05_binary_models_cont.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TyYr72VvvOUD"
   },
   "source": [
    "# Final Binary Image Classifier, Part 2\n",
    "The previous `SFW` and `NSFW` image classifier only included the photos from the `explicit`, `neutral`, and `sexy` classes. This image classifier will include the additional `breastfeeding` and `violent` categories as well. The `violent` photos will be considered `NSFW` while the `breastfeeding` photos will be categorized as `SFW`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "2Em8FXhlQMmB",
    "outputId": "61fd5ac4-d30f-499f-9545-5b966f08cfe9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "# Mount google drive to notebook to pull pickle files\n",
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 73,
     "resources": {
      "http://localhost:8080/nbextensions/google.colab/files.js": {
       "data": "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",
       "headers": [
        [
         "content-type",
         "application/javascript"
        ]
       ],
       "ok": true,
       "status": 200,
       "status_text": ""
      }
     }
    },
    "id": "g8Tkyg7GQR8C",
    "outputId": "b218328b-dd45-4dda-bdbc-a19ebc4119af"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "     <input type=\"file\" id=\"files-42272d7f-54cf-414c-88a0-4397ca97d2e0\" name=\"files[]\" multiple disabled\n",
       "        style=\"border:none\" />\n",
       "     <output id=\"result-42272d7f-54cf-414c-88a0-4397ca97d2e0\">\n",
       "      Upload widget is only available when the cell has been executed in the\n",
       "      current browser session. Please rerun this cell to enable.\n",
       "      </output>\n",
       "      <script src=\"/nbextensions/google.colab/files.js\"></script> "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving helpers_py.py to helpers_py (1).py\n"
     ]
    }
   ],
   "source": [
    "# Import libraries\n",
    "import time, scipy\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.metrics import confusion_matrix, plot_confusion_matrix, classification_report, recall_score, f1_score, precision_score, accuracy_score\n",
    " \n",
    "# Modeling libraries\n",
    "from keras import models\n",
    "from keras import layers\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Flatten, MaxPooling2D, Activation, Conv2D\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras.optimizers import SGD\n",
    "import pickle\n",
    "import keras\n",
    "\n",
    "# Import 'helpers.py' file\n",
    "from google.colab import files\n",
    "src = list(files.upload().values())[0]\n",
    "open('helpers.py','wb').write(src)\n",
    "import helpers as hp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "NbXA3aXaQgNB"
   },
   "outputs": [],
   "source": [
    "# Open Pickle files containing train, test, val data\n",
    "infile = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/X.pickle','rb')\n",
    "X = pickle.load(infile)\n",
    "\n",
    "infile2 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/y.pickle','rb')\n",
    "y = pickle.load(infile2)\n",
    "\n",
    "infile5 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/X_val.pickle','rb')\n",
    "X_val = pickle.load(infile5)\n",
    "\n",
    "infile6 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/y_val.pickle','rb')\n",
    "y_val = pickle.load(infile6)\n",
    "\n",
    "\n",
    "# Import new classes to append to X and y\n",
    "infile3 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/X_extra.pickle','rb')\n",
    "X_extra = pickle.load(infile3)\n",
    "\n",
    "infile4 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/y_extra.pickle','rb')\n",
    "y_extra = pickle.load(infile4)\n",
    "\n",
    "infile7 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/X_val_extra.pickle','rb')\n",
    "X_val_extra = pickle.load(infile7)\n",
    "\n",
    "infile8 = open('/content/drive/MyDrive/nsfw_image_classification/Pickles/y_val_extra.pickle','rb')\n",
    "y_val_extra = pickle.load(infile8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qHCqX51cQiR5",
    "outputId": "bebc03ae-a67f-42ec-f566-33497e403102"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 0, 1, 1])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Replace breastfeeding (0) with class number 1 and violence (1), with class number 0 before appending to y\n",
    "y_extra_num = np.where(y_extra == 0, 1, 0) \n",
    "y_extra_num[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "92Vet_U8Qj-S",
    "outputId": "e5f494c8-8b25-46a9-b911-c494a222f8d8"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 1, 0, 1])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Replace breastfeeding (0) with class number 1 and violence (1), with class number 0 before appending to y\n",
    "y_val_extra_num = np.where(y_val_extra == 0, 1, 0) \n",
    "y_val_extra_num[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "uTUD7aIlQlh9",
    "outputId": "04efbb74-dbaa-4a7a-940f-2cb824a3dcd4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_new Size: 65267\n",
      "y_new Size: 65267\n",
      "X_val_new Size: 9299\n",
      "y_val_new Size: 9299\n"
     ]
    }
   ],
   "source": [
    "# Concatenate X and y values for train and validation data to create new arrays\n",
    "X_new = np.concatenate((X, X_extra), axis=0)\n",
    "y_new = np.concatenate((y, y_extra_num), axis=0)\n",
    "\n",
    "X_val_new = np.concatenate((X_val, X_val_extra), axis=0)\n",
    "y_val_new = np.concatenate((y_val, y_val_extra_num), axis=0)\n",
    "\n",
    "print(f'X_new Size: {len(X_new)}')\n",
    "print(f'y_new Size: {len(y_new)}')\n",
    "print(f'X_val_new Size: {len(X_val_new)}')\n",
    "print(f'y_val_new Size: {len(y_val_new)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "jDB-yjPqRD0e",
    "outputId": "ce9f2cf8-5607-4cf2-c4f4-a56337ab84bd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Explicit Images In y train: 31436\n",
      "NSFW Images In y_train_binary: 31436\n"
     ]
    }
   ],
   "source": [
    "# Relabel explicit data to 1 for 'NSFW' and sexy and neutral data to 0 for 'SFW'\n",
    "print(f'Explicit Images In y train: {sum(y_new==0)}')\n",
    "y_train_binary = np.where(y_new > 0, 0, 1)\n",
    "print(f'NSFW Images In y_train_binary: {sum(y_train_binary==1)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "EGLewTOERTzG",
    "outputId": "df7fe394-802b-426c-bc66-baa223ac4d8f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Explicit Images In y_val: 4488\n",
      "NSFW Images In y_val_binary: 4488\n"
     ]
    }
   ],
   "source": [
    "# Repeat the process for y_val\n",
    "print(f'Explicit Images In y_val: {sum(y_val_new==0)}')\n",
    "y_val_binary = np.where(y_val_new > 0, 0, 1)\n",
    "print(f'NSFW Images In y_val_binary: {sum(y_val_binary==1)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "Jwa8Hx-XRWH8"
   },
   "outputs": [],
   "source": [
    "# Data Augmentation with horizonatal and vertical flips\n",
    "data_augmentation = Sequential([\n",
    "  layers.experimental.preprocessing.RandomFlip(\"horizontal_and_vertical\"),\n",
    "  layers.experimental.preprocessing.RandomRotation(0.2),\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HODfG2qwwQcU"
   },
   "source": [
    "I have reused the best model from the [3-class CNN model]('02_best_3_class_model.ipynb') to build this binary model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "vxrx8mazRpMd",
    "outputId": "5b9b3e14-6928-48b0-a53c-a1a10715f57f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1632/1632 [==============================] - 400s 244ms/step - loss: 0.6757 - accuracy: 0.5859 - val_loss: 0.6480 - val_accuracy: 0.6288\n",
      "Epoch 2/50\n",
      "1632/1632 [==============================] - 401s 245ms/step - loss: 0.6177 - accuracy: 0.6687 - val_loss: 0.5960 - val_accuracy: 0.6922\n",
      "Epoch 3/50\n",
      "1632/1632 [==============================] - 400s 245ms/step - loss: 0.5845 - accuracy: 0.6956 - val_loss: 0.5721 - val_accuracy: 0.7009\n",
      "Epoch 4/50\n",
      "1632/1632 [==============================] - 404s 248ms/step - loss: 0.5553 - accuracy: 0.7225 - val_loss: 0.6752 - val_accuracy: 0.6462\n",
      "Epoch 5/50\n",
      "1632/1632 [==============================] - 398s 244ms/step - loss: 0.5393 - accuracy: 0.7342 - val_loss: 0.5310 - val_accuracy: 0.7395\n",
      "Epoch 6/50\n",
      "1632/1632 [==============================] - 400s 245ms/step - loss: 0.5271 - accuracy: 0.7422 - val_loss: 0.6583 - val_accuracy: 0.6213\n",
      "Epoch 7/50\n",
      "1632/1632 [==============================] - 400s 245ms/step - loss: 0.5117 - accuracy: 0.7496 - val_loss: 0.5239 - val_accuracy: 0.7448\n",
      "Epoch 8/50\n",
      "1632/1632 [==============================] - 396s 243ms/step - loss: 0.4877 - accuracy: 0.7661 - val_loss: 0.4962 - val_accuracy: 0.7586\n",
      "Epoch 9/50\n",
      "1632/1632 [==============================] - 401s 246ms/step - loss: 0.4722 - accuracy: 0.7769 - val_loss: 0.4863 - val_accuracy: 0.7757\n",
      "Epoch 10/50\n",
      "1632/1632 [==============================] - 402s 246ms/step - loss: 0.4583 - accuracy: 0.7844 - val_loss: 0.4723 - val_accuracy: 0.7793\n",
      "Epoch 11/50\n",
      "1632/1632 [==============================] - 405s 248ms/step - loss: 0.4501 - accuracy: 0.7902 - val_loss: 0.4807 - val_accuracy: 0.7677\n",
      "Epoch 12/50\n",
      "1632/1632 [==============================] - 399s 244ms/step - loss: 0.4364 - accuracy: 0.7987 - val_loss: 0.4506 - val_accuracy: 0.7919\n",
      "Epoch 13/50\n",
      "1632/1632 [==============================] - 403s 247ms/step - loss: 0.4308 - accuracy: 0.7994 - val_loss: 0.5673 - val_accuracy: 0.7402\n",
      "Epoch 14/50\n",
      "1632/1632 [==============================] - 404s 247ms/step - loss: 0.4206 - accuracy: 0.8064 - val_loss: 0.4461 - val_accuracy: 0.7963\n",
      "Epoch 15/50\n",
      "1632/1632 [==============================] - 404s 247ms/step - loss: 0.4120 - accuracy: 0.8124 - val_loss: 0.4731 - val_accuracy: 0.7850\n",
      "Epoch 16/50\n",
      "1632/1632 [==============================] - 402s 246ms/step - loss: 0.4060 - accuracy: 0.8148 - val_loss: 0.4646 - val_accuracy: 0.7906\n",
      "Epoch 17/50\n",
      "1632/1632 [==============================] - 398s 244ms/step - loss: 0.3992 - accuracy: 0.8194 - val_loss: 0.4397 - val_accuracy: 0.7927\n",
      "Epoch 18/50\n",
      "1632/1632 [==============================] - 403s 247ms/step - loss: 0.3967 - accuracy: 0.8214 - val_loss: 0.4880 - val_accuracy: 0.7863\n",
      "Epoch 19/50\n",
      "1632/1632 [==============================] - 400s 245ms/step - loss: 0.3905 - accuracy: 0.8245 - val_loss: 0.4692 - val_accuracy: 0.7932\n",
      "Epoch 20/50\n",
      "1632/1632 [==============================] - 401s 246ms/step - loss: 0.3876 - accuracy: 0.8268 - val_loss: 0.4248 - val_accuracy: 0.8082\n",
      "Epoch 21/50\n",
      "1632/1632 [==============================] - 402s 246ms/step - loss: 0.3831 - accuracy: 0.8280 - val_loss: 0.4659 - val_accuracy: 0.7776\n",
      "Epoch 22/50\n",
      "1632/1632 [==============================] - 399s 245ms/step - loss: 0.3822 - accuracy: 0.8277 - val_loss: 0.4226 - val_accuracy: 0.8095\n",
      "Epoch 23/50\n",
      "1632/1632 [==============================] - 399s 245ms/step - loss: 0.3784 - accuracy: 0.8309 - val_loss: 0.4279 - val_accuracy: 0.8057\n",
      "Epoch 24/50\n",
      "1632/1632 [==============================] - 400s 245ms/step - loss: 0.3730 - accuracy: 0.8337 - val_loss: 0.4492 - val_accuracy: 0.8087\n",
      "Epoch 25/50\n",
      "1632/1632 [==============================] - 402s 246ms/step - loss: 0.3689 - accuracy: 0.8357 - val_loss: 0.4506 - val_accuracy: 0.8045\n",
      "Epoch 26/50\n",
      "1632/1632 [==============================] - 403s 247ms/step - loss: 0.3671 - accuracy: 0.8377 - val_loss: 0.5977 - val_accuracy: 0.7371\n",
      "Epoch 27/50\n",
      "1632/1632 [==============================] - 403s 247ms/step - loss: 0.3621 - accuracy: 0.8398 - val_loss: 0.4890 - val_accuracy: 0.7950\n",
      "Epoch 28/50\n",
      "1632/1632 [==============================] - 399s 245ms/step - loss: 0.3601 - accuracy: 0.8415 - val_loss: 0.4695 - val_accuracy: 0.7928\n",
      "Epoch 29/50\n",
      "1632/1632 [==============================] - 399s 244ms/step - loss: 0.3585 - accuracy: 0.8410 - val_loss: 0.4100 - val_accuracy: 0.8183\n",
      "Epoch 30/50\n",
      "1632/1632 [==============================] - 396s 243ms/step - loss: 0.3543 - accuracy: 0.8438 - val_loss: 0.4079 - val_accuracy: 0.8187\n",
      "Epoch 31/50\n",
      "1632/1632 [==============================] - 401s 246ms/step - loss: 0.3542 - accuracy: 0.8434 - val_loss: 0.4061 - val_accuracy: 0.8262\n",
      "Epoch 32/50\n",
      "1632/1632 [==============================] - 399s 244ms/step - loss: 0.3503 - accuracy: 0.8466 - val_loss: 0.4109 - val_accuracy: 0.8164\n",
      "Epoch 33/50\n",
      "1632/1632 [==============================] - 397s 244ms/step - loss: 0.3496 - accuracy: 0.8465 - val_loss: 0.4222 - val_accuracy: 0.8158\n",
      "Epoch 34/50\n",
      "1632/1632 [==============================] - 392s 240ms/step - loss: 0.3476 - accuracy: 0.8477 - val_loss: 0.7709 - val_accuracy: 0.7133\n",
      "Epoch 35/50\n",
      "1632/1632 [==============================] - 396s 243ms/step - loss: 0.3421 - accuracy: 0.8503 - val_loss: 0.4151 - val_accuracy: 0.8209\n",
      "Epoch 36/50\n",
      "1632/1632 [==============================] - 396s 242ms/step - loss: 0.3414 - accuracy: 0.8508 - val_loss: 0.4081 - val_accuracy: 0.8253\n",
      "Epoch 37/50\n",
      "1632/1632 [==============================] - 395s 242ms/step - loss: 0.3398 - accuracy: 0.8521 - val_loss: 0.3925 - val_accuracy: 0.8338\n",
      "Epoch 38/50\n",
      "1632/1632 [==============================] - 399s 244ms/step - loss: 0.3377 - accuracy: 0.8525 - val_loss: 0.4668 - val_accuracy: 0.8077\n",
      "Epoch 39/50\n",
      "1632/1632 [==============================] - 398s 244ms/step - loss: 0.3366 - accuracy: 0.8550 - val_loss: 0.5224 - val_accuracy: 0.7867\n",
      "Epoch 40/50\n",
      "1632/1632 [==============================] - 396s 243ms/step - loss: 0.3336 - accuracy: 0.8549 - val_loss: 0.4803 - val_accuracy: 0.8006\n",
      "Epoch 41/50\n",
      "1632/1632 [==============================] - 395s 242ms/step - loss: 0.3362 - accuracy: 0.8539 - val_loss: 0.4279 - val_accuracy: 0.8142\n",
      "Epoch 42/50\n",
      "1632/1632 [==============================] - 393s 241ms/step - loss: 0.3305 - accuracy: 0.8555 - val_loss: 0.4073 - val_accuracy: 0.8299\n",
      "Epoch 43/50\n",
      "1632/1632 [==============================] - 393s 241ms/step - loss: 0.3273 - accuracy: 0.8575 - val_loss: 0.4045 - val_accuracy: 0.8286\n",
      "Epoch 44/50\n",
      "1632/1632 [==============================] - 399s 245ms/step - loss: 0.3281 - accuracy: 0.8576 - val_loss: 0.3949 - val_accuracy: 0.8289\n",
      "Epoch 45/50\n",
      "1632/1632 [==============================] - 396s 242ms/step - loss: 0.3281 - accuracy: 0.8593 - val_loss: 0.4090 - val_accuracy: 0.8148\n",
      "Epoch 46/50\n",
      "1632/1632 [==============================] - 397s 243ms/step - loss: 0.3240 - accuracy: 0.8597 - val_loss: 0.4123 - val_accuracy: 0.8134\n",
      "Epoch 47/50\n",
      "1632/1632 [==============================] - 398s 244ms/step - loss: 0.3235 - accuracy: 0.8613 - val_loss: 0.4343 - val_accuracy: 0.8230\n",
      "Epoch 48/50\n",
      "1632/1632 [==============================] - 395s 242ms/step - loss: 0.3220 - accuracy: 0.8623 - val_loss: 0.4382 - val_accuracy: 0.8119\n",
      "Epoch 49/50\n",
      "1632/1632 [==============================] - 399s 245ms/step - loss: 0.3208 - accuracy: 0.8601 - val_loss: 0.4194 - val_accuracy: 0.8308\n",
      "Epoch 50/50\n",
      "1632/1632 [==============================] - 394s 241ms/step - loss: 0.3186 - accuracy: 0.8630 - val_loss: 0.4459 - val_accuracy: 0.8156\n"
     ]
    }
   ],
   "source": [
    "model = Sequential([ # Model steps are performed one after another rather than at the same time\n",
    "  data_augmentation, # Apply data augmentation to the model\n",
    "  layers.Conv2D(filters=48, kernel_size=3, activation='relu', input_shape=X.shape[1:]),\n",
    "  layers.MaxPooling2D(pool_size=2, strides=2) \n",
    "]) # Max pooling to reduce noise and calculate the maximum value in each patch of each feature map\n",
    "\n",
    "model.add(layers.Conv2D(filters=48, kernel_size=3, activation='relu'))\n",
    "model.add(layers.MaxPool2D(pool_size=2, strides=1)) # More max pooling, decreased stride\n",
    "model.add(layers.Conv2D(filters=32, kernel_size=3, activation='relu'))\n",
    "model.add(layers.MaxPool2D(pool_size=2, strides=1)) # More max pooling, decreased stride\n",
    "model.add(layers.Conv2D(filters=32, kernel_size=3, activation='relu'))\n",
    "model.add(layers.MaxPool2D(pool_size=2, strides=1)) # More max pooling, decreased stride\n",
    "model.add(layers.Flatten()) # Transform features into a single column\n",
    "model.add(layers.Dense(128, activation='relu')) # First hidden layer\n",
    "model.add(layers.Dense(72, activation='relu')) # Second hidden layer\n",
    "model.add(layers.Dense(64, activation='relu')) # Third hidden layer\n",
    "model.add(layers.Dense(30, activation='relu')) # Fourth hidden layer\n",
    "model.add(layers.Dense(8, activation='relu')) # Fifth hidden layer\n",
    "model.add(layers.Dense(3, activation='softmax')) # Sixth hidden layer\n",
    "\n",
    "# Compile and train CNN\n",
    "opt = SGD(learning_rate=0.06) # Small learning rate to avoid overfitting\n",
    "model.compile(loss = 'sparse_categorical_crossentropy', # Sparse_categorical_crossentropy because the labels are not one hot encoded, and to reduce comp time\n",
    "               optimizer = opt, \n",
    "               metrics=['accuracy'])\n",
    "\n",
    "# Fit and score model performance\n",
    "model_score = model.fit(X_new,\n",
    "                        y_train_binary,\n",
    "                        epochs=50,\n",
    "                        validation_split=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "XvT2vJTfRr0i"
   },
   "outputs": [],
   "source": [
    "# Score perofmance on unseen data\n",
    "val_score = model.evaluate(X_val_new, y_val_binary)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "-YfIPA-MThqH",
    "outputId": "7f063fe0-7d52-45ee-f9cc-3aef54b1dd8a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This model has a 83.2% Accuracy\n"
     ]
    }
   ],
   "source": [
    "# Print accuracy\n",
    "print(f'This model has a {round(((val_score[1])* 100),1)}% Accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 573
    },
    "id": "k_1h2D4X8-_k",
    "outputId": "47a730f0-2b92-4555-e510-3ad2b0037b55"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOydd3zV9fX/nyd7kh02SdhLZYnbojhArfNbxVG1tdJlv1qtrdo6an/ddbSOOlprrV/ciyogqDgBBVxsEiBAAtl7j/v+/fG+N7kk9yY3kJvc3Jzn45HHzf2s+75RPq/P2WKMQVEURVE6EtLfC1AURVECExUIRVEUxSMqEIqiKIpHVCAURVEUj6hAKIqiKB5RgVAURVE8ogKhDHpEJFNEjIiE+XDstSLycV+sS1H6GxUIZUAhIrki0iQiqR22f+G8yWf2z8oOWUuciNSIyPL+XouiHAkqEMpAZA9wueuNiBwFxPTfcjpxCdAInCkiw/ryg32xghTFV1QglIHIf4Cr3d5fAzzjfoCIJIjIMyJSLCJ7ReRXIhLi3BcqIn8RkRIR2Q2c6+Hcf4rIQRHJF5H/JyKhPVjfNcBjwNfAVR2ufbKIrBGRChHZLyLXOrdHi8h9zrVWisjHzm3zRCSvwzVyReQM5+/3iMjLIvKsiFQB14rIXBFZ6/yMgyLysIhEuJ0/TURWiUiZiBSKyB0iMkxE6kQkxe24Wc6/X3gPvrsSRKhAKAORdcAQEZnivHEvAp7tcMxDQAIwFvgGVlC+49x3PXAeMBOYA/xPh3OfBlqA8c5jzgK+58vCRCQDmAf8n/Pn6g77ljvXlgbMAL507v4LMBs4EUgGfg44fPlM4ALgZSDR+ZmtwE+BVOAEYD7wI+ca4oF3gBXACOd3fNcYUwC8D1zqdt1vA88bY5p9XIcSZKhAKAMVlxVxJrANyHftcBON240x1caYXOA+7A0P7E3wQWPMfmNMGfB7t3OHAucANxljao0xRcADzuv5wreBr40xW4HngWkiMtO57wrgHWPMc8aYZmNMqTHmS6dl813gRmNMvjGm1RizxhjT6ONnrjXGvG6McRhj6o0xG40x64wxLc7v/jhWJMEKY4Ex5j5jTIPz7/Opc9+/cVo8zr/h5di/szJIUX+lMlD5D/AhkEUH9xL2yTkc2Ou2bS8w0vn7CGB/h30uMpznHhQR17aQDsd3xdXAkwDGmHwR+QDrcvoCGA3s8nBOKhDlZZ8vHLI2EZkI3I+1jmKw/843Ond7WwPAG8BjIpIFTAIqjTGfHeaalCBALQhlQGKM2YsNVp8DvNphdwnQjL3ZuxhDu5VxEHujdN/nYj82wJxqjEl0/gwxxkzrbk0iciIwAbhdRApEpAA4DrjCGTzeD4zzcGoJ0OBlXy1uAXjnk31ah2M6tmT+O7AdmGCMGQLcAbjUbj/W7dYJY0wD8CLWivg2aj0MelQglIHMdcDpxpha943GmFbsje63IhLv9P3fTHuc4kXgf0VklIgkAbe5nXsQWAncJyJDRCRERMaJyDfonmuAVcBUbHxhBjAdiAYWYuMDZ4jIpSISJiIpIjLDGOMAngLuF5ERziD6CSISCewEokTkXGew+FdAZDfriAeqgBoRmQz80G3fm8BwEblJRCKdf5/j3PY/A1wLnI8KxKBHBUIZsBhjdhljNnjZ/RPs0/du4GNgCfYmDNYF9DbwFfA5nS2Qq4EIYCtQjg0AD+9qLSIShY1tPGSMKXD72YO90V5jjNmHtXhuAcqwAepjnJf4GbAJWO/c90cgxBhTiQ0w/wNrAdUCh2Q1eeBn2HhHtfO7vuDaYYypxsZtvgkUANnAaW77P8EGxz93WmnKIEZ0YJCiKO6IyHvAEmPMP/p7LUr/ogKhKEobInIs1k022mltKIMYdTEpigKAiPwbWyNxk4qDAmpBKIqiKF5QC0JRFEXxSNAUyqWmpprMzMz+XoaiKMqAYuPGjSXGmI61NUAQCURmZiYbNnjLeFQURVE8ISJe05nVxaQoiqJ4RAVCURRF8YgKhKIoiuKRoIlBeKK5uZm8vDwaGhr6eyl+JyoqilGjRhEerrNdFEXpHYJaIPLy8oiPjyczMxO31s1BhzGG0tJS8vLyyMrK6u/lKIoSJAS1i6mhoYGUlJSgFgcAESElJWVQWEqKovQdQS0QQNCLg4vB8j0VRek7gtrFpCiKMtBobGklMizUp2Mr6ppYtbWQ5lbDFceN6f6EHqIC4WcqKipYsmQJP/rRj3p03jnnnMOSJUtITEz008oURekPmlocZBdVs7OwmgMVDeRX1HOwop4DFQ0cqKynuqGFzJQYjh+b0vYzLCGq7fyi6gZWbilkxeYC1u4updVhmDkmUQViIFJRUcGjjz7aSSBaWloIC/P+51+2bJm/l6Yoih8wxtDU6qCusZW65lYKKhvYeqCSzflVbDlYyc6CGppaHW3HJ8WEMyIxmtHJMRw3NpmkmAi2HKjirU0HeX69HTeelRrLnIwk9pbWsX5vGcbYbYtPHcvC6cM4amSCX76LCoSfue2229i1axczZswgPDycqKgokpKS2L59Ozt37uTCCy9k//79NDQ0cOONN7J48WKgvXVITU0NCxcu5OSTT2bNmjWMHDmSN954g+jo6H7+ZooyeGlobmV7QTWb8yvZcqCSLQeqKKxqaBOFVkfnLtlJMeFMG5HAd07OZNqIBKYOj2dkYgzREZ7dSa0Ow7aDVazbXcq63aWs3FrI8IQobpw/gYXThzNxaJzfY4+DRiB+/d8tbD1Q1avXnDpiCHd/s+tZ9n/4wx/YvHkzX375Je+//z7nnnsumzdvbktHfeqpp0hOTqa+vp5jjz2WSy65hJSUlEOukZ2dzXPPPceTTz7JpZdeyiuvvMJVV13Vq99FUQY7xhh2FdeyIbeMTfmVNDQ7MMbgMAaHAYcxtLQa9pTUklNc0yYCCdHhTBsxhCkT04mJDCUmIpSYiDDnayjJsZFMHTGEEQlRPbqhh4YI00cmMH1kAt87Zay/vnaXDBqBCBTmzp17SK3C3/72N1577TUA9u/fT3Z2dieByMrKYsaMGQDMnj2b3NzcPluvogQr9U2tbD1YyYbcctbnlrNxbxnldc0ADIkKIz4qnJAQCBEhRAQRCBVhdHIMZ00byrQRCUwbMYRRSdFBm0U4aASiuyf9viI2Nrbt9/fff5933nmHtWvXEhMTw7x58zzWMkRGRrb9HhoaSn19fZ+sVVECheZWBxv3lrN6RxFrckqpa2ohNEQIDQkhNMTeuENDhIRo688fkRjNSOfrcGeAd3tBNdsPVrG9oJptB6vYU1qLa15aVmosZ0wZyrGZyczOTGJsamzQ3vR7gl8FQkQWAH8FQoF/GGP+0GH/GODfQKLzmNuMMctEJBPYBuxwHrrOGPMDf67VX8THx1Nd7Xl6Y2VlJUlJScTExLB9+3bWrVvXx6tTlMCkudVBcXUjH2eXsHpHER9nl1Dd2EJ4qDA7I4kxKTG0thpajaHV0f5TVN3Il/sr2iwBT2SkxDB5WDznzxjBlOFDmDUmibT4SK/HD2b8JhAiEgo8ApwJ5AHrRWSpMWar22G/Al40xvxdRKYCy4BM575dxpgZ/lpfX5GSksJJJ53E9OnTiY6OZujQoW37FixYwGOPPcaUKVOYNGkSxx9/fD+uVFH8T2VdM7tLathdXMueEvuTX1FPXVMLtY2t9rWplaaW9iyfoUMiOffo4cyblM5J41OIj+q+31hdU4tNG62o52BlPa0OmDw8nklD44mNHDSOkyPGn3+puUCOMWY3gIg8D1wAuAuEAYY4f08ADvhxPf3GkiVLPG6PjIxk+fLlHve54gypqals3ry5bfvPfvazXl+fovQG+8vqWL2jiPzyeqobW6hpaKHG+VrV0ExRdSNltU1tx4eGCGOSYxiVZN1AMRFhxEaGEh0RSmxEGEOiwpiblcKU4fE9dvfERIQxPj2O8elxvf01BxX+FIiRwH6393nAcR2OuQdYKSI/AWKBM9z2ZYnIF0AV8CtjzEcdP0BEFgOLAcaM6f0iEUVRvNPqMHyxr5x3thXx7rZCsotqAIgKDyEuMpz4qDDiIu3P6OQYZo5JZGxqHFmpsWSlxTI6KYaIsKDv9jOg6W9b63LgaWPMfSJyAvAfEZkOHATGGGNKRWQ28LqITDPGHJKnaox5AngCYM6cOZ0TjxVF6RUq65rJLa0lt7SWvaV1ZBfV8HF2MeV1zYSFCHOzklk0dwzzJ6eTmRrb/QWVAYE/BSIfGO32fpRzmzvXAQsAjDFrRSQKSDXGFAGNzu0bRWQXMBHQodOK0ks4HIZ1e0p59fN8vtpfYdM5Q6QtK0hEaHUY9pfXUdEh6Ds8IYp5k9KZPyWdUyemMcSHuIAy8PCnQKwHJohIFlYYFgFXdDhmHzAfeFpEpgBRQLGIpAFlxphWERkLTAB2+3GtijJo2FVcw6uf5/H6FwfIr6gnLjKM48cmExoitDpwFobZrKAQEY4elUBmSiwZKTFkpsYyJjmGqHDfmskpAxu/CYQxpkVEbgDexqawPmWM2SIi9wIbjDFLgVuAJ0Xkp9iA9bXGGCMipwL3ikgz4AB+YIwp89daFSUYyK+o5+3NBbyzrZDaxhainRW90eGhzt9D+Sqv0mktwCkT0vj5gkmcNXWY13YPyuDGrzEIY8wybOqq+7a73H7fCpzk4bxXgFf8uTZFGQg0NLeSV15PTEQo8VFhxEaEERLSntGzq7iGFZsLeHtLAV/nVQIwaWg8wxKiqG9qpai6gfqmVuqbbI+g4QnR/PKcKVwwYwTpQ6K8fayiAP0fpA56DrfdN8CDDz7I4sWLiYmJ8cPKlEDD1fphc34Vm/Ir2ZxfSXZRzSGN30QgLiKM+KgwRIT8CltVf8zoRG5bOJmzpw0jS4PESi+hAuFnvLX79oUHH3yQq666SgUiSCmsauCzPWVsyC1jfW452wuqcGlBalwE00cmcMaUoYxLj6Wx2UF1QwvVjS1UNzRT3dBCQ3Mr3zsli7OnDWNEonb3VXofFQg/497u+8wzzyQ9PZ0XX3yRxsZGLrroIn79619TW1vLpZdeSl5eHq2trdx5550UFhZy4MABTjvtNFJTU1m9enV/fxXlCHA4DNlFNWzcW24FYW8Z+8vs039MRCizxiRxw2njOWpUItNHDmHYkJ51/lQUfzB4BGL5bVCwqXevOewoWPiHLg9xb/e9cuVKXn75ZT777DOMMZx//vl8+OGHFBcXM2LECN566y3A9mhKSEjg/vvvZ/Xq1aSmpvbuuhW/U9fUwpf7K9iYW87GfeV8vrecqoYWwFoHczKSufbELI7NTGLK8CGEh2rBmBJ4DB6BCABWrlzJypUrmTlzJgA1NTVkZ2dzyimncMstt/CLX/yC8847j1NOOaWfV6r0lMKqBjbklrNhbxkb95az5UBVW+xg4tA4zj16OLMzkpmTkURGSoxaB8qAYPAIRDdP+n2BMYbbb7+d73//+532ff755yxbtoxf/epXzJ8/n7vuusvDFZRAobaxhfd3FPPutkI+yy0jr9y6i6LCQzhmVCI/+MZY5mQmM2t0EgkxWkTmlX2fwp4P4NRbbQReCSgGj0D0E+7tvs8++2zuvPNOrrzySuLi4sjPzyc8PJyWlhaSk5O56qqrSExM5B//+Mch56qLKTCorG/m3W12WPwHO4tpbHGQFBPOcVkpXHtiJnMyk5k2Qt1FPrNrNTx3ObTUw7j5MGp2f69I6YAKhJ9xb/e9cOFCrrjiCk444QQA4uLiePbZZ8nJyeHWW28lJCSE8PBw/v73vwOwePFiFixYwIgRIzRI3U/UNLbw9uYCln51gDW7SmhuNQwbEsXlc8dw9rRhHJuZRJgKQs/JXgXPXwkp46FkJ2x5VQWiOza9DOlTYGjfDT8TY4Kjx92cOXPMhg2Htmratm0bU6ZM6acV9T2D7fv6i5ZWBx/llPDa5/ms3FpAQ7OD0cnRnDN9OGdPH8aMUYmHFKspPWT7MnjpGnuz+/br8NoPoGgr3LSpb91M29+Ct26BxAxInQBpkyB1ov1JHAMSAq1N0FwPLQ321dEKKeP63h1Wkg0PH2vX9sNPILT33JYistEYM8fTPrUglKDHGENxdSO5pXXObqS1NDY7iAwPITIslMiwEPsTHsqOgmre/PoAJTVNJMaE8z+zR3HRzJHMGpOkgeXeYOsb8PJ3YfgxcNWrEJ0I0y6C7LchfyOM8nif6n2aG2xmY0g4hITCzhXwxX/a94eEWTHAwwP0sdfDuX/pm3W6WPOQFaWSHbDhKTiucxzTH6hAKEGFMYZdxbWs3VXCuj1l7CqqYW9pHfXNrW3HhIYIkWEhNDS34ujw7z8iNIT5U9K5aOZI5k1KD655BbUlYAzEpfXP529+BV653orAlS9BVILdPmmhvVFvea3vBOKzx6FyH1z9BoydZ7fVldkn9ZIdULbHCkdYFIRHt7/ufh/WPwlHfQvGdBxv4yeqC+Cr52D2tVCaA6t/Zz8/JtnvHx30AmGMGRRPfsHiKjwc8srrWJNTyppdJazZVUpRdSNgW1JPHT6EE8elkpkaQ0ZKLJkpMYxIjG4LJLe0OmhqddDY7KCxxUGcc8hN0LF3LTx/hXVNXLcKkjL67rNbGuGzJ2HVnTD6eLjyRYiMb98fnQjj51vr4qz/53/3TV0ZfHgfTDirXRzA3nDHHNf1jX/K+ZD7Cbx5E3z/w1519Xjl08fA0QIn3GBdXY+dDO//Hs75s98/Ogj/JbQTFRVFaWkpKSkpQS0SxhhKS0uJiho8zdeKqhv471cHeePL/LYmdalxEZwwLpUTx6Vw4rgUxiR3X28QFhpCWGgIMRF9sep+YtPL8PoPIWEU1JXCs5fAdSv9/wTa0mjdNh/dD1X5MP5MuPTfEOGhV9TUC62bpy/cTB/8CZqq4cx7e35uZByc8ycrtmsfgZNv6v31udNQBeufssKUMs5um30trP8nzPmujeP4kaAWiFGjRpGXl0dxcXF/L8XvREVFMWrUqP5ehl9xZRS9/mU+n+SU4DAwfeQQfnnOFL4xKY0J6XFB/SDQY4yBj+6D934DY06ERf8HxdvhmQvhuUXWvRLuhx5OLU1uwpAHo+bCBQ/D2NO8WweTFkJohP/dTKW7rIto1tWHf3OdfC5MOhfe/4ONnxyONfb1i1aoLn0Ghk71ftzn/4bGSjjpxvZtp/0SNr0Cb99h4zh+/H8+qLOYlIFNbaNtV+GqUF6fW9aWUXThjJFcMGME49Pju7/QYKS1Gd78qb1RH/UtuOARCIu0+7a8Di9da290lz5jfe29xbY3YcVtULkfRh0L826Hcaf7dhNbchkUbLbZTCF+iv288G3IeRf+9wuIH3r416nMg4fnQubJcMULPbtJ71xpBdq0QvJYuH61dbN1pKUJ/noMpI6Ha/576L61j1iBuOJFmHj24X8PNItJGUB8uruU5ZsL2LC3jG0Hq2l1GETsjINFx47hm8cM14yi7qivgBevdlYo/xxOu+PQG9i0C6H69/ZGvuI2WPin3nsKfetmiBwCV71ii996ct1pF7W7mUYf2zvrcWffOti21D6BH4k4gHXXnXYHrPylvebUC3w7b/96m+I7bDqcfhc8dxm89n1Y9FxnUdz0ElQfgAse6nydY6+32Uxv32EtszD/+EhVIJSAYFNeJX96ezsfZZcQHR7KjNGJ/GjeOOZkJjNzTKLOPO5IcwOsuguqD0JTLTTX2demWqgrsa8XPAozr/R8/vE/tE/Bax+GISN7x5fucEBtsXXfjD+j5+e7u5l6WyCMgbd/CfHD4YQf9841j/sBfP08LP+FvUlHDen6+OIdsORbED8MrnzFZpMt+AMs+xl8+GeY94v2Yx0OWPM3GHqUFdqOhEXA2b+DJZdal1lvfaeOH+OXqyqKj+wuruG+VTt56+uDJMaE86tzp3DV8Rk687g7diyzqZop4+0Te0QsDBkB4TH29xlXQMaJXV/jzN9YgXnnbhuLGHOCTT2NGmKv2VPXU305GAfEHGZrmKgEezN0ZTP1pptpy2uQv8G62jwFyQ+H0DA470H4xxmw+rew8I/ej63Mg/9cZAXwqlfbU42P/Z61mN7/PYyY0e4uyl5p40UX/8O7FTbhLPv3ev+PcPRlENv7LXlUIJR+4UBFPQ+9l8OLG/YTGRbC/54+nu+dOlYtBV/Z8irEDYUff3b4MYSQELjw71BTBMt/3nl/5BBrXXxnmW8ZT3Wl9jUm5fDWA04303J7Mx899/Cv405LI7xzDwydDsdc3jvXdDFqjs0m+uwJu/bRx3W+odeVwX8uhsZquPYtSM5q3ycC5z0AhZvh1ettPCJlHHzyICSMse5Ab4hYK+LvJ1qBOu+B3v1uqEAofURTi4MNe8v4cGcJH+4sZuvBKsJDhW8fn8GPTxtPWnxkfy9x4NBQZQOds6898gBzWKR9os37zMYuGirbf4q3w9bXoWgbZHYaHd+ZuhL7GnsEAjFpgdPN9HrvCcSnj0PFXvs9ezMg72L+XbD9TXjqbIhKtNlRaZOdr5Pgvd9Cea6Nyww/uvP54dFw2bPwxDwbRD/7t7BvLSz4Y/d1FumTYe5iKz7G9HpGkwqE4jeaWx28tCGPd7cVsnZ3KXVNrYSFCLMzkrj17Emcf8wIRifrONUes2M5tDbC9Et653phETYbpyMFm6xAuG783VHrPO5wXUxg3Uzjz7Cf2xtupop9Nh11wlm2GM8fRCfaupIdK6B4m401bHkNNv7LeYDYbLGsLua8JGXCJf+0NSpLLoPoJJj1bd8+/+zf+S3rSwVC8Qs5RTXc/OKXfJ1XSUZKDJfMGsWpE9M4YVxKcFYq9yVbXoUho2waqT+JdfrJa32sI+oNFxPYorkdy47czWQMvHmz/f3c+45sTd2RlAnH/+DQz64ptNZXdJKNL3TH+Pkw/05491449kbfYyX+SglGBULpZRwOwzNrc/n98u1ER4Ty6JWzOOeo4f29LN+oKbL5+yNmBe7wmvpym8d/3Pf9emMA2m/0tT5aEG0upiMMlk5aCKGRzmymIxCIza9AziqbKZQ45sjW1FNEbLZS/LCenXfyzTBiJmR4sOj6gSDqRKb0NwWVDVzzr8+4579bOWFcCitvOrV/xSH7HetL95U3fwpPng6PnQKf/8emknqjJMdWwj5zoe0z1NWxvcm2N8HR3Hvupa4IDbdPv75aELWlEBHfXpB3uEQNsU/TW1636Z6HQ12ZTT8dOdv66AcKIraw0E91DT1FBULpFZZ+dYCzHviADbnl/Pai6fzr2mNJH9KPvaGy34H/uwTWPOzb8S1NtlPn6ONshevSG+CBqdbcrzpgj6nYBx8/CI+fCg/Ptl01y3bbPPa/Hm0/q6nWb18JsO6lpEz7lNkXxKT2zII4kgC1O9MuskVi+9Ye3vlv/xIaKuD8h/wTmB4kqItJOSIaW1q5Z+kWnvtsPzPHJHL/pTPISu2lPPPDpaXJVgiDzSc//Zfdn5O3Hppq4MT/tS0ocj+CdY/ZfkKf/BVSJ0HRFnvsyNk2MDj1Qlt7kPuRLXRa+Uv4+H44/kcw9/r2dta9RW0J7P7A9uXpKxdYbFoPBKL0yOMPLiafC5EJtlrYlwwqd3a9B18tgVNu6dPpa8GICoRy2BRWNfDDZzfy+b4KfjhvHLecOTEwxm9+9gSUZkPmKfbmXVMEceldn7PrPZBQm2kiAlmn2p/yXOtCyv8c5t9tn2zd89ih/dh9n8JHf7HN8T75m+3Rk3FC732vrW9Y66Yv3EsuYlNtVo4v1JbYSuXeICLWVoF/9iTU/L77/34umurgvzfZAsJTPdR2KD3Cr/+aRWSBiOwQkRwRuc3D/jEislpEvhCRr0XkHLd9tzvP2yEiR9aNSul1Nu4t57yHPmZ7QTWPXjmLXyyYHBjiUFMEH/zRtpY+6//ZbTnvdn/ervdsVlDHp/6kTJuX/t3lcMrNncXBnTHH2UE4iz+wcYItrx321/DIltfsyMm+fCqOTetZFlNvVvPO+a79O37+b9/Pef93tubhm3+F8MHT/t5f+O1ftIiEAo8AC4GpwOUi0rGv7a+AF40xM4FFwKPOc6c6308DFgCPOq+nBABLPt3HoifWEhMRyms/OimwspTe/bXtS7Tg9zDsaIhNt5ksXVFXBge+sMHB3mDEDNuls2Jv71wP7FSx3I9h2sV9m2EVm2ozp1pbuj7OGGtB9JaLCeyc6LHzYMPT3X8+wIEvbZfTWdd4rutQeow/H/nmAjnGmN3GmCbgeaBjy0MDuDpcJQDOaCAXAM8bYxqNMXuAHOf1lH6iudVBUVUDd7y2iTte28QJ41JZ+uOTmTQsgNpt538OX/yfbaKWOsGmgY4/w1oHjlbv5+1+HzC9JxAAiRlQ3osCseV1wMD0i3vvmr4Qm2Y/t76s6+Oaam3xXm8KBNheRVV5dmZ1V7S2wNKf2PUeziAgxSP+jEGMBPa7vc8DOs7yuwdYKSI/AWIBVwvIkcC6DueO7PgBIrIYWAwwZkwf5zkHMS+s38eqrUWU1zVRVttEaU0jVQ3tT3A/mjeOW86aRGhIANUKGGPTGmNT4RtuvucJZ9iAZf5G7zn1u96zAdHezAxKyoDdq3uv/cGWV20vobRJR36tnuByGdUWdx0H6K0aiI5MXAjxI2wsYvK53o/79DEo+Bq+9bTn2QrKYdHfQerLgaeNMfeJyAnAf0Rkuq8nG2OeAJ4AOzDIT2scVPzt3WzuX7WzbXbz1BFDSImNIDk2gpTYCKYMH8KcTP8PS+8xX79o+wld8MihcYSxp4GEQM47ngXCGGtBjD3VdufsLRIzrKurtqS9c+fhUrEf9n8Kp9/ZO2vrCb5WU9e6qqh7WSBCw2DOd2wzupIcOzynIxX77P4JZ9vMMqXX8KdA5AOj3d6Pcm5z5zpsjAFjzFoRiQJSfTxX6UWMMdy/aicPvZfDxTNH8udvHeO7hXDgS3h1sX1662p8or9orLEtq0fMgmOuOHRfTLINPmevsgNeOlKaY6unT7m5d9fkGkNZsffIBcIV7O5r9xK03/C7S3XtrTYbnph1jU082PAULPjdofuMgbd+Zn8/9y+BWwE/QPFnDGI9MEFEskQkAht0Xn2uAB0AACAASURBVNrhmH3AfAARmQJEAcXO4xaJSKSIZAETgM/8uNZBjTGGP6zYzkPv5XDZnNE9EweAPR9CyQ47xayx2n8L9cZH99m5Bgv/5Ln9xPgzbRDa001u13v2tTfjD2AtCLBpsr6w5mH4Y6Zt1Lb2ETj4dXsV8ZZXrfsreWzvrtEX2iyI7gSiFzq5eiN+KEw5H7581qaxurP1DRufOO2Xfd9OYxDgN4EwxrQANwBvA9uw2UpbROReETnfedgtwPUi8hXwHHCtsWwBXgS2AiuAHxtjuogyKoeLMYZ739zK4x/s5qrjx/D7i4/qeWyhNBvCoqFslw0UHu6c86Za6+4p3OL7Ofs/s1PRjl7kfQrZ+PmA8Zzuuus9e+NNyjyMBXeB62blaybTng/s360k246RfPwU+PNYWLLIitu0frAewLbakBAfXEy90Mm1K479nm2bsvnl9m0NlTbuNOxom5ig9Dp+jUEYY5YByzpsu8vt962AxzJJY8xvgd/6c32DHYfDcNfSzTy7bh/fOSmTu86bamc9f3SfbZZ24g2+Xagk2z7hTjzLDmYZffyhnS290VgD+9dB7iew9xMbSHa02BvSvNttJWxXbRI2/tu2uYgfDmf+2vtxw2fYJ+GcVXDMZe3bW5pgz0cwo5eHyABExtmbpa+ZTCXZMO4066arzLcFfns+stZZeEz/uJfAWmQxqd0LRF2JneMQ6aestowTIW2KDVbP/LZ1Jb17L9QWweXP9W78SGlD/6qDlPqmVu58YzMvb8zj+98Yy20LJltxaKiyTeiGjOyBQOyEyefBiTfaJ/qVv4SRs7xnDRXvhBW/sG0jTCuEhFmBOeEGO/Zy88s26Jj7EVz8ZOeOmK5WGhv+aYPQ//NU1xPPQkLsaMbslTbd1SU6eZ9Bc23vu5dcJGX4ZkG0NNnjjvqWfZ8wEo5ZZH/ApnD25w0wNrU9xuANV5sNf8UARODY6+wDQf5GO9p0/T+t5TByln8+U1GBGIx8tqeMn7/8Fbmlddw4fwI3nTHBigPYyVgtDVC+x3Yo7a4ata7M3hxcdQcXPgqPfwNeuha+/+GhaY8tTbav0Yd/sk/FJ99kC5pGzbVP3C4mng1Z34Blt8LfT4KLH7f1DADVhTbWsX+d7Zs0/27fbp4TzrQD5g98YcdEQnt7jcwuBrkcCYkZcODz7o8r32NveCkeMnSg/5+OY32wIGpL/edecnH0ZdZCXfd3KNpq+2D50mdLOWwCoDeC0lfUNbVwz9ItXPbEWlqNYcn3juOnZ05sFweAr1+wr8ZhLYPuKMm2r6kT7Wt0kp2eVVsCr3yvvUAtb6Mdqbj6/1lr44b1dlTjuNMPFQewT4uzvg2LV1vX0LOX2BvDvnX2Gge/stO3zvqN7zfPcadb11W2W1X1rveslRM1xPt5R0JShh1W31WRHthMKvCcwhkI+NJuozc7uXojaoi1qja/bAXinL/4z6WlACoQg4Z1u0tZ8OBHPL0ml6uPz2DFjady4vgOT3xVB63bZ4ozh6B4e/cXdolI6oT2bSNmwDl/soVi794LK+6Af55hWzYseg6+9S/fmq+lT4Hr37Npjh8/YGf+hobB91bBUf/j2xd3EZNsu7C62m7Ultr0XH+5l8BaEI4WqOomQ9slst4siP4mNq29zsEbtSX+tyAA5lxnXyefB5PP6fpY5YhRF1OQ09Dcyu+WbeOZtXvJSInh+cXHc/xYL096m18GjK0X2LHMjkvsjpKdNjjpSut0Mesa2930kwft+znXwRl397wFdkQMnP8325NnzwfWpdRVvKErxp8J7//e3sz2fECvt9foiKsWonxv1ymYpTm2Z1RvtwfvLWJTobESWhq9DwOqK/NPDURHhk6Fq9/ou3kYgxwViCCmrLaJxc9sYMPecr5zUia3nj2JmIgu/pN//YJ9yk6fYp9mfRKIbEge1znbSMTOAY5NtSMkM048si8z/eIjz+SZcIbt9rnrPSsQUb3cXqMjiW7FcnQR5yjNCVzrAQ4tlkvo1PHGxpYaK3u/zYY3xs7rm89RVCCClT0ltXznX59xoLKBR66YxblHd9NxtXArFGyyxWYAaZPh4Jfdf1BpNqR7qZ6OiLFxgkBh+Ez7lJu9yqbVjp3n32ljCaMB6T7VtTTHimig4t5uw5NA+LOKWulXNAYRhKzPLeOiRz+hqqGF564/rntxANj0os3ocRVkpU+xN7aOlavutDRB2Z72AHWg40p33fqGjQv4070Edq7wkJFdp7rWV9gbbyBbEC6BqPNSTa0CEbSoQAQZb3yZz5VPfkpyTASvXzed2QUvwTMX2ICsNxwO+PolW3Hs6huUNhkwtoWGN8r32DoG9wB1oDPhTNuWGmwNhb9J6qbtd+ku+xrQAtFNPyZ/dXJV+h0ViCDBGMPD72Vz4/NfcsHwUpaPfZkx/5oFy38Oe9fa2oH6Cs8n71tje+4f7VZlnD7FvhZ1kcnUluI6gARi3HxAbNwkKaPbw4+YpMyuLYhSVwZTAP8N3Vt+e8LfbTaUfkNjEEHC397ZyZ7VT7M66QOySrZARbRNBT32Omhthn8thDd+DJc927na9avnISIOJrmlDSaPtdlJxV0Eql0proF8c+tIbIqtsfAWN+ltEjNsI0FvRYelOda119u9oHqTyCH2/wVvAuFyMakFEXSoQAQBK7cU0Pj+fTwY8QImehyc+nvbXyg6qf2gM+6Blb+yg1WO/2H79uYG65Of8k0bVHYRGm5v/N1ZEPHD/Vdo5i/Of6jvPstlpVTu92xplWTbY8Ii+m5NPUWk61qIulJADv3/TQkKVCAGONmF1dz5whrejXiT1nFnEXrFC55bXp9wA+xdAyvvtK0tRs2223eugMaqQ91LLtInQ9567x9esjOwfeeBQKJbLYQngSjdNTD+hl2126gtseLgz4wwpV/QGMQAprK+mcX/2ch3w5YTZ2oJnf8rz+IA9inwgkfsE/9L19qqZrCT2OKGQdapnc9Jm2KndTXWdN5njPWfD5QMpv6ibXBQbud9DoezBmIAuOi66uhaV6LupSBFBWKA0uow3Pj8F1SVFXFd6HLbemD4MV2fFJNs21xUH4TXf2xdBtkrbazC09Nf+mT76imTqbbY9uNXgeiauGG2dbqnTKbqA9BSDynj+n5dPSU2zXsWU1806lP6BRWIAcr9q3bw/o5inp78GWHN1Z7HaXpi1Bw4817Y8RY8dxk4mj27l8BaEOC5otpTDyalMyEhkDjacybTQMoCi03tug7icNufKAGNCsQA5K2vD/LI6l1cNyue6XnPwbSLYOg03y9w/A+txZG33orAsKM8H5ecZZ9+VSCOjEQvtRCuLq4DIgaRBs11dupfR9TFFLRokHqAsb2gip+99BWzxiRye8IKpKnWTl/rCa54RF0ZzPmO9yEvIaHWheSpq2tJjh0zOmRUz7/EYCPJy1yI0hwIj7VxoUDHvRYiIrZ9u8PhbNSnAhGMqAUxgDDGcPurm4iNDOPxi0YTtuEfdgpZ2qSeXyw6Eb67HI6+tOvj0id7TnUt2WnnF3gLiivtJGbYpICGqkO3l+bY+IO/prD1Jm39mDq4mRoqbDW9ttkISvRf9wBi1dZCvthXwc/OmkjaV4/Z9svzbvPvh6ZNtlXWHW9uJTs1QO0rSe5dXd0ozRk4Ljpv7Ta0SC6oUYEYILQ6DH9ZuYOxqbH8z8RQO4/5mEX+z4Bxtdwodstkaq636a8DIT0zEGirhcht39bS6PwbDoD4Axza0dWdtjYbakEEIyoQA4Q3vsxnZ2ENN581kbA1D9pJZafe6v8PbhMIt0B12W7ADJyn3/7G1UbDPVBd1s0c6kAjxks/Jm3UF9SoQAwAmlocPPDOTqYOH8I5o1th49Mw40qbZeRvEjNtMNo9DtGWwaQuJp+IToKI+ENdTKUBPma0IxExNqDuzcWkFkRQogIxAHh+/T72l9Vz64JJhHzwB1vFfOrP+ubDQ0IgbeKhFkTbDOUBUOAVCIh0bvs9kFJcXXiqhdBOrkGNCkSAU9fUwt/ezWFuZjLzIrbDl8/CCT/uesZxb5M2pbMFkTD60HRHpWsSMw61IEpyIG7owGp0GJvmwcVUajsBe+pUqwx4VCACnH99kktJTSO/OCMTefOn1p/t78yljqRPtm0hXPMkSrI1/tBTkjJsUNoY+z7Q51B7wlPDvtoSdS8FMSoQAUxlXTOPf7CL0yenM3vfU/amct4DEB7dtwtxtdwo3m5vcCXapK/HJGXaSmTXDbY0e4AKRIeW33WlKhBBjF8FQkQWiMgOEckRkU6PvSLygIh86fzZKSIVbvta3fYt9ec6A5XHPtxFVUMLdxwr8PEDtmeSv+coe8LVtK9oG1QdgObagXdz62/c237Xldkb60D7G7pcTC4rCLTNRpDTbasNEfkm8JYxxtGTC4tIKPAIcCaQB6wXkaXGmK2uY4wxP3U7/ifATLdL1BtjZvTkM4OJoqoG/vXJHi44ehjj190BkXFw9u/6ZzEJYyA8xloQrpRNtSB6hqdiuYHmpotNs80dGyptJT5YiyK9B33AlAGFLxbEZUC2iPxJRCb34NpzgRxjzG5jTBPwPHBBF8dfDjzXg+sHLcYY/vz2DlpaDXeO2AD711lx6K8ntZAQ286jaFt79o0KRM9wJRWU5w7MDCZwq4Vwy2TSTq5BTbcCYYy5Cvtkvwt4WkTWishiEYnv5tSRwH6393nObZ0QkQwgC3jPbXOUiGwQkXUicqGX8xY7j9lQXOxlmMkAo7GllZtf/IqXNubxk7nxpK79rR3mc8zl/buwtCnWgijZaXP644f173oGGhGx9gm8Yq+NPwT6HGpPuB5QXKmuTbV2noW6mIIWn2IQxpgq4GWsFTAcuAj43OkW6g0WAS8bY1rdtmUYY+YAVwAPikinpHtjzBPGmDnGmDlpaWm9tJT+o7SmkSuf/JTXvsjnljMn8r9N/4CWBjjvwf5v6JY+GWoKYf+ntklff69nIOJq+12aY8UhNLy/V9QzOrbb0BqIoKdbgRCR80XkNeB9IByYa4xZCBwD3NLFqfnAaLf3o5zbPLGIDu4lY0y+83W387Nndj4teMgurObCRz9hU34lD18xk5+M2oVsfR2+cWtgFKSlT7WvB79S99LhkuSshSgZQE363OkoENpmI+jxxYK4BHjAGHOUMebPxpgiAGNMHXBdF+etByaISJaIRGBFoFM2kjOukQSsdduWJCKRzt9TgZOArR3PDRY+2FnMxY+uob7JwQvfP4Hzjh4BH/7ZNsM78cb+Xp4lzS38NBBvboFAYgZU5kHZroEXf4D2dFaX5VBXduh2JejwZWDQPcBB1xsRiQaGGmNyjTHvejvJGNMiIjcAbwOhwFPGmC0ici+wwRjjEotFwPPGuOfOMQV4XEQcWBH7g3v2UzDxn7W53PPfrUxIj+Of1x7LyERnjUP1QRg7D8Ii+nF1biSMsrGHpmq1IA6XpAzbZNHREhhWYU8Ji4CohHaB0E6uQY8vAvEScKLb+1bntmO7O9EYswxY1mHbXR3e3+PhvDWAlzmYwYExhvtX7eSh93KYPzmdv14+k7hIt/8c9eW2yVugIGIzmfI3aJvvw8VVCwED92/o3m5DXUxBjy8upjBnmioAzt8D5LF2gOBwwLb/2lfA4TDcvXQLD72Xw6VzRvH4t2cfKg7NDbbqNpAEAmygWkIgeWx/r2RgkuQuEAPQxQSHCkRtCYSEQ+QA6iel9AhfBKJYRM53vRGRC4CSLo5XOrJvDbxwFeR+SHOrg5+++CXPrN3L4lPH8sdLjiYstMN/hgZnQXmgCcSJN8JFj2tjtsMlYbQV2Ii4gZsmHJPiFoNwttnQjLagxRcX0w+A/xORhwHB1jZc7ddVBRtVBwBoqizkB//ZyHvbi/j5gkn88BvjEE//uOrL7WugCUTaRPujHB6h4TBkpC0sG6g31dg0m+oMViDUvRTUdCsQxphdwPEiEud8X+P3VQUbNUUAPPveF6wuieO3F03nyuMyvB8fqAKhHDnHXmcDvQOV2DQrDI5W7eQ6CPDFgkBEzgWmYaubATDG3OvHdQUVjRUHiQRqKor426KZfPOYEV2foAIRvJz80+6PCWRi0+yo1PpyG6QePmjbpQ0KfCmUewzbj+knWBfTt4AuHn8Vd4wxfL51BwAXT47pXhxABUIJXGJdtRDF6mIaBPgSpD7RGHM1UG6M+TVwAqCOaB95ek0uzZUFAIyKrPftJBUIJVBxVVNXH7RdXbXNRlDji0A0OF/rRGQE0Iztx6R0w9d5Ffxu2TbGRjnDNvVlvp1YX26buUV21w9RUfoYl0AU73C+1xhEMONLDOK/IpII/Bn4HDDAk35dVRBQ1dDMDUu+IC0ukhGh1VZWXZZBd7iK5AZqposSvLQJhHNGuQapg5ouBUJEQoB3jTEVwCsi8iYQZYyp7JPVDVCMMdz+yibyK+p58frZhPzbOaaxrgcWhLqXlEAkOgmQdgtCXUxBTZcuJucUuUfc3jeqOHTPs5/u461NB7n17EnMTnEABsKiem5BKEqgERJqrYaibfa9BqmDGl9iEO+KyCXisaJL6ciWA5X85s2tzJuUxuJTxkKtrYEgdQI0VkFrc/cXUYFQApnYtPZqf7UgghpfBOL72OZ8jSJSJSLVIlLl53UNSGoaW7hhyRckxYRz37eOISRE2orkSJtiX32xIlQglEDG3WrQ/0+DGl8qqTWVxkf+9fEecktref7640mJi7Qbawrta7pznkJdGcSld32h+gr9h6cELi6BiE6CUJ9qbZUBSrf/dUXkVE/bjTEf9v5yBjZvbTrInIwkjhvrltnhEgjXwJ3uUl1bm60rSgVCCVRcmUzqXgp6fJH/W91+jwLmAhuB0/2yogFKbkkt2wuqufO8qYfuqCmy7ZCHOCuou3MxNThzAFQglEClTSA0xTXY8cXF9E339yIyGnjQbysaoKzYYqulF0zv0Ma5ptC6lKKT7fvuUl21iloJdFwuJs1gCnp8CVJ3JA87ElRxY/nmAo4eldA+MtRFTRHEDbUtnqF7F5MKhBLouFxLakEEPb7EIB7CVk+DFZQZ2IpqxcmBinq+2l/BzxdM6ryzphCGTrdDYkLC1YJQBj4uF5NaEEGPLzGIDW6/twDPGWM+8dN6BiQrNlv30sLpHlpU1RTDuKG2bUZMcg8siMReXqWi9BIagxg0+CIQLwMNxphWABEJFZEYY0ydf5c2cFixuYDJw+LJSo09dEdzPTRWtqe1RierBaEMfJIyYdbVMOHs/l6J4md8qqQG3B3r0cA7/lnOwKO4upH1e8s6B6ehvUgubqh9jU6yNQ5dUV8OyMCeOqYEN6FhcP5DkDq+v1ei+BlfBCLKfcyo8/cY/y1pYLFyawHGeMhegs4C4auLKSrB9rxRFEXpR3wRiFoRmeV6IyKzAR8n3wQ/KzYXkJUay6ShHgrOXUVybS6mJN9cTOpeUhQlAPAlBnET8JKIHMCOHB2GHUE66Kmoa2LtrlKuP3UsHnsZtglEBwvCGO+zHlQgFEUJEHwplFsvIpMBVw7nDmOMDy1Jg59VWwtpcRgWenIvgdPFJG69a5KhtQmaaiEyzvM5KhCKogQI3bqYROTHQKwxZrMxZjMQJyI/8uXiIrJARHaISI6I3OZh/wMi8qXzZ6eIVLjtu0ZEsp0/1/TkS/UVKzYXMDIxmqNGegko1xZZqyE03L73pVhOBUJRlADBlxjE9c6JcgAYY8qB67s7SURCscOGFgJTgctF5JBGRcaYnxpjZhhjZgAPAa86z00G7gaOw/Z+ultEAuquWdPYwkfZJZw9bZhn9xK0V1G7cN34u4pD1Je3C4miKEo/4otAhLoPC3Le+CN8OG8ukGOM2W2MaQKeBy7o4vjLgeecv58NrDLGlDkFaRWwwIfP9D8OB6y4g5LnfkRTq4OFR3lxL0F7HyYXrn5M3hr2ORza6ltRlIDBF4FYAbwgIvNFZD72Jr7ch/NGAvvd3uc5t3VCRDKALOC9np7bpzgc8OZNsO4Rhu1dSlpcBLPHdHEzryk81ILozsXUWAkYFQhFUQICXwTiF9gb9w+cP5s4tHCuN1gEvOyq1vYVEVksIhtEZENxcXEvL6kDDge89VP4/N840qYSZer51sQQOzXOE8Y4XUweLAhvLibXdhUIRVECgG4FwhjjAD4FcrFuo9OBbT5cOx8Y7fZ+lHObJxbR7l7y+VxjzBPGmDnGmDlpaWk+LOkwcTjgrZth49Nw8s1smPILAM4ZVu39nMYqaGnwHIPw5mJyVVmrQCiKEgB4FQgRmSgid4vIdmwAeR+AMeY0Y8zDPlx7PTBBRLJEJAIrAks9fM5kIAlY67b5beAsEUlyBqfPcm7re9rE4V9w8k9h/l28edAWxU0JP+j9vI5V1ABhERAR792C0D5MiqIEEF1ZENux1sJ5xpiTjTEPAT67gIwxLcAN2Bv7NuBFY8wWEblXRM53O3QR8LwxxridWwb8Bisy64F7ndv6FocDlt3iJg5309RqeC27mbqQOEJLd3o/t2MVtYvoJO8xCBUIRVECiK4K5S7G3rxXi8gKbBaSF4e7Z4wxy4BlHbbd1eH9PV7OfQp4qief1+u8cxdseApOugnm3w0ifLKrmOqGVppHToCSbO/nuiyI2A4CEZPUhYtJBUJRlMDBqwVhjHndGLMImAysxrbcSBeRv4vIWX21wH7DGNj4DEy9AM64p601xopNBcRHhhE3agoU7/B+vicXE3Td8tslEFE6C0JRlP7HlyB1rTFmiXM29SjgC2xmU3BTX27TTkfNbROHllYHK7cWcPqUdELTJ9tKaW/WQE0hhIR1tga66uhaXw6RQ2w7ZUVRlH6mRzOpjTHlzsyh+f5aUMBQtse+Jme1bfpsTxnldc0smDYMUp2tqYq9xCFqiqx7KaTDn7g7C0InySmKEiD0SCAGFeUugRjbtmn55gKiwkP4xqQ0SJtoN5Z4E4jCzgFqsBZEQyU4PMT7tQ+ToigBhAqEN1wWRFImAA6H4e0tBcybmE5MRBgkZkBoJJR4iUN0rKJ2EZ0MGM+T5VQgFEUJIFQgvFG2G+KHQ7gtGv9ifzlF1Y3tvZdCQiFlfNcuJk8WRFuxnAc3kwqEoigBhAqEN8r3QFJ7/GH5pgIiQkM4fbLbTT9tomcLwuGA2mLPFkRMFw37VCAURQkgVCC8UbanLUBtjGH55gJOnpBKfFR4+zGpk6B8LzQ3HHpufRmYVi8WhJd+TMaoQCiKElCoQHiiqQ5qCtosiM35VeRX1LOg4+S41AmAgdKcQ7d7q6IGWygHnV1MjdVWVFQgFEUJEFQgPFGea1+dFsTyzQcJDRHOnNLBZZTmTHXt6GbqOIvaHW8WhFZRK4oSYKhAeKK8vQbCGMOKzQUcPzaZpNgOc5JSxgPSOVDtrYoaICoBJLSzBaECoShKgKEC4Ymy3fY1KYudhTXsLqllwfThnY8Lj4akjM61EF25mERsMZxaEIqiBDgqEJ4o22Of9GOSWbG5ABE4e5oHawAgdaIHgSiC8BiIiPN8TnRy5ywmFQhFUQIMFQhPuKW4Lt98kDkZSaTHR3k+NnWi7erqXhntqqIWL81vPfVjUoFQFCXAUIHwRNkeSB5Lbkkt2wuqPbuXXKRNgtZGqNjbvs1bFbWL6GSo82JBaCdXRVECBBWIjrS2QOV+SM5i+eYCgM7pre64mva5z4aoKYbYLkagerMgwmMg3IuloiiK0seoQHSkcj84WiApi7e3FHDMqARGJkZ7Pz51gn11nw3RrQWR5CFIXaHuJUVRAgoViI64MpiSs8gpqmHmmG5u2jHJ1lpw1UK0NFnroDuBaKmH5vr2bVpFrShKgKEC0RFnDURt7GhqGlsYOsQHl0/qpPZaiNpi++opxdVFjIdiORUIRVECDBWIjpTtgbAoDjrszXpYQmT356Q5U12N6bqK2kW0h4Z9KhCKogQYKhAdKc+FpEwKq5sAfLQgJkJDhbUeuqqidtHW0VUtCEVRAhcViI6U2RqIgkrboXWYrwIBNlDdVRW1i479mLSTq6IoAYgKhDvG2BhEchYFVU6BSPBBINqa9u2EWpcF4UMMwmVBNNfbWgoVCEVRAoiw/l5AQFFTCM11kJRFUUED8VFhdrxodwwZCeGxzjiEw7bpCOsiduESApcFoVXUiqIEIGpBuFPW3sW1oKrBN/cS2JYaqRPaXUxdxR/ANvkLi24XBhUIRVECEBUId9rafI+loKrRN/eSi7RJ1oKoKepeIMBZTa0CoShK4KIC4U7ZHpAQSBhNYWWDbxlMLlInQlW+LbTrKv7gIjpZXUyKogQ0fhUIEVkgIjtEJEdEbvNyzKUislVEtojIErftrSLypfNnqT/X2Ub5HkgYRWtIOMU1jb67mKA9UO2Liwns6NF6FQhFUQIXvwWpRSQUeAQ4E8gD1ovIUmPMVrdjJgC3AycZY8pFxP3Ru94YM8Nf6/NI2W5IyqKkppFWh2FoT1xMrlRX8N2CKNxif1eBUBQlAPGnBTEXyDHG7DbGNAHPAxd0OOZ64BFjTDmAMabIj+vpnjJnimtPaiBcJI+FEKfe+mJBRHewIEIjbfBaURQlQPCnQIwE9ru9z3Nuc2ciMFFEPhGRdSKywG1flIhscG6/0NMHiMhi5zEbiouLj2y1DZX2hp08tr0GoicCERpuRQJ8syBcQWqHo71IztuAIUVRlH6gv4PUYcAEYB5wOfCkiLgm5mQYY+YAVwAPisi4jicbY54wxswxxsxJS+ti/oIvuFJck7IodArEUF/6MLnjcjPF+uhiMg5orNQqakVRAhJ/CkQ+MNrt/SjnNnfygKXGmGZjzB5gJ1YwMMbkO193A+8DM/24VrcUV+tiCg0RUmMPUyB8TXMFKw4qEIqiBCD+FIj1wAQRyRKRCGAR0DEb6XWs9YCIpGJdTrtFJElEIt22nwRsxZ+0WRCZFFQ1kB4fSUhIkckMewAACmZJREFUD10+s74N8+/yPUgNdvSoDgtSFCUA8ZtAGGNagBuAt4FtwIvGmC0icq+InO887G2gVES2AquBW40xpcAUYIOIfOXc/gf37Ce/ULbbDv6JjKewqoc1EC6Sx8Ipt/gWS3Dvx6QWhKIoAYhfezEZY5YByzpsu8vtdwPc7PxxP2YNcJQ/19aJ8lxIygKgoLKBCenx/v08935M9eUQndj18YqiKH1MfwepA4eyPW1ZSIU9bbNxOLhcTDUF0FyrFoSiKAGHCgRAS6Ntk5GcRU1ji++jRo+E6ERAoHSX870KhKIogYUKBED5XsAckuLq06jRIyEk1LYFL9tt36tAKIoSYKhAwCEproXOKmq/WxBgA9VqQSiKEqCoQED7U3xS1uFVUR8u0clQfcD5uwqEoiiBhQoE2AB1RDzEpvZs1OiR4i4KKhCKogQYKhDgnEOdCSIUVvZg1OiR4qqFABUIRVECDhUIsBaEqwaiJ6NGjxRXqquEQqSf6y4URVF6iAqEoxUq9kKySyD6oAbChcuC0E6uiqIEICoQ1QXQ2tRmQfR41OiR4HIrubuaFEVRAoQ+cLQHOAkj4Q6bSdTqMD0fNXokuFsQiqIoAYYKBEBELAAlVQ121OgQPxfJuXAJgwqEoigBiLqY3CjoyyI5aA9Sq0AoihKAqEC40ac1EKAuJkVRAhoVCDcK+7KKGiAmBSTEviqKogQYGoNwwzVqNCWuj2IQEbFwxUswwr/TVBVFUQ4HFQg3CqsaSY+PJLSno0aPhAln9N1nKYqi9AB1Mblx2KNGFUVRghAVCDf6tM2GoihKgKMC4UZhZUPfZTApiqIEOCoQTmobW6jui1GjiqIoAwQVCCcFfTVqVFEUZYCgAuGkT0eNKoqiDABUIJz06ahRRVGUAYAKhJM+b7OhKIoS4KhAOOnTUaOKoigDABUIJwVaJKcoinIIfhUIEVkgIjtEJEdEbvNyzKUislVEtojIErft14hItvPnGn+uE5yjRlUgFEVR2vCbP0VEQoFHgDOBPGC9iCw1xmx1O2YCcDtwkjGmXETSnduTgbuBOYABNjrPLffXegsrGxg/PtVfl1cURRlw+NOCmAvkGGN2G2OagOeBCzoccz3wiOvGb4wpcm4/G1hljClz7lsFLPDXQttGjWoNhKIoShv+FIiRwH6393nObe5MBCaKyCcisk5EFvTgXERksYhsEJENxcXFh73Q0ppGWh1GXUyKoihu9HeQOgyYAMwDLgeeFJFEX082xjxhjJljjJmTlpZ22ItwpbhqkFpRFKUdfwpEPjDa7f0o5zZ38oClxphmY8weYCdWMHw5t9dwzaLWGghFUZR2/CkQ64EJIpIlIhHAImBph2Nex1oPiEgq1uW0G3gbOEtEkkQkCTjLuc0v9PmoUUVRlAGA37KYjDEtInID9sYeCjxljNkiIvcCG4wxS2kXgq1AK3CrMaYUQER+gxUZgHuNMWX+WmtBVR+PGlUURRkA+LVs2BizDFjWYdtdbr8b4GbnT8dznwKe8uf6XBRU9sOoUUVRlACnv4PUAYGOGlUURemMCgQ6alRRFMUTKhDoqFFFURRPDHqB0FGjiqIonhn0AtHY4uCbx4xg2ogh/b0URVGUgGLQDz9Ijo3goctn9vcyFEVRAo5Bb0EoiqIonlGBUBRFUTyiAqEoiqJ4RAVCURRF8YgKhKIoiuIRFQhFURTFIyoQiqIoikdUIBRFURSPiO24PfARkWJg7xFcIhUo6aXlDCT0ew8u9HsPLnz53hnGGI8zm4NGII4UEdlgjJnT3+voa/R7Dy70ew8ujvR7q4tJURRF8YgKhKIoiuIRFYh2nujvBfQT+r0HF/q9BxdH9L01BqEoiqJ4RC0IRVEUxSMqEIqiKIpHBr1AiMgCEdkhIjkiclt/r8efiMhTIlIkIpvdtiWLyCoRyXa+JvXnGnsbERktIqtFZKuIbBGRG53bg/17R4nIZyLylfN7/9q5PUtEPnX+//6CiET091r9gYiEisgXIvKm8/1g+d65IrJJRL4UkQ3ObYf9//qgFggRCQUeARYCU4HLRWRq/67KrzwNLOiw7TbgXWPMBOBd5/tgogW4xRgzFTge+PH/b+/uQqyqwjCO/x/UQDKyrEScZAiFILIpQvrwwoQiSjIosjCQEAIvwi76oG6CyIu66MPqpk+8sEIqy6tQUiooKiyzwi5KhJLRUWoqIazs6WKvqYMdozOeMzvPfn4wnL3XGYb3ZdbMu9faZ69Vfsf9nvdhYLHtC4Ah4GpJlwAPA4/Zngv8AKysMcZeWg3sajlvSt4AV9geann+Ydx9vdEFAlgAfG17t+1fgVeApTXH1DO23wW+P6p5KbCuHK8Drp/QoHrM9rDtT8rxz1T/NGbT/3nb9qFyOqV8GVgMvFra+y5vAEkDwLXAc+VcNCDvfzHuvt70AjEb+Lbl/LvS1iQzbQ+X433AzDqD6SVJg8CFwIc0IO8yzbIDGAG2AN8Ao7Z/L9/Sr/39ceAe4I9yPoNm5A3VRcBmSdsl3V7axt3XJ3c7ujhx2bakvvzcs6RpwGvAnbZ/qi4qK/2at+0jwJCk6cBG4NyaQ+o5SUuAEdvbJS2qO54aLLS9V9JZwBZJX7W+2Wlfb/oIYi9wdsv5QGlrkv2SZgGU15Ga4+k6SVOoisN626+X5r7Pe4ztUWAbcCkwXdLYhWE/9vfLgesk7aGaMl4MPEH/5w2A7b3ldYTqomABx9HXm14gPgbmlU84nATcDGyqOaaJtglYUY5XAG/WGEvXlfnn54Fdth9teavf8z6zjByQNBW4kur+yzbgxvJtfZe37ftsD9gepPp73mp7OX2eN4CkkyWdMnYMXAV8wXH09cY/SS3pGqo5y0nAC7bX1BxSz0h6GVhEtQTwfuAB4A1gAzCHarn0m2wffSP7hCVpIfAe8Dl/z0nfT3Ufop/znk91Q3IS1YXgBtsPSjqH6sr6dOBT4Fbbh+uLtHfKFNNdtpc0Ie+S48ZyOhl4yfYaSTMYZ19vfIGIiIj2mj7FFBERx5ACERERbaVAREREWykQERHRVgpERES0lQIR0QFJR8pKmWNfXVvkT9Jg60q7EXXLUhsRnfnF9lDdQURMhIwgIrqgrMP/SFmL/yNJc0v7oKStknZKelvSnNI+U9LGsl/DZ5IuKz9qkqRnyx4Om8tT0BG1SIGI6MzUo6aYlrW896Pt84GnqJ7OB3gSWGd7PrAeWFva1wLvlP0aLgK+LO3zgKdtnweMAjf0OJ+IY8qT1BEdkHTI9rQ27XuoNujZXRYH3Gd7hqSDwCzbv5X2YdtnSDoADLQu91CWI99SNnZB0r3AFNsP9T6ziH/KCCKie3yM4060rg90hNwnjBqlQER0z7KW1w/K8ftUq4oCLKdaOBCqrR9XwV8b+5w6UUFG/Fe5OonozNSyS9uYt2yPfdT1NEk7qUYBt5S2O4AXJd0NHABuK+2rgWckraQaKawChon4H8k9iIguKPcgLrZ9sO5YIrolU0wREdFWRhAREdFWRhAREdFWCkRERLSVAhEREW2lQERERFspEBER0dafrQz7a4+KKIIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# View change in accuracy and loss with each epoch\n",
    "hp.plot_model_train_val_accuracy_loss(model_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zocMFoV4eTje",
    "outputId": "b22cd346-39ed-4e53-98a0-b604c4470161"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predicted     0     1   All\n",
      "Actual                     \n",
      "0          4401   410  4811\n",
      "1          1148  3340  4488\n",
      "All        5549  3750  9299\n"
     ]
    }
   ],
   "source": [
    "# Calculate predicted values\n",
    "y_pred = model.predict(X_val_new)\n",
    "# Convert predicted 3-d array into 1-d array\n",
    "y_pred = np.argmax(y_pred, axis=1)\n",
    "# Convert numpy arrays to pandas series\n",
    "y_actu = pd.Series(y_val_binary, name='Actual')\n",
    "y_pred = pd.Series(y_pred, name='Predicted')\n",
    "\n",
    "\n",
    "# Print confusion matrix\n",
    "df_confusion = pd.crosstab(y_actu, y_pred, rownames=['Actual'], colnames=['Predicted'], margins=True)\n",
    "print(df_confusion)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 264
    },
    "id": "MZ2JfQZ0ekfJ",
    "outputId": "d6c64f52-13b8-4e4c-c3f8-00a3228c8b5e"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot confusion matrix\n",
    "conf = confusion_matrix(y_actu, y_pred)\n",
    "labels = ['SFW', 'NSFW']\n",
    "sns.heatmap(conf, annot=True, fmt=\"d\", cmap='Blues', xticklabels=labels, yticklabels=labels);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "YwJf932LewTu",
    "outputId": "9fffd658-abe8-46d5-c6cc-e0bd9a57ae5f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         SFW       0.79      0.91      0.85      4811\n",
      "        NSFW       0.89      0.74      0.81      4488\n",
      "\n",
      "    accuracy                           0.83      9299\n",
      "   macro avg       0.84      0.83      0.83      9299\n",
      "weighted avg       0.84      0.83      0.83      9299\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print classification report\n",
    "print(classification_report(y_val_binary, y_pred, target_names=labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "41o3y-Fje6TS",
    "outputId": "845c361a-0dac-4850-e5e9-ebb5dc6061c1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "sequential (Sequential)      (None, 90, 90, 1)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 88, 88, 48)        480       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 44, 44, 48)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 42, 42, 48)        20784     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 41, 41, 48)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 39, 39, 32)        13856     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 38, 38, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 36, 36, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 35, 35, 32)        0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 39200)             0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 128)               5017728   \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 72)                9288      \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 64)                4672      \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 30)                1950      \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 8)                 248       \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 3)                 27        \n",
      "=================================================================\n",
      "Total params: 5,078,281\n",
      "Trainable params: 5,078,281\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# View model summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sef0t7_oxY8M"
   },
   "source": [
    "### Model Analysis"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "machine_shape": "hm",
   "name": "05_binary_models_cont.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "name": "python3"
  },
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}