiRJWPH78iE2bfqVUKTu1ruDVq9egYcPGzJs3k5Ejx1C8eHHWrv2FUqVs1dq1xo0biatrPSpUqIhMJuPMmVPs3v0HHTp0xt6+TIb11gaRVHLA2FAfEyMDorLY58WLKL74wo9//30GgJdXRebPb0mpUmZZHFV0HDjgj7W1DW5u7hmWe3l1ZObMb7l27QouLrVwd2/A4sU/sXHjembP/p6kpERKl7Zn6NAv093pGBoasnDhUv76608CAvazb58vSUlJ2NmVpnHjpjnqAJBdBgYGLF68jKVLF/PDD98ik0HDhk0YPXqC2iOLlJQUkpOT0z2W2LFjC8+fP8fY2Ji6dd1Yvfo3tWk/0hw4EIC+vj6tW7dLV5Yd1arV4MABf54/f45MlvoLvGvXj+nVqw+Ghqnr85iammJmZsbGjeuJiAjHwMAAZ+cq6abWsbOzJzQ0hJ9+8vl/TyUzatd2Zdas+ekezyxePE9tOpe5c1OT0/seL2akVClbli1bxfLlPnz33VQMDY1o2LAxI0eOSdfz633vm7NzFXx8VvDrr2uYM+d7UlJSKF++PAsW/Kg2ZUq3bp9gYGDI7t1/4Oe3F1NTU6pVc+GHH+ap7mbNzMz56KNSbN68gdDQUIoVM8He3oGvv56Gp6d6r7Po6GhWrFhKeHgYVlYladGiFYMGDUvXFXzGjNksX/4jixfPJzFRiatrPWbOnKc2mr5s2fLs37+HkJDXJCcnU6aMA198MSpdZwttk0l58ZAtHwsLiyYlJXtvwZGLTzEzM6G+c+ZdXqOjlTRvvpG4uCTmzWtJx47Ome6bXS9fPsLWNv1jkNwmVscrGkSdi4as6pzVd4yenoySJTP/cazTwY/BwcEMHjyYOnXq4OHhwcyZM4mLy/qRUhpfX1/atWuHi4sL3t7e+Pn55XK0/zl36zUnrzxLt/3GjRCio1OnbjEzM+K33zpz8uQArSYUQRCE/ExnSUWhUNC/f39iYmJYunQpkydPZt++fUydOvW9xwYEBDBp0iTatGnDmjVraNCgAePGjeP48eN5EHl6CQlJzJv3D23abGbOnJOq7TVq2FBCh8sOC4Ig5DWdtals27YNhUKBr68vVlZWQGqXvQkTJjB8+HAqVcq8d9TSpUtp164d48en9g338PDg4cOH/PTTTzRr1ixP4k9z7txzxo49wN274chkqT033u3qJwiCUFTo7E4lMDAQDw8PVUIB8PT0xMjIiMDAwEyPe/LkCQ8ePMDbW33mXm9vb65du0Z4eHiuxfw2ZXwS06cfpUOHbdy9G07FiiX466+ezJ7dQiQUQRCKLJ3dqQQFBdG9e3e1bUZGRjg6OvLgQeZzI6WVOTk5qW2vWLGiqvztRJUb4mIS2exzgciwePT1ZYwc6cb48R6YmIjOdIIgFG06+xZUKBTI5emnq5DL5URGRmZ6XFrZu8emTdOd1bEZyaoXQ2ZWTGtN2LU33LkTxvr1nahTx+79B2nR69d66OvLdHJHZGBQ9Ca2FnUuGkSdU0mShJ6eHjY2WU/5kuk5PzQobdO0PeLdfXI6kWBOuhQDrF7dkejoOAwN9QkJyWrEivbJZHrExcVjZKTdNV3eR3S7LBpEnYuGzOqsVCYgk+ll+r2Wb7sUy+VytWma00RFRWV4B5MmszuStHNldaw2yeXGOptR2MzMkjdvQlAqE/JkLh9BEAo/SZJQKhN48yYEMzPLHJ9HZ3cqTk5OBAUFqW1TKpU8fvyYbt26ZXIUVKiQOk/PgwcP1NpV0s6VVl6YFSuWutRuZGRounmdcpOent57F6cqbESdiwZR51T6+gaYm5dQfcfkhM6SStOmTfn555/VJn87ePAgSqUyy27BDg4OVKhQAT8/P9q0aaPavm/fPlxcXHK9kT6/KFas+Ad98DlhY2Oe54/6dE3UuWgQddYenT3+6tWrF+bm5gwfPpwTJ07g6+vLzJkz8fLyUvXkApg6dSrVqqlPwPjVV1/h7++Pj48PZ86cYc6cOfzzzz+MGjUqr6shCIIgvEVndypyuZwNGzYwa9YsRo0ahbGxMd7e3kycOFFtv7QJ997Wvn174uPjWbVqFevWrcPR0ZHFixfn+cBHQRAEQZ2YUDKHvb/E7XLRIOpcNIg6ay7f9v4SBEEQCp98N04lr+np5XwA4YccW1CJOhcNos5FQ07q/L5jivzjL0EQBEF7xOMvQRAEQWtEUhEEQRC0RiQVQRAEQWtEUhEEQRC0RiQVQRAEQWtEUhEEQRC0RiQVQRAEQWtEUhEEQRC0RiQVQRAEQWtEUnlHcHAwgwcPpk6dOnh4eDBz5kzi4uI0OtbX15d27drh4uKCt7c3fn5+uRytduSkztHR0fz000/06NGDevXq0aBBA4YNG8bt27fzKOoP8yGfc5rr169TtWpV6tSpk0tRateH1Dk6Opo5c+bQtGlTatSoQcuWLVm6dGkuR/zhclrn2NhYFi1aROvWralVqxZt27Zl+fLlKJXKPIg65x49esS3335L586dqVatGh06dND4WG19fxX5ub/eplAo6N+/P6VLl2bp0qWEh4czd+5cwsPD8fHxyfLYgIAAJk2axLBhw2jUqBGHDh1i3LhxFC9ePF9PyZ/TOj9//pzt27fTvXt3Ro8eTUJCAuvXr6dnz5788ccfVKpUKQ9rkT0f8jmnSUlJYcaMGVhZWREbG5vLEX+4D6lzfHw8/fv3JyYmhrFjx2Jvb8/Tp095/vx5HkWfMx9S5xkzZnDo0CHGjh1LpUqVuHr1KsuWLUOhUDB16tQ8qkH23bt3j+PHj1OrVi1SUlI0Xm5cq99fkqDyyy+/SLVq1ZLCwsJU2/bs2SM5OztLd+/ezfLYdu3aSV999ZXatkGDBkndu3fPlVi1Jad1jomJkWJjY9W2RUdHS/Xr15d++OGHXItXGz7kc06zdetWqW3bttLixYul2rVr51aoWvMhdV66dKlUr149KSQkJLfD1Kqc1jkxMVFycXGRli5dqrb9u+++kxo2bJhr8WpDcnKy6v8nTZokeXt7a3ScNr+/xOOvtwQGBuLh4aG2JLGnpydGRkYEBgZmetyTJ0948OAB3t7eatu9vb25du0a4eHhuRbzh8ppnU1NTSlWrJjatuLFi+Po6EhYWFiuxasNOa1zmrRfulOnTsXQ0DA3Q9WaD6nzjh07aN++PdbW1rkdplbltM6SJJGcnIy5ubnadrlcnu/XsdfTy/5Xura/v0RSeUtQUJDaUsYARkZGODo68uDBg0yPSytzcnJS2552rqyO1bWc1jkjCoWCe/fupTtffvOhdV60aBGurq75+rHmu3Ja56dPnxISEoKtrS0TJ06kVq1a1KlTh3HjxuXrH0uQ8zobGhrSrVs3Nm3axJUrV4iJieH06dPs2LGDPn365HbYeU7b31+iTeUtCoUCuVyebrtcLicyMjLT49LK3j3WwsJCrTw/ymmdM7Jw4UIAunbtqpXYcsuH1PnSpUvs37+fffv25VZ4uSKndQ4NDQVgzZo1uLu7s2LFCkJCQliwYAFjxoxh48aNuRbzh/qQz3nGjBl89913fPLJJ6ptAwcOZOTIkVqPU9e0/f0lkooGJElCJnv/Yjbv7iP9v5FMk2PzG03rnGbXrl3s2LGD2bNnY29vn4uR5Z731Tk5OZnvv/+egQMH4uDgkIeR5R5N6gypXzjLli3DyMgISH3UOWrUKK5cuUKtWrXyJFZt0eRve9GiRRw7doxZs2ZRrlw5Ll++zIoVK7C2tmbo0KF5FGne0tb3l3j89Ra5XI5CoUi3PSoqKsNfPGkyy+hp58rqWF3LaZ3fdvz4cb799luGDx/Oxx9/rO0QtS6ndd6xYwchISF8+umnKBQKFAoFCQkJQOpnHR8fn2sxf6ic1tnS0hIAV1dXVUIB8PDwAOD+/fvaDVSLclrnu3fvsn79er7//nt69OiBm5sbQ4cO5fPPP2fZsmVER0fnZth5TtvfXyKpvMXJyYmgoCC1bUqlksePH1OhQoVMj0sre/fZY9q5sjpW13Ja5zSXL19m9OjRdO7cmdGjR+dWmFqV0zo/ePCA0NBQmjZtipubG25ubqxZs4bY2Fjc3NxYvHhxboeeYzmts4ODg1oyeVdaUs2PclrntERZtWpVte3VqlVDqVTy6tUr7QerQ9r+/hJJ5S1Nmzbl9OnTREREqLYdPHgQpVKZZaOsg4MDFSpUSDdYaN++fbi4uKj1PslvclpnSP3H9/nnn+Ph4cEPP/yQ26FqTU7r3LdvXzZu3Kj2X9euXTE2Nmbjxo307ds3L8LPkZzW2cjIiEaNGnHhwgW1gX+nTp0CoEaNGrkX9AfKaZ3THt/euHFDbfv169eRyWSULl06dwLWEa1/f2W7E3IhFhkZKTVp0kTq1auXFBgYKO3evVtyd3eXxowZo7bflClTpKpVq6pt8/PzkypXriwtWbJEOn36tDR79mypcuXK0rFjx/KyCtmW0zqHhoZKTZs2lZo0aSKdOnVKunTpkuq/Gzdu5HU1suVDPud3LVu2rECMU/mQOl+7dk2qXr26NGTIEOnYsWPSjh07JHd3d2nQoEF5WYVsy2mdk5KSpI8//lhq0KCBtHXrVunUqVPSqlWrpJo1a0pTp07N62pkS2xsrOTv7y/5+/tLffv2lZo1a6Z6/fTpU0mScv/7SzTUv0Uul7NhwwZmzZrFqFGjMDY2xtvbm4kTJ6rtl5KSomrATNO+fXvi4+NZtWoV69atw9HRkcWLF+f7bqc5rfP9+/d5+fIlkNor5m329vYcOXIk12PPqQ/5nAuqD6lzjRo1WLt2LYsXL2bkyJGYmZnRvn17JkyYkJdVyLac1llfX59Vq1axdOlSVq9eTWhoKHZ2dgwaNIjPP/88r6uRLWFhYekeQ6e9njt3Lt26dcv17y+ZJGk4jl8QBEEQ3kO0qQiCIAhaI5KKIAiCoDUiqQiCIAhaI5KKIAiCoDUiqQiCIAhaI5KKIAiCoDUiqQgFSuXKlfnpp590HUauOnPmDJUrV+bMmTMa7d+yZUsmT56cy1EJgmbE4Ech1/35559MmTIlw7KePXvm+yle+vXrx9mzZ1WvjY2NKVeuHD169KBPnz45Whgpu44ePcr169cZNWpUrl9LU+9+rvr6+pQsWZJGjRoxduxYSpUqlaPzrly5EmdnZ1q3bq2tUIU8JJKKkGdGjRqVbsr48uXL6yia7LGxsVGNxI6IiOCvv/5i1qxZhIaGMnbsWK1ey83NjatXr6qtKnn06FG2b9+eYVIJCAjQ6fIKaZ+rUqnk8uXL7N69m/Pnz7Nv3z5MTEyyfb6ff/4Zb29vkVQKKJFUhDzTuHFjateureswcsTMzIzOnTurXvfs2ZN27dqxceNGRo0ahYGB9v4p6enpYWxsrPH+Wc0inBfe/lx79OiBpaUla9eu5ciRI3h5eek0NiHviTYVQefevHnD/Pnz6dixI3Xq1KFOnTr069ePCxcuvPfYpKQkVq5ciaenJzVr1sTd3Z1evXoREBCgtl9wcDBjx47Fw8ODGjVq0LFjR3bt2pXjmIsVK0atWrWIjY0lLCwMgOfPnzN+/Hjc3d1xcXGhc+fO+Pr6pjvWz8+P7t274+rqSt26denYsSPLly9Xlb/bpjJ58mS2b98OpLYppf339OlTQL1NJTQ0lGrVqvHjjz+mu25GZUqlkhUrVuDp6UmNGjVo1KgR3377bYbrkGjKzc0NSF37/O3rLF26lG7duuHm5kbNmjXp0aMHhw8fVju2cuXKKJVKdu/erapnv379VOXR0dHMnz+fli1bUqNGDZo3b86iRYvUZlAWdEvcqQh5JioqKt265iVKlODJkyf8/fffeHp64ujoSFRUFDt37mTgwIHs2rULZ2fnTM+5fPlyVq1axccff0zNmjWJjY3l1q1bXLlyhXbt2gGp60L07t2bEiVK8Nlnn2Fubs7x48eZOnUqUVFR6SbE1NTTp0/R19dHLpcTHh5O7969iYyMpG/fvnz00Uf4+fkxadIkIiMjGTBgAJA6Zfy4cePw8PBg/Pjx6Onp8fDhQ86fP5/pdXr27MmLFy84ffo0CxYsUG3PaEpya2tr6tevj7+/P2PGjFErCwgIIDk5GW9vbyB1Zb+RI0dy+vRpevTogbOzM48fP2bz5s3cuHGDbdu2qT2C09SzZ8+A/xZ/gtRksG3bNry8vOjevTsJCQns3buXESNGsGbNGpo0aQLAggULmDZtGrVq1VIt5WttbQ1AfHw8/fr14+nTp3zyySc4Ojpy69Ytfv31Vx48eMDKlSuzHauQCz58smVByNquXbskZ2fnDP8LCwuTEhISpKSkJLVjIiIiJA8PD2natGlq252dnaVly5apXnfu3FkaNmxYltf/7LPPpPbt20uxsbFq20ePHi3VqVNHiomJyfL4vn37Sm3atJHCwsKksLAw6f79+9KsWbMkZ2dn6YsvvpAkSZLmzZsnOTs7S6dOnVIdl5CQIH388cdS7dq1JYVCIUmSJM2ePVtydXVNV9+3nT59WnJ2dpZOnz6t2vbNN99Izs7OGe7fokULadKkSarXO3bskJydnaVr166p7derVy+pQ4cOqtd79uyRKleuLP37779q+x07dkxydnaWdu/eneX7kva5BgYGSmFhYdKLFy+kgIAAycPDQ6pZs6b08uVL1b5JSUlSQkKC2vEJCQlS+/btpYEDB6ptr1Gjhlp90vz8889SzZo1pfv376tt//333yVnZ2fp3LlzWcYr5A1xpyLkmenTp+Pk5KS2zdzcXO3XcEJCArGxsUiSRM2aNdMtlPQuMzMz7t27x8OHDzNs9I+MjOTUqVOMGDGCuLg44uLiVGVNmzbF39+f69evU79+/Syv8+jRIxo0aKB6raenR8eOHZk2bRqQ2pBerVo1tX2MjIwYMGAA48eP58yZM7Ru3RozMzPi4uI4efJkri2L0LZtW77//nv8/PxUi2i9fPmSS5cuqU2L7u/vT7ly5XB2dla7g3RxccHU1JQzZ87QpUuX915vyJAhaq8dHR1ZsmSJWu8vfX199PX1gdRHYbGxsaSkpODm5oa/v79G9fL398fV1ZUSJUqoxduwYUMATp8+Tb169TQ6l5B7RFIR8oyLi0uGDfUpKSmsXbuW7du3q9oJ0pQpUybLc44aNYqRI0fSrl07KlasSOPGjfH29qZmzZpAajKQJInly5ertVu8La1NJCt2dnbMmTMHmUyGXC7HwcFBbe3uZ8+e0bZt23THVaxYUVUO0Lt3b/z9/Rk2bBgfffQRDRs2pE2bNrRq1UprPbgsLCxo3Lgx/v7+TJw4EZlMhp+fH5IkqR59QWo708OHD9US4ds0eV/gvx8LUVFR/PHHH5w/f16VQN62c+dOfvvtN4KCgpDeWnFD03oHBwdz+/btD45XyF0iqQg6t2bNGpYsWULXrl0ZM2YMlpaW6Ovr88svv6g19mbE3d2dQ4cOcfToUU6ePImvry8bNmxg7NixfP7556SkpACpC4lldmeQVZtNGhMTE9Uv4uyQ3lmuyMbGhr/++otTp05x4sQJTpw4ga+vL02bNmX16tVaSyze3t4cPXqUS5cu4erqyv79+6lRowaOjo6qfVJSUqhYsaLqbutdJUqU0Ohab/9YaN26NX369GH8+PEEBARQvHhxIHVp2unTp9OyZUuGDh2KlZUVBgYG7Nq1i3379ml0nZSUFDw8PDJdKCun42IE7RJJRdA5Pz8/6tevz7x589S2L1u2TKPjLSws6NKlC126dCE+Pp6hQ4eyfPlyhgwZohoXo6+vn6OkoCl7e3sePHiQbnvatrR1zyH1sVjz5s1p3rw5kiSxePFi1qxZw8WLF6lbt26G589usmnVqhXFihXDz88Pa2trrl+/zqRJk9T2cXR05MaNG3h4eGhtAKe+vj4TJkygT58+bNq0iS+++AJI/YwdHBxYuXKlWl0y6oGXWV0dHR2JiYnJ1c9R+HCiS7Ggcxk9Krl48SKXL19+77ERERFqr01MTKhQoQJKpZK4uDhKliyJh4cHO3bs4MWLF+mOf7c3Wk61aNGCmzdvcvr0adW2xMRENm7cSLFixXB3d88wXplMRrVq1QCy7MZramoKpLYRacLU1JTmzZsTEBDA3r17kclktG/fXm0fLy8vQkND+f3339Mdn5SUpPG13lWvXj3q1KnDhg0biI+PB/77jN++c3vy5AmHDh1Kd3yxYsUyvLaXlxfXrl1L1w0ZUnuGxcTE5CheQbvEnYqgcy1btuSnn37i66+/pl69egQHB7Njxw4qVqxIbGxslsd6eXnh5uZGjRo1KFGiBHfu3OGPP/6gefPmmJmZATBjxgx69+5Np06d6NGjB+XKlSMiIoIbN25w8uRJLl68+MF1GDp0KPv37+fLL7+kX79+2NjY4Ofnx+XLl5kyZQrm5uZAavtDREQEDRo0wNbWllevXrF582ZsbGxU4zsyktbg/sMPP9C0aVMMDAxo0aKFKtlkxNvbG39/f9atW4erqyt2dnZq5Z06deLgwYPMnj2bc+fO4ebmhkwm49GjR/z9999MnjxZrQ0mOwYNGsSoUaPYuXMn/fr1o2XLlhw4cIAvv/ySli1b8urVK7Zs2UK5cuW4fft2urr++++/rFu3DltbW6ysrGjQoAGDBw/m+PHjjBo1ik6dOuHi4kJiYiIPHjwgICCA1atXF9jBtYWJSCqCzg0bNoy4uDj27t1LQEAAlSpVYsmSJfj5+anNuZWRAQMGcOTIEU6fPk18fDx2dnYMHTqUoUOHqvYpX748f/75JytWrGDv3r1ERERgaWlJxYoV0z0SyikrKyu2bdvG4sWL2b59O7GxsZQvX5758+er9aDq1KkTO3bsYNu2bURGRmJtbU2zZs0YMWKEKglmpF27dly6dAl/f3/279+PJEkcPnw4y6TSrFkzzM3NiYqKyjA56OnpsWzZMjZt2sTu3bs5fvw4RkZGlC5dmk6dOn1QT6rWrVtTtmxZ1q9fT69evejatSthYWFs3bqVU6dOUbZsWaZMmcLjx4/TJZVp06YxY8YMli9fTmxsLPXr16dBgwaYmJiwYcMG1q5dy/79+9m3bx/FixenTJky9O/fnwoVKuQ4XkF7ZNK7LYmCIAiCkEOiTUUQBEHQGpFUBEEQBK0RSUUQBEHQGpFUBEEQBK0RSUUQBEHQGpFUBEEQBK0RSUUQBEHQGpFUBEEQBK0RSUUQBEHQGpFUBEEQBK35H76sBNN1vPi3AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve, auc, roc_auc_score\n",
    "\n",
    "# calculate scores\n",
    "auc = roc_auc_score(y_test, predictions)\n",
    "# summarize scores\n",
    "print('ROC AUC=%.3f' % (auc))\n",
    "# calculate roc curves\n",
    "\n",
    "ns_fpr, ns_tpr, _ = roc_curve(y_test, predictions)\n",
    "\n",
    "# plot the roc curve for the model\n",
    "plt.plot(ns_fpr, ns_tpr, linestyle='--', label=\"AUC = \"+ str(auc))\n",
    "plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')\n",
    "# axis labels\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "# show the legend\n",
    "plt.legend()\n",
    "# show the plot\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 8.210492,
     "end_time": "2021-03-19T05:02:03.512316",
     "exception": false,
     "start_time": "2021-03-19T05:01:55.301824",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Save Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
    "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:02:19.764526Z",
     "iopub.status.busy": "2021-03-19T05:02:19.763709Z",
     "iopub.status.idle": "2021-03-19T05:02:19.842448Z",
     "shell.execute_reply": "2021-03-19T05:02:19.841588Z"
    },
    "papermill": {
     "duration": 8.199177,
     "end_time": "2021-03-19T05:02:19.842601",
     "exception": false,
     "start_time": "2021-03-19T05:02:11.643424",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File name is severstal-classification-CNNbase-2021-03-19-05-02-19.h5\n"
     ]
    }
   ],
   "source": [
    "from datetime import *\n",
    "filename1 = \"severstal-classification-CNNbase\" + \"-\" + datetime.now().strftime(\"%Y-%m-%d-%H-%M-%S\")+\".h5\"\n",
    "print (\"File name is \"+filename1)\n",
    "model.save(filename1, save_format=\"h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "papermill": {
     "duration": 8.060276,
     "end_time": "2021-03-19T05:02:36.161405",
     "exception": false,
     "start_time": "2021-03-19T05:02:28.101129",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Predict submission file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:02:52.486994Z",
     "iopub.status.busy": "2021-03-19T05:02:52.485189Z",
     "iopub.status.idle": "2021-03-19T05:03:42.336985Z",
     "shell.execute_reply": "2021-03-19T05:03:42.336495Z"
    },
    "papermill": {
     "duration": 57.887448,
     "end_time": "2021-03-19T05:03:42.337112",
     "exception": false,
     "start_time": "2021-03-19T05:02:44.449664",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicting...\n",
      "173/173 [==============================] - 49s 286ms/step\n"
     ]
    }
   ],
   "source": [
    "STEP_SIZE_TEST = compute_steps_per_epoch(submission_df.shape[0])\n",
    "print('predicting...')\n",
    "\n",
    "submit_test = model.predict(\n",
    "    submission_gen,\n",
    "    steps=STEP_SIZE_TEST,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:03:59.273534Z",
     "iopub.status.busy": "2021-03-19T05:03:59.272516Z",
     "iopub.status.idle": "2021-03-19T05:03:59.274966Z",
     "shell.execute_reply": "2021-03-19T05:03:59.275529Z"
    },
    "papermill": {
     "duration": 8.730118,
     "end_time": "2021-03-19T05:03:59.275688",
     "exception": false,
     "start_time": "2021-03-19T05:03:50.545570",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "submission_df['defect_label'] = (submit_test > 0.5).astype(\"int32\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:04:15.373465Z",
     "iopub.status.busy": "2021-03-19T05:04:15.372726Z",
     "iopub.status.idle": "2021-03-19T05:04:15.376332Z",
     "shell.execute_reply": "2021-03-19T05:04:15.375925Z"
    },
    "papermill": {
     "duration": 8.128859,
     "end_time": "2021-03-19T05:04:15.376449",
     "exception": false,
     "start_time": "2021-03-19T05:04:07.247590",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "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>ImageId</th>\n",
       "      <th>defect_label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0000f269f.jpg</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>000ccc2ac.jpg</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>002451917.jpg</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>003c5da97.jpg</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0042e163f.jpg</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ImageId  defect_label\n",
       "0  0000f269f.jpg             0\n",
       "1  000ccc2ac.jpg             1\n",
       "2  002451917.jpg             1\n",
       "3  003c5da97.jpg             1\n",
       "4  0042e163f.jpg             1"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-03-19T05:04:32.241351Z",
     "iopub.status.busy": "2021-03-19T05:04:32.240646Z",
     "iopub.status.idle": "2021-03-19T05:04:32.243924Z",
     "shell.execute_reply": "2021-03-19T05:04:32.244405Z"
    },
    "papermill": {
     "duration": 8.613271,
     "end_time": "2021-03-19T05:04:32.244593",
     "exception": false,
     "start_time": "2021-03-19T05:04:23.631322",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    3588\n",
       "0    1918\n",
       "Name: defect_label, dtype: int64"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission_df['defect_label'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2021-03-19T05:04:48.203533Z",
     "iopub.status.busy": "2021-03-19T05:04:48.202715Z",
     "iopub.status.idle": "2021-03-19T05:04:48.338426Z",
     "shell.execute_reply": "2021-03-19T05:04:48.337937Z"
    },
    "papermill": {
     "duration": 8.216662,
     "end_time": "2021-03-19T05:04:48.338547",
     "exception": false,
     "start_time": "2021-03-19T05:04:40.121885",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "#pd.DataFrame(history.history).to_csv('CNN_BaseModel_history.csv', index=False)\n",
    "df.to_csv('train_binClass.csv', index=False)\n",
    "submission_df.to_csv('submission_df_classification_result.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "papermill": {
     "duration": 8.683003,
     "end_time": "2021-03-19T05:05:05.162133",
     "exception": false,
     "start_time": "2021-03-19T05:04:56.479130",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.0"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 2805.618516,
   "end_time": "2021-03-19T05:05:16.226094",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2021-03-19T04:18:30.607578",
   "version": "2.2.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}