ch20.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 20장 전이 학습을 통해 딥러닝의 성능 극대화하기"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[<img src=\"https://raw.githubusercontent.com/taehojo/taehojo.github.io/master/assets/images/linktocolab.png\" align=\"left\"/> ](https://colab.research.google.com/github/taehojo/deeplearning/blob/master/colab/ch20-colab.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. 소규모 데이터셋으로 만드는 강력한 학습 모델"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 실습: 치매 환자의 뇌인지 일반인의 뇌인지 예측하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 160 images belonging to 2 classes.\n",
      "Found 120 images belonging to 2 classes.\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 148, 148, 32)      896       \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 148, 148, 32)      0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 74, 74, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 72, 72, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 72, 72, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 36, 36, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 34, 34, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 34, 34, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 17, 17, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 18496)             0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                1183808   \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 65        \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 1)                 0         \n",
      "=================================================================\n",
      "Total params: 1,212,513\n",
      "Trainable params: 1,212,513\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Activation, Dropout, Flatten, Conv2D, MaxPooling2D\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras import optimizers\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "#학습셋의 변형을 설정하는 부분입니다. \n",
    "train_datagen = ImageDataGenerator(rescale=1./255,          # 주어진 이미지의 크기를 설정합니다.\n",
    "                                  horizontal_flip=True,     # 수평 대칭 이미지를 50% 확률로 만들어 추가합니다.\n",
    "                                  width_shift_range=0.1,    # 전체 크기의 15% 범위에서 좌우로 이동합니다.\n",
    "                                  height_shift_range=0.1,   # 마찬가지로 위, 아래로 이동합니다.\n",
    "                                  #rotation_range=5,        # 정해진 각도만큼 회전시킵니다.\n",
    "                                  #shear_range=0.7,         # 좌표 하나를 고정시키고 나머지를 이동시킵니다.\n",
    "                                  #zoom_range=1.2,          # 확대 또는 축소시킵니다.\n",
    "                                  #vertical_flip=True,      # 수직 대칭 이미지를 만듭니다.\n",
    "                                  #fill_mode='nearest'      # 빈 공간을 채우는 방법입니다. nearest 옵션은 가장 비슷한 색으로 채우게 됩니다.\n",
    "                                  )      \n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "       './data/train',   # 학습셋이 있는 폴더의 위치입니다.\n",
    "       target_size=(150, 150),\n",
    "       batch_size=5,\n",
    "       class_mode='binary')\n",
    "\n",
    "# 테스트셋은 이미지 부풀리기 과정을 진행하지 않습니다.\n",
    "test_datagen = ImageDataGenerator(rescale=1./255)  \n",
    "\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "       './data/test',   # 테스트셋이 있는 폴더의 위치입니다.\n",
    "       target_size=(150, 150),\n",
    "       batch_size=5,\n",
    "       class_mode='binary')\n",
    "\n",
    "\n",
    "# 앞서 배운 CNN 모델을 만들어 적용해 보겠습니다.\n",
    "model = Sequential()\n",
    "model.add(Conv2D(32, (3, 3), input_shape=(150,150,3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(1))\n",
    "model.add(Activation('sigmoid'))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: SciPy in c:\\users\\gilbut\\anaconda3\\envs\\deep_learning\\lib\\site-packages (1.7.3)\n",
      "Requirement already satisfied: numpy<1.23.0,>=1.16.5 in c:\\users\\gilbut\\anaconda3\\envs\\deep_learning\\lib\\site-packages (from SciPy) (1.19.5)\n"
     ]
    }
   ],
   "source": [
    "!pip install SciPy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "32/32 [==============================] - 3s 68ms/step - loss: 0.7053 - accuracy: 0.5063 - val_loss: 0.6896 - val_accuracy: 0.5000\n",
      "Epoch 2/100\n",
      "32/32 [==============================] - 2s 66ms/step - loss: 0.7028 - accuracy: 0.5125 - val_loss: 0.6904 - val_accuracy: 0.4800\n",
      "Epoch 3/100\n",
      "32/32 [==============================] - 2s 74ms/step - loss: 0.6913 - accuracy: 0.5188 - val_loss: 0.6883 - val_accuracy: 0.4800\n",
      "Epoch 4/100\n",
      "32/32 [==============================] - 2s 71ms/step - loss: 0.6854 - accuracy: 0.5375 - val_loss: 0.6786 - val_accuracy: 0.8200\n",
      "Epoch 5/100\n",
      "32/32 [==============================] - 2s 75ms/step - loss: 0.6756 - accuracy: 0.5750 - val_loss: 0.6573 - val_accuracy: 0.6400\n",
      "Epoch 6/100\n",
      "32/32 [==============================] - 2s 76ms/step - loss: 0.6825 - accuracy: 0.5813 - val_loss: 0.6378 - val_accuracy: 0.8400\n",
      "Epoch 7/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.6558 - accuracy: 0.6313 - val_loss: 0.6667 - val_accuracy: 0.5000\n",
      "Epoch 8/100\n",
      "32/32 [==============================] - 2s 63ms/step - loss: 0.6345 - accuracy: 0.6500 - val_loss: 0.5991 - val_accuracy: 0.6400\n",
      "Epoch 9/100\n",
      "32/32 [==============================] - 2s 67ms/step - loss: 0.6429 - accuracy: 0.6313 - val_loss: 0.6015 - val_accuracy: 0.8200\n",
      "Epoch 10/100\n",
      "32/32 [==============================] - 2s 73ms/step - loss: 0.5748 - accuracy: 0.6938 - val_loss: 0.4762 - val_accuracy: 0.9200\n",
      "Epoch 11/100\n",
      "32/32 [==============================] - 3s 77ms/step - loss: 0.5332 - accuracy: 0.7437 - val_loss: 0.4443 - val_accuracy: 0.8400\n",
      "Epoch 12/100\n",
      "32/32 [==============================] - 2s 75ms/step - loss: 0.4228 - accuracy: 0.8625 - val_loss: 0.4070 - val_accuracy: 0.8400\n",
      "Epoch 13/100\n",
      "32/32 [==============================] - 2s 65ms/step - loss: 0.3653 - accuracy: 0.8500 - val_loss: 0.2481 - val_accuracy: 0.9600\n",
      "Epoch 14/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.3076 - accuracy: 0.8813 - val_loss: 0.3884 - val_accuracy: 0.8400\n",
      "Epoch 15/100\n",
      "32/32 [==============================] - 2s 66ms/step - loss: 0.2986 - accuracy: 0.9062 - val_loss: 0.2137 - val_accuracy: 0.9200\n",
      "Epoch 16/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.2292 - accuracy: 0.9438 - val_loss: 0.1383 - val_accuracy: 0.9600\n",
      "Epoch 17/100\n",
      "32/32 [==============================] - 2s 66ms/step - loss: 0.2399 - accuracy: 0.9062 - val_loss: 0.1318 - val_accuracy: 0.9800\n",
      "Epoch 18/100\n",
      "32/32 [==============================] - 2s 65ms/step - loss: 0.2025 - accuracy: 0.9000 - val_loss: 0.1201 - val_accuracy: 0.9800\n",
      "Epoch 19/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.1659 - accuracy: 0.9375 - val_loss: 0.1257 - val_accuracy: 0.9800\n",
      "Epoch 20/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.1465 - accuracy: 0.9438 - val_loss: 0.1550 - val_accuracy: 0.9000\n",
      "Epoch 21/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.2073 - accuracy: 0.9312 - val_loss: 0.1501 - val_accuracy: 0.9400\n",
      "Epoch 22/100\n",
      "32/32 [==============================] - 2s 63ms/step - loss: 0.1289 - accuracy: 0.9688 - val_loss: 0.1103 - val_accuracy: 0.9200\n",
      "Epoch 23/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.1045 - accuracy: 0.9688 - val_loss: 0.0534 - val_accuracy: 0.9800\n",
      "Epoch 24/100\n",
      "32/32 [==============================] - 2s 63ms/step - loss: 0.0545 - accuracy: 0.9875 - val_loss: 0.1120 - val_accuracy: 0.9600\n",
      "Epoch 25/100\n",
      "32/32 [==============================] - 2s 67ms/step - loss: 0.0969 - accuracy: 0.9563 - val_loss: 0.0927 - val_accuracy: 0.9600\n",
      "Epoch 26/100\n",
      "32/32 [==============================] - 2s 65ms/step - loss: 0.1004 - accuracy: 0.9625 - val_loss: 0.1100 - val_accuracy: 0.9600\n",
      "Epoch 27/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.0948 - accuracy: 0.9563 - val_loss: 0.0291 - val_accuracy: 1.0000\n",
      "Epoch 28/100\n",
      "32/32 [==============================] - 2s 63ms/step - loss: 0.1080 - accuracy: 0.9625 - val_loss: 0.0329 - val_accuracy: 1.0000\n",
      "Epoch 29/100\n",
      "32/32 [==============================] - 2s 64ms/step - loss: 0.1362 - accuracy: 0.9563 - val_loss: 0.0802 - val_accuracy: 0.9800\n",
      "Epoch 30/100\n",
      "32/32 [==============================] - 2s 73ms/step - loss: 0.0909 - accuracy: 0.9625 - val_loss: 0.0423 - val_accuracy: 0.9800\n",
      "Epoch 31/100\n",
      "32/32 [==============================] - 2s 65ms/step - loss: 0.0631 - accuracy: 0.9875 - val_loss: 0.1227 - val_accuracy: 0.9600\n",
      "Epoch 32/100\n",
      "32/32 [==============================] - 2s 65ms/step - loss: 0.1099 - accuracy: 0.9563 - val_loss: 0.0388 - val_accuracy: 1.0000\n"
     ]
    }
   ],
   "source": [
    "#모델 실행의 옵션을 설정합니다. \n",
    "model.compile(loss='binary_crossentropy', optimizer=optimizers.Adam(learning_rate=0.0002), metrics=['accuracy'])\n",
    "\n",
    "# 학습의 조기 중단을 설정합니다.\n",
    "early_stopping_callback = EarlyStopping(monitor='val_loss', patience=5)\n",
    "\n",
    "#모델을 실행합니다\n",
    "history = model.fit(\n",
    "       train_generator,\n",
    "       epochs=100,\n",
    "       validation_data=test_generator,\n",
    "       validation_steps=10, \n",
    "       callbacks=[early_stopping_callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEGCAYAAABo25JHAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAABF0klEQVR4nO3dd3iTVfvA8e/pLnsjAgpuwUIZFiMiZYsogiKigPCi4gZERYYoG3HhQpQf4npRVFAE4RUVWkCoskRlKCKiFlQQZBSh8/79cdJJR5L2adP2/lxXrjTJyZNzmjZ3njPuY0QEpZRS5VdASVdAKaVUydJAoJRS5ZwGAqWUKuc0ECilVDmngUAppcq5oJKugLdq1aoljRo18um5J06coGLFikVboWKmbfAfZaEd2gb/UBxt2Lx5898iUju3x0pdIGjUqBGbNm3y6bmxsbFER0cXbYWKmbbBf5SFdmgb/ENxtMEY82tej2nXkFJKlXMaCJRSqpzTQKCUUuVcqRsjUEr5n+TkZOLj4zl16lSxv3bVqlXZuXNnsb9uUSrKNoSFhdGgQQOCg4M9fo4GAqVUocXHx1O5cmUaNWqEMaZYX/v48eNUrly5WF+zqBVVG0SEQ4cOER8fT+PGjT1+nnYNKaUK7dSpU9SsWbPYg4DKzhhDzZo1vT4zKzeBIC4O5s8/i7i4kq6JUmWTBgH/4Mv74GjXkDHmKuB5IBCYKyJP5Hh8JtDBfbMCUEdEqhV1PeLioEMHSExszJtvwrhxcNllULOmvdSoAVWqgDG2bGwsREeDy1XUNVFKKf/jWCAwxgQCs4AuQDyw0RizRER2pJcRkQeylL8faOFEXWJjISkJwJCcDBMmnF4mKAgqV4YjR0AEQkJg2TLo3NmJGimllP9wsmsoCtgtIntEJAlYAFyXT/mbgXedqEh0NISFQUBAGmFh8M478OWX8PHH8Prr8PTT8PDDcN55NgiADRzdu8MNN8AHH8C//zpRM6VUUTh06BCRkZFERkZyxhlnUL9+/YzbSfZbYL5iY2NZv369T6+9d+9e3nnnnQKPf8011/h0/OLgZNdQfeD3LLfjgTa5FTTGnA00Blbl8fhQYChA3bp1iY2N9boyTz1VhQ0bwomKOkm9esdITrbdQVWqQHrqovr1q/Ddd81JTjYEBoLL9TcxMdX48MMQwsJSufzyv+nY8QCXXnqYn36qzNat1YiMPELTpse8ro+vEhISfGq/PykLbYCy0Y6iakPVqlU5fvy4V88J+Pprgr78kpQrriCtTa4fDR5JTU0lJCSEtWvXAjBt2jQqVarEsGHDAEhMTCQxMTHfY6xYsYJKlSoRERHh9evv2LGDt956i2uvvTbPMv/++y8pKSl5/o5SU1O9/v3l59SpU169r/4yfbQfsFBEUnN7UETmAHMAWrduLb7k5IgOjaP71tmc03IIuKLtV//kZPvV332JPieJlkdiiV2RSHTvGrjub02qCWL1aliwIJBFi+qyalVdKlaEUycFSRNCQuCzLwJo164QrfeC5lXxH2WhHUXVhp07d2ZOfxwxArZuzf8JR4/Cd99BWhqhAQHQrBlUrZp3+chIeO65XB/KOfUyNDSU0NBQdu3axciRI0lISKBWrVq88cYb1KtXjxdeeIFXXnmFoKAgmjRpwhNPPMHrr79OYGAgH3zwAS+++CJ//vknEydOJDAwkKpVq7JmzRpSU1MZPXo0sbGxJCYmcu+993LnnXcyadIkdu7cSbt27Rg0aBAPPPDAaXWsUKECQUFBVK5cmcOHDzNkyBD27NlDhQoVmDNnDo0bN2bLli0MHz4csAO+a9asISEhgZtuuoljx46RkpLC7NmzaefBh01YWBgtWnje0+5kINgHNMxyu4H7vtz0A+51rCZxcdC+PY2Tk2HuXDsgkJKSa1GX+8JaYCQEVqpEx2rV6Fi1KrMurMkXKdGM+2EA36SdBwRwKgk6RKcR0SyAiAj79xwRYS/16sFXX+ngs1KnOXoU0tLsz2lp9nZ+gcBLIsL999/Pxx9/TO3atXnvvfcYN24c8+bN44knnuCXX34hNDSUI0eOUK1aNe666y4qVarEQw89BEBERAQrVqygfv36HDlyBIDXXnuNqlWrsnHjRhITE2nbti1du3bliSee4Omnn+aTTz7xqG6PP/44LVq0YPHixaxatYpbb72VtWvX8vTTTzNr1izatm1LQkICYWFhzJkzh27dujFu3DhSU1P516E+aicDwUbgfGNMY2wA6AfckrOQMeYioDrg3MTO2FhITcXYF4S2baF9ezsiHBpqr0NC4PPPYfFi+4cZEGCnGjVrZkeQjx4l+MgRuh/5hGp8RSc+IokQAkmlT9BSDv1zCSuXN+Ltt8MyXrZKFUhIyBx8XrnSvrSv0qfAhoZqUFF+LI9v7tnExUGnTvZsPCQE5s8v0j/qxMREtm3bRpcuXQDb9VKvXj0AmjVrRv/+/enVqxe9evXK9flt27Zl8ODB9O3bl+uvvx6Azz77jO+++46FCxcCcPToUX766SdCQkK8qtuXX37JokWLAOjYsSOHDh3i2LFjtG3blpEjR9K/f3+uv/56GjRowKWXXsqQIUNITk6mV69eREZG+vDbKJhjgUBEUowx9wErsNNH54nIdmPMJGCTiCxxF+0HLBBJH6Z1QHQ0hIaSlphIQGgoTJ+e+x9d8+bwv/9l/nFOnpxrOVdcHCujryY2uS3RAWtwRSbZU+GkJA5VPIvvW9zK9w268/bO1mz8NhgwJCZCt2528LlHD+jaFapVK7jqhw/D5s2waJE9mUlNbczbb0NMjAYDVYq5XPabkUOnyyJC06ZNictl4dCyZctYs2YNS5cuZerUqXz//fenlXnllVf4+uuvWbZsGa1atWLz5s2ICC+++CLdunXLVraoxolGjx5Njx49WL58OW3btmXFihVceeWVrFmzhmXLljF48GBGjhzJrbfeWiSvl5WjYwQishxYnuO+x3LcnuBkHYCMP7q98+ZxzpAhef/RefrH6XLhip2OKzYWop+05U6cgFWrqLlsGdHL3yT6yym05jI6sZIkggkkjcub/ssnn1TnrbcgMNCeHVx9tQ0Mx47BZ59BnTpw8iRs3AibNsHPP+d8cRtUnnkG3F9MlCqdXC7Hvs2EhoZy8OBB4uLicLlcJCcns2vXLi6++GJ+//13OnTowBVXXMGCBQtISEigcuXKHDuWOenj559/pk2bNrRp04b//e9//P7773Tr1o3Zs2fTsWNHgoOD2bVrF/Xr16dy5cpeDfS2a9eO+fPnM378eGJjY6lVqxZVqlTh559/JiIigoiICDZu3MgPP/xAeHg4DRo04I477iAxMZEtW7aUvkDgV1wufktM5JyC/vA8/ePMWa5iRbj2WnsRge3bcT38MCs/7UQs0UQTi6tdW1LXP83XX9s1CsuXw+jR9pLTWWdB69Zw++32OjUVeveGxMQ0IIBFi+Dee2HmTHvyopTKFBAQwMKFCxk2bBhHjx4lJSWFESNGcMEFFzBgwACOHj2KiDBs2DCqVavGtddeS58+ffj444958cUXmTlzJj/99BMiQqdOnWjevDnNmjVj7969tGzZEhGhdu3aLF68mGbNmhEYGEjz5s0ZPHhwroPFWU2YMIEhQ4bQrFkzKlSowJtvvgnAc889R0xMDAEBATRt2pTu3buzYMECnnrqKYKDg6lUqRJvvfWWM78wESlVl1atWomvYmJifH6uT9avFwkPFwkIEAGRkBCR6dNFkpIyisTHi/TubR8GW3Ts2LwPd/vtP8uaNSIPPmjLt20rsn9/MbWniBT7++CQstCOomrDjh07iuQ4vjh27FiJvXZRKeo25PZ+YLvkc/1cLTe5hkpEelfTlCl29do118CYMfYr/oYNANSvbxezhYfb7qLQUFssr8P17/8b7drZRXDvvgvffAOtWkG2tTCrV8PYsWhiJaWUJzQQOM3lsh/+PXvaEd+PPoJDh2yyo+HD4fjxjHgxebK99rTbtF8/+1kfHm6HNGbPBvlosZ3tNH06dOyowUCpYrRixYqMFc3pl969e5d0tQpUfsYI/EWvXvYDetw4ePFF+PBDePllXLVq4SIWiMa9ksEjzZrZQeX+/eGee2BT8DFmSQhhJEJioh341ulFShWLbt26nTarqDTQQFASqlSxQeCWW+COO+zZQmCgfSx9wYEXH97Vq8PSoUuZ8Pl2piSPJo5LuZYl9JLFuByad6yUKju0a6gkuVywZYtdVJCaai+nTtmFbd544QUCb+jF5MhFTB97jJ1cxJOMIppY1r+715GqK6XKDg0EJS0kxObFDnOvSBaxnf2LFmWmQs1LaioMG2bHGnr2hNWrkUpVCAgwgCGJUG79b1d2b/zH6VYopUoxDQT+wOWCVatg2jR44QWoVQv69LFL8HNZ9QjY3BW9etkupgcftKvLKlRIX0RNYCAEB6Xxh9Ql4vJKPPGEzbHni7g4O/as485KlU0aCPxF+uyi+++3c0JnzYJvv7VZF++91840Srd/P1x5pV2R9vLLdi6pe4wh6wyk1WsC2HXNg1zN8pyzVj22dq2dhDR+vI1LGgyUPyrMfgSbNm3KSFldVN544w3279+fb5no6Gg2bdpUpK/rKx0s9kdBQXYKUL9+8Pjjtqvo3XfhP//hvD17YN06m4di6VKboyKHbIuep93Hok+asfjG+dy3/hYuu8z2Jk2ebHdkyyk52eY2io21l1WrMs8kkpJ0EpIqOkW5LWzNmjXZ6k59PWHChGyZRAFSUlIICsr9465169a0bt26cBXI4Y033uCSSy7hzDPPLNLjOkUDgT+rUcN2/dx5JwweDM8+SwOwGVTfeCPXIHCaiAi44QZ6rbibjt91Z+xT1XnhBTtrddgwOzZdt6494YiNtTu3nThhn9q0qR16WLzYDkcEBdl/WqXy4+V2BBRyO4I8DR48mLCwML755hvatm1Lv379GD58OKdOnSI8PJzXX3+dCy+8kNjY2Iw00hMmTOC3335jz549/Pbbb4wYMYJhw4Zx4sQJ+vbtS3x8PKmpqYwfP56bbrqJzZs3n7bnwbp169i0aRP9+/cnPDycuLg4wsPD863rBx98wMyZMxERevTowYwZM0hNTeW2225j06ZNGGMYMmQIDzzwwGn7KSxYsMC7X0wuNBCUBpdcAtdfb7uM0v9z9uW1tUMuHnsMFi2iyrzneOmlidxyCwwYYFc0Z9W0qY030dG256lOHXv/Z5/ZgOBgjjBVzji8HUGG+Ph41q9fT2BgIMeOHWPt2rUEBQXxxRdfMHbs2Ix00Fn98MMPxMTEcPz4cS688ELuvvtuPv30U84880yWLVvmrv9RkpOT89zz4KWXXuLpp5/26Exj//79PP7442zZsoXq1avTtWtXFi9eTMOGDdm3bx/btm0DyNgXIed+CkVBA0Fp0aFDZirtkBDvvpo3a2YDyXPPwYgRXH55dYYMsfFBxMaV0aNh6tTcn961K4waZbuTtm2zcUmpvPjBdgQZbrzxRgLd42dHjx5l0KBB/PTTTxhjSM5j9kSPHj0ydjmrU6cOf/31FxERETz44IM88sgjXHPNNbRr145t27blueeBNzZu3MgVV1xB7dq1Aejfvz9r1qxh/Pjx7Nmzh/vvv58ePXrQtWtXwLP9FLylg8WlRXoq7SFDvF5wBthP/WPH4PnnAftPGBZWcH6jdCNGQKVKNm2SUoXla1oVb1WsWDHj5/Hjx9OhQwe2bdvG0qVLOXXqVK7PCQ0Nzfg5MDCQlJQULrjgArZs2UJERASPPvookyZNytjzYOvWrWzdupXvv/+ezz77rMjqXr16db799luio6N55ZVXuP322wG7n8K9997Lli1buPTSS0nJY7dFb2ggKE1cLn7r39+3/5rmze100+eegyNHvP5HrFHDTmh6/33YscOXyiuVXfpEueLqbjx69Cj169cH7GCuN/bv30+FChUYMGAADz/8MFu2bOHCCy/M2PMAIDk5me3btwN4tUdBVFQU69at4++//yY1NZV3332X9u3b8/fff5OWlsYNN9zAlClT2LJlC2lpaRn7KcyYMYOjR4+SkJDgVVtyo4GgPHnsMdsZ+8ILgPf/iCNHQoUKeXchKeXPRo0axZgxY2jRooXX36K///57oqKiiIyMZOLEiTz66KOEhISwcOFCHnnkEZo3b05kZCTr3WmABw8ezF133UVkZCQnT57M99j16tVjwoQJdOjQgebNm9OqVSuuu+469u3bR3R0NJGRkQwYMIDp06eTmprKgAEDiIiIoEWLFhn7KRRaXvmp/fVSqvYjcECh23DddSLVqokcOeLT00eNsnsm/PCD71UoC++DSNloh+5H4B90PwJVvB57DI4cyTgr8NaDD9oxhWnTirZaSqmSo4GgvGnZ0m6nOXOmHTz2Up06cPfddpbH7t0O1E+pMqZ3796n7VGwYsWKkq5WNo4GAmPMVcaYH40xu40xuezMC8aYvsaYHcaY7caYd5ysj3J7/HH45x+7WM0HDz8MwcF6VqCyk4KSJJZTH330UcbMovSLk3sW+PI+OBYIjDGBwCygO9AEuNkY0yRHmfOBMUBbEWkKjHCqPiqLVq3sfNEZM2xQyCuBUFIS/PmnTW8xeXJGuTPOsIud33oLfvmlGOut/FZYWBiHDh3SYFDCRIRDhw4Rlp7N2ENOLiiLAnaLyB4AY8wC4Dog6+TDO4BZIvIPgIgccLA+KqveveGTT2DSJDsNqH17m7ri0CE4fNhesk5LM8amIHXPNR01Cl55xd41Z07JNUP5hwYNGhAfH8/BgweL/bVPnTrl9QefvynKNoSFhdGgQQOvnuNkIKgP/J7ldjzQJkeZCwCMMeuAQGCCiHya80DGmKHAUIC6desSGxvrU4USEhJ8fq6/KKo2nBUXR2PAAJKaSvLmzZxs0IDkKlVIuegie12lCpV++IFa69djREhLTGTvvHn8lpgIQPfu5/P66/Xo2PFrzjgjsdjbUNLKQjvKShsqVapU0tUolKJuw6+//urdE/KaTlTYC9AHmJvl9kDgpRxlPgE+AoKBxtjAUS2/4+r00ZiiOdD69SLh4SKBgfZ6/fq8ywUHi4BIWFi2cr/9Zh+6+27vXrosvA8iZaMd2gb/UBxtoISmj+4DGma53cB9X1bxwBIRSRaRX4BdwPkO1kml83RpscsFb75pfx42LFu5hg1hyBB47TWIjy+GOiulHOFkINgInG+MaWyMCQH6AUtylFkMRAMYY2phu4r2OFgnlZWnS4v79YMGDWDP6W/NmDE2e+SMGQ7VUSnlOMcCgYikAPcBK4CdwPsist0YM8kY09NdbAVwyBizA4gBHhaRQ7kfUZUYY6BzZ3vmkJqa7aGzz7apq//v/+zGaUqp0sfRdQQislxELhCRc0Vkqvu+x0RkiftnEZGRItJERCJEpPA7LChndOli1x58881pD40ZAykp8MADurexUqWRrixWnunUyV5//vlpD51zDnTrZjOT6t7GSpU+GgiUZ+rWtRvcfPFFrg9fdJG9Tk3N3NtYKVU6aCBQnuvSxW5q/O+/pz3Up49NOwF2SEH3Nlaq9NBAoDzXubP9ur927WkPuVz2LKB1azte4M2WykqpkqWBQHmuXTu7wWwe3UOXX25PGFwuGDQIvvuumOunlPKJBgLluYoV7ad9LgPG6UJDYdEiqFbN7ox5SCcDK+X3NBAo73TpAt9+Cwfyzg9Yrx58+KHtHrrpJttVpJTyXxoIlHe6dLHXK1fmW6xNG5uddOVKeOSRYqiXUspnGgiUd1q2hOrV8xwnyOo//4H774dnn4W33y6GuimlfKKBQHknMBA6drTjBB5sQvLMM9ChA9xxB2zaVAz1U0p5TQOB8l7nzvD777BrV4FFg4Phvffsrma9e8NffxVD/ZRSXtFAoLyXPk7gQfcQQO3asHixnUHUpw8kJxvn6qaU8poGAuW9c8+Fxo3znUaaU2QkzJtn1xkMHx6puYiU8iMaCJRvOneGmBiv5oaefTYEBcHOnVXp0EET0ynlLzQQKN906QLHjsHGjR4/JTY2c3xZE9Mp5T80ECjfdOxos8t50T0UHW0zVIBk3FZKlTwNBMo3NWvaNQUeDhhD5jbJl1/+NyJQo4aD9VNKeUwDgfJdly62o//4cY+f4nLB8OG7AViScwdrpVSJ0ECgfNe5sx0sXrPGq6fVqZNIixYaCJTyFxoIlO/atoWwMK/GCdL17Anr18PBgw7USynlFUcDgTHmKmPMj8aY3caY0bk8PtgYc9AYs9V9ud3J+qgiFhZm9yjwMRCkpcHy5Q7USynlFccCgTEmEJgFdAeaADcbY5rkUvQ9EYl0X+Y6VR/lkC5dYMcO2L/fq6e1aAH162v3kFL+wMkzgihgt4jsEZEkYAFwnYOvp0qCl+km0hljzwpWrIBTpxyol1LKY0EOHrs+8HuW2/FAm1zK3WCMuRLYBTwgIr/nLGCMGQoMBahbty6xPq5ESkhI8Pm5/sLv2pCWxuXVqnH4v//lh7PO8ugp6W1o1KgGJ0404/nnv6NNm8MOV7To+d174QNtg38o8TaIiCMXoA8wN8vtgcBLOcrUBELdP98JrCrouK1atRJfxcTE+Pxcf+GXbejXT6RePZG0NI+Kp7fh5EmRihVF7rrLwbo5yC/fCy9pG/xDcbQB2CR5fK462TW0D2iY5XYD931Zg9AhEUl035wLtHKwPsopnTvDH3/YsQIvhIVBt252nMCDrQ2UUg5xMhBsBM43xjQ2xoQA/YBsQ4PGmHpZbvYEdjpYH+WU9HECH2cP7d8PW7YUcZ2UUh5zLBCISApwH7AC+wH/vohsN8ZMMsb0dBcbZozZboz5FhgGDHaqPspBZ50FF1zgUyC4+moICIClSx2ol1LKI46uIxCR5SJygYicKyJT3fc9JiJL3D+PEZGmItJcRDqIyA9O1kc5qHNnWLUKpkzxKr907dpw+eU6jVSpkqQri1XRaNjQzgN9/HHo1MmrYNCzJ3zzjd39UilV/DQQqKJx8qS9TkvzerOBnu6OQu0eUqpkaCBQReOqq+w0IIDUVDjzTI+feuGFdohBu4eUKhkaCFTRcLnsGMHIkVCnDowY4XX30KpVdtMzpVTx0kCgio7LBc88Axs2QK1adlrpqlUePbVnT0hOhs8+c7iOSqnTaCBQRe/ss+0eBY0b2/mhn3xS4FNcLrtjmXYPKVX8NBAoZ9SrZweMIyKgd2947718iwcFQY8esGyZ3etGKVV8NBAo59Ssmb5JMdx8M7z2Wr7Fe/aEw4fthjVKqeKjgUA5q0oV+N//bFKh22+H55/Ps2i3bhASot1DShU3DQTKeRUqwOLFcMMNMGIEzR56KNev/ZUrQ4cOGgiUKm4aCFTxCA2F4cMhMJAamzdDdHSu00t79oSffoIffyz+KipVXmkgUMXnyy8zf05OhmefPa3Itdfaaz0rUKr4aCBQxSc6GkJCSAsIsClHP/oIPv00W5GGDe1+xhoIlCo+GghU8XG5YOVK9g4ZYjcrbtYM+vSBjRuzFevZ0w4hHDxYQvVUqpzRQKCKl8vFb/3727TVy5fbdBRXX20HBtx69rS565YvL8F6KlWOaCBQJeeMM+yZAdi5o3/+CdiuoVq14KmnvEpXpJTykQYCVbLOP99+9T9wALp3h2PH+Oor+Ocf2L7d660NlFI+0ECgSt6ll8LChbBtG1x/PbFfpGRsZn/qlFdbGyilfKCBQPmHq66CefNg5Uqi10wiNFQwBkTslshKKedoIFD+Y+BAmDED1xeTWekaz9j2X1KpQgqvv07GGYJSqug5GgiMMVcZY340xuw2xozOp9wNxhgxxrR2sj6qFHj4YejbF9eqqUxZfSXTk0excqXNUKGUcoZjgcAYEwjMAroDTYCbjTFNcilXGRgOfO1UXVQpYgw0b25/FuGu1FlcUvcADz5oxwuUUkXPyTOCKGC3iOwRkSRgAXBdLuUmAzMA/TdXVocONjcREBRseH7cQX75xW5+ppQqekEOHrs+8HuW2/FAm6wFjDEtgYYisswY83BeBzLGDAWGAtStW5dYH6eRJCQk+Pxcf1Fe2lBt6lSaP/QQf3TqREDEQa688iBTptTgggs2ULt2YvFUtADl5b3wd9qGIiAijlyAPsDcLLcHAi9luR0AxAKN3LdjgdYFHbdVq1biq5iYGJ+f6y/KVRuaNhW5+moREfnlF5GwMJGbb3asWl4rV++FH9M2eAbYJHl8rjrZNbQPaJjldgP3fekqA5cAscaYvcBlwBIdMFYZoqJgwwYQoVEjGDUK3n0X1q714VhxcTB9uq5OUyoXTgaCjcD5xpjGxpgQoB+QkVNSRI6KSC0RaSQijYCvgJ4issnBOqnSJCoK/v4b9u4F4JFHbHbSYcMgNdWL48TFwZVXwrhxulRZqVx4FAiMMcONMVWM9ZoxZosxpmt+zxGRFOA+YAWwE3hfRLYbYyYZY3oWvuqqzGvjHlLasAGwG509/TRs3Qpz53pxnC++gJQUuxghKUmXKiuVg6dnBENE5BjQFaiO7e9/oqAnichyEblARM4Vkanu+x4TkdOyzYtItJ4NqGwuuQTCwjICAcCNN0L79vbL/T//eHicWrUyfw4JsfsiKKUyeBoIjPv6auBtEdme5T6lnBEcDC1bZgsExsALL9gg8PjjHh4na8R46SW7L4JSKoOngWCzMeYzbCBY4V4EluZctZRyi4qCzZtt145bs2Zw113w8ss2T12B1q6F6tXtz14NLihVPngaCG4DRgOXisi/QDDwH8dqpVS6qCg4edLmpM5i0iSoUgWGDy8gD1FqKqxbB337QuXK8M03ztZXqVLI00DgAn4UkSPGmAHAo8BR56qllFtUlL3+OnsGkpo1YcoUWLXK5qrLcyLQt9/C8eN21lBkpAYCpXLhaSCYDfxrjGkOPAj8DLzlWK2USnfOOVCjRrZxgnQREXbMYP78fGaFpi86aNfOjjd8+612DymVg6eBIMW9Mu067OrgWdgFYUo5y5jMhWU5fPmlfRjy2cBm7Vpo1MguQGjRwnYz/fijkzVWqtTxNBAcN8aMwU4bXWaMCcCOEyjlvKgoO0aQkJDt7ujojNx0iMDFF+d4ngisWWPPBsAGAtDuIaVy8DQQ3AQkYtcT/IlNF/GUY7VSKqs2bSAtDbZsyXa3ywUrV9otDEJC4L33cjxv1y44eNCOD4CNFKGhGgiUysGjQOD+8J8PVDXGXAOcEhEdI1DF49JL7XUu3UMuFzz5JIweDQsW5MhDlHV8AOy6hIgIDQRK5eBpiom+wAbgRqAv8LUxpo+TFVMqQ+3a0LhxroEgXXoeouHDs4wFr10LderABRdkFmzRwgYC3ftSqQyedg2Nw64hGCQit2I3nRnvXLWUyiGPAeN0FSrAU0/Zz/h589x3rlkDV1yROaIMNhD88w/8+quz9VWqFPE0EASIyIEstw958VylCi8qyn54//VXnkX69rXDAWPHwpHt+2zW0vTxgXQ6YKzUaTz9MP/UGLPCGDPYGDMYWAYsd65aSuWQvrAsn7MCY+D55+HwYZj48HF7Z/r4QLpmzSAgQAOBUll4Olj8MDAHaOa+zBGRR5ysmFLZtGgBgYH5BgKwi4fvuANeWnE+Oyu0gubNsxeoUAEuukgDgVJZeNy9IyKLRGSk+/KRk5VS6jQVK9q01AUEAoDJk6EiJxgR/ioSEHh6gfQBY6UUUEAgMMYcN8Ycy+Vy3BhzrLgqqRRg1xO4t67MT+2AQ0xMG89nh1qxdGkuBVq0gH374MCBXB5UqvzJNxCISGURqZLLpbKIVCmuSioF2HGCI0dg9+78y61bxz28TJNGJxg5EhITczyuA8ZKZaMzf1Tp4cGAMQBr1hAcEsBzLwXz888wc2aOxzUQKJWNBgJVejRpYscKCgoEa9dCmzZ06RHCddfZdNX792d5vHp1m4hOA4FSgAYCVZoEBkKrVvkHghMnbE4i97TRZ56B5GSbgiKbFi1Oy12kVHmlgUCVLlFR9pt8UlLuj3/1ld3W0h0Izj0XHnwQ3n7bTivN2LOgRQs71nBM5zwo5WggMMZcZYz50Riz2xiT8zsZxpi7jDHfG2O2GmO+NMY0cbI+qgyIirKjv999l/vja9bYBWOXX55xV6dOdrHZ3LnQti1cfz18dKo7R6hqN6pRqpxzLBAYYwKBWUB3oAlwcy4f9O+ISISIRAJPAs86VR9VRhQ0YLx2rV1VViVzUtuGDTY2gJ15umwZXD+tNTU5xGUDz2f8eFi92p5kxMXB9On5bH2pVBnk5BlBFLBbRPaISBKwALvDWQYRyXpeXhHQlJAqf2edZTOK5hYIkpJs11COtBLR0Xa/gsBACA+Hzz+HNauFRyvMJODfBKZPt2WqVrU56h59NJ+tL5Uqg4IcPHZ94Pcst+OBNjkLGWPuBUYCIUDH3A5kjBkKDAWoW7cusbnuSViwhIQEn5/rL7QNcMl55xEeG8vGHMeosn07LU+eZFv16vyd47GnnqrC1q3ViIw8Qlqa/f4xvOn7PHxoNrHz3mDr1uq8+25DduyoAhgSE9OYN28viYm/OdYOf6Bt8A8l3gYRceQC9AHmZrk9ELvfcV7lbwHeLOi4rVq1El/FxMT4/Fx/oW0QkcmTRYwROXIk+/0zZoiAyJ9/enac0aNFgoJETp0SEZH160WCg+0hQkPt7fzoe+EftA2eATZJHp+rTnYN7QMaZrndwH1fXhYAvRysjyoroqJsZ//mzdnvX7sWLrwQ6tb17DgtW9oZRtu2AXa3sxUr7FKF5s3tbaXKAycDwUbgfGNMY2NMCNAPWJK1gDHm/Cw3ewA/OVgfVVa0bm2vs44TpKXBl1+ennY6P7msMO7QAR57zB5648YiqKtSpYBjgUBEUoD7gBXATuB9EdlujJlkjOnpLnafMWa7MWYrdpxgkFP1UWVIjRpw/vnZA8G2bTYPkTeB4JxzoHLl01YY3323XXw8dWrRVFcpf+fkYDEispwcG9iIyGNZfh7u5OurMiwqCrIOruXcqN4TAQF2qmmOFcaVK8OIEfD443a5QrNmha2sUv5NVxar0ikqyqaS3ucedlq7Fho0sDmEvNGypf20z9jx3rr/fhsQpk0rmuoq5c80EKjSKevCMhEbCNq1y75RvSdatIB//4Vdu7LdXb063HcfvP8+/PhjEdVZKT+lgUCVTpGREBxsA8GePTa9qDfdQunySUn9wAMQFmZXGitVlmkgUKVTWJid47lhQ+b4wJVXen+ciy+G0NBcA0Ht2nDXXfDf/8IvvxSyvkr5MQ0EqvSKirJzPFevtjOJLr7Y+2MEB9u9kPNISf3QQzY1xYwZhayrUn5MA4EqvaKi4Phx+OADmyQowMc/55Yt7RlBLnshn3km3HYbvP46xMcXsr5K+SkNBKr0Sh8wPnHCt/GBdC1awD//wG+55xUaNcquV3v6ad9fQil/poFAlV4XXggVKtifq1f3/TgF7GHcqBEMHAhz5sBff/n+Mkr5Kw0EqvT6+ms4dcr+fP/9vueNbtbMdivls4fx6NF2P5yZM317CaX8mQYCVXplXVmclJT9tjcqVICLLsp3D+MLLoC+fWHWLDh82LPD6iY3qrTQQKBKr+hoO/UzMNDuPBMd7fuxWrTI94wAYOxYSEiAF14o+HDr1tkEduPG6SY3yv9pIFCll8sFK1fC5Mn2ujB5o1u0sOkqDh7Ms0hEBPTqBc8/n/ue98nJthr33Qfdu9uuJBHbexUT43vVlHKao0nnlHKcy1U0GwdkHTDu2jXPYuPGweLF8MgjkJJyFmCDwocfwtKlttsoPBwuvdTumpmcbIPB77/neUilSpwGAqXA40DQujW0aQOvvALQmLlz7f3VqsG110Lv3tCtmx12iIuzZwKrV9vybdrA4MEOt0MpH2ggUArs9NNGjfIdME4XGWknLIHBGPvh/uqrdpFyVuknKw89BNdcA7ffDnXqwNVXF331lSoMHSNQKp0HA8YAgwbZVEcBAWmEhcEdd5weBLIKCYFFi2xqpBtvTA8iSvkPDQRKpWvRAn76yaatyIfLBatWwZAhez0eo65cGZYvhzPOgB49Tst6rVSJ0kCgVLr0cYKHHipwvqfLBf37/+bVOHXdurBihV271q0b/PFHIeqqVBHSQKBUurQ0e/1//+fY5P/zzrNnBgcP2immR48W+Uso5TUNBEql27bNXosUbqVyAVq3tmMG27fD9dfb9QYlIS4O5s8/Sxe7KWcDgTHmKmPMj8aY3caY0bk8PtIYs8MY850xZqUx5mwn66NUvjp0yBz1DQoq3ErlAnTrBvPm2bGGHj3s3sjF+YG8fDm0bQuvvdZYVz4r5wKBMSYQmAV0B5oANxtjmuQo9g3QWkSaAQuBJ52qj1IFcrng88+hUiW7R0FRLFTLx8CBcO+9djXyo48WbyqKyZPtiY+IcfLkR5USTp4RRAG7RWSPiCQBC4DrshYQkRgR+dd98yuggYP1Uapg7dvbDQji4mDnTsdf7swzwZjMVBTF8YG8a5fd2C2dwyc/qhRwckFZfSDrwvp4oE0+5W8D/pfbA8aYocBQgLp16xLr439LQkKCz8/1F9oG5wVHRHBZSAh/jRrFrgcfzLNcUbSjatUqhIQ0JzExABHDd9/9SmyssxskP/ZYU0JCqvPQQz8ybdpFREb+Q2LitlJ7VuDvf0+eKPE2iIgjF6APMDfL7YHAS3mUHYA9Iwgt6LitWrUSX8XExPj8XH+hbSgmQ4eKhIWJHDiQZ5Giasf69SKTJ4u4XCIBASILFxbJYXO1dq3tEJo0yd7u2TNeQkNFDh507jWdVir+ngpQHG0ANkken6tOdg3tAxpmud3AfV82xpjOwDigp4iU0PwJpXIYMcL21dikQo5yuewYwRdfwGWXwS232HGDoiZil0iceSaMHGnv69VrH4mJdsasKr+cDAQbgfONMY2NMSFAP2BJ1gLGmBbAq9ggcMDBuijlnYsvtkmBXnopcxc0h1WoAJ98YjfB6dULNm0q2uN/8IFNbzF5MlSsaO9r3PhfOnWCl1+GlJSifT1VejgWCEQkBbgPWAHsBN4Xke3GmEnGmJ7uYk8BlYAPjDFbjTFL8jicUsVv5Eg4cADefbfYXrJ6dbv6uFYtu+Dshx+K5riJiTBmjN1TYdCg7I8NGwbx8Ta9tiqfHF1HICLLReQCETlXRKa673tMRJa4f+4sInVFJNJ96Zn/EZUqRh072v2Mn33W9qsUkzPPtLNYAwJsRuz4+MIf8+WXYc8eeOopu6FbVj16QOPGnu28psomXVmsVF6MsWcF27bZDvxidN558OmnNgVF165w6JDvx/rnH9sd1KWLXciWU2CgXc+wdq1HyVdVGaSBQKn89Otns8U9+2yxv3SLFrBkif0mf/XVdr9kX0ydCkeO2LOBvAwZYscoXnzRt9dQpZsGAqXyExpqNyH+9FObHKiYtW8PCxbYgeNOnWDKFO9WH//yi/1wHzTI7oeQl+rV4dZb4Z134O+/C19vVbpoIFCqIHfdZXeiee65Enn5Xr3sHskbNsD48XbowtNgMHas7fqZPLngsvffj04lLac0EChVkFq17Ffqt9+2s4hKQOXKdsgC7GzWMWMK7irauNGeTYwcCQ08SN7SpAkZU0mTkwtfZ1V6aCBQyhMjRtivy7Nnl8jLR0fbk5LAQHtZvdquN3j77cxtFDLExSHTpvPQ0GPUrm1TJ3lKp5KWTxoIlPLERRfZeZazZhXbArOsXC672njyZDu7Z/16+y3/1lttOukNG9wF166F9u1ZMu5r1mytwsT2q6jyz6+5RIvcpU8l1UHj8sXJpHNKlS0jR9q+k/nz4bbbiv3lXa7smbG/+greegtGj4Y2bWBw/2SmrR9NrWThEZ7gQn7g9oXdYGGKPZ047zx7GpF+SUnhrA0b7IC4+8CBgXZs/MEH7VTS9N07VdmmgUApT3XoYKfePPusnW9ZwgICYPBgu8vZ1DHHmTk7jIXyKS3Yyo9cxIzAsQS/8LzdbGfXLnvZsQOWLs0YBGgMNrCtXJkRDIYMsYPSL75oN89RZZ92DSnlqfQFZjt2wGeflXRtMlT5bRszljZle/ilNG+SzFraAcKEoMnEtbgH7rjDLiL4+GO7x8K//9qv/MZgwI59ZEmBXK2aTiUtbzQQKOWNfv2gXj147DHOmj+/5Pd4XLnSDhKkpHD+ujfoMaAGAQEAhqSUwNz3GAgKghtugLAwBOz4QfoWnW46lbR80UCglDdCQqBnT9iwgcZz59oVX6+9ViIDyLz1Flx1FTRsaAcMIiOJjrZd/oGBtqp57jzmHn3eO2iQHS+YOBG++y7j4SZNoHNnnUqam3Xr7K+rpL8DFCUNBEp5q04dANutkpwMt99uJ/o3bw7/+Y/tXP/yS/ttffr0ov/EELHThwYNgiuvtK911llA9tlFWbr9c+dy8evgwbBqFVSpAtdeC3/+mfGwTiU93fLl9lc+YYJ3C/v8nQ4WK+Wt7t3h6adJS0wkICTEjqyeOGGn2SxfDm+8kb18UBC8+abdcaawkpPtSud582xH/v/9n/3qn0XO2UUFql/fDiC3a2eXMcfEQHg4V19te8FGjbJTVb06Zhm0YwcMGJA5E/fUKQ+CbSmhZwRKeSu9W2XIEPtteuxYm9lt+XL7jXrfPvshnb4UOCUF+ve3ZwxPPAF79/r2ul98Yfts5s2zweeNN04LAj5r2RL++1+7c81//gMibNhgB4v37rU9YOvWFc1LlUaffpr5gR8ainscpuj2iyhpGgiU8oXLxW/9+5/+ddAYu6FAen6iwEB7PWKETe85ZoxdsXX55bYL6a+/bP9C1i6kpCSb4O6DD2xndN++cM45No/07t12YLd798xAU1R697aB6r33YOJEYmMzv/0mJ9v48NdfRfuS/k7Evk3pC+22brUnTFOm2JOn+fPhf/8r5IusWMFZb71Vov1M2jWklBPSO+tjY+2IbXrA+OUXmwDo3XdtJ/zw4fYDXcR+zWzQwJ5RpO8baYwNAmFhmeXS0uxxneiTGDXKfs2dOJHoCZcTEtKVpCRbtV9/tSc1b71l90go65KT7dsze7adHzB/PlSqZIdjXC44edLuMX3rrTZA1K/vw4vExsJVV9n1HAsWlFhfk54RKOUUl8ueAWT9x27c2N733Xd2w5v27e0Huwikptqzhocftt00W7bYsYfdu+3MpPQzjHynAxWSMfDqq3Dllbim92Tlc98zebLNbbR5s82/162bjRdJSc5UwR8cOWL3gJg9274dH35og0BW4eH25OnkSTv849Oez7NmAe6JB0lJ5D7f13kaCJQqKU2bwrRp9hMlMNBev/aava9/f5vfITzclvVqOlAhhYTAokXQoAGu0e0Z888oXMRxySU2o+ndd9v1aW3b2hjllXXrbPuKqhskLq7I13Ps3m2/6a9ebYdjnnzy9O090110kQ0Wa9Z4lur7NFn3uAgIyDfA5+xBLFIiUqourVq1El/FxMT4/Fx/oW3wH0XWjvXrRaZNs9fFLN82vPOOiD1XEQkPz1a/Dz8UqV5dpFIlkbff9vDF1q8XCQy0xwsOFlm7tlB1l5UrRYKDJS0g4LT6+WL9epHbbxepXFmkZk2R1as9f+7gwSLG2Cp5bONG+7sYOVJO1aolcs45ImlpuRZdt04kKMj++nxtKrBJ8vhcdfSMwBhzlTHmR2PMbmPM6Fwev9IYs8UYk2KM6eNkXZTyW7l1IfmDvXszp8ecOpWt26J3b/j2W3vSMnCgHbueMKGAb6tvvmm7v8B2wPfta7OleuvUKbtJUM+ekJyMSUsrVLeKiD0Ra9cO5s61+zzMnm3XC3jqpZfgwgvtiZzHA+qzZtn+pscftwv79uyxCwNz8cQTtuspNdWZHiTHAoExJhCYBXQHmgA3G2Oa5Cj2GzAYeMepeiilfJS+TDl9kPqcc7I93LChnUFz++12euXEiTY5a67BIC3N9rUYkznOkZpqP2379vVsSm1Skh2/OO88eOABuOgi4gLaMo3RxAVe4fW4yeHDdkZQ8+a2DekxKiDA+y6vihXh/fft2MLAgR5k/f77bzthYOBAqFKFAx072qAwZ85pRbdts7/fgADnhoicPCOIAnaLyB4RSQIWANdlLSAie0XkO8CzZOlKqeKTPi4xerQdxH7zTRsQsggMtPEhjxOHTG+/bWcjjR9vO9NjY+0MqokTYdky29n+6KMZ265l6w9PSeHEnPnsOrc7MXct4L8V7uDJO36i7zmbaMcaHmUq7VO+YPZWF4cP598kERuPBgyws3yHDbOxbtSozKEaXz9oIyLghRfg889hxowCCs+bZ5M53XsvAKkVKtjTiffes9HE7eRJuPlmqFEDlixxbojISI43tsgObLt6rhKR2923BwJtROS+XMq+AXwiIgvzONZQYChA3bp1Wy1YsMCnOiUkJFAp59B/KaNt8B9loR2etqHB++9z3uzZfD91KocuvzzbY9u3V+HBB5uTlBSAiOGuu3Zz003xGY8HnjxJ1MCBJNauzZZZszKjhlvowYOcM2cOdb/4gsSaNfmw+yRuffdOUlIDMAgV+JcTnF7H4OBUkpMDcM+5yVCjRiKNGv1Lo0YnaNToBI0bn+DQoRCWLavH3r0VOXgwjIoVU+jS5S969PiD885LyGjH1q3ViIw8QtOmxzz9FWYjAlOmXExsbB2ee24rERFHTy+UmkqbAQNIrFuXre59sBMSEjhj/35a33knPw0bxr7evQF44YXz+OijBsyY8R1RUQVEuQJ06NBhs4i0zqPizgzqAn2AuVluDwReyqPsG0AfT46rg8UxJV2FQisLbRApG+3wuA1JSSIXX2wHNE+ePO3h9etFJk2yD9erJ3L4cJYHx42zg6Jxcfm/RlycSFSU9OXdjDFqSJXLQjbLtP7b5M030uSLL0R27hQ5dsy+Zni4SKBJkTD+lWcGfStPPWUHblu3FqlQQbIcx14CAkTGjxc5ccLjX5HXjh4VOfdckdq1RR59NJeB3aVLbWXefz/jroz3oXVrkUsuEUlLk08+scVGjCiaepHPYLGTgcAFrMhyewwwJo+yGgg8pG3wH2WhHV614fPP7UfGlCl5Ftm82c5uGTjQfcfevSKhoSK33OLRS3yzOVVCAhIlgBQJJEnCOSHrh76eZ/n160VuH/KTrK/bS6Rjx2yPpaaK7NkjcuutdkYP2Fk306Z5VJVCef31PCdciVx1lciZZ9rg6pbxPsyZIwKyf8lGqVVLpHlzkVOniqZO+QUCJ8cINgLnG2MaG2NCgH7AEgdfTynlpM6d7T4GU6fCb7/lWqRlSxg3zg4JfPwxtvM9IMBOeynA4cNwfZ8AatcUloTcyGQzgZUhV+MafGGez3G5oP/AeFzDo2zep507Mx4LCLDr97Jm+3ByLV5Wf/yR2QN28iQsTO/03r3bjvwOHXraHhAA9OtHWsXKDL4rlBMn7OZAoaHO19exQCAiKcB9wApgJ/C+iGw3xkwyxvQEMMZcaoyJB24EXjXGbM/7iEqpEvfMM/b6oYfyLDJ2LERGwp1Dkjj0/hfwyCN2ilE+UlPtoOi+fbBoaSg9Yh9mzNRKuGKnezYyettt9oP1lVdOe6jI1+J5sLIrfcJVejCYM8cOIjN7ts1GO3Ro7k+sXJnnm73GZ/sjeHbKCZrknGfplLxOFfz1ol1DMSVdhUIrC20QKRvt8KkNEyfaPo98Vk9t3ZIqwSZJbg7/yKMO+TFj7CHnzPG+OhltuOUWkSpVRBISvD+Ip9avFwkJsX1NBazsSl8nuHCh7fYPDEyTF8MflrQb+55WNr0N33wjEhKcKtfxkaQ9/0KRVp2SWlCmlCqDHn4YGjWycy/z2L6s+dY3eUwm8u7JXiz6X4V8D/fhh/YL9h132IvP7rkHjh2z/SlOWbTIrmcQsXNlY2LyLJq+TvCGG2D9eugR8Rv3n3ySuxNn5vpr+/dfe1ZUq3YAc5u/hPm/OadN13WKBgKllHfCw2HmTJsn5+WXT3/8+HEYO5ZHomJp1Uq4+244eDD3Q+3YYTdaa9PGLu4qlMsvt6vDZs1y5gM0OdnuOQGZi+y+/96jp1auJHzE9YyuNZdXl5xJ165w6FD2MiNHwo8/2uyute69ya4kK6bU1BoIlFLeu+46m4v68cfhwIHsj02fDn/+SfCLz/LGG4ajRzPWTWVz9KhNVVGhgh1MLfSgqDH2rODbb535AJ00yQ5GT5liL3362NTRM2cW/NyvviJg6xamT07h7bdt9aKiMnPOrV1bi1dftSdbnTphTw3yWGnsiLz6jPz1omMEMSVdhUIrC20QKRvtKFQbdu60c0WHDMm8b88eO100Y/6oyPTptv//vfcyi6WmivTsaZ/uTXK33GRrw/Hjdpygf//CHTSnNWvsIoTBgzPvS0kRuf760xuXm/79bb2OHxcRu2Sibl2b4G7MGJHQ0BS58EKRxMQsz7nzTpGwsByLMnyHjhEopYrcRRfZndfmzYMNG+x9o0bZeZrTp2cUe+gh++33nnsyE7JNnWpTJjzzjHfJ3QpUqZLta/rgg9PPVHx15IjNSdGokc0hkS4w0O4b0batzRm0enXuzz9wwNZn8OCMTQ0uu8ym9K5Xz/6qEhMD+PVXu+dDhjvvtOMQ//1v0bQjHxoIlFK+Gz/efprdd5/NH7Rwoc1NlGW7rqAgu71yQoLNLzdoEDz2mP1svf9+B+p0zz12QPe114rmePfea+e1zp8PlStnfyw83Ea0c86xe1duz2UG/Ny5tj733JPt7oYN7YY2dsdRQ3JyjjxNLVpA69Y20Z7Dg8YaCJRSvqtSxe7csnEjXHUV1KmT6xqDiy+2GT7XrLGDocbAkCFFv+0yYM9UOna0awrSU4r6av58Owvp8cft1/jc1KhhF4mFh9vfQXxmniVSUmw9One2eapz6NrVLnYLCEjLfbHbnXfa4OLwoLEGAqVU4aSnH01MtN0oW7fmWqxevcyfAwLyTL1fNO65x65+Tp/l44u9e+1x2ra180Dzc/bZ9rWOHrV7XB51J5tbuhR+/z330XIyF7sNGbI398Vu/frZ7qRXX/W9HR7QQKCUKpysfeOpqXnumtKxY+FTPXusZ0+bZzq36a2eSEmxfVdg++iDggp+TmSkXWewc6edDpWYaKeyNmwI11yT59NcLujf/7fcVzxXqmTTU7//Pvzzj09N8YQGAqVU4aTnUyjgE744t10mONimcfj0U/j5Z++f/8QTdn/ll1+2g8Se6tLFDp7HxNgtz1autGcIngSSvBTDoLEGAqVU4XjxCV+su3LecYf9AJ4927vnff213Xfzllvst3FvDRxoM91t3Ghvv/lm4fr40weNZ86EadMcGS/QQKCUKjx/3Hf5zDNtF828eTYFqCeOH7cf/g0a2G4dXzVsmDkSftp0IB907mx3dBs/Pp/9QH2ngUApVXbdc4/tW3/vvYLLxsVBhw52E/m334Zq1Xx/3Q4dijb3dfqy67Q0R3avL0THlVJK+bn27aFJEzvFdf9+28Vyxhl2Jk/Wy7ZtdraTiO1OKkyfPmR2l8XG2iBQ2DOlbt1sG5KSHBlp10CglCq7jLFz+5991u6Yk1NgoF38lp5EDux1bGzhP7xdrqLrKivqwJKDBgKlVNlWuXLmB70xcNNNNoV2w4Z2cUNgoO0W6tTJsW/cRaIoA0sOOkaglCrbunXL7K8PC7NBwOWyA8KBgbZMsc5t9T96RqCUKts87VZx8Bu3v9NAoJQq+8rxh7wntGtIKaXKOUcDgTHmKmPMj8aY3caY0bk8HmqMec/9+NfGmEZO1kcppdTpHAsExphAYBbQHWgC3GyMaZKj2G3APyJyHjATmOFUfZRSSuXOyTOCKGC3iOwRkSRgAXBdjjLXAW+6f14IdDLGkQzlSiml8mDEoZ1vjDF9gKtE5Hb37YFAGxG5L0uZbe4y8e7bP7vL/J3jWEOBoQB169ZttWDBAp/qlJCQQCX3VnGllbbBf5SFdmgb/ENxtKFDhw6bRaR1bo+VillDIjIHmAPQunVrifZxsUdsbCy+PtdfaBv8R1loh7bBP5R0G5wMBPuAhlluN3Dfl1uZeGNMEFAVOJTfQTdv3vy3MeZXH+tUC/i7wFL+TdvgP8pCO7QN/qE42nB2Xg84GQg2AucbYxpjP/D7AbfkKLMEGATEAX2AVVJAX5WI1Pa1QsaYTXmdGpUW2gb/URbaoW3wDyXdBscCgYikGGPuA1YAgcA8EdlujJkEbBKRJcBrwNvGmN3AYWywUEopVYwcHSMQkeXA8hz3PZbl51PAjU7WQSmlVP7K28riOSVdgSKgbfAfZaEd2gb/UKJtcGz6qFJKqdKhvJ0RKKWUykEDgVJKlXPlJhAUlACvNDDG7DXGfG+M2WqM2VTS9fGEMWaeMeaAexV5+n01jDGfG2N+cl9XL8k6FiSPNkwwxuxzvxdbjTFXl2QdC2KMaWiMiTHG7DDGbDfGDHffX2rei3zaUGreC2NMmDFmgzHmW3cbJrrvb+xOvLnbnYgzpFjrVR7GCNwJ8HYBXYB47BqHm0VkR4lWzEvGmL1A65wpOPyZMeZKIAF4S0Qucd/3JHBYRJ5wB+XqIvJISdYzP3m0YQKQICJPl2TdPGWMqQfUE5EtxpjKwGagFzCYUvJe5NOGvpSS98KdS62iiCQYY4KBL4HhwEjgQxFZYIx5BfhWRGYXV73KyxmBJwnwlANEZA12jUhWWZMNvon9Z/ZbebShVBGRP0Rki/vn48BOoD6l6L3Ipw2lhlgJ7pvB7osAHbGJN6EE3ofyEgjqA79nuR1PKfsDchPgM2PMZncivtKqroj84f75T6BuSVamEO4zxnzn7jry2y6VnNz7frQAvqaUvhc52gCl6L0wxgQaY7YCB4DPgZ+BIyKS4i5S7J9P5SUQlBVXiEhL7B4P97q7LEo1d0qR0tg/ORs4F4gE/gCeKdHaeMgYUwlYBIwQkWNZHyst70UubShV74WIpIpIJDb/WhRwUcnWqPwEAk8S4Pk9Ednnvj4AfIT9IyqN/nL396b3+x4o4fp4TUT+cv9DpwH/Ryl4L9x90ouA+SLyofvuUvVe5NaG0vheAIjIESAGcAHV3Ik3oQQ+n8pLIMhIgOceje+HTXhXahhjKroHyDDGVAS6Atvyf5bfSk82iPv64xKsi0/SPzzdeuPn74V7kPI1YKeIPJvloVLzXuTVhtL0Xhhjahtjqrl/DsdOYNmJDQh93MWK/X0oF7OGANxTyp4jMwHe1JKtkXeMMedgzwLA5oh6pzS0wRjzLhCNTbP7F/A4sBh4HzgL+BXoKyJ+OxibRxuisV0RAuwF7szS1+53jDFXAGuB74E0991jsX3speK9yKcNN1NK3gtjTDPsYHAg9ov4+yIyyf3/vQCoAXwDDBCRxGKrV3kJBEoppXJXXrqGlFJK5UEDgVJKlXMaCJRSqpzTQKCUUuWcBgKllCrnNBAoVYyMMdHGmE9Kuh5KZaWBQCmlyjkNBErlwhgzwJ03fqsx5lV3orAEY8xMdx75lcaY2u6ykcaYr9xJzz5KT3pmjDnPGPOFO/f8FmPMue7DVzLGLDTG/GCMme9eMatUidFAoFQOxpiLgZuAtu7kYKlAf6AisElEmgKrsSuMAd4CHhGRZthVr+n3zwdmiUhz4HJsQjSwWTNHAE2Ac4C2DjdJqXwFFVxEqXKnE9AK2Oj+sh6OTcaWBrznLvNf4ENjTFWgmoisdt//JvCBOy9UfRH5CEBETgG4j7dBROLdt7cCjbAblChVIjQQKHU6A7wpImOy3WnM+BzlfM3PkjWHTCr6f6hKmHYNKXW6lUAfY0wdyNjX92zs/0t6hshbgC9F5CjwjzGmnfv+gcBq9w5a8caYXu5jhBpjKhRnI5TylH4TUSoHEdlhjHkUuxtcAJAM3AucAKLcjx3AjiOATRv8ivuDfg/wH/f9A4FXjTGT3Me4sRiboZTHNPuoUh4yxiSISKWSrodSRU27hpRSqpzTMwKllCrn9IxAKaXKOQ0ESilVzmkgUEqpck4DgVJKlXMaCJRSqpz7f6kSK9zlscUzAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 검증셋과 학습셋의 오차를 저장합니다.\n",
    "y_vloss = history.history['val_loss']\n",
    "y_loss = history.history['loss']\n",
    "\n",
    "# 그래프로 표현해 봅니다.\n",
    "x_len = np.arange(len(y_loss))\n",
    "plt.plot(x_len, y_vloss, marker='.', c=\"red\", label='Testset_loss')\n",
    "plt.plot(x_len, y_loss, marker='.', c=\"blue\", label='Trainset_loss')\n",
    "\n",
    "# 그래프에 그리드를 주고 레이블을 표시하겠습니다.\n",
    "plt.legend(loc='upper right')\n",
    "plt.grid()\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 전이 학습으로 모델 성능 극대화하기"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 실습: 전이 학습 실습하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 160 images belonging to 2 classes.\n",
      "Found 120 images belonging to 2 classes.\n",
      "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
      "58892288/58889256 [==============================] - 5s 0us/step\n",
      "58900480/58889256 [==============================] - 5s 0us/step\n",
      "Model: \"vgg16\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 150, 150, 3)]     0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
      "=================================================================\n",
      "Total params: 14,714,688\n",
      "Trainable params: 0\n",
      "Non-trainable params: 14,714,688\n",
      "_________________________________________________________________\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "vgg16 (Functional)           (None, 4, 4, 512)         14714688  \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 8192)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 64)                524352    \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 1)                 65        \n",
      "_________________________________________________________________\n",
      "activation_6 (Activation)    (None, 1)                 0         \n",
      "=================================================================\n",
      "Total params: 15,239,105\n",
      "Trainable params: 524,417\n",
      "Non-trainable params: 14,714,688\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras import Input, models, layers, optimizers, metrics\n",
    "from tensorflow.keras.layers import Dense, Flatten, Activation, Dropout\n",
    "from tensorflow.keras.applications import VGG16\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# 학습셋의 변형을 설정하는 부분입니다. \n",
    "train_datagen = ImageDataGenerator(rescale=1./255,          # 주어진 이미지의 크기를 설정합니다.\n",
    "                                  horizontal_flip=True,     # 수평 대칭 이미지를 50% 확률로 만들어 추가합니다.\n",
    "                                  width_shift_range=0.1,    # 전체 크기의 15% 범위에서 좌우로 이동합니다.\n",
    "                                  height_shift_range=0.1,   # 마찬가지로 위, 아래로 이동합니다.\n",
    "                                  #rotation_range=5,        # 정해진 각도만큼 회전시킵니다.\n",
    "                                  #shear_range=0.7,         # 좌표 하나를 고정시키고 나머지를 이동시킵니다.\n",
    "                                  #zoom_range=1.2,          # 확대 또는 축소시킵니다.\n",
    "                                  #vertical_flip=True,      # 수직 대칭 이미지를 만듭니다.\n",
    "                                  #fill_mode='nearest'      # 빈 공간을 채우는 방법입니다. nearest 옵션은 가장 비슷한 색으로 채우게 됩니다.\n",
    "                                  )      \n",
    "\n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "       './data/train',\n",
    "       target_size=(150, 150),\n",
    "       batch_size=5,\n",
    "       class_mode='binary')\n",
    "\n",
    "# 테스트셋의 정규화를 설정합니다.\n",
    "test_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "       './data/test',\n",
    "       target_size=(150, 150),\n",
    "       batch_size=5,\n",
    "       class_mode='binary')\n",
    "\n",
    "# VGG16 모델을 불러옵니다.\n",
    "transfer_model = VGG16(weights='imagenet', include_top=False, input_shape=(150, 150, 3))\n",
    "transfer_model.trainable = False\n",
    "transfer_model.summary()\n",
    "\n",
    "# 우리의 모델을 설정합니다.\n",
    "finetune_model = models.Sequential()\n",
    "finetune_model.add(transfer_model)\n",
    "finetune_model.add(Flatten())\n",
    "finetune_model.add(Dense(64))\n",
    "finetune_model.add(Activation('relu'))\n",
    "finetune_model.add(Dropout(0.5))\n",
    "finetune_model.add(Dense(1))\n",
    "finetune_model.add(Activation('sigmoid'))\n",
    "finetune_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "32/32 [==============================] - 7s 217ms/step - loss: 0.7590 - accuracy: 0.5688 - val_loss: 0.5409 - val_accuracy: 0.7800\n",
      "Epoch 2/20\n",
      "32/32 [==============================] - 7s 224ms/step - loss: 0.5509 - accuracy: 0.7125 - val_loss: 0.5420 - val_accuracy: 0.7600\n",
      "Epoch 3/20\n",
      "32/32 [==============================] - 7s 208ms/step - loss: 0.4324 - accuracy: 0.8188 - val_loss: 0.4058 - val_accuracy: 0.8200\n",
      "Epoch 4/20\n",
      "32/32 [==============================] - 7s 217ms/step - loss: 0.4192 - accuracy: 0.8313 - val_loss: 0.3407 - val_accuracy: 0.9200\n",
      "Epoch 5/20\n",
      "32/32 [==============================] - 7s 208ms/step - loss: 0.3139 - accuracy: 0.8813 - val_loss: 0.2570 - val_accuracy: 0.9600\n",
      "Epoch 6/20\n",
      "32/32 [==============================] - 7s 220ms/step - loss: 0.2795 - accuracy: 0.9125 - val_loss: 0.2508 - val_accuracy: 0.9200\n",
      "Epoch 7/20\n",
      "32/32 [==============================] - 7s 228ms/step - loss: 0.2519 - accuracy: 0.9125 - val_loss: 0.3089 - val_accuracy: 0.8600\n",
      "Epoch 8/20\n",
      "32/32 [==============================] - 7s 231ms/step - loss: 0.2198 - accuracy: 0.9375 - val_loss: 0.1758 - val_accuracy: 0.9800\n",
      "Epoch 9/20\n",
      "32/32 [==============================] - 7s 230ms/step - loss: 0.2399 - accuracy: 0.9312 - val_loss: 0.1661 - val_accuracy: 0.9400\n",
      "Epoch 10/20\n",
      "32/32 [==============================] - 7s 228ms/step - loss: 0.2002 - accuracy: 0.9375 - val_loss: 0.1326 - val_accuracy: 0.9800\n",
      "Epoch 11/20\n",
      "32/32 [==============================] - 8s 240ms/step - loss: 0.2483 - accuracy: 0.9125 - val_loss: 0.1493 - val_accuracy: 0.9400\n",
      "Epoch 12/20\n",
      "32/32 [==============================] - 8s 248ms/step - loss: 0.1724 - accuracy: 0.9438 - val_loss: 0.1812 - val_accuracy: 0.9400\n",
      "Epoch 13/20\n",
      "32/32 [==============================] - 8s 265ms/step - loss: 0.1952 - accuracy: 0.9500 - val_loss: 0.1734 - val_accuracy: 0.9200\n",
      "Epoch 14/20\n",
      "32/32 [==============================] - 8s 247ms/step - loss: 0.1618 - accuracy: 0.9625 - val_loss: 0.1516 - val_accuracy: 0.9400\n",
      "Epoch 15/20\n",
      "32/32 [==============================] - 10s 300ms/step - loss: 0.1693 - accuracy: 0.9563 - val_loss: 0.2223 - val_accuracy: 0.9200\n"
     ]
    }
   ],
   "source": [
    "# 모델의 실행 옵션을 설정합니다. \n",
    "finetune_model.compile(loss='binary_crossentropy', optimizer=optimizers.Adam(learning_rate=0.0002), metrics=['accuracy'])\n",
    "\n",
    "# 학습의 조기 중단을 설정합니다.\n",
    "early_stopping_callback = EarlyStopping(monitor='val_loss', patience=5)\n",
    "\n",
    "# 모델을 실행합니다.\n",
    "history = finetune_model.fit(\n",
    "       train_generator,\n",
    "       epochs=20,\n",
    "       validation_data=test_generator,\n",
    "       validation_steps=10, \n",
    "       callbacks=[early_stopping_callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEGCAYAAABo25JHAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAAA9oklEQVR4nO3dd3hUZfbA8e9JSKGJSJcuvVfBiEoQpayIoKCyoCAiq4uC6IoovUhTLCDCKiIWFMUG+5MVlKqABRCQolKVgNJcShASkpzfH+8EAiQhZSaTyZzP88yTmcm9d04myT1z33JeUVWMMcYErxB/B2CMMca/LBEYY0yQs0RgjDFBzhKBMcYEOUsExhgT5PL5O4DMKl68uFaqVClL+548eZKCBQt6NyAfCqR4AylWCKx4AylWCKx4AylWyF6869atO6yqJVL9pqoG1K1JkyaaVcuWLcvyvv4QSPEGUqyqgRVvIMWqGljxBlKsqtmLF1iraZxXrWnIGGOCnCUCY4wJcpYIjDEmyAVcZ7ExJvc5c+YMMTExnD59miJFirBt2zZ/h5QhgRQrZCzeyMhIypUrR1hYWIaPa4nAGJNtMTExFC5cmEqVKhEbG0vhwoX9HVKGnDhxImBihUvHq6ocOXKEmJgYKleunOHjWtOQMSbbTp8+TbFixRARf4cS1ESEYsWKcfr06UztFzSJYM0amDOnAmvW+DsSY/ImSwK5Q1Z+D0HRNLRmDbRqBXFxlZkzB5Ysgagof0dljDG5Q1BcESxfDvHxAEJ8vHtsjDHGCYpEEB0NEREAioh7bIzJO44cOULDhg1p2LAhpUuXpmzZsmcfx7tPgelavnw5q1evztJr79mzh3ffffeSx+/QoUOWjp8TgiIRREXB0qVQuvRpype3ZiFjcoU1a2D8eLzRcVesWDE2bNjAhg0bePDBBxk4cODZx+Hh4Zfc39eJILcLij4CcCf/O++MYcqUamzbBrVq+TsiY/KmiCefhK1b09/o2DHYtAmSkiAkBOrXhyJF0t6+YUN48cVMxbFu3Toee+wxYmNjKV68OLNnz6ZMmTJMmTKFGTNmkC9fPqpVq8bkyZOZMWMGoaGhvPPOO0ydOpU//viDUaNGERoaSpEiRVi5ciWJiYkMHjyY5cuXExcXR79+/fjHP/7B4MGD2bZtGw0bNqRnz54MHDgw3bj+/PNPevfuza5duyhQoACvvvoq9evXZ8WKFQwYMABwHb4rV64kNjaWu+66i+PHj5OQkMDkyZNp27Ztpt6HjAiaRABw/fWHmDq1GvPmwfDh/o7GmCB27JhLAuC+HjuWfiLIJFXlkUceYf78+ZQoUYL333+fIUOGMGvWLCZMmMDu3buJiIhg7969lC9fngcffJBChQrxr3/9C4B69eqxaNEiypYty9GjRwF4/fXXKVKkCN9//z1xcXG0aNGCNm3aMGHCBJ577jn+7//+L0OxjRgxgkaNGvHpp5+ydOlS7r33XjZs2MBzzz3HtGnTaNGiBbGxsURGRvLqq6/Stm1bhgwZQmJiIgcOHPDae5RSUCWC4sXjadECSwTG+FDcxImEX2qS1po10Lq1G8URHg5z5ni1zTYuLo7Nmzdz8803A5CYmEiZMmUAqF+/Pt27d6dTp060bt061f1btGhBr169uPPOO7n99tsBWLx4MZs2beLDDz8E4NixY2zfvj1DTU8pff3113z00UcA3HjjjRw5coTjx4/TokULHnvsMbp3787tt99OuXLluPrqq+nduzdnzpyhU6dOVKlSJUvvx6UERR9BSl27wubN8NNP/o7EmCAWFeXGcY8Z45Px3KpKnTp1zvYT/PjjjyxevBiAzz77jH79+rF+/Xqio6NJSEi4aP8ZM2YwduxY9u7dS5MmTThy5AiqytSpU88ec/fu3bRp08ZrMQ8ePJiZM2dy6tQpWrRowU8//cQNN9zAypUrKVu2LL169fJZX0TQJYI77nBfPUndGOMvUVHw1FM+Gb0RERHBoUOHWOPpiD5z5gxbtmwhKSmJvXv30qpVKyZOnMjx48fPlsQ4ceLE2f137txJ8+bNGT16NCVKlGDv3r20bduW6dOnc+bMGQB++eUXTp48edG+l3L99dczZ84cwHVSFy9enMsuu4ydO3dSr149nnzySa6++mp++uknfv31V0qVKsUDDzxAnz592LhxoxffpXOCqmkIoGxZuPZa1zw0dKi/ozHG+EJISAgffvgh/fv359ixYyQkJPDoo49SvXp1evTowbFjx1BVHnzwQS6//HJuvfVWunTpwvz585k6dSovvPAC27dvR1Vp3bo1DRo0oH79+uzZs4fGjRujqpQoUYJPP/2U+vXrExoaSoMGDejVq9clO4tHjhxJ7969qV+/PgUKFODNN98E4MUXX2TZsmWEhIRQp04d2rdvz9y5c3n22WcJCwujUKFCvPLKK755w9JasSa33ryxQtkLL6iC6s8/Z/lQOSKQVk8KpFhVAyveQIh169atZ+8fP37cj5FkTiDFqprxeFP+PpJhK5Sdr0sX99Wah4wxJgj7CADKlXPNkvPm+TsSY0xesmjRorMzmpNvnTt39ndYlxR0fQTJunSBxx+HHTugalV/R2OMyQvatm3rkwlfvhaUVwRwrnnIrgqMMcEuaBNBhQrQvLn1ExhjTNAmAnCTy9avh127/B2JMcb4j08TgYi0E5GfRWSHiAxO5fsviMgGz+0XETnqy3gulDy5zJqHjDHBzGeJQERCgWlAe6A20E1EaqfcRlUHqmpDVW0ITAU+9lU8qalUCa6+2hKBMYEuO+sRrF27lv79+3s1ntmzZ7N///50t4mOjmbt2rVefd2s8uWooWbADlXdBSAic4HbgLTq03YDRvgwnlR17QqDBsHu3VC5ck6/ujHBa80at1pgdHT2q0wkr0cAbuZuykqiAAkJCeTLl/rprmnTpjRt2jR7AVxg9uzZ1K1blyuvvNKrx/UVXyaCssDeFI9jgOapbSgiFYHKwNI0vt8X6AtQqlQplmdxrcnY2NiL9i1XLhK4hokTd3L33XtT3c9fUos3twqkWCGw4g2EWIsUKXK23s6gQeFs3nxxIbeUjh+HzZtDzy5HULduIpddlvb29eolMXFiXIZiiYuLIywsjO7duxMZGcnGjRu55ppruOOOO3jyySeJi4sjMjKS6dOnc9VVV7Fw4UKmTJnCvHnzGDduHDExMezZs4eYmBgeeughHnroIU6ePEnPnj3Zv38/iYmJDBo0iDvuuIMffviBp59+mpMnT3LFFVcwY8YMvvnmG9auXUu3bt3Inz8/X375Jfnz578ozsTERE6ePMmJEyeYN28ekydPRlVp27Yto0ePJjExkX79+vHDDz8gIvTo0YOHHnqISZMmMWvWLPLly0eNGjWYPXv2Rcc+ffp0pv5mcss8gruBD1U1MbVvquqrwKsATZs21egsrjW5fPlyUtt38mRYv74KM2b4psRrVqUVb24USLFCYMUbCLFu27aNwp7S0yLxhIamf2o5ceL85QhOnMhH0aJpbx8eDoULZ6zcc0REBBEREYSFhXHgwAG+/fZbQkNDOX78OKtXryZfvnx8+eWXPPPMM8yePZsCBQqQL18+ChcuTEREBDt37mTZsmWcOHGCGjVqMHDgQBYvXkyFChVYtGgR4EpQR0ZGMnjw4PPWPBg/fjyzZs3i9ddf57nnnkv3SiM0NJSCBQty4sQJRo4cybp16yhatCht2rRhyZIllC9fnoMHD7LVs8jP0aNHCQ0N5cUXXzy7nsLRo0fPvu8pRUZG0qhRowy9X+DbRLAPKJ/icTnPc6m5G+jnw1jS1bUrDB4Me/a4fgNjTNZNnBh3yZO2j5cjOKtr166EhoYC7uTds2dPtm/fjoicrSJ6oVtuueVsMilZsiQHDhygXr16PP744zz55JN06NCB66+/ns2bN6e55kFmfP/990RHR1OiRAkAunfvzsqVKxk2bBi7du3ikUce4ZZbbqFNmzacPHnyvPUUOnXqlLU35gK+HDX0PVBNRCqLSDjuZL/gwo1EpCZQFMj+wqVZ1LWr++pZK8IY42M+Xo7grIIFC569P2zYMFq1asXmzZv5z3/+w+nTp1PdJyIi4uz90NBQEhISqF69OuvXr6devXoMHTqU0aNHp7vmgTcULVqUjRs3Eh0dzYwZM+jTpw9w/noKV199darrKWSWzxKBqiYADwOLgG3AB6q6RURGi0jHFJveDcz1VMfzi6uugsaNbfSQMTnJh8sRpOrYsWOULVsWINV29fTs37+fAgUK0KNHD5544gnWr19PjRo1Ul3zAMjUGgXNmjVjxYoVHD58mMTERN577z1atmzJ4cOHSUpK4o477mDs2LGsX7/+ovUUjh07RmxsbKZ+ltT4tI9AVRcCCy94bvgFj0f6MoaM6tIFnn4afvvNzTo2xuQtgwYNomfPnowdO5ZbbrklU/v++OOPPPHEE4SEhBAWFsb06dMJDw9Pdc2DOnXq0KtXLx588EHy58/PmjVrUu0sTlamTBkmTJhAq1atUFVuueUWbrvtNjZu3Mh9991HkqczZfz48SQmJp63nkL//v25/PLLs/O2OGnVp86tN2+sR5Ca7dvdGgWTJ2f58F4XCHXokwVSrKqBFW8gxGrrEeQMW4/Ax6pWhYYNrfaQMSb4WCJIoWtXN5phb+6aTmCMCWCdO3e+aI2C5GGouUVumUeQK3TpAkOGuNFDjz7q72iMCSyqioj4O4xc55NPPsnR19MsjLuxK4IUqleH+vVt9JAxmRUZGcmRI0eydBIy3qOqHDlyhMjIyEztZ1cEF+jaFYYNg5gYt6SlMebSypUrR0xMDIcOHeL06dOZPhH5SyDFChmLNzIyknKZPHlZIrhAly4uEXz8MXi5IKExeVZYWBiVPVUbly9fnqnyBv4USLGC7+K1pqEL1KwJdeta85AxJnhYIkhF166wahVcopy4McbkCZYIUtG1K6ha7SFjTHCwRJCKWrWgdm2bXGaMCQ6WCNLQtSt89RX8/ru/IzHGGN+yRJCG5Oahj3N0FWVjjMl5lgjSUKeOayKy0UPGmLzOEkE6unSBlSvhwAF/R2KMMb5jiSAd1jxkjAkGlgjSUbcu1KhhzUPGmLzNEkE6RNxVwYoVcPCgv6MxxhjfsERwCV26QFIS5HAlWWOMyTGWCC6hfn2oVs2ah4wxeZclgktIbh5atgwOHfJ3NMYY432WCDKga1drHjLG5F0+TQQi0k5EfhaRHSIyOI1t7hSRrSKyRUTe9WU8WdWgAVSpYrWHjDF5k88SgYiEAtOA9kBtoJuI1L5gm2rAU0ALVa0DPOqreLIjuXlo6VI4fNjf0RhjjHf58oqgGbBDVXepajwwF7jtgm0eAKap6v8AVDXXDtLs2hUSE+HTT/0diTHGeJf4arFpEekCtFPVPp7H9wDNVfXhFNt8CvwCtABCgZGq+nkqx+oL9AUoVapUk7lz52YpptjYWAoVKpSlfVWhR4/mXHnlKZ59dlOWjpFZ2Yk3pwVSrBBY8QZSrBBY8QZSrJC9eFu1arVOVZum+k1V9ckN6ALMTPH4HuDlC7b5P+ATIAyoDOwFLk/vuE2aNNGsWrZsWZb3VVUdNEg1NFT18OFsHSbDshtvTgqkWFUDK95AilU1sOINpFhVsxcvsFbTOK/6smloH1A+xeNynudSigEWqOoZVd2Nuzqo5sOYsiW5eWj+fH9HYowx3uPLRPA9UE1EKotIOHA3sOCCbT4FogFEpDhQHdjlw5iypUkTqFTJJpcZY/IWnyUCVU0AHgYWAduAD1R1i4iMFpGOns0WAUdEZCuwDHhCVY/4KqbsSh499OWX8Oef/o7GGGO8w6fzCFR1oapWV9UqqvqM57nhqrrAc19V9TFVra2q9VQ1a73AOahLF0hIgAUXXtsYY0yAspnFmXT11VCxojUPGWPyDksEmSTirgq++AKOHvV3NMYYk32WCLKga1c4c8ZGDxlj8gZLBFnQrBmUL2+1h4wxeYMlgixIbh5avBiOHfN3NMYYkz2WCLKoa1eIj7fRQ8aYwGeJIIuaN4dy5Wz0kDEm8FkiyKKQELjjDli0CI4f93c0xhiTdZYIsiG5eeiBB2DNGn9HY4wxWWOJIJtE4IMPoHVrSwbGmMBkiSAbVq48dz8uDpYv91soxhiTZcGTCNasocKcOd772J6YSPS18USGJQKupvf113vn0MYYk5Py+TuAHLFmDURHUzk+HmbNgqZNoVAhVz3uzBn3Na37aX1flShgCdcwlYd5T7uzecEurrvuKn//tMYYkynBkQiWL4czZxCApCTYv98tLJAvHxQsCGFh7n6+fJe+n/K5VauI+uILrtFv2E9Zhk5rxl1PQdGi/v1xjTEmM4IjEURHQ2QkSXFxhEREuN7dqKjsH3fNGvjqK+TUKV6SgTQ+vZ4RI2DKlOwf2hhjckpw9BFERcGSJezp3RuWLPFOEkhxXDp3poFuoG+733jlFdi82TuHN8aYnBAciQAgKorfunf3XhJIcVzefx8qV2bMvvspXFh59FFQ9e7LGGOMrwRPIvClsDAYOpTiG5cw+s7NLFliJaqNMYHDEoG33HMPVKnCQ9/1pk4d5bHH4PRpfwdljDGXZonAW8LCYNgw8m1Yy0t3rmL3bnj+eX8HZYwxl2aJwJu6d4dq1Wj98cN07qSMGwf79vk7KGOMSZ8lAm/Klw+GD4eNG5l88+ckJMCTT/o7KGOMSZ8lAm/r1g1q1KDy9EH863FlzhxYvdrfQRljTNp8mghEpJ2I/CwiO0RkcCrf7yUih0Rkg+fWx5fx5IjQUBgxAjZv5qkaH1O2LPTv7yY0G2NMbuSzRCAiocA0oD1QG+gmIrVT2fR9VW3ouc30VTw56s47oVYtCk4czqQJSaxbB2+84e+gjDEmdb68ImgG7FDVXaoaD8wFbvPh6+UeoaEwciRs3Uq30A9o0QKeftoWujfG5E6iPpoCKyJdgHaq2sfz+B6guao+nGKbXsB44BDwCzBQVfemcqy+QF+AUqVKNZk7d26WYoqNjaVQoUJZ2jfTkpJo2qcPkpjIO4Pn8VC/q+nSJYZ//nNnhg+Ro/FmUyDFCoEVbyDFCoEVbyDFCtmLt1WrVutUtWmq31RVn9yALsDMFI/vAV6+YJtiQITn/j+ApZc6bpMmTTSrli1bluV9s2TePFVQfecdvf9+1Xz5VH/6KeO753i82RBIsaoGVryBFKtqYMUbSLGqZi9eYK2mcV71ZdPQPqB8isflPM+lTEJHVDXO83Am0MSH8eS822+H+vVh1CjGjU6gQAEYONDfQRljzPl8mQi+B6qJSGURCQfuBhak3EBEyqR42BHY5sN4cl5IiOsr2L6dkl++y4gR8N//wmef+TswY4w5x2eJQFUTgIeBRbgT/AequkVERotIR89m/UVki4hsBPoDvXwVj9906gSNGsHo0Tz8YAI1arirgvh4fwdmjDGOT+cRqOpCVa2uqlVU9RnPc8NVdYHn/lOqWkdVG6hqK1X9yZfx+IWIuyrYuZPw99/mxRdh+3Z46SV/B2aMMY7NLM4Jt94KTZrAmDG0a32GDh1gzBj44w9/B2aMMZYIcoYIjBoFu3fDm2/y/POuRPVTT/k7MGOMsUSQc/72N2jWDMaOpVrFeAYOhNmz4bvv/B2YMSbYWSLIKclXBb/+CrNnM3QolC5tdYiMMf5niSAntW0L11wDY8dSODyOCRPg22/hnXf8HZgxJphlKBGIyAARuUyc10VkvYi08XVweY4IjB4Ne/fCrFncc49rLRo8GE6c8HdwxphgldErgt6qehxoAxTFlYuY4LOo8rKbboIWLeCZZwiJP82UKfD77zBunL8DM8YEq4wmAvF8/RvwtqpuSfGcyYzkvoJ9+2DmTJo3h3vvdesb79jh7+CMMcEoo4lgnYgsxiWCRSJSGLAuzqy68Ua44QYYPx5OnWLCBAgPh8cf93dgxphglNFEcD8wGLhaVf8CwoD7fBZVXpd8VbB/P7z6KmXKwNChsGABLF7s7+CMMcEmo4kgCvhZVY+KSA9gKGDLrGRHdDS0agUTJsBff/Hoo1C1Kjz6KJw54+fYjDFBJaOJYDrwl4g0AB4HdgJv+SyqYDFqlKszMWMGERGun2DbNpg2zd+BGWOCSUYTQYJnYYPbcIvLTAMK+y6sIHH99dC6NUycCCdP0qGDm2owciQcOuTv4IwxwSKjieCEiDyFGzb6mYiE4PoJTHaNGgUHD8IrryACL74IJ09Cnz4wZ04F1qzxd4DGmLwuo4ngLiAON5/gD9xqY8/6LKpg0qIFtGkDkyZBbCw1a8Idd7iO49dfr8yNN8KqVf4O0hiTl2UoEXhO/nOAIiLSATitqtZH4C2jRsHhw2c7B6pXd0+rCqdPnxttOmCAK1S3aZN1KBtjvCejJSbuBL4DugJ3At+KSBdfBhZUrrkG2rd3VwUnTtC+PeTPDyEhSliYW84gMRFefx3uuw8aNIDChaFpU+jbF6ZPh2++gb/+8vcPYowJRPkyuN0Q3ByCgwAiUgL4EvjQV4EFnZEjoXlzmDqVqKefZskSmDVrN717X0VUlNskMdGtbvbDD+62fj189BG89pr7fkgI1KzpVsZs3Nh9bdgQihY99zJr1sDy5W70avJxjTHBLaOJICQ5CXgcwSqXelezZtChAzz3HPTrR1RUEeLifiMq6qqzm4SGuhN9zZrQrZt7TtXVsFu//lxyWL4c5sw5d+hKlVxiKFYM3noLEhLcTOYlSywZGGMyngg+F5FFwHuex3cBC30TUhAbOdK190yZAsOGZWgXEahQwd06dTr3/MGD5185/PDD+bWM4uNdwrBEYIzJUCJQ1SdE5A6gheepV1X1E9+FFaSaNIHbbnMzyx55JFuHKlnSzUlo2/bcc19+6R4nJUFYmGseMsaYDDfvqOpHqvqY52ZJwFdGjoSjR92EAi+76Sb48EPXLFS3ruujNsaYdBOBiJwQkeOp3E6IyPFLHVxE2onIzyKyQ0QGp7PdHSKiItI0Kz9EntKwIdx+Ozz7LJVmzcLbM8o6d3Y5Zu1aeOMNrx7aGBOg0k0EqlpYVS9L5VZYVS9Lb18RCQWmAe2B2kA3EamdynaFgQHAt1n/MfKYTp3gr7+o+PbbrgSFl5PBP/4BLVvCY4+5ZRGMMcHNlyN/mgE7VHWXqsYDc3G1ii40BpgInPZhLIElJgZE3Mo/p065xn0vCgmBmTNdh/FDD7mRR8aY4OXLRFAW2JvicYznubNEpDFQXlU/82EcgSc6GiIjUfEsAvef/7gCRF5UtSqMHesOPXeuVw9tjAkwoj76OOiZedxOVft4Ht8DNFfVhz2PQ4ClQC9V3SMiy4F/qeraVI7VF+gLUKpUqSZzs3jmio2NpVChQlnaN6ddtmUL+b/7jvyJiVR87z2O167Nj+PHk+DF+BMT4ZFHGrNvX35mz/6OokWzXrcikN5bCKx4AylWCKx4AylWyF68rVq1WqeqqffDqqpPbrjFbBalePwU8FSKx0WAw8Aez+00sB9omt5xmzRpolm1bNmyLO/rD2fj/eAD1bAw1UaNVA8e9OprbN6sGh6uetdd2TtOwL63ASCQYlUNrHgDKVbV7MULrNU0zqu+bBr6HqgmIpVFJBy4G1iQIgEdU9XiqlpJVSsB3wAdNZUrgqDXtSvMn+9WrbnhBq/28NapA8OHw/vvw6efeu2wxpgA4rNEoKoJwMPAImAb8IGqbhGR0SLS0Vevm2e1bw+ff+6SwPXXw65dXjv0oEFu1OpDD8H//ue1wxpjAoRP6wWp6kJVra6qVVT1Gc9zw1V1QSrbRtvVwCW0bOkKBB07BtddB1u3euWwYWEwa5ZbFe2xx7xySGNMALHCcYHm6qthxQo35rNlS1dIyAsaNYInn3TrHXz+uVcOaYwJEJYIAlHduvDVV1CwILRqBV9/7ZXDDhsGtWq5NQ6OX3LeuDEmr7BEEKiqVnXJoHRpt9Tl4sXZPmRkpFv8JiYGBqdZEMQYk9dYIghk5cvDypVubctbb4VPsl8LMCoKHn3UrXq2YkX2QzTG5H6WCAJdqVKwbJlbeaZrV3j77WwfcuxYqFIF7r/flr80JhhYIsgLihaFL75wncf33us+zmdDgQKuFtHOnW6OgTEmb7NEkFcUKgSffeaaiP75T5g4MVuHi46GBx+EF16Ab77xTojGmNzJEkFeEhnpVrPv1s319g4Zkq3SohMnQtmy0Ls3xMV5MU5jTK5iiSCvCQtz/QQPPADjxkH//m5tyiy47DL4979dZYuxY70cpzEm17BEkBeFhroz+OOPw8svu17fhIQsHap9e+jZE8aPhw0bvBumMSZ3sESQV4nAs8/CqFFuunC3bm4lmix4/nkoXtw1EZ3JeqVqY0wuZYkgLxNxw36ef96tWn/bbW6o6fjxmVr+8oor3ECkH35wucUYk7fk83cAJgcMHAiFC7t+g8WLXYIID3cF7KKiMnSIzp3dNIVRo9z9WrV8HLMxJsfYFUGw6NMH7r7bdRwnJrpmouXLM3WIqVNdPund2x3CGJM3WCIIJv37u1FF4K4KoqMztXupUvDSS25ewZQp3g/PGOMflgiCSVSUuwpo2tSNIvrtt0wf4u9/hw4d3BSFnTu9H6IxJudZIgg2114Lq1a5hW169YJ16zK1uwjMmOEuLPr0yfIUBWNMLmKJIBiFh7sZyCVLupFEv/+eqd3LloXJk93FxWuv+SZEY0zOsUQQrEqWhPnz3SLFnTvD6dOZ2v3++6F1a3jiCdi710cxGmNyhCWCYNawoStH8e23blmyTNQlEnFXA4mJ8I9/ZKukkTHGzywRBLvbb3eTA95+27X3ZELlyjBhAvz3vzBqVO3MzFEzxuQilgiMW6y4a1cYNAgWLszUro0bQ0gIrFhRghtucEkh11uzhgpz5mRqdrUxeZklAuPaeWbPdk1F3bq5cqMZtHKl2x2EhATo1AmeegoOH/ZNqNm2ahVER1N51izXyWHJwBjfJgIRaSciP4vIDhG5aDl0EXlQRH4UkQ0i8rWI1PZlPCYdBQq4zuP8+d3iNn/+maHdoqPdIKSQkCQiI+GGG9w6BpUq5cKEcPKk6wuJj0eSkrI0u9qYvMhniUBEQoFpQHugNtAtlRP9u6paT1UbApOA530Vj8mA8uXhk0/cMKA778xQqdGoKFeyqHfvPSxd6lbM3LwZOnY8lxAGD4ZDh3wffrr27YPrr3dXOyGeP/t8+TI9u9qYvMiXVwTNgB2quktV44G5wG0pN1DV4ykeFgRs7Im/RUXBq6+6s/tjj2V4l+7dfztbv652bXj3XdiyxU1TmDTJdSz7LSGsWwfNmsH27fCf/8AXX5AQGQl162a46J4xeZmoj8b9iUgXoJ2q9vE8vgdorqoPX7BdP+AxIBy4UVW3p3KsvkBfgFKlSjWZO3dulmKKjY2lUKFCWdrXH/wZb5Xp0yn/wQf8/Nhj/H7rrZfcPr1Yf/21AG+/XZGlS0sSEZFEp077uOuuvVx+ue8XNyj+1VfUGjeOM0WK8OMzz3CyShUASrz7LnVee40NkydztHFjn8eRHfZ36zuBFCtkL95WrVqtU9WmqX5TVX1yA7oAM1M8vgd4OZ3t/w68eanjNmnSRLNq2bJlWd7XH/wab0KCart2qvnyqS5ffsnNMxLrtm2q3burhoSoFiigOmiQ6sGDXog1NUlJqhMmqIJq8+aqf/xx3rdXLFqkWq6calSU2zYXs79b3wmkWFWzFy+wVtM4r/qyaWgfUD7F43Ke59IyF+jkw3hMZoSGwnvvQZUqcMcdsHt3tg9Zsya8845rMurcGZ57zvUhDBoEBw9mP+Sz4uNdrezBg13p7WXLXOnUFJLCw92w2TVrMj1k1pi8xpeJ4HugmohUFpFw4G5gQcoNRKRaioe3ABc1Cxk/uvxy16aemOga+0+c8MphkxPC1q1uPtvkya4PwSsJ4fBhuPlmNxx2xAjXWZE/f+rb3ncfXHWVSwhWPc8EMZ8lAlVNAB4GFgHbgA9UdYuIjBaRjp7NHhaRLSKyAddP0NNX8ZgsqlYNPvjAnbXvucerJ8waNdyE5m3b3EVHckJ44oksJoSffoJrrnElM+bMgZEjkyc5pC4szCWLH36Ajz/O6o9hTMDz6TwCVV2oqtVVtYqqPuN5briqLvDcH6CqdVS1oaq2UtUtvozHZNHNN7t1j+fPd2sge1n16vDWW+cSwvPPuyajv/8dnn46g3O+kpfdPHHCNQX9/e8Ze/Hu3d0lyvDhtuyaCVo2s9hkzCOPuAUInnkGsjhq61JSJoToaNdFMX68G/4/ZUo6FyP//je0bQvlyrmrgcwMCQ0NhdGj3Yu++643fgxjAo4lApMxIjBtmjsr33cfrF3rs5eqXt29TPK8r8REGDDAPT9xYopmo8REGDgQHnwQ2rRx5SMqVcr8C95xhyuvMXJkhibRGZPXWCIwGZe8oE2pUq6oUCYXtMmM6GiIiHAf2PPnd+focuXcQKBy5eDO28+wpMVwkl58yWWJBQvgssuy9mIhITBmDOzaBW+84c0fw5iAYInAZE6JEu6ke/SoSwaZXNAmo5JLV4wZ476OGOHKAm3bBo/0PM6SBSe56dtnqFHiTyZd+SIH/8yXvRe85RbX0TxmjM9+JmNyK0sEJvPq13fDfb77LtML2mRGVJQrXJeyyb/msW+Z/J/q7CtQnTlPbuLKWpfz5JPuKuGuu1zSyNLAJhEYOxZiYlyfgzFBxBKByZrOnV0n69tvu5lhOeH9912bUcGCRH6znL9PqM+KFe4q4eGH4csv4aab3LDUSZOyMAS1dWto1QrGjXOVSo0JEpYITNYNHeqqlA4aRM1nnoGvv/bN66i6pHP33dC0qRsZVPtcIduaNd2Q03373ES1MmU47yph6dJMXCWMHesyyNSpvvlZjMmFstmwaoKaiBuxM28epb/80n0kL1rUnYFLl3adyhfekp8vXtyVgb6U06fh/vvd0M5773WVUSMiUt00MtJNC+je3V0lvPoqvPmmmw9XtaprxerVC3bsgDlzKhARkcpI02uvhb/9zV1SPPQQFCmS7bfJmNzOEoHJnm++caNuEhNdYqha1X0kP3DAlX0+cABOnbp4PxGXDC5MEClve/e6T+i//eaaawYPTn+mcAq1asELL7h5CB9+6JLCoEGuzwEgKakyc+acm4d2njFjoEkTd5kxalT23p8gs2ZNOknW5FqWCEz2eJYoS4qLIyQiAl566fwzgCrExsIff7ikkPKW8rnVq93Xv/66+DXCw93rZDAJpBQZCT16uNvWrfDAA+6lQDh1CmbMcIOFzjt048ZubsELL7iJdMWLZ/p1g9Hq1XDjjXDmTDpJ1uRKlghM9njGee6ZNYureve++D9fBAoXdrdq1VI/RkrJSeO55+C111zjfmKiGzuazbNK7drusK1bw+nTCghvvQXr17s+hbvucuWHAHcl8PHHrolo0qRsvW4wOHzYNb3FxQHI2VVALREEBussNtkXFcVv3bt757++UCHXvNSz57kZZclXBF6QPD/h/vt3s2KF60NQdfX0qlZ1pSxOngTq1HGdDS+/7NOJc5eS3NSSoXpLfvL551Cvnqv5FxoKoKjCddf5OzKTUZYITO504YwyL360TF5a8/rrXf/zpk2u2nb58m6ScsWK7oLgSP9Rbm2DceO89toZlZjo+jVuuAFmzqzMjTdmsPheDvrrL+jXD9q3d61n69bBV19BVNQRkpLcyGIfTTExXmaJwOReqc0o84GQEOjQwY1+/fprN3Bo5EioEH0Vj9b6nN9mLIRff/VpDABHjrjBUT16uL7yf/wDEhIAhNOn3bIJqXWh+MP330OjRvDKK25p6++/hwYN3K9q3LjNDBniWvZ8UKzW+IAlAmNSaNHCVdD48Ufo0gWm/dSaKgk/ce9N+9i82buvpQobNrgLjhYtoGRJ1xq1eLGreDF6tKuzFBKihIa6C6Natdy8On990k5IcBdp117rktKSJW4dicjI87cbM8YVqx071rWumdzNEoExqahb1/Uf7NwpPNxwFR/taEC9enDrrdmbN3fiBHzyiRu9VK6c+1Q9ZIhrgRo61M2V++MP99rDhrkTbe/eu/nqK9f5WrSom1d3ww2uKSYn7djhqsIOH+7mEW7a5EYJpUYEpk93C9v17+/mcpjcyxKBMemoUAFe+G9Nfouswah6H7JmjTsZXned61e41IxlVfj5Zzcl4aaboFgxtzznBx+4q4A33nAn/u+/d/0SzZqdK78N5/ozoqKgZUt38n/1VXfMq692c+3++MO374Gqa+Zp2NB1CL/3nlsArmjR9PfLl89t26KFa+768kvfxmmyzhKBMZdSujTFBvRg+OY7+XXhFqZMcbXpOnZ0o2XefNN1ko4f7zp0T51yI2keecSNRKpZEx5/3J2wBw50n+wPH3bJoFcv1x+QUaGh7mpi+3bXNv/2226dhkmTkoduetfBg67IbN++br7Fjz+6K5KMyp/fNbXVqOHKU61f7/0Yg8brr1PhnXd8M2pAVQPq1qRJE82qZcuWZXlffwikeAMpVtUsxHv4sOpll6l27qyqqvHxqm+/rVq3riqoirhbSIhqeLh7Ln9+1Q4dVF95RXXPHt/E+vPP7jVAtUoV1U8/VU1KyvprpbRggWqJEqoREaovvqiamJj1ePftU61YUbVkSdXt270TnzcEzN/t+PGqoEki7g9r9epMHwJYq2mcV+2KwJiMKFbMfQT/5BNYt46wMNfcsWmTm/LgTsWuqahJE/jvf90ooP/8x5UsqljRN2FVr+5e4/PP3XSLTp3cEtPZ6diOjXVXAB07QtmyrjlqwIDzm6wy68orYdEiNyy2TRvfN2flKZ995jqSAFHl7Gw9L7JEYExGDRwIV1zhenU9RNwwz/z5z62mNnkytGvn7ueUtm1h40Y3IW79ejeUs18/l4wyY80a1xcwc6Yr7fTtt25unTfUqAELF7rmpnbt4Ngx7xw3T1u0yHUqVasG+fOTFBLi1QmWySwRGJNRl13malF8/vl5Q4d8OPctU8LCXL/E9u3uKuTf/3bnjylTLr0U85kzbpTSdde5IaIrVrg+j/Bw78bYrJlb7XTLFp8ucJc3LF3q3qRatVwhpyVL2NO7t0/+yHyaCESknYj8LCI7RGRwKt9/TES2isgmEVkiIj66gDbGS/r1c727Q4eeN5g/h+a+ZUixYm7s/oYNrplqwAB3hfD556lv//PPbl7A2LGu1MamTW5klK+0bQuzZ7vWjR49XHORucBXX7mxylWqwBdfuCtRb5ZyuYDPEoGIhALTgPZAbaCbiNS+YLMfgKaqWh/4ELDqXiZ3K1jQtdeuWOE+meVideu6yWnz57tP/O3buxnUv/zivq8K06a5uQy7d7ty3bNnuwsfX+ve3RV3/egjt7qclaJIYc0atyZG+fLub6xECZ+/pC+vCJoBO1R1l6rGA3OB21JuoKrLVDV50vw3QDkfxmOMd/Tt6/5JhwzJ9WcwEdfpu3mzG2K6cqVr87/rLtdm//DDbn7Cjz+6yts56dFHXUvbjBluFrUB1q51HSilS7umocyMLc4GXyaCssDeFI9jPM+l5X7gvz6MxxjviIhw02u/+w7+7//8HU2GRETAE0+4/oN27dwchu3bXb/CsGFuLSF/GD/ezaUYOdIlhJyU6yq7/vCDG/JVrJhLAldemWMvnSvWIxCRHkBToGUa3+8L9AUoVaoUy7M4dCo2NjbL+/pDIMUbSLFC9uOVSpVoduWVJD76KGsLFsze2MpL8PZ7W7p0BUQqoyokJibxxht7iI//zWvHz2y8PXoIP/9ch3/+sxgHDmyhZcvDXoslLd98cwXDhtUlIaEyb72VyPPPb6ROneM+f920FNy1i4YDB5IYGcmGceM4vXMn7Nx50XY++z9La4JBdm9AFLAoxeOngKdS2e4mYBtQMiPHtQlluVMgxarqpXjfecdNH3j//ewfKx3efm9Xr3ZzkkJDszw3KV1ZiffkSdVrr3WT8Xz1p/T7725yX6tWbvLfudkfqtddp7p/v29e95K2bnUz7a688pKz7bLzt4CfJpR9D1QTkcoiEg7cDSxIuYGINAL+DXRU1YM+jMUY77v7brfs2fDhyfWiA0JuGe6aUoECbmJc1aquUN2GDd457u+/uw7x6GjX0vLPf7rnevVyzWXJlV1XrYLKlV2fyW/euzi6tF9+cZX7RFxzUNWqOfji5/gsEahqAvAwsAj3if8DVd0iIqNFpKNns2eBQsA8EdkgIgvSOJwxuU9oqDub/vyzq8IWQHLTcNdkV1zh5k8VKeL6MXbtytpx9u+HqVNdhdayZd3J/dAhl69//NGtXT1rFixbdq6y6y+/uKGzr77qRmz26eOqrfrUrl0uCSQmuiRQo4aPXzBtPu0jUNWFwMILnhue4v5Nvnx9Y3yuc2e32P3IkdCtm/dnYAWZcuVcMrjuOleKYtUqKLVztSvt0aaN60xNRUyMG4o6b56be6Xqhs+OGAFdu7oLtwtFRUFc3G9ERV0FuAqrw4bBs8+6+2+84X6lTz+d+v7Z8uuvLgmcOuUyktdfIHNsZrEx2SHiZmLt2eOK9OeaISiBq1YtV15n/37lb432c7xFe3juOZcIChZ041/btmXv3U/wYrvPaVHtAOXLu+GoJ44lMmoUbNvmPv2PGJG5c2yFCu5qYvduV1rq009dQunSxQ3q8YqYGJcEjh1zk8Xq1/fSgbMuV4waMiagFSniRg3Nn++qzb3yirtSuOIKf0cWmE6e5JrvXufDwuvp+PtMWvMlHZnPTSzlypql+ehMR+atasI3J+sB0JAfGMsUujKP6pu3w94i8EF5N9ejXDn3tfwFjwsUgDVrqDBnjussuKCNrEwZd2Xw5JPw0kuuTMdHH7mV44YOdSW5s+T3310SOHTILdDQuHE23yzvsERgTHatWHHufny8a2Du08clgqpVz92qVTt3v1gxdzVhzjl0yH0cnzYN/vyTv113HU/X+44xS65lLU0YySiS1ocCbjb0uK7QpWM81QoWhb3tIKYu7N177hYT40qnHkxlHErhwhAbS2VVeOstNx8klWan4sVdN9Djj7uwXnjB5YzWrV1CaNkyE7/Ggwfdjvv3uynfzZpl6u1JnveQSt7KNksExmRXdLT7VBkf72ZojRrllufascPN2lq9GubOPX85syJFLk4OyY9LlDh3dknnU2uesWuXK9k6a5arQtepk5v9du215B8PslRRDSEJpW1bd0KuUiV553CgElSqlPbxT592J9+USWL+fPj2WwTc761dO5cIOnd2w5ZKlz7vEJdf7iaSDxjgivk9+yy0auVWXxs61NVPSjchHD7slqjbs8ddNV577SXfFlVXPTYmxl08PP00JCRUZs4c74/2skRgTHYlj8dcvtwlhdT+Q+Pi3Elg+3aXIJKTxHffuWm+KZNE4cIuKRQtCitXUjkxEZ/89/vb+vWu7sW8eW4E1r33wr/+5ZZ084iOhshIIT4ewsOFESNSJoEMioyEq65yt5QHbt2apLg4QsLCXH2N776DBx90pVujolxS6Nz5vBcsVMhdHfzzny5vTZzoajg1aeISQseOqcwt/N//XJLZvt1debRsSWIiHDjgTvIpb/v2nX//4lXn5OxyBJYIjMltoqLS/8+MiHDDA1MbIhgf70aRpEwSO3a4xQASEtyn1lOnXIfp3LnuqiNQqbqPt5Mmua+XXeZO/gMGpFpSISM5Nks8B94zaxZX9e7tHqu6+tiffOJuTzzhbvXqnUsKDRqACPnzu0K0DzzglgsdN859u25dN0opNtblncvD/yJmxFvs238fMdfeSczQ0sT0cl0FF1ZdDQ93XRhly0Lz5u5+8uP//c+VGI+PTyI8PMTbyxHYUpW5WSDFG0ixqgZIvJ4pwEnJ62CCW+/xlVdUT53yd3RpSvW9PXNG9b33VBs1cj9HmTKqkyapHj2a4/GllO7fwe7dqi+8oHrDDW4NUlCtVEl14EDVlStVExLObnrmjFu6tGLF82csJ98KRp7RGjVUW7dW7dlTdcgQ1enT3XKg69erHjx46SVGV69W7dNnZ5ZngpPOzGK/n9gze7NEkDsFUqyqARTv6tW6s08f1VWrVBcuVI2Kcv+2V17pTlInT/o7wouc996ePKk6dao7gYJqzZqqr7+uevq03+JLKcN/BwcOqM6cqXrLLecWpS5ZUrVPH9XPPjv78zzzzLmcEUKCDpAX9ehb8722jnQglpgwxmRX8mIk117rGqNXrXJtJdWru6UzK1VyDdUnTvg70vMdPuw6zStUcG0aZcq4QflbtkDv3q6pLJCULAn33+/a+A8fhvffd8NA33/fjSktUQLuvptWiV8SkS+RUBKIII67RtWhyD0dc/0AMUsExgQSEXcCWrbMrWLVuLFbXLhSJTfO8ehR/8b34Yc06tfPNW6PHOkS2Ndfu5FTt93m0yqtOaZwYTd58L333JDXhQvdAg9LlxI1/GaWxF/HGIaxJF87om4q6O9oMyQP/FaMCVLXXefWn/zuO3d/+HCoWNENXzns+1LOgBsJ9dZbbt5EuXLQtSuXbd3qivDNmQMLFrgxlnlVRIS7UnvtNdcD3LcvUfItTzGBKF3terkDgCUCYwLd1Ve7cfEbNrgB7ePGuYTwr3/BH39473VUXYG9115zFdoqVnQlO3v2hI8/diOARDjbCvLrr9577UAQGurKmkZGuvvh4Xh/eI9vWCIwJq9o0MDNSdiyBW6/3U2DrVwZ+vd3k6gyKynJrWT/8suuKaRMGTfGv29fVyOneXM3E3jTJncF8vrrEBlJUkhIQJ0EvSo31vjOAJtHYExeU6uWG9w+YgRMmADTp7t1IO+7z/UnVK6c+n4JCa6y2sqV7vbVV24AO7j6PDff7Go7t2zpZkBf2AOa2tj8YHSpOSW5kCUCY/KqqlVh5kxXW3nSJPeJ/fXXoUcPV1Jh+3Y32uXPP92Jf9UqNxMK3In+9tvdSf+GG1wzUEZERfFbXBxXBdiJMNhZIjAmr6tY0RXoGTLEzU6eNg3efPP8berWdSUeWraE66/332r2xi8sERgTLK68Ep5/HvLnh/HjXedvSIirZjZmjL+jM35kncXGBJsOHc6NbImIgL/9zd8RGT+zKwJjgo3PKrmZQGWJwJhgFIAjW4zvWNOQMcYEOUsExhgT5CwRGGNMkLNEYIwxQc4SgTHGBDlLBMYYE+TErWAWOETkEJDV+rbFgRwq1O4VgRRvIMUKgRVvIMUKgRVvIMUK2Yu3oqqWSO0bAZcIskNE1qpqU3/HkVGBFG8gxQqBFW8gxQqBFW8gxQq+i9eahowxJshZIjDGmCAXbIngVX8HkEmBFG8gxQqBFW8gxQqBFW8gxQo+ijeo+giMMcZcLNiuCIwxxlzAEoExxgS5oEkEItJORH4WkR0iMtjf8aRFRMqLyDIR2SoiW0RkgL9jyggRCRWRH0Tk//wdS3pE5HIR+VBEfhKRbSKSq2sxi8hAz9/BZhF5T0Qi/R1TSiIyS0QOisjmFM9dISJfiMh2z9ei/owxWRqxPuv5W9gkIp+IyOV+DPGs1GJN8b3HRURFpLi3Xi8oEoGIhALTgPZAbaCbiNT2b1RpSgAeV9XawDVAv1wca0oDgG3+DiIDXgI+V9WaQANyccwiUhboDzRV1bpAKHC3f6O6yGyg3QXPDQaWqGo1YInncW4wm4tj/QKoq6r1gV+Ap3I6qDTM5uJYEZHyQBvgN2++WFAkAqAZsENVd6lqPDAXuM3PMaVKVX9X1fWe+ydwJ6qy/o0qfSJSDrgFmOnvWNIjIkWAG4DXAVQ1XlWP+jWoS8sH5BeRfEABYL+f4zmPqq4E/rzg6duANz333wQ65WRMaUktVlVdrKoJnoffAOVyPLBUpPG+ArwADAK8OsonWBJBWWBviscx5PKTK4CIVAIaAd/6OZRLeRH3x5nk5zgupTJwCHjD04w1U0QK+juotKjqPuA53Ke/34FjqrrYv1FlSClV/d1z/w+glD+DyYTewH/9HURaROQ2YJ+qbvT2sYMlEQQcESkEfAQ8qqrH/R1PWkSkA3BQVdf5O5YMyAc0BqaraiPgJLmn2eIinrb123AJ7EqgoIj08G9UmaNufHquH6MuIkNwzbJz/B1LakSkAPA0MNwXxw+WRLAPKJ/icTnPc7mSiIThksAcVf3Y3/FcQgugo4jswTW53Sgi7/g3pDTFADGqmnyF9SEuMeRWNwG7VfWQqp4BPgau9XNMGXFARMoAeL4e9HM86RKRXkAHoLvm3olVVXAfCDZ6/tfKAetFpLQ3Dh4sieB7oJqIVBaRcFyH2wI/x5QqERFcG/Y2VX3e3/Fciqo+parlVLUS7n1dqqq58lOrqv4B7BWRGp6nWgNb/RjSpfwGXCMiBTx/F63JxZ3bKSwAenru9wTm+zGWdIlIO1yzZkdV/cvf8aRFVX9U1ZKqWsnzvxYDNPb8TWdbUCQCT2fQw8Ai3D/SB6q6xb9RpakFcA/uk/UGz+1v/g4qD3kEmCMim4CGwDj/hpM2z5XLh8B64Efc/2uuKokgIu8Ba4AaIhIjIvcDE4CbRWQ77qpmgj9jTJZGrC8DhYEvPP9rM/wapEcasfru9XLvlZAxxpicEBRXBMYYY9JmicAYY4KcJQJjjAlylgiMMSbIWSIwxpggZ4nAmBwkItG5vUKrCT6WCIwxJshZIjAmFSLSQ0S+80wy+rdnvYVYEXnBsz7AEhEp4dm2oYh8k6KmfVHP81VF5EsR2Sgi60WkiufwhVKsiTDHM2vYGL+xRGDMBUSkFnAX0EJVGwKJQHegILBWVesAK4ARnl3eAp701LT/McXzc4BpqtoAVyMouSJnI+BR3NoYV+FmkxvjN/n8HYAxuVBroAnwvefDen5c4bQk4H3PNu8AH3vWOLhcVVd4nn8TmCcihYGyqvoJgKqeBvAc7ztVjfE83gBUAr72+U9lTBosERhzMQHeVNXzVqsSkWEXbJfV+ixxKe4nYv+Hxs+saciYiy0BuohISTi7Bm9F3P9LF882fwe+VtVjwP9E5HrP8/cAKzyry8WISCfPMSI8NeWNyXXsk4gxF1DVrSIyFFgsIiHAGaAfbiGbZp7vHcT1I4ArtTzDc6LfBdznef4e4N8iMtpzjK45+GMYk2FWfdSYDBKRWFUt5O84jPE2axoyxpggZ1cExhgT5OyKwBhjgpwlAmOMCXKWCIwxJshZIjDGmCBnicAYY4Lc/wOFz50+hhSuQgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 검증셋과 학습셋의 오차를 저장합니다.\n",
    "y_vloss = history.history['val_loss']\n",
    "y_loss = history.history['loss']\n",
    "\n",
    "# 그래프로 표현해 봅니다.\n",
    "x_len = np.arange(len(y_loss))\n",
    "plt.plot(x_len, y_vloss, marker='.', c=\"red\", label='Testset_loss')\n",
    "plt.plot(x_len, y_loss, marker='.', c=\"blue\", label='Trainset_loss')\n",
    "\n",
    "# 그래프에 그리드를 주고 레이블을 표시하겠습니다.\n",
    "plt.legend(loc='upper right')\n",
    "plt.grid()\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}