Step3_Modeling.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1764fbb0",
   "metadata": {},
   "source": [
    "# Step 3: Modeling\n",
    "\n",
    "In this notebook I train a series of convolutional neural networks and evaluate their performance on validation data:\n",
    " 1) Baseline model with 3 convolution layers and 3 dense layers\n",
    " 2) Baseline model with image augmentation added to pipeline\n",
    " 3) Model 2 but with max pooling and drop out layers added\n",
    " 4) Transfer learning model leveraging Resnet architecture and weights\n",
    " \n",
    "All models are evaluated based on validation accuracy (and model stability). Training is for 50 epochs with early stopping if the model does not improve after 10 epochs. The weights for the best performing epoch are then saved.\n",
    "\n",
    "This notebook was run on a Google Cloud VM with GPU access to speed up training/iteration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "b0aa7179",
   "metadata": {},
   "outputs": [],
   "source": [
    "#imports\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import os\n",
    "import tensorflow as tf\n",
    "\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.preprocessing import image_dataset_from_directory\n",
    "from tensorflow.keras.layers.experimental.preprocessing import Rescaling\n",
    "from tensorflow.keras.layers import Dense, Conv2D, Flatten, MaxPooling2D, BatchNormalization, Activation, Dropout, Concatenate\n",
    "from tensorflow.keras import Model, Input\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping\n",
    "from tensorflow.keras.applications.resnet50 import ResNet50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "56f5934f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# paths\n",
    "DIR = ('image_data')\n",
    "train_path = os.path.join(DIR, 'train_another')\n",
    "val_path = os.path.join(DIR, 'validation_another')\n",
    "test_path = os.path.join(DIR, 'test')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e2bc612",
   "metadata": {},
   "source": [
    "## Load Training/Validation/Test Data\n",
    "\n",
    "### Simple Loading\n",
    "First I load the training and validation sets as-is with no image augmentation. I'll use this for the first baseline model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "546323bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 10000 files belonging to 2 classes.\n",
      "Found 2000 files belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "image_size = (128, 128)\n",
    "batch_size = 32\n",
    "\n",
    "train_ds = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    train_path,\n",
    "    seed=42,\n",
    "    image_size=image_size,\n",
    "    batch_size=batch_size,\n",
    ")\n",
    "val_ds = tf.keras.preprocessing.image_dataset_from_directory(\n",
    "    val_path,\n",
    "    seed=42,\n",
    "    image_size=image_size,\n",
    "    batch_size=batch_size,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "74310e98",
   "metadata": {},
   "source": [
    "### Loading with Image Augmentation\n",
    "Below I leverage the keras ImageDataGenerator to add an image augmentation pipeline. AS the training dataset is rather small (10,000 images), this will create more variation in the training dataset to reduce model overfitting and improve performance.\n",
    "\n",
    "I limited image augmentation to a few types that are similar to the kinds of variation that might be caused by differences between instances of structures, as well as differences due to the how imagery aircraft fly.\n",
    "\n",
    "Note that we do not do any image augmentation for the validation data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "d84cbb55",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 10000 images belonging to 2 classes.\n",
      "Found 2000 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "image_size = (128, 128)\n",
    "batch_size = 100\n",
    "\n",
    "# TRAINING\n",
    "# specify training image transformations \n",
    "train_datagen = ImageDataGenerator(\n",
    "        rescale=1./255,\n",
    "        rotation_range=0.1,\n",
    "        zoom_range=0.1,\n",
    "        horizontal_flip=True)\n",
    "\n",
    "# keras generator for training data\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "        train_path,\n",
    "        target_size=image_size,\n",
    "        batch_size=batch_size,\n",
    "        class_mode='binary')\n",
    "\n",
    "\n",
    "# VALIDATION\n",
    "# specify validation image transformations (rescale only)\n",
    "val_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "# keras generator for validation data\n",
    "validation_generator = val_datagen.flow_from_directory(\n",
    "        val_path,\n",
    "        target_size=image_size,\n",
    "        batch_size=batch_size,\n",
    "        class_mode='binary')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7b131efc",
   "metadata": {},
   "source": [
    "# Functions for evaluating model performance"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a413ca9",
   "metadata": {},
   "source": [
    "As this is a binary classification problem with equal sized classes, I will use accuracy to evaluate the performance of the models. Below I define a function which plots the training and validation accuracy and loss over all epochs. This will enable me to inspect the training process visually. If training accuracy increases close to 1.0 but validation accuracy stays poor, I will know the model is overfitting. If validation accuracy "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ed1e4cda",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_nn_training(history):\n",
    "    \n",
    "    '''Given model training history, plots validation accuracy and loss'''\n",
    "    \n",
    "    # plot history for accuracy\n",
    "    plt.plot(history.history['accuracy'])\n",
    "    plt.plot(history.history['val_accuracy'])\n",
    "    plt.title('model accuracy')\n",
    "    plt.ylabel('accuracy')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.legend(['train', 'val'], loc='upper left')\n",
    "    plt.show()\n",
    "    \n",
    "    # plot history for loss\n",
    "    plt.plot(history.history['loss'])\n",
    "    plt.plot(history.history['val_loss'])\n",
    "    plt.title('model loss')\n",
    "    plt.ylabel('loss')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.legend(['train', 'val'], loc='upper left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "484135e1",
   "metadata": {},
   "source": [
    "## Baseline Model - No Image Augmentation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dca128b8",
   "metadata": {},
   "source": [
    "I start by training a simple baseline model using the sequential API.The model architecture includes 3 convolution layers (with batch normalization and ReLu activation), followed by 3 dense layers with ReLu activation, and then an ouput layer with Softmax activation. Training data is used as-is with no image augmentation. \n",
    "\n",
    "The baseline model is already performing well validation data, suggested that with some improvements this model could be highly accurate. Training accuracy is reaching 0.999 but validation accuracy is 0.9365. This suggests that the model is overfitting the training data. It is also training quite slowly. I will address both of these issues in later models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "c8b27275",
   "metadata": {},
   "outputs": [],
   "source": [
    "# create model\n",
    "model = Sequential()\n",
    "\n",
    "# scaling \n",
    "model.add(Rescaling(1.0 / 255))\n",
    "\n",
    "# convolution layers\n",
    "model.add(Conv2D(32, kernel_size=5, strides=2, padding=\"same\", input_shape=(128,128,3)))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Activation(\"relu\"))\n",
    "\n",
    "model.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Activation(\"relu\"))\n",
    "\n",
    "model.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Activation(\"relu\"))\n",
    "\n",
    "# flatten\n",
    "model.add(Flatten())\n",
    "\n",
    "# dense layers\n",
    "model.add(Dense(512, activation='relu'))\n",
    "model.add(Dense(256, activation='relu'))\n",
    "model.add(Dense(128, activation='relu'))\n",
    "\n",
    "# output layer\n",
    "model.add(Dense(2, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "b0cb4fb9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model1/best_model.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "4412d416",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x16ae98700> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x336ac9b80> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50450, saving model to models/Model1/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.50450 to 0.76150, saving model to models/Model1/best_model.h5\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.76150 to 0.88850, saving model to models/Model1/best_model.h5\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.88850 to 0.89600, saving model to models/Model1/best_model.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.89600\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy did not improve from 0.89600\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.89600\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy improved from 0.89600 to 0.90750, saving model to models/Model1/best_model.h5\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.90750\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy improved from 0.90750 to 0.93650, saving model to models/Model1/best_model.h5\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.93650\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.93650\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.93650\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.93650\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.93650\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.93650\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.93650\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.93650\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.93650\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.93650\n",
      "Epoch 00020: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results1 = model.fit(\n",
    "    train_ds,\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=val_ds,\n",
    "    verbose=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aa0dcb9f",
   "metadata": {},
   "source": [
    "**Accuracy:** 0.93650"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "28647cb3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "#plot performance by epoch \n",
    "plot_nn_training(results1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1358fbba",
   "metadata": {},
   "source": [
    "## Baseline Model - With Image Augmentation\n",
    "\n",
    "The difference betwen training and validation accuracy may be because the model is overfitting to the relatively small training dataset (there are only 10,000 images). This can be addressed by image augmentation (see above for the image augmentation pipeline).\n",
    "\n",
    "Below I train the same baseline model, but use the image augmentation pipeline developed above to create variation in the training data as it is fed into the model each epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "d8e54eb2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# create model\n",
    "model_img = Sequential()\n",
    "\n",
    "# scaling \n",
    "#model_img.add(Rescaling(1.0 / 255))\n",
    "\n",
    "# convolution layers\n",
    "model_img.add(Conv2D(32, kernel_size=5, strides=2, padding=\"same\", input_shape=(128,128,3)))\n",
    "model_img.add(BatchNormalization())\n",
    "model_img.add(Activation(\"relu\"))\n",
    "\n",
    "model_img.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
    "model_img.add(BatchNormalization())\n",
    "model_img.add(Activation(\"relu\"))\n",
    "\n",
    "model_img.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
    "model_img.add(BatchNormalization())\n",
    "model_img.add(Activation(\"relu\"))\n",
    "\n",
    "# flatten\n",
    "model_img.add(Flatten())\n",
    "\n",
    "# dense layers\n",
    "model_img.add(Dense(512, activation='relu'))\n",
    "model_img.add(Dense(256, activation='relu'))\n",
    "model_img.add(Dense(128, activation='relu'))\n",
    "\n",
    "# output layer\n",
    "model_img.add(Dense(2, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "9e7296da",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model_img.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model1/best_model_img.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "9ee1bf8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x14843dd30> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x14bd10550> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50000, saving model to models/Model1/best_model_img.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.50000\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.50000\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.50000 to 0.63900, saving model to models/Model1/best_model_img.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.63900\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.63900 to 0.89650, saving model to models/Model1/best_model_img.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.89650\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.89650\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.89650\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy improved from 0.89650 to 0.92600, saving model to models/Model1/best_model_img.h5\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.92600\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.92600\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy improved from 0.92600 to 0.94400, saving model to models/Model1/best_model_img.h5\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.94400\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.94400\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.94400\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.94400\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.94400\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.94400\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.94400\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.94400\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.94400\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.94400\n",
      "Epoch 00023: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results_img = model_img.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=20, # num_samples//batch_size\n",
    "    verbose=4) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0c9328a",
   "metadata": {},
   "source": [
    "**Accuracy:** 0.94400"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "975d50bd",
   "metadata": {},
   "source": [
    "## Improving Model Performance"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0bfa3c0a",
   "metadata": {},
   "source": [
    "### Adding Max Pooling and Drop Layers\n",
    "\n",
    "During training of the baseline model, training validation scores regularly exceeded 0.99, while the validation accuracy was substantially lower. This suggested that the model was overfitting, even with variation in training data introduced by the image augmentation. Two common ways to reduce overfitting for convolutional neural networks are to 1) add max pooling layers after each convolution layer and 2) add dropout layers after each dense layer. As an added benefit, both these techniques also increase training efficiency. \n",
    "\n",
    "Below I try a few iterations with different hyperparameters for max pooling and dropout. From this I learn that increasing convolution kernel size decreases performance, but reducing the number of nodes in the dense layers by 50% does not substantially decrease performance. \n",
    "\n",
    "The models achieve much higher accuracy than the baseline models but are unfortunately not converging, which suggests instability/lack of generalizability. I address that is subsquent iterations further down."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "16d487e7",
   "metadata": {},
   "source": [
    "### Model 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "28402af5",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_model2(input_shape, num_classes):\n",
    "    \n",
    "    '''CNN model architecture with max pooling + dropout layers'''\n",
    "    \n",
    "    inputs = Input(shape=input_shape)\n",
    "    \n",
    "    #x = Rescaling(1.0 / 255)\n",
    "    \n",
    "    # convolution (with max pooling)\n",
    "    x = Conv2D(32, kernel_size=5, strides=2, padding=\"same\")(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x) \n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "\n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    #flatten\n",
    "    x = Flatten()(x)\n",
    "    \n",
    "    #dense layers (with dropout to avoid overfitting)\n",
    "    x = Dense(1024, activation='relu')(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    x = Dense(512, activation='relu')(x)\n",
    "    x = Dropout(0.2)(x)\n",
    "    x = Dense(256, activation='relu')(x)\n",
    "    x = Dropout(0.1)(x)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(2, activation='softmax')(x)\n",
    "    \n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "ecc998e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model2/best_model.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "model2 = generate_model2(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "model2.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "91a1dd90",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x14838fdc0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x16c3a6af0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50000, saving model to models/Model2/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.50000\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.50000\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.50000 to 0.50050, saving model to models/Model2/best_model.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy improved from 0.50050 to 0.73950, saving model to models/Model2/best_model.h5\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy did not improve from 0.73950\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.73950 to 0.88800, saving model to models/Model2/best_model.h5\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.88800\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.88800\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.88800\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.88800\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.88800\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.88800\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.88800\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.88800\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy improved from 0.88800 to 0.94300, saving model to models/Model2/best_model.h5\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.94300\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.94300\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.94300\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy improved from 0.94300 to 0.95000, saving model to models/Model2/best_model.h5\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.95000\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.95000\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.95000\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.95000\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.95000\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy did not improve from 0.95000\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.95000\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.95000\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.95000\n",
      "Epoch 30/50\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.95000\n",
      "Epoch 00030: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results2 = model2.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=20, # num_samples//batch_size\n",
    "    verbose=4) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f81362a4",
   "metadata": {},
   "source": [
    "**Accuracy:** 0.9500"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "1ca4629e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEWCAYAAABollyxAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8/fFQqAAAACXBIWXMAAAsTAAALEwEAmpwYAAA8w0lEQVR4nO2deZwcZbX3v6dnz0z2zAyThCwgEpJAFkLYRBFR2QQFBFQQ0Vf06r2K20W9rvfq1Xtfry/uCoKCbGKAiwLKIpsEQpZJWJOwZk+Ymawzk9m6+3n/OFUzPZPpme6erl4q5/v5zKeqq6rreWq6+1enznPOecQ5h2EYhhFeIvnugGEYhhEsJvSGYRghx4TeMAwj5JjQG4ZhhBwTesMwjJBjQm8YhhFyTOgNAxCR34vI91I8doOInD7S8xhGrjChNwzDCDkm9IZhGCHHhN4oGjyXyVdE5DkRaReR60WkXkT+KiKtIvKwiIxPOP5cEXlRRPaIyGMiclTCvgUi0ui9749A5YC2zhGRNd57nxKRYzLs8ydF5FUR2SUifxaRyd52EZH/JyJNIrLXu6a53r6zROQlr29bReTLGf3DDMPDhN4oNi4A3g28FXgf8Ffg68Ak9Pv8OQAReStwG3AVUAvcD/xFRMpFpBz4X+APwATgT9558d67ELgB+BQwEfgN8GcRqUinoyJyGvAD4CKgAdgI3O7tfg/wdu86xgEXAzu9fdcDn3LOjQbmAo+k065hDMSE3ig2fuace9M5txX4B/CMc261c64LuBtY4B13MXCfc+4h51wP8COgCjgJOAEoA65xzvU455YAKxLa+CTwG+fcM865mHPuRqDLe186fAS4wTnX6PXva8CJIjID6AFGA7MAcc6tdc5t997XA8wWkTHOud3OucY02zWMfpjQG8XGmwnrHYO8rvHWJ6MWNADOuTiwGZji7dvq+lf025iwPh34kue22SMie4BDvfelw8A+tKFW+xTn3CPAz4FfAG+KyLUiMsY79ALgLGCjiDwuIiem2a5h9MOE3ggr21DBBtQnjor1VmA7MMXb5jMtYX0z8H3n3LiEv1HOudtG2Idq1BW0FcA591Pn3LHAHNSF8xVv+wrn3HlAHepiuiPNdg2jHyb0Rli5AzhbRN4lImXAl1D3y1PA00AU+JyIlIrI+cDihPdeB3xaRI73Bk2rReRsERmdZh9uBa4Qkfmef/8/UVfTBhE5zjt/GdAOdAIxbwzhIyIy1nM57QNiI/g/GIYJvRFOnHPrgUuBnwEt6MDt+5xz3c65buB84GPAbtSff1fCe1eifvqfe/tf9Y5Ntw9/B74J3Ik+RRwOXOLtHoPeUHaj7p2d6DgCwGXABhHZB3zauw7DyBixiUcMwzDCjVn0hmEYIceE3jAMI+SY0BuGYYQcE3rDMIyQU5rvDiQyadIkN2PGjHx3wzAMo2hYtWpVi3OudqhjCkroZ8yYwcqVK/PdDcMwjKJBRDYOd4y5bgzDMEKOCb1hGEbIMaE3DMMIOQXlox+Mnp4etmzZQmdnZ767EiiVlZVMnTqVsrKyfHfFMIyQUfBCv2XLFkaPHs2MGTPoX2wwPDjn2LlzJ1u2bGHmzJn57o5hGCGj4F03nZ2dTJw4MbQiDyAiTJw4MfRPLYZh5IeCF3og1CLvczBco2EY+aEohL6gifVAx+5898IwDCMpJvTDsGfPHn75y18mP6C9GXZvgHj/uSHOOuss9uzZE2jfDMMwUsGEfhiSCX0s5gl7T4cuXX+hv//++xk3blzAvTMMwxieQKNuRGQD0IpOhRZ1zi0Ksr0g+OpXv8prr73G/PnzKSsro6amhoaGBtasWcNLL73E+y+9ks1bt9EZFT5/1VVceeWVQF85h7a2Ns4880ze9ra38dRTTzFlyhTuueceqqqq8nxlhmEcLOQivPKdzrmWbJzou395kZe27cvGqXqZPXkM337fnKT7f/jDH/LCCy+wZs0aHnvsMc4++2xeeOEFDYOMx7jhR99kwvixdIyaynFveycXXHABEydO7HeOV155hdtuu43rrruOiy66iDvvvJNLL7XZ4QzDyA0FH0dfaCxevLgv1j3ayU9vuI27//oolJazefNmXnnllQOEfubMmcyfPx+AY489lg0bNuS204ZhHNQELfQOeFBEHPAb59y1Aw8QkSuBKwGmTZs25MmGsrxzRXV1de/6Y488zMP/WM7Tf/k9oxpmcepZ5w8aC19RUdG7XlJSQkdHR076ahiGAcEPxp7snFsInAl8VkTePvAA59y1zrlFzrlFtbVDllTOC6NHj6a1tXXQfXt37WT82NGMqqpi3bp1LFu2LMe9MwzDGJ5ALXrn3DZv2SQidwOLgSeCbDPbTJw4kZNPPpm5c+dSVVVFfX19774zTj2eX1//O445/SKOPHIWJ5xwQh57ahiGMTjinAvmxCLVQMQ51+qtPwT8u3Pub8nes2jRIjdw4pG1a9dy1FFHBdLHEbPjBaiogY49UFMLY6aM6HQFfa2GYRQkIrJquIjGIC36euBuL7W/FLh1KJEvOuJRiPdAaSVESg5ImDIMwygUAhN659zrwLygzp93ol26LKsCMaE3DKNwsczYTPEzYn2L3pnQG4ZRmJjQZ0q0EyQCJeVm0RuGUdCY0GdKT6da8yJm0RuGUdCY0GdKtEOFHrzB2Hh++2MYhpEEE/pMiEU16qbME/oE101NTU0eO2YYhnEgJvSZEPXKHJR6FSgjJUAcnFn1hmEUHlbUbBiuvvpqpk+fzmc+8xkAvvOd7yA9+3ni8cfY3d5DTzTK977xFc475WgbkDUMoyApLqH/61dhx/PZPechR8OZP0y6+5JLLuGqq67qFfo77riDv/3xBr5w+XmMOeIkWnbu5ITjF3PuE0sQG5A1DKMAKS6hzwMLFiygqamJbdu20dzczPjx42mYNIYvfOP7PLHin4hEImzdtp03m3dySK0JvWEYhUdxCf0QlneQXHjhhSxZsoQdO3ZwySWXcMsdd9G8ex+rVq2irKyMGdOn09nVba4bwzAKEhuMTYFLLrmE22+/nSVLlnDhB85j79691NXVUVZWxqOPPsrGTZv0QHPdGIZRgJjQp8CcOXNobW1lypQpNNSO5yPnn8nK1c+zaNEibrnlFmbNOlIPNIveMIwCpLhcN3nk+ee9QeC2ZiZNGM/TTy2FkjLdFo/BjufAxWhra8tfJw3DMAbBLPp0iXZoglQk4R4p3r/RLHrDMAoQE/p06enUjFits6+IqPhbwpRhGAVIUQh9ULNgpY1zmhXrZ8QmMsLJRwrmGg3DCB0FL/SVlZXs3LmzMIQw3qORNX4xs0SkROvfZIBzjp07d1JZOch5DcMwRkjBD8ZOnTqVLVu20NzcnO+uqNumvQlqgNKW/vvamnTZ1J3RqSsrK5k6derI+mcYhjEIBS/0ZWVlzJw5M9/dUJ7+BTzwdfjKa1A9qf++2/4D9myEf1qan74ZhmEkoeBdNwVF00swatKBIg9QORY69+a+T4ZhGMNgQp8OTeug7qjB91WOhc59ue2PYRhGCpjQp4pz0DyU0I+Brn0205RhGAWHCX2q7N0M3W1DW/Q46G7NabcMwzCGw4Q+VZrW6bI2idBXjNGl+ekNwygwTOhTpeklXdbNGnx/5VhdmtAbhlFgmNCnSvM6GN0AVeMH398r9DYgaxhGYWFCnypNa6E2iTUPOhgLZtEbhlFwmNCnQjwOzeuhbnbyY8x1YxhGgWJCnwp7Nmh54mT+eYDKcbrsMteNYRiFReBCLyIlIrJaRO4Nuq3AGC7iBqBitC7NojcMo8DIhUX/eWBtDtoJDj/ipvbI5MeUlEFZtQm9YRgFR6BCLyJTgbOB3wbZTuA0r4Oxh/YNuCbD6t0YhlGABG3RXwP8K1DcdQGGi7jxqRxjQm8YRsERmNCLyDlAk3Nu1TDHXSkiK0VkZUHUnB9ILAotLycvfZCIWfSGYRQgQVr0JwPnisgG4HbgNBG5eeBBzrlrnXOLnHOLamtrA+xOhux+A2LdqQu9Rd0YhlFgBCb0zrmvOeemOudmAJcAjzjnLg2qvcDoLX2QgtBXmOvGOAiIdkPH7nz3wkgDi6MfjqZ1gMCkISJufKwmvXEwsPQa+PUp+e6FkQY5EXrn3GPOuXNy0VbWaV4L46dD+ajhj/UHYwthInPDCIqWV7yy3fvz3RMjRcyiH46mtUOXPkikcizEe6CnI9g+GUY+aW/qvzQKHhP6oYh2w85XUwuthL56NzYga4SZ9hZdtpnQFwsm9EOx81WIR1O36G3yEeNgwBf4tjfz2w8jZUzoh6LZq9wwVDGzRPzCZib0RliJx2G/b9Gb0BcLJvRD0bQOJAITj0jteJt8xAg7HbvBeYnurSb0xYIJ/VA0vQQTDoeyytSO7518ZE9gXTKMvJI4AGsWfdFgQj8UzetSd9uATT5ihJ/2hDIlNhhbNJjQJ6OnE3a9PnQN+oFY1I0RdnxxHzPVLPoiwoQ+GS0vqy8yldIHPqWVECkzi94IL35oZf0cs+iLCBP6ZDR7s0qlI/QiVgbBCDftzSAlOglP25uWBV4kmNAno+kltc4nHJ7e+6wmvRFm2pugehKMbtAscCtuVhSY0CejaR1MfAuUlqf3PqtJb4SZ9haoroOaOn1t7puiwIQ+Gc1r04u48bGa9EaYafMt+kO81zYgWwyY0A9G+07YvQEOOTr991pNeiPMtDdDdS3U1Otrs+iLAhP6wdi4VJfT35b+e811Y4SZ9hZ12/S6bsyiLwZM6Adj41IorYLJC9J/r0XdGGGlux162tV1UzFGw4lN6IsCE/rB2LAUDl2c/kAsqND3tEOsJ/v9Mox84mfFVtdpKHFNnQl9kWBCP5D9u+DNF2BGBm4bsMJmRnhp84W+Vpc19Sb0RYIJ/UA2PQ24kQt9l/npjZDRa9FP0mVNvQ3GFgkm9APZsFR9j1OOzez9NvmIEVZ8ofcHYs11UzSY0A9k45Mw9Tgorcjs/VbB0ggrfoniUQkW/f6dNh5VBJjQJ9KxB7Y/l7nbBhJq0puP3ggZ7S1QMbZvfgbfsk8sXWwUJCb0iWxaBjiYfnLm5zCL3ggrflasT2/SlLlvCh0T+kQ2Pgkl5TB1UebnsJr0Rljxs2J9LDu2aDChT2SD558vq8r8HOWjATGL3ggf7S1QM5jQm0Vf6JjQ+3Tug+3PjsxtAxCJWL0bI5y0Nw2w6K0MQrFgQu+z+RmdUWrGCIUerAyCET5iUU0mrK7r21ZaAZXjzHVTBJjQ+2z4h040MnXxyM9lk48YYWP/TsD1H4wFy44tEkzofTYs1SSp8lEjP5dVsDTCRvuA8gc+NXXQakJf6JjQA3S1wbbV2XHbgDf5iAm9ESIGZsX6mEVfFAQm9CJSKSLLReRZEXlRRL4bVFsjZvMycLGRD8T62GCsETaSWvRW76YYCNKi7wJOc87NA+YDZ4jICQG2lzkblkKkFA49PjvnM9eNETaGct30tOtTsVGwBCb0TvE//TLvzwXV3ojYuFQnGamoyc75KsdAVyvE49k5n2Hkm7YmTSb0EwJ9LJa+KAjURy8iJSKyBmgCHnLOPTPIMVeKyEoRWdncnIeaGd3tsLUxe24b0B+Di0O3WTlGSGhvUWtepP/23lh6c98UMoEKvXMu5pybD0wFFovI3EGOudY5t8g5t6i2tvaAcwTO5uUQ7xlZIbOBWBkEI2y0Nx8YWglm0RcJOYm6cc7tAR4DzshFe2mxcSlISfb882A16Y3wMTAr1sfq3RQFQUbd1IrIOG+9CjgdWBdUexmzYSk0zOsrL5wNrIKlETbaW/pnxfqMmqiGkln0BU2QFn0D8KiIPAesQH309wbYXvr0dMDWldmLn/exeWONMOHcgSWKfSIRm2mqCCgN6sTOueeABUGdPytsWQGxbphxSnbPaxa9ESa6WiHWdWCylE9NnbluCpyDOzN2w1KQCEzLcni/Cb0RJpLF0PvU1EPbjtz1x0ibg1zon4RDjj4wNnik+IOxVgbBCAO9Qj+I6wbMoi8CDl6h7+lU1830LIZV+pSWQ2mVWfRGOOgV+mSuG68MgiUIFiwHr9BvXaV+x2zGzydiZRCMsOBb60O5blwMOnblrk9GWhy8Qr/hSUBg+onBnL9yjEXdGOGgvUWXQ7luwCJvCpiUhF5EPi8iY0S5XkQaReQ9QXcuUDY+CfVzoWp8MOc3i94IC+1N+jspKRt8v2XHFjypWvQfd87tA94D1AJXAD8MrFdBE+2CzSuCc9uACb0RHtqbk7ttwLJji4BUhd6vZHQW8Dvn3LMJ24qPrY0Q7ch+olQiFWOs1o0RDpJlxfqY66bgSVXoV4nIg6jQPyAio4HiHWLf+KQup50UXBtm0RthIVlWrE95DZSNMou+gEk1M/YT6OQhrzvn9ovIBNR9U5xsWAp1c6B6YnBtVI61wVgjHLQ3J8+KBS1dbFMKFjSpWvQnAuudc3tE5FLgG0BxmquxHtj8TLBuG9Com1iXxusbRrES7YbOPUP76MGEvsBJVeh/BewXkXnAvwIbgZsC61WQbFsNPfuzO9HIYFgZBCMM7B8mtNLHsmMLmlSFPuqcc8B5wE+ccz8BRgfXrQDZ4PnnAxf6cbq0AVmjmBkuK9anph5ard5NoZKqj75VRL4GXAacIiIl6BywxcfGpVA7C2oCns3KJh8xwkDbMAXNfGrq1cUT7YLSisC7ZaRHqhb9xUAXGk+/A5gC/N/AehUUsShsWha8NQ8Jrps9wbdlGEHhW/TDGUY2d2xBk5LQe+J+CzBWRM4BOp1zxeej3/6sTtgd9EAs2OQjRjhoH6bOjY8lTRU0qZZAuAhYDnwQuAh4RkQuDLJjgbBxqS6DqFg5kEpz3RghoL0ZSis1Vn4oLGmqoEnVR/9vwHHOuSbQ+WCBh4ElQXUsEDYtgwmHwej64NuyqBsjDPhZsTJMIrzVuyloUvXRR3yR99iZxnsLA+c0fv7QLM8mlYyyUTppskXdGMXMcFmxPr5rx1w3BUmqFv3fROQB4Dbv9cXA/cF0KSB2vqYxwdOOz017IlYGwSh+2pthzOThjysth6oJZtEXKCkJvXPuKyJyAXAyWszsWufc3YH2LNtsXqbLQ3Mk9GBlEIzip70ZGo5J7VjLji1YUrXocc7dCdwZYF+CZdMyFd5JR+auzcoxZtEbxYtzXoniYZKlfCw7tmAZUuhFpBVwg+0CnHNuTCC9CoLNz6g1H8nh0IK5boxipnMPxKPDh1b6jD5EDSqj4BhS6J1zxVnmYCD7d0HLy3DMxbltt3Ksjg0YRjGSalasj2/ROzd8lI6RU4orciZTNi/X5bQcRdz4VJhFbxQxqWbF+tTU64Q+FmlWcBwkQr8MIqUweWFu2zXXjVHMpJoV62PZsQXLwSH0m56BhnlQPiq37VaO1ZILsWhu2zWMbNDulyhOYzAWLPKmAAm/0Ee7YVtj7hKlEvHLINijrFGMtDcDAqMmpHa8ZccWLOEX+u3PQrQzd4lSiVgZBKOYaWuCURMhUpLa8ea6KVjCL/S9iVJ5sOgrzKI3ipjh5oodSOU4iJSZRV+ABCb0InKoiDwqImtF5EUR+XxQbQ3JpmUwfkZuCpkNxCx6o5hpb06tzo1PJGJJUwVKkBZ9FPiSc+4o4ATgsyIyO8D2DiTXhcwGYkJvFDPpZMX61NSZRV+ABCb0zrntzrlGb70VWIvOTJU7dr2uX9Z8+OchoSZ9mq6bllfh6V/ojcow8kV7S+qhlT5B17uJx+DeL8CbLwbXRgjJiY9eRGYAC4BnBtl3pYisFJGVzc3N2W14s9dcLguZJZKpRb/8N/DA180yMvJHT6eOLaXjuoHgXTc7X4WVN8CaW4NrI4QELvQiUoMWQ7vKOXeAaeucu9Y5t8g5t6i2NssTdm9aptmptUdl97ypkulg7NZGXZrVYuSL3qzYdF03h+h747Hs9wmgeZ0ut60O5vwhJVChF5EyVORvcc7dFWRbg7J5ORx6XG4LmSUSKYHy0elZ9NFu2PG8rje9FEy/DGM40s2K9ampAxeH/Tuz3yeA5pd1uW1NcDeTEBJk1I0A1wNrnXM/DqqdpHTshua1+RuI9Um3DELTSxDr0vU3TeiNPJFuVqyPH0vfuiO7/fFpWa/LnnZoXh9MGyEkSFP3ZOAy4DQRWeP9nRVge/3ZvEKX+RqI9UlX6P1H0gmHQZO5bow84btu0vbRB5w01bwexk3X9W2NwbQRQoKMunnSOSfOuWOcc/O9v9xNP7h5mc7ZOuXYnDU5KOlOPrKtEarGw1vP1C+1PZ4a+aBtBK4bCCaQIB6HllfgyLN0/GurCX2qhDczdtMzOgVaeXV++5GuRb91NUxeAPVztHTDrteD61sydr0O3ftz365ROLS3QHlN+oUAgxT6vZu0DHLdLC1SaBZ9yoRT6GM9sHVV/v3zoJZHqlE33fvVRz95IdR7uWW5jrxpfRN+cQIs/Ulu2zUKi/am9N02oIZV+ehgXDf+QGztLJiyEHa8ANGu7LcTQsIp9Nuf0zt/vv3zkJ5F/+YL4GL6Ja6dBRLJfeTN6j/oYPCW5blt1ygsMsmK9QkqO9YPrZz0VjWG4j36mzGGJZxCn89CZgOpHKuZsalkufo+x8kLoaxKB2RzadHHY7Dq97q+/VnLzD2YySQr1qemPhiLvmW99mnUBDWGwPz0KRJOod+0DMZNgzEN+e6JDsa6GHS3D3/stkZNOPH7XTc7txb9qw/D3s0w4xSNg963NXdtG4VFW4auGwjQon9Zn3QBxh4KoyZZ4lSKhE/o813IbCDplEHY2thnqYAOyO56I7WbRDZYcb1aY+/8ur7e/mxu2jUKi3gc9reknxXrE4RF75xGoU16q74W0d+KWfQpET6h37NRrYlC8M9Dn9APNyDbuQ92vtJ/Xtu62YDr800Gye6N8MqDsPCj0DBfxwe2rQm+XaPw6Nil2a0Zu27qoGsv9HRkr09tb+o5a4/s2zZ5obpzutqy105ICZ/Qb8pzIbOB+PVuhrPot6/R5ZQFfdvq5+gyF376xhvVSlp4uYbUTTrSLPqDld5kqQyFfvQhusymVe9nwfoWPahF7+L2PU2B8An95mUqrnW5LX2flMpxuhxO6P1H0IYEoR8/E8pGBV8KIdoNjTfBEe+FcYfqtsnz7Qd0sDJSoQ9i7tiWhNBKH//p1+LphyV8Qr/pGZi6KPV5LoOm10c/jOtmW6OmdldP7NsWiegXO+hSCOv+oj/u4z7Rt61hHrTtCK5miVG4+JZ4xj76AJKmmtepAec/LQDU1MLYaeanT4FwCX3HHo1SKZSBWEiYfGTP0MdtW91/INanfnbwFv3K32mU0uGn9W1rmKfL7c8F27ZRePQWNCsgi94fiBXpv33KArPoUyBcQr9lJeAKZyAWUvPRt7fAnk39B2J96uZoBESQRaI2/AOOvaL/U9AhRwPSN3ZgHDy0N2mdKN/tmC6jJgGS3e9sy8v9B2J9Ji+E3Rtg/67stRVCwiX0vYXMFuW7J32UVUJJxdBRN34s8OQFB+4LuhTCyhsgUgYLLuu/vWI0THyL+ekPRtqb1ZrPdB6HklKNwc+WRd+xW881mNBPMT99KoRL6Dctg0PmQkVNvnvSn+HKIGxtBEQHQAdS50XeBJE41b0f1twGs89Vf+dAGuaZ0B+MjCQr1iebsfR+jZtJgwh9w3xAtBigkZTwCH0hFTIbyHBCv221+h8rRh+4r6ZWf3RB+OlfuFNjkxd9YvD9DfM0U7Y9oNmCjMKkrWnwG386ZDM71p9spPatB+6rHAOTjjCLfhjCI/Q7noee/YXln/epHJM86sY5/ZIONhDrUzc7mMiblddrVM/0kwbf3zsguyb7bRuFi++6GQlZtejXQ2ll34QjA5m8UI08q82UlPAI/WY/UarILPp929TyGcw/71M/B5rWZXcSkq2N+iSx6OMHRjL4NByjS3PfHFxkReg9iz4b4tu8HiYekTxkespCbWvftpG3FVLCJfRjD4WxU/LdkwMZSuj9R87BIm586mZr2eXdG7LXp5U3aDLWvEuSH1M1HsbPMKE/mOhu1yfjbFj0se7hw4pToWX94G4bH0ucGpZwCL1zmihVKGUPBjLU5CPbVkOk1AtnTEK2SyF07FH//NwL+hK6kmEDsgcXI82K9cnW3LHd7bBn8+ADsT6HHK2/IUucSko4hH7vZmjdBtMK0G0DQ1v0WxvVYi+rTP7+2lmAZC/y5rk/qtV2XJJB2EQa5sHuN/TmYISfNk/oM82K9clWdmzLK4AbPLTSp6xSf0Nm0SclHELfW8hscX77kYzKsTr/68Bpz5xTi34o/zxokbEJh2VnNh3ntBzx5IXDtwt9A7I7CjBDNh7Pdw/CR69Fn2Etep8ar1RB60iF3q9xM4TQg/rpt622AdkkhEPoNy/TiYz9mPNCI1m9m12vqw9zqIgbn2yVQtj4lPo8F308teMb5uuy0Nw3XW3wk2Pgvi/bjzubtHuulmwMxsLILfrm9ZoEOeHwoY+bvFCfmne9PrL2Qko4hN4vZFZSmu+eDE6yyUd6M2JTEPq6Ofol7t4/sr6svF77M/eC1I6vngRjphae0L90j7rsVlwHD33TxD5bZMtHXzlWM8JH7LpZr0+zpeVDH2dTCw5J8Qt9tAtatxdmWKVP7+Qjgwh9aSXUHTX8OeqzMAlJWxO89GeY92F1B6VKIQ7IrrlVrbzjPglP/Qye+FG+exQO2lugYiyUVozsPCLZiaVvXj+82wag9igorTI/fRIK1AROg9IK+PIrEOsa/th8kayw2dZGOOQYKCkb/hyJpRBScfUMxuqbId4Di65I730N82D9/dDVOnj2bq7Z9QZsfBJO+ya87YvQ3QaPfk9LX5zwT/nuXXGTjaxYn5Fmx8Z69Cl21jnDH1tSqnkfZtEPSvFb9KDFl8qq8t2L5AzmuonH1EpOZUAUYMJMtVgy9dPHY7DqdzrxdyoWUiIN8wAHO7IwGJwNnr0dEM0BiETg3J/DUe+Dv30VGv+Q794VN9lIlvIZqUW/63WIR1P/vk5eqL+pWDTzNkNKOIS+0OmtSZ8wGNvyMvS0p26dR0r0C59pKYTXHtFSyKkOwibiF1srBPdNPA7P3gqHnQpjp+q2klK44Ho4/F3wl8/BC3fltYtFTXvzyCNufEZq0ftuylSFfspCTSzMxRzLRYYJfS4YzKLfmkJG7EDq52Ru0a/8nVpqqTwGD2T0IWqdFYLQb1yqN6z5H+m/vbQCLr5Zx2ru+iS8/GB++lfstDdD9Qhj6H1q6mH/TnXBZEJv1cohsmITmXKsLs1PfwAm9LmgvAYk0l/otzVCuVfzPVXqZmv4mz8DUKq07oCX/wbzPzx89EIyGuYVRnGzNbfqmMessw/cVz4KPny73hDvuAw2PJn7/hUzsahO4JE1100d4NL/vvq0rNeyJuXVqR0/4TA1qsxPfwAm9LlA5MAyCFsb1SWSzuQOmZZCWH0zuBgsvDy99yXSME8fiUca3jkSuto0rHLOB5JHDVWOhUvv1ho9t14MW1bltItFzf6dgMviYOwIpxRMNeLGR0THvMyiP4DAhF5EbhCRJhEpkBG8PJNYBiHarVmuqQ7E+tRnMAlJPA6NN+kg7MRhkk6GomE+uHhmZRicg6U/GXmtnpfu0XGNgW6bgVRPhMv+V33NN58f3OxcYSNbMfQ+I6l3E49r+YOhatwMxuSF+nn3dKbfZogJ0qL/PXBGgOcvLirH9g3GNr2olf3SDZOsqdP5ONMRrjcehz0bR2bNw8hq0296Gh76Ftz9qZGVWl5zi8bOp1LqYkwDfPQejca66f2w87XM2z1YyFZWrE9vduyO9N+7d5MOrA5VtXIwpizUSJ1slAsJEYEJvXPuCcBm7PVJtOgzGYj1qZ+dntA33qjlho96X/ptJTJ2KlRNyGxA9ulfaBr7jufVx54Ju17Xgdj5H05eP38g42eo2LsY3HAGPHOtWXpD4fvSszUYO2ayTjD++mPpv3eo6QOHwv9NmZ++H3n30YvIlSKyUkRWNjc357s7wZEo9NsaVTTHTUv/PHVz1FeeSkGv9hZYey8cc8nQ1TFTQUSt+m1r0nvfrjdg3X1w8udg6mJ45D808SpdEmPn06H2SLj8L+q2+utX4KfzYdmvoacj/T6EHd/Fkq3wypIyOOZiWPsXHeRNh3RDK33GTFaXkfnp+5F3oXfOXeucW+ScW1Rbm6VHxkKkcmzfYOy2NfqImaplmkj9bC0xvPuN4Y9dc6tmwh47QreNT8M8aFp7YBXOoVh+reYALL4SzviBDsw9eU167cbjOon54e/si51Ph/o5cMVf4aN/1siMv10NP5kPT//SBD+R9mYoKR9+joJ0WHiZuimfuyO997WsVxfSqAnpvU+kb2pBo5e8C/1BQ8UYtei796tYZuK2gf6lEIbCOR2EPfT41GrppMLk+XrjaFqb2vGd+zRTdc4H1NKaugiO/iA8/XOdTCJVNj6pPtvhBmGHQgQOewdccT9cfq9OKP3A1+CaY+Cpn+c3mqhQaG9Rcc3EAEnGIUdr0EHjTekVnmt+OX23jc+UhTqQm2ye5oMQE/pc4Vv029eozzjTejV13iQkwyVObXwKdr4y8kHYRHoHZFP006/+A3S3wgmf6dv2rm/r8uHvpN7uULHzmTDzFPjYvfCx+9Q18OC/acnjp36mMxodrLQ3ZW8gNpEFl2kAQqruFOfSD61MZPJCwBVG3keBEGR45W3A08CRIrJFRFKYzijE+GUQXn9cl+mGVvqUV+sg43ClEBpvVHGc8/7M2hmM8TO1smEqP6B4DJ75NUw7sf9NbdyhcNLn4IUlsHnF8OfpatWwyrnnZ7+e0Yy3qeBf8VdNRnvwG2rhb1qW3XaKhWzWuUnk6Au1TlOqdYja3tRKrxkLvffbsgHZXoKMuvmQc67BOVfmnJvqnLs+qLaKAt/v+fqjMHqylhXIlOFKIXTsVnE8+oOpZxWmgohWCEzFol93n5YqGKya5Mmf1xmIHvja8I/zL92jYxIjcdsMx/ST4PI/w8cfUB/1o98Prq1Cpi0goa8cqwbH80tSe2JqXq/LVEsfDKR6IoybbgOyCZjrJlf4Qr9lZeZuG5+62bDrteQDic/doVMXZmsQNpGGeVrFcrj6Jct+qVFFg9XWqaiBd30LtqzQScqHYs2tWiZi6nGZ9zlVpp2g8+i+8QS0vBp8e4WEc2rRZysrdiALP6puvJfuGf7YVKcPHIopC2Hr6szfHzJM6HOFX5PexTJ32/jUz9EsVd/yScQ5WHWjZrL6PvVs0jBfa//7P8bB2NqoSVLHf1ojbgZj3oe0fw99O/lAaCax8yNlwWUQKdWSzgcTO1/Tz3X05GDOP+1EvWE33jT8sc3r9PcyuiHz9iYv1AH8TOvshAwT+lyRGLI2Uot+qFIIW1ep/z4Iax5SG5Bd9ist2LbgsuTHRCLw3h/Avi2aUDUYfuz8MWnGzo+E0fU66Lvm1oMruWrZL9VtNecDwZxfRL8Pm57WiJihaF6vbpuR3NxtasF+mNDnikSh9yfczpQJh+kUhINlyK76PZSNgrkXjqyNZEw8HMqqkydO7dsGL96l8dP+AHQyZpysGbtP/j/Yt73/vn6x81Oy0vWUOfYK6NiliT4HA+0tWl5i3iV6owuKeR/SDOnhrPqWl0fmtgHPIJHi8NOvvgXu+edAczpM6HOFL/TjZ6afBDKQ3klIBlj0Xa066cbc84cX2ZG0PdSA7PLrNOJm8ZWpne/d/66x+Y98r//2bMTOZ8rMd+jNdOUNuW87VbL5tLH8Oh3TOfGfs3fOwRhdD0eeCc/elnyMp2O3Rt2MVOgrRus5Ct2i379Lo712vqqTqQeECX2u8H30I3Xb+NQNEnnz/BKt7rjwY9lpIxkN87RuzcACZd371bc962yd+jAVJhymvvw1t/R/Slh9S3Zj59MhEoFjPwabnoKmApytaPuz8H/fok9CI6V7P6y4Dt565sjFNRUWXKaDvi//bfD9mda4GYwpizRUtnUEs1wFzUPf0vyas3+cXsnyNDGhzxUlpbDoE0P7rdOhfrZWBWzf2bet8Ua9AUxdlJ02ktEwT28oAytCPnubWmQnfja98739yzBqIjzwbzqY3LkvuNj5VJn/EfVZF9qg7P5d8MfLNILl8f/WSWVGwrO3ah36k/4lO/0bjrecroOsyWLqW7wAg3SrVg7GSf+i5RfuvjK12lC5ZtMyTSo84TP6ew4QE/pccs6P1eecDeq8L4afOLX9Odi2Wgdhg45QGaxkcTyuCVIN8zXCIh0qx8I7v67umnX3qshHO/LjtvGpngRHnas3r0IpjxCPw11X6jjI+b9V98dIYv7jMR0In3Ks5hLkgpJSjaJ69SG9joE0r1cXxrjpI2+rbhac+V9aPfPJH4/8fNkk1gP3fhHGTIV3XB14cyb0xUrvbFOe+6bxRh2gPeai4NuedKS2leinf+3vOoh24mczu9EsvBxqj4IHv6nXkqvY+aFYdIXWJ3rx7vz2w+fxH6pAnvlfcMwHdRxk9c2ZT6yy7j4NYT3pX3IXvgqw4FIND15zy4H7/IibZGG56bLwozD3Anj0P7OT8excdgZNn/m1Gmln/pfmlQSMCX2xUlOvpY6bXlSL87k7YPZ5Wns+aEpKoX5uf6F/+hea7Tr7/Zmf873f16qcW1bkNnY+GdNPVtEpBPfN+r/B4/8F8z4Miz6u297+ZR10fOhbmZ3zqZ+p5XzUudnrZypMOExnPGv8w4EulZb12XHb+IjAOddo6Y0ln0i/XHIiXW1w07lwzdHDh4gOxd4t8OgP4K1n5GwMyoS+WBHpK4Xw4t06oJPNAmbD0TBPhT4e1z68/igs/mTmk48DvOVdcMR7dCL1XMbOJ0NEQy23rNDB53yx63X1Mx9yjLr//BvgqAnw9n+FVx+GV/+e3jk3LYMtyzXSJlvWczosvFxnPtvwj75t3e1a1TQbA7GJVI6BC3+n0Tx//pf0qmj6dLXCLRfqhPOxHvjDBwZ3PaXC376qTzRn/nfOjBkT+mKmbraWDF71e5h4RO78rKBC37VPLfBnfqVFq3xLcyR84DdaRjjXsfPJmHeJuqlW5smq796vg68IXPyHAwenF39SrfKHvpXeNI1P/Uyf/hbkaRzkqHN0bCYxpr7lFcAFE/0zZSG8+7s6BrT82vTe27lXhX3zcrjgevjo/2rQwc0X6DIdXn5Q8zPe8RUYn4VxiBQxoS9m6mdr9MuW5eqLzKWrwx+Qfe0RePaPKogjzQ8APceMk0d+nmwxaoJmiz53hz665xLn4N6r1Ad/wfVatXQgpRVw+nd0jtRnb0vtvC2vqn/+uP+T3aJ36VBWdeDsU9mocTMUJ3xG3SUPfiP1Utsdu+Gm8zT096IbNRJs8gK4+Ga9Md32odR99t374f4v6xPLiTmKcvIwoS9m6ufqMlKmPu1cUjdb233ke1ojZbAqlWHh2Cs0nPGFJbltd8Vv4bk/akTSEacnP27OBzRm/JHvpVYd8umfa+hoqkltQbHgMv3uPP8nfd28XjNnJxweTHsicN4vYdQk+NMVw09p2b4Tbnyf3mgvvrn/vMuHvxPOv1ZdYEs+DrHo8O3/43/UXXX2/4zMxZkBJvTFTO0s9WfPOjt783ymSmm5PlF07tHY6Fwk2+SLQxdrfkIuM2U3PaO+3LeeAad8eehjRXQgu3V78rpBPm3NWsdn3iVQk6VJwDOl4RgNx/Vnn2pZr4l2QYpg9US44LfqcrzvS8n99W3NKvItr8CHboMjzzjwmLnnq599/f365DWU77/5ZVj6Ex17mnlKVi4lHUzoi5mKGrU03vuf+Wnfd98kziAVRkQ01HL7s7lJqW99E/50OYw9VMcsUsmYnHaCRs88ec3QmaArrlMrOlcJUsOx8DJ1O21b7c0qNSv4NmecDO/4qj4trbn1wP2tO+D3Z+sg+If/qIZMMo6/Et7+FU18GljGw8c5uO+LUD4K3pPkmIAxoS92Zp2dv4HLYz+ms0Udflp+2s8lx1ykxeKCtupjPbDkCujYozfxqnGpv/f076iIP/aDwfd379e6NkeepXPmFgJzvdmnVt6gwprpZCPp8vYva4jn/V/uK7sAGknz+7M1BPLSJXDYqcOf653/plFE//gRPPObA/c//yeNLnrXt4Or9z8MJvRG5kw5Ft7zH/mPd88FlWM18eaFOzUKIx1aXlWLteUVFZLOvckjZB7+jtbgP/encMjc9NqZeLgOsDbeOHiNnjW3aFXOkz6X3nmDpGqc5n+suRXi0dy5ACMlcP51Oii85AodUN2zGX53lj4RXXaXTjWZCiJaq2bWOfDXq7XmlE/HHnjg6/pbOfaKQC4lFUrz1rJhFBuLPq6P6M/doWGNw9G+U8Me19w8+P7SKo16Ka/WxKfSSti6EhZ/KvMM57f/q5Z3fuhb8JE7+rb3ljtYpG6eQmLhR+G523U9l2M9YxrUNXbLhXDPZzVfomOvhk+mWy+qpFQjo24+H+7+tEZrHX4aPPIfWkvo0jsDLVo2HCb0hpEqUxbquMTK36nlnOxJJh7XG8LD39bIjpM/D4eeoBEx3W3esl0jeXrXvX3HfmxkftzqifD2L6nQv/5Yn+th7V90APLd3y28J7DpJ2mkza7Xcue68Tni3Tpe8dTPoHIcXH5P5jPAlVXCJbeq6+f2S+G934MV18Pxnwpmtrc0CIXQd/bEqCzLQ3afcfBx7BUaYbFlhUbjDOTNF+HeL8DmZ2DaSZrJWndUbvu4+FOw/LcaL37lEyrsT/1U50IYbA7ffCMCp/0bvPZofuL6T/uWzoh21Dl9NaQypWocfGQJ3PAe/R7UHKI+/DxT9D76aCzOGdc8wWdvaeSFrWn6Tg0jXY6+UEVh4KBsV5sK669P0UkkzvslXHF/7kUe1LI8/dtatuG5P+r0fVtXacG5fJQ7SIW5F8B5P89P26XlcOrVIxd5nzENcOnd+mTwvp8ENwlQGojLpO5DQCxatMitXLkyrffs747y80de5Q9Pb6S1K8opR0ziM6e+hRMOm4AU2iOqEQ7u/YIOHn5xrZYRWHefDsLt26L+5tO/m50s4ZEQj8Nv36WhgnWzNLPzCy9qiJ8RKkRklXNuyEGFohd6n32dPdyybBPXP/kGLW1dzD90HJ859XBOP6qeSMQE38giO56HX79Nfbstr+hsSXVz1E1TSAOdG5bC78/S9XdcrRm2Rug4qITep7Mnxp9WbeHaJ15j864Ojqir4dPvOJxz50+mrKToPVVGoXDduzRCpqwa3vk1nQ6xpCzfvTqQ2z+ilS2vej5vMdxGsByUQu8TjcW57/nt/Oqx11i3o5Up46r45Ckzufi4aVSVF6if0igeNj6l8dKnfBHGTs13b5LT1arum0JJkDKyzkEt9D7OOR5d38QvH32NlRt3U1YijKkso6aylOryUmoqS6mp0L/qilJGJ2wfP6qMiTUVTKwuZ1JNBROqyykvtacCwzAKh1SEPhThlUMhIpw2q57TZtWz/I1d/H3dm7R2RmnvitLWGaWtK8qb+zp5vUvXWzujdEWTTyQ8prKUSTUVTKwpZ2K1LidUlzO2qowxVWWMqSzz1kt7t9WUl9o4gWEYeSP0Qp/I4pkTWDxz+GiInlicts4oezp62NnWRUtbNzvbu9jZ1q2v23X5WnMbyzd0s3t/95CF6yICoytV/MsiERz6pOG/xTlwOF16G0u9J4+xVWWMHeUtq8oYV9W37m8vjUSIO0fc6Tl0He+1tx53iAijyku8v1JGVZQwqqyEUhu7MIxQE6jQi8gZwE+AEuC3zrkfBtletigriTC+upzx1eXMnDR8Akc87mjtirKvo4e9HT3s6+zpW++I9tsW9QQXQNBcEV0K4m8EYnHHXu8c2/Z2sHe/rkfj2Xe1lZdGGFVeQnV5KVXlJVR7N4LqigHL8hJGVZT23iiqy0uoKIv03qAc9N5s/BtZ72ugvETbqSovoapMl6MS1stLIhYSaxgBEJjQi0gJ8Avg3cAWYIWI/Nk591JQbeaLSER6rexDA2zHOcf+7hh7Onp6hX9vRzdxp08NIkJEhIhARATxlv62mHN0dMfY3/sXHWQ9SntXjI7uGNv3drK/O0Z7l+5r745mNN1mqpREhKqyEipKVfD9m2DitQC9633Xp+/1r1XX9XMp8a8/AqWRCCUR6f0rTVhPfN17t+37zw/Ra0m4WYP0ey2920oiff0pjUjvekmJt4xI7zWkcq/zb6b+kxsMfJLr2y9ARVmEilL93/ZbLy2h0ntdXhqhtGSw/0vkgP+RQG9bsbgj5hzxuLYXi7ve7X5fevtM341/4FOtiBoDZSURykqE0pKI99prc8A/JhZ3tHerG1b/9Lva3u0vo0RjrveaK0tLqCzz/wd9111ZFqGyrISyYdorZoK06BcDrzrnXgcQkduB84DQCX2uEBGqvUHjKeOqhn9DlnHO0dkTp707yv4uFf6uaLxX5HwhThThRBHsisbp7Om70fjrHd0xOnr0JtPRHaczGvPEoc8VlfjEQMK675ZKFJZEsendHoeOWIxo3BGLx4nGdJ++7vtL9sSU7CffK1gDxKv3qSahr9G4imEQT2VhR4ReIS6JCF3RGJ09ycfSstleWYl4N58I5aURzxhQBgazuAEvDvhODHDR+vsmVJdz3+eCm5AkSKGfAmxOeL0FOD7A9oyAERF1u5SXQE2+e1PcxD0ruPcm41nEsTRuAv4TDELvU9zAm2xEhLhzdEfjdEXjdEVjdPUkrEfj3msVzmg8fsCNz1/Ge1/HiTs8q5feJ5JkT1PiPW1CwpNPwtNPIj0xR08sTk8sTnc0TjTu6Il6r7190VicijJ1NVZXlPQaP9Xl3rq3vaailJKI0B2L09mjRoZvbCSud/WocdEdjfdvPxanJzrgdUyvXxJv/QOuIfGl75IdzEWb+PQ3ujLYHIwghX4wI+iAb7GIXAlcCTBt2rQAu2MYhUMkIkQQclGLrwS1SKsrgm/LKEyCDLfYAv1c1lOBbQMPcs5d65xb5JxbVFtrmXuGYRjZJkihXwEcISIzRaQcuAT4c4DtGYZhGIMQmOvGORcVkX8GHkDDK29wzr0YVHuGYRjG4AQaR++cux+4P8g2DMMwjKGxlEjDMIyQY0JvGIYRckzoDcMwQo4JvWEYRsgpqHr0ItIMbMzw7ZOAlix2J9+E7XogfNcUtuuB8F1T2K4HDrym6c65IZOQCkroR4KIrByu+H4xEbbrgfBdU9iuB8J3TWG7Hsjsmsx1YxiGEXJM6A3DMEJOmIT+2nx3IMuE7XogfNcUtuuB8F1T2K4HMrim0PjoDcMwjMEJk0VvGIZhDIIJvWEYRsgpeqEXkTNEZL2IvCoiX813f7KBiGwQkedFZI2IrMx3f9JFRG4QkSYReSFh2wQReUhEXvGW4/PZx3RJck3fEZGt3ue0RkTOymcf00FEDhWRR0VkrYi8KCKf97YX7ec0xDUV5eckIpUislxEnvWu57ve9rQ/o6L20XsTkL9MwgTkwIeKfQJyEdkALHLOFWWih4i8HWgDbnLOzfW2/Tewyzn3Q++GPN45d3U++5kOSa7pO0Cbc+5H+exbJohIA9DgnGsUkdHAKuD9wMco0s9piGu6iCL8nETnX6x2zrWJSBnwJPB54HzS/IyK3aLvnYDcOdcN+BOQG3nEOfcEsGvA5vOAG731G9EfYNGQ5JqKFufcdudco7feCqxF53ku2s9piGsqSpzS5r0s8/4cGXxGxS70g01AXrQfbAIOeFBEVnlz6oaBeufcdtAfJFCX5/5ki38Wkec8107RuDkSEZEZwALgGULyOQ24JijSz0lESkRkDdAEPOScy+gzKnahT2kC8iLkZOfcQuBM4LOe28AoPH4FHA7MB7YD/5PX3mSAiNQAdwJXOef25bs/2WCQayraz8k5F3POzUfn3F4sInMzOU+xC31KE5AXG865bd6yCbgbdVEVO296PlTfl9qU5/6MGOfcm94PMQ5cR5F9Tp7f907gFufcXd7mov6cBrumYv+cAJxze4DHgDPI4DMqdqEP3QTkIlLtDSQhItXAe4AXhn5XUfBn4HJv/XLgnjz2JSv4PzaPD1BEn5M30Hc9sNY59+OEXUX7OSW7pmL9nESkVkTGeetVwOnAOjL4jIo66gbAC5W6hr4JyL+f3x6NDBE5DLXiQef0vbXYrklEbgNORcupvgl8G/hf4A5gGrAJ+KBzrmgGN5Nc06moO8ABG4BP+b7TQkdE3gb8A3geiHubv476tIvycxrimj5EEX5OInIMOthaghrldzjn/l1EJpLmZ1T0Qm8YhmEMTbG7bgzDMIxhMKE3DMMIOSb0hmEYIceE3jAMI+SY0BuGYYQcE3rDyAIicqqI3JvvfhjGYJjQG4ZhhBwTeuOgQkQu9Wp8rxGR33hFo9pE5H9EpFFE/i4itd6x80VkmVcM626/GJaIvEVEHvbqhDeKyOHe6WtEZImIrBORW7xMTcPIOyb0xkGDiBwFXIwWjZsPxICPANVAo1dI7nE06xXgJuBq59wxaLalv/0W4BfOuXnASWihLNBqiVcBs4HDgJMDviTDSInSfHfAMHLIu4BjgRWesV2FFoSKA3/0jrkZuEtExgLjnHOPe9tvBP7k1SGa4py7G8A51wngnW+5c26L93oNMAOdLMIw8ooJvXEwIcCNzrmv9dso8s0Bxw1VF2Qod0xXwnoM+30ZBYK5boyDib8DF4pIHfTOvTkd/R1c6B3zYeBJ59xeYLeInOJtvwx43KtvvkVE3u+do0JERuXyIgwjXcziMA4anHMvicg30Nm7IkAP8FmgHZgjIquAvagfH7QE7K89IX8duMLbfhnwGxH5d+8cH8zhZRhG2lj1SuOgR0TanHM1+e6HYQSFuW4MwzBCjln0hmEYIccsesMwjJBjQm8YhhFyTOgNwzBCjgm9YRhGyDGhNwzDCDn/H7HgBgET9u+1AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_nn_training(results2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "475028ed",
   "metadata": {},
   "source": [
    "### Model 3 - less dense nodes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "3d7808b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# model with less dense nodes - will train faster, but will it be as accurate?\n",
    "def generate_model3(input_shape, num_classes):\n",
    "    \n",
    "    '''CNN model architecture with max pooling + dropout layers'''\n",
    "    \n",
    "    inputs = Input(shape=input_shape)\n",
    "    \n",
    "    #x = Rescaling(1.0 / 255)\n",
    "    \n",
    "    # convolution (with max pooling)\n",
    "    x = Conv2D(32, kernel_size=5, strides=2, padding=\"same\")(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x) \n",
    "    x = BatchNormalization()(x) # why?\n",
    "    x = Activation(\"relu\")(x)\n",
    "\n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    #flatten\n",
    "    x = Flatten()(x)\n",
    "    \n",
    "    #dense layers (with dropout to avoid overfitting)\n",
    "    x = Dense(512, activation='relu')(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    x = Dense(256, activation='relu')(x)\n",
    "    x = Dropout(0.2)(x)\n",
    "    x = Dense(128, activation='relu')(x)\n",
    "    x = Dropout(0.1)(x)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(2, activation='softmax')(x)\n",
    "    \n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "d7c8a3be",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model3/best_model.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "model3 = generate_model3(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "model3.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "87112a03",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x33418e040> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x2bd735310> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50000, saving model to models/Model3/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.50000\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.50000\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy did not improve from 0.50000\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy improved from 0.50000 to 0.78400, saving model to models/Model3/best_model.h5\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.78400 to 0.94400, saving model to models/Model3/best_model.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.94400\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.94400\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.94400\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.94400\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.94400\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.94400\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.94400\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy improved from 0.94400 to 0.95350, saving model to models/Model3/best_model.h5\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.95350\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.95350\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.95350\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.95350\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.95350\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.95350\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.95350\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.95350\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.95350\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.95350\n",
      "Epoch 00024: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results3 = model3.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e600fdbe",
   "metadata": {},
   "source": [
    "**Accuracy:** 0.95350"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "f2211027",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0304c954",
   "metadata": {},
   "source": [
    "### Model 4 - larger convolution kernel size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "6eef176c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_model4(input_shape, num_classes):\n",
    "    \n",
    "    '''CNN model architecture with max pooling + dropout layers - larger convo kernel size'''\n",
    "    \n",
    "    inputs = Input(shape=input_shape)\n",
    "    \n",
    "    # convolution (with max pooling)\n",
    "    x = Conv2D(32, kernel_size=10, strides=2, padding=\"same\")(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x) \n",
    "    x = BatchNormalization()(x) # why?\n",
    "    x = Activation(\"relu\")(x)\n",
    "\n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    #flatten\n",
    "    x = Flatten()(x)\n",
    "    \n",
    "    #dense layers (with dropout to avoid overfitting)\n",
    "    x = Dense(1024, activation='relu')(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    x = Dense(512, activation='relu')(x)\n",
    "    x = Dropout(0.2)(x)\n",
    "    x = Dense(256, activation='relu')(x)\n",
    "    x = Dropout(0.1)(x)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(2, activation='softmax')(x)\n",
    "    \n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "1967145f",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model4/best_model.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "model4 = generate_model4(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "model4.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "63c1756c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x320cc50d0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x320ec7d30> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50000, saving model to models/Model4/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.50000 to 0.50350, saving model to models/Model4/best_model.h5\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.50350\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.50350 to 0.79500, saving model to models/Model4/best_model.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.79500\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy did not improve from 0.79500\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.79500 to 0.87550, saving model to models/Model4/best_model.h5\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.87550\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.87550\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.87550\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.87550\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.87550\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy improved from 0.87550 to 0.89000, saving model to models/Model4/best_model.h5\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.89000\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.89000\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.89000\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.89000\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.89000\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.89000\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.89000\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.89000\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy improved from 0.89000 to 0.92300, saving model to models/Model4/best_model.h5\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.92300\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.92300\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.92300\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy did not improve from 0.92300\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.92300\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.92300\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.92300\n",
      "Epoch 30/50\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.92300\n",
      "Epoch 31/50\n",
      "\n",
      "Epoch 00031: val_accuracy did not improve from 0.92300\n",
      "Epoch 32/50\n",
      "\n",
      "Epoch 00032: val_accuracy did not improve from 0.92300\n",
      "Epoch 00032: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results4 = model4.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a5d0ded9",
   "metadata": {},
   "source": [
    "**Accuracy:** 0.92300"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "cd972c4d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d7955d2d",
   "metadata": {},
   "source": [
    "## Transfer Learning\n",
    "\n",
    "After training and evaluating the models described above, I next leveraged transfer learning to find out if pre-trained deep learning models could produce better and/or more stable results. Out of the many deep learning models available I decided to use Resnet50 because a) it offers a good balance of accuracy and training efficiency and b) it was used in the baseline model for the xView2 competition, which suggested that it would perform well for a similar aerial imagery classification task."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "56e2e120",
   "metadata": {},
   "source": [
    "### TF Model 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c9d3aa2f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_TL_model(input_shape, num_classes):\n",
    "    \n",
    "    weights = 'imagenet'\n",
    "    inputs = Input(input_shape)\n",
    "\n",
    "    #what does include_top=False remove?\n",
    "    base_model = ResNet50(include_top=False, weights=weights, input_shape=(128, 128, 3))\n",
    "\n",
    "    #freezing the resnet layers\n",
    "    for layer in base_model.layers:\n",
    "        layer.trainable = False\n",
    "\n",
    "    #these convolution layers come before the resnet layers\n",
    "    x = Conv2D(32, (5, 5), strides=2, padding='same', activation='relu', input_shape=(128, 128, 3))(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Flatten()(x)\n",
    "\n",
    "    base_resnet = base_model(inputs)\n",
    "    base_resnet = Flatten()(base_resnet)\n",
    "\n",
    "    concated_layers = Concatenate()([x, base_resnet])\n",
    "\n",
    "    # dense layers come after the resnet layers\n",
    "    # dropout to reduce overfit\n",
    "    concated_layers = Dense(1024, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.3)(concated_layers)\n",
    "    concated_layers = Dense(512, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.2)(concated_layers)\n",
    "    concated_layers = Dense(256, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.1)(concated_layers)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(num_classes, activation='softmax')(concated_layers)\n",
    "\n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "eb11092e",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model5/best_model.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "TL_model = generate_TL_model(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "TL_model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "7c13f4d0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x16d9f24c0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x16d9f24c0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x10eb42310> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x10eb42310> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.94650, saving model to models/Model5/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.94650\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.94650 to 0.95550, saving model to models/Model5/best_model.h5\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.95550 to 0.95800, saving model to models/Model5/best_model.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.95800\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.95800 to 0.95900, saving model to models/Model5/best_model.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.95900 to 0.96050, saving model to models/Model5/best_model.h5\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.96050\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy improved from 0.96050 to 0.96600, saving model to models/Model5/best_model.h5\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.96600\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy improved from 0.96600 to 0.97000, saving model to models/Model5/best_model.h5\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy improved from 0.97000 to 0.97050, saving model to models/Model5/best_model.h5\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.97050\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.97050\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.97050\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.97050\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy improved from 0.97050 to 0.97200, saving model to models/Model5/best_model.h5\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.97200\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.97200\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.97200\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.97200\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.97200\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.97200\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy improved from 0.97200 to 0.97650, saving model to models/Model5/best_model.h5\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.97650\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy did not improve from 0.97650\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.97650\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.97650\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.97650\n",
      "Epoch 30/50\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.97650\n",
      "Epoch 31/50\n",
      "\n",
      "Epoch 00031: val_accuracy did not improve from 0.97650\n",
      "Epoch 32/50\n",
      "\n",
      "Epoch 00032: val_accuracy did not improve from 0.97650\n",
      "Epoch 33/50\n",
      "\n",
      "Epoch 00033: val_accuracy did not improve from 0.97650\n",
      "Epoch 34/50\n",
      "\n",
      "Epoch 00034: val_accuracy did not improve from 0.97650\n",
      "Epoch 00034: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results_TL = TL_model.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "794e0137",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results_TL)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e1583ed8",
   "metadata": {},
   "source": [
    "### TF Model 2 - Increased kernel size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "6296779a",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_TL_mode_2(input_shape, num_classes):\n",
    "    \n",
    "    '''TL model, Conv kernel=10 '''\n",
    "    \n",
    "    weights = 'imagenet'\n",
    "    inputs = Input(input_shape)\n",
    "\n",
    "    #what does include_top=False remove?\n",
    "    base_model = ResNet50(include_top=False, weights=weights, input_shape=(128, 128, 3))\n",
    "\n",
    "    #freezing the resnet layers\n",
    "    for layer in base_model.layers:\n",
    "        layer.trainable = False\n",
    "\n",
    "    #these convolution layers come before the resnet layers\n",
    "    x = Conv2D(32, kernel_size=10, strides=2, padding='same', activation='relu', input_shape=(128, 128, 3))(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Flatten()(x)\n",
    "\n",
    "    base_resnet = base_model(inputs)\n",
    "    base_resnet = Flatten()(base_resnet)\n",
    "\n",
    "    concated_layers = Concatenate()([x, base_resnet])\n",
    "\n",
    "    # dense layers come after the resnet layers\n",
    "    # dropout to reduce overfit\n",
    "    concated_layers = Dense(1024, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.3)(concated_layers)\n",
    "    concated_layers = Dense(512, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.2)(concated_layers)\n",
    "    concated_layers = Dense(256, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.1)(concated_layers)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(num_classes, activation='softmax')(concated_layers)\n",
    "\n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "1e7b38dd",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model5/best_model_10.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "TL_model_2 = generate_TL_mode_2(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "TL_model_2.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "63c86697",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x2cea85280> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x14dba6a60> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.92000, saving model to models/Model5/best_model.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.92000 to 0.95650, saving model to models/Model5/best_model.h5\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.95650\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.95650 to 0.96450, saving model to models/Model5/best_model.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.96450\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.96450 to 0.96650, saving model to models/Model5/best_model.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.96650\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.96650\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.96650\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.96650\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy improved from 0.96650 to 0.97200, saving model to models/Model5/best_model.h5\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.97200\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.97200\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.97200\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.97200\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.97200\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy improved from 0.97200 to 0.97300, saving model to models/Model5/best_model.h5\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.97300\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.97300\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.97300\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.97300\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.97300\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.97300\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy improved from 0.97300 to 0.97450, saving model to models/Model5/best_model.h5\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.97450\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy did not improve from 0.97450\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.97450\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.97450\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.97450\n",
      "Epoch 30/50\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.97450\n",
      "Epoch 31/50\n",
      "\n",
      "Epoch 00031: val_accuracy did not improve from 0.97450\n",
      "Epoch 32/50\n",
      "\n",
      "Epoch 00032: val_accuracy did not improve from 0.97450\n",
      "Epoch 33/50\n",
      "\n",
      "Epoch 00033: val_accuracy did not improve from 0.97450\n",
      "Epoch 34/50\n",
      "\n",
      "Epoch 00034: val_accuracy did not improve from 0.97450\n",
      "Epoch 00034: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results_TL = TL_model_2.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "229b3a15",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results_TL_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b23785bc",
   "metadata": {},
   "source": [
    "### TF Model 3 - Reduced Learning Rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "caf589e1",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Customized adam optimizer\n",
    "adam = tf.keras.optimizers.Adam(lr=0.0001,\n",
    "                                beta_1=0.9,\n",
    "                                beta_2=0.999,\n",
    "                                decay=0.0,\n",
    "                                amsgrad=False)\n",
    "\n",
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model5/best_model_lr.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model (model 3 is same as 2 but with different learning rate for adam optimizer)\n",
    "TL_model_3 = generate_TL_mode_2(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "TL_model_3.compile(optimizer=adam, loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "b884404b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x296e17670> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x797e72dc0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.94850, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.94850 to 0.95000, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.95000 to 0.95400, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy did not improve from 0.95400\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.95400\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy did not improve from 0.95400\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.95400\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.95400\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy improved from 0.95400 to 0.96550, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.96550\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.96550\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.96550\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.96550\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.96550\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.96550\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.96550\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy improved from 0.96550 to 0.96850, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.96850\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy improved from 0.96850 to 0.97400, saving model to models/Model5/best_model_lr.h5\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.97400\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.97400\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.97400\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.97400\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.97400\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.97400\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy did not improve from 0.97400\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.97400\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.97400\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.97400\n",
      "Epoch 00029: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results_TL_3 = TL_model_3.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "70594a80",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8/fFQqAAAACXBIWXMAAAsTAAALEwEAmpwYAABDOklEQVR4nO3dd3iUZdb48e9JIySQSgsJkFCkN+kWwN67q2AviL29Wyz7urrvFt3frq6uumJdG4qKdV1WBBUQ6VV6byEJhIQU0pO5f3/cT2AIk2RSJpNkzue6uJh52pzHwefM3cUYg1JKKVVVkL8DUEop1TxpglBKKeWRJgillFIeaYJQSinlkSYIpZRSHmmCUEop5ZEmCKUAEXlbRP7o5bG7ReRsX8eklL9pglBKKeWRJgilWhERCfF3DKr10AShWgynaufXIvKziBSIyJsi0llE/isi+SIyV0Ri3Y6/VEQ2iEiOiMwTkf5u+4aLyCrnvI+A8CqfdbGIrHHOXSQiQ7yM8SIRWS0ieSKyT0SeqrL/NOd6Oc7+W5ztbUXkWRHZIyK5IrLQ2TZRRFI9/Hc423n9lIjMFJH3RSQPuEVERovIYucz0kXkJREJczt/oIjMEZFsETkgIo+LSBcRKRSReLfjRohIpoiEenPvqvXRBKFamquAc4CTgEuA/wKPAx2w/54fABCRk4APgYeAjsAs4N8iEuY8LL8A3gPigE+c6+KcezLwFnAnEA+8CnwlIm28iK8AuAmIAS4C7haRy53rdnfifdGJaRiwxjnvb8AI4BQnpt8ALi//m1wGzHQ+czpQATyM/W8yDjgLuMeJoT0wF/gG6Ar0Br4zxmQA84Br3K57AzDDGFPmZRyqldEEoVqaF40xB4wx+4EfgaXGmNXGmBLgc2C4c9y1wH+MMXOcB9zfgLbYB/BYIBR43hhTZoyZCSx3+4w7gFeNMUuNMRXGmHeAEue8Ghlj5hlj1hljXMaYn7FJaoKz+3pgrjHmQ+dzs4wxa0QkCLgNeNAYs9/5zEXOPXljsTHmC+czi4wxK40xS4wx5caY3dgEVxnDxUCGMeZZY0yxMSbfGLPU2fcONikgIsHAZGwSVQFKE4RqaQ64vS7y8L6d87orsKdyhzHGBewDEp19+83xM1XucXvdA/ilU0WTIyI5QDfnvBqJyBgR+cGpmskF7sL+kse5xg4Pp3XAVnF52ueNfVViOElEvhaRDKfa6c9exADwJTBARHpiS2m5xphl9YxJtQKaIFRrlYZ90AMgIoJ9OO4H0oFEZ1ul7m6v9wF/MsbEuP2JMMZ86MXnfgB8BXQzxkQD04DKz9kH9PJwziGguJp9BUCE230EY6un3FWdkvkVYDPQxxgTha2Cqy0GjDHFwMfYks6NaOkh4GmCUK3Vx8BFInKW08j6S2w10SJgMVAOPCAiISJyJTDa7dzXgbuc0oCISKTT+Nzei89tD2QbY4pFZDRwndu+6cDZInKN87nxIjLMKd28BTwnIl1FJFhExjltHluBcOfzQ4H/BWprC2kP5AFHRKQfcLfbvq+BLiLykIi0EZH2IjLGbf+7wC3ApcD7XtyvasU0QahWyRizBVuf/iL2F/olwCXGmFJjTClwJfZBeBjbXvGZ27krsO0QLzn7tzvHeuMe4P9EJB/4HTZRVV53L3AhNlllYxuohzq7fwWsw7aFZAN/AYKMMbnONd/Aln4KgON6NXnwK2xiyscmu4/cYsjHVh9dAmQA24Az3Pb/hG0cX+W0X6gAJrpgkFLKnYh8D3xgjHnD37Eo/9IEoZQ6SkRGAXOwbSj5/o5H+ZdWMSmlABCRd7BjJB7S5KBASxBKKaWqoSUIpZRSHrWqib06dOhgkpOT/R2GUkq1GCtXrjxkjKk6tgZoZQkiOTmZFStW+DsMpZRqMURkT3X7tIpJKaWUR5oglFJKeaQJQimllEetqg3Ck7KyMlJTUykuLvZ3KD4VHh5OUlISoaG6totSqnG0+gSRmppK+/btSU5O5vjJO1sPYwxZWVmkpqaSkpLi73CUUq1Eq69iKi4uJj4+vtUmBwARIT4+vtWXkpRSTavVJwigVSeHSoFwj0qpptXqq5iUUspbxWUVrN6bw4a0XIZ3j+Xk7jEB/eNLE4SP5eTk8MEHH3DPPffU6bwLL7yQDz74gJiYGN8EppQit6iMlXuyWbbrMMt3Z/Nzag5lFcfmpxuSFM3N45K5eGgCbUKCmySmkvIK3lu8h283HsDbufKi24bxxs0jGz0WTRA+lpOTwz//+c8TEkRFRQXBwdX/g5s1a5avQ1Mq4BzMK2bZ7myW78pm6a5sthzIxxgIDRYGJ0Zz22kpjE6OY0DXKOZuOsg7i3bzy0/W8udZm7huTHeuH9ODLtHhPonN5TL8++c0/jp7C6mHixiUGEVUuHe9EkODfVPK0QThY48++ig7duxg2LBhhIaG0q5dOxISElizZg0bN27k8ssvZ9++fRQXF/Pggw8ydepU4Ni0IUeOHOGCCy7gtNNOY9GiRSQmJvLll1/Stm1bP9+ZUs2bMYY9WYUs253Nsl3ZLN+dzZ6sQgAiwoI5uXssFwxKYFRKLMO7xdI27PgfbDeO7cENY7rz0/Ys3l60i5d+2M4r83Zw/qAu3HJKMiN6xDZa9dOiHYd4etZm1u3PpX9CFO/eNpjxJ3mcHqlJtarpvkeOHGmqzsW0adMm+vfvD8Dv/72BjWl5jfqZA7pG8eQlA6vdv3v3bi6++GLWr1/PvHnzuOiii1i/fv3R7qjZ2dnExcVRVFTEqFGjmD9/PvHx8ccliN69e7NixQqGDRvGNddcw6WXXsoNN9xwwme536tSLdXBvGIO5JUQ1y6M+MgwwkO9q9pxuQybM/JZvjv7aFLIzC8BIDYilJHJcYxOjmNUShwDu0YRGly3Pjp7swp5b8luZizfR35xOQO7RnHLKclcMrSr1zFWtSUjn2f+u4kftmTSNTqcX53Xl8uHJRIU1HTtHiKy0hjjsX5KSxBNbPTo0ceNVfjHP/7B559/DsC+ffvYtm0b8fHxx52TkpLCsGHDABgxYgS7d+9uqnCV8jljDFsO5DN34wHmbDzA2tTc4/a3axNCvJMs4tu1oUO7MOIiw4iPbEN8uzDScopZvtuWEPKLywFIiA7nlF7xjEqOY3RKHL07tmvwQ7d7fAS/vWgAD59zEp+v3s/bP+3m1zN/5k+zNjGyRywDEqIY0DWKAQnRdItrW2PpIiO3mOfmbGHmylQi24Tw2AX9uPmU5HonGl8JqARR0y/9phIZGXn09bx585g7dy6LFy8mIiKCiRMnehzL0KZNm6Ovg4ODKSoqapJYlfKV8goXy3cfZs7GA8zZlMG+bPtvemi3GH59Xl96d2rH4YJSsgpKOXSkhKwjpWQXlLIvu5A1+3LILiilwnWs9qNnx0guHpLAqOQ4RiXHkRRb8wO6ISLCQrh+TA+uG92dxTuymLkqlfX7c/l+80EqQ2rfJoT+RxOG/btP53aUlLt4df4O3ly4C5cLbjs1hXvP6E1sZJhPYm0onyYIETkfeAEIBt4wxjxTZX8s8BbQCygGbjPGrHf2PQxMAQywDrjVGNPiRoK1b9+e/HzPqzfm5uYSGxtLREQEmzdvZsmSJU0cnWqpjDH8d30Gg7pG0z0+wt/heOVISTkLtmYyZ+MBvt98kNyiMsJCgji1Vzx3T+jN2f070SnKuwZgl8uQW1RGVkEJMRFhdGjXpvaTGpmIcErvDpzSuwNgu8huychnY3oeG9Py2Jiex8cr9lFYWgFASJDQJiSIgtIKLhvWlV+d25ducc37u/NZghCRYOBl4BwgFVguIl8ZYza6HfY4sMYYc4WI9HOOP0tEEoEHgAHGmCIR+RiYBLztq3h9JT4+nlNPPZVBgwbRtm1bOnfufHTf+eefz7Rp0xgyZAh9+/Zl7NixfoxUtRQul+F3X63n/SV7CQsJ4q7xPbl7Yu8TGln9raCknFV7D7N8l20TWLUnh9IKFzERoZzVvxPnDujM6X06Etmm7o+hoCAhNjKsWf3yDg8NZmi3GIZ2izm6zeUy7MkudBJGLpn5Jdw4NpnBSdH+C7QOfNZILSLjgKeMMec57x8DMMY87XbMf4CnjTELnfc7gFOwiWsJMBTIA74A/mGM+bamz6ytkbq1C6R7DVRlFS5+/clavliTxq2nJpN1pJSv1qaRFNuW3108gHMGdG5w1UqFyxBcj/r67IJS2xbg9Bhan5ZHhcsQJLYzx9iUeM4e0JmRPWIJqWMDsfIdfzVSJwL73N6nAmOqHLMWuBJYKCKjgR5AkjFmpYj8DdgLFAHfVpccRGQqMBWge/fujXsHSjUjxWUV3PfBauZuOsCvz+vLvWf0BmDy6O48+dV6pr63kol9O/LUJQNJ7hBZy9WOV1RawbcbM5i5MpWF2w8RFhxEdNtQYiJCiW4bSnTbsKOvY9qGEu28rnAZVuyxpYRtB48AEBYSxLBuMdw9oRejUuI4uXsM7b3sz6+aF18mCE8/QaoWV54BXhCRNdh2htVAudM2cRmQAuQAn4jIDcaY90+4oDGvAa+BLUE0WvRKNSMFJeXc8e4KFu3I4g+XDeTGcclH943rFc9/Hjiddxbt5vm52zj37wuYOr4n955Rc7WTMYZVe3OYuXIfX69NJ7+knKTYttxxek8EyCksI7eojJyiUvbnFLExLZecorKjdeqV2rcJYURyLJcPT2R0ShxDkqKbbNSx8i1fJohUoJvb+yQgzf0AY0wecCuA2HLxLufPecAuY0yms+8zbNXTCQlCqdYup7CUW/61nHX7c/n7tUO5YnjSCceEBgcx5fSeXDq0K3+etYmXftjO56v388TF/TlvYJfjqp0ycov5dFUqn65MZeehAtqGBnPB4C78YkQ3xqTE1dodtLTcRW6RTR4uY+jVsV29qqRU8+fLBLEc6CMiKcB+bCPzde4HiEgMUGiMKcX2WFpgjMkTkb3AWBGJwFYxnQUc37igVAA4mF/MTW8uY2dmAf+8/mTOG9ilxuM7RYXz/KThTrXTBu56fxWn9+nA4xf2Z9vBI7YKaVsmLgOjk+O4a2IvLhycQLs6NBSHhQTRsX0bOrZv+p5Dqmn5LEEYY8pF5D5gNrab61vGmA0icpezfxrQH3hXRCqAjcDtzr6lIjITWAWUY6ueXvNVrEo1R6mHC7nhjaUczC/hX7eO4lSnO6U3xvSM5+v7T+PdxXv4+5ytXPDCjwAkxrTlvjN6c+XJSXVup1CBx6fjIIwxs4BZVbZNc3u9GOhTzblPAk/6Mj6lmqvtB49w45tLKSgp573bxzCiR2ydrxESHMRtp6Vw8dAEPlu1n8GJ0YzrGd+k0zioli2gRlK3BO3atePIkSP+DkP50fr9udz81jJEYMbUcQzoGtWg63VqH85dE3o1UnQqkGiCUMoHjDFkFZSyJ6uQvdkFFJRUEBIkBAcJocFBBAcJIUFCSHDQ0e0hwcLhgjIe/exnosJDeX/KGFK0Gkj5kSYIH3vkkUfo0aPH0fUgnnrqKUSEBQsWcPjwYcrKyvjjH//IZZdd5udIVV2VV7hIzy1mT1Yhe7IL2JtV6LwuZG9WAQVVuoN6q2eHSN6bMobEGJ3SXflXYCWI/z4KGesa95pdBsMFz1S7e9KkSTz00ENHE8THH3/MN998w8MPP0xUVBSHDh1i7NixXHrppQG9tGFLUlru4okv1vPpqlTK3SaMCwsOoltcW3rERzImJY4e8RH0iI+ge1wkUW1DqHAZyisM5S5DhctFudv78gr73uUyDOkWU6deRUr5iv4r9LHhw4dz8OBB0tLSyMzMJDY2loSEBB5++GEWLFhAUFAQ+/fv58CBA3TpUnMXRuV/BSXl3PX+Sn7cdojJo7sxNCmG7vER9IiPpEtUeMsfD7B/JSx4Fi76G0R19Xc0ys8CK0HU8Evfl66++mpmzpxJRkYGkyZNYvr06WRmZrJy5UpCQ0NJTk72OM23al6yjpRw29t2wNpfrhrMtaNa2dQuh7bD9F9AYRaER8MVr/g7IuVnOmNWE5g0aRIzZsxg5syZXH311eTm5tKpUydCQ0P54Ycf2LNnj79DVLXYl13I1dMWszkjn1dvHNn6kkN+Brx/BSAw+BpY+yGkr/V3VMrPAqsE4ScDBw4kPz+fxMREEhISuP7667nkkksYOXIkw4YNo1+/fv4OUdVgU3oeN7+1jOKyCqZPGcPI5Dh/h9S4inPh/auhIAtu+TfE94Yd38Hs38LN/wZtGwtYmiCayLp1xxrHO3TowOLFiz0ep2MgmpclO7O4490VRIaFMPPuUzipc3t/h9S4ykvgoxsgcxNM/ggSR9jtEx+DWb+CrbOh7/n+jVH5jVYxKVWNb9anc9Nby+gcFc6n97TC5OBywed3wq4FcNnL0OfsY/tG3ALxfWDOE1BR5rcQlX9pglDKg+lL93DP9FUM7BrFJ3eOa31jEoyB2Y/Bhs/hnP+DoZOO3x8carcf2gor3/ZLiMr/AiJB+GrVvOYkEO6xKRhjeGHuNn77+Xom9u3E9CljmtWylo3mp+dh6TQYew+c8oDnY/peAMmnw7ynbTuFCjitPkGEh4eTlZXVqh+gxhiysrIID/duwXflWYXL8MSX6/n73K1cdXISr944goiwVthMt+YDmPsUDLoKzv1T9Y3QInDuH2y314V/b9IQvVJyBDZ8Aek/+zuSVqsV/us/XlJSEqmpqWRmZvo7FJ8KDw8nKenEhWRUzUrKK1iyM5s5GzOYu/EgGXnF3D2xF785r2/rHNm+bQ58eR+kTIDLX4GgWn4jdh0OQybB4n/CyNsgxs/de8uKYNu3sP5T2PotlBeBBMM5v4dx97WeHlc5e2HXjzDsOr/eU6tPEKGhoaSkpPg7DNWINmfk8bfZW4iPbMOArlEM7BpFv4Qor6enyC0s44ctB5mz8QDzt2ZypKSctqHBjD+pA08OG8AFgxN8fAeNoLQQlrwMCHQaAJ36Q0yPmh/4qSvh45ug80C49n0I8XLBn7OegI1fwHd/gKteb4zo66a8BHZ8D+s/gy2zoPQIRHaE4TdA/4th+Rvw7f/CvqW2sT08uuljbEyF2fDuZZC9E4qy4ZT7/RZKq08QqnXZdiCf619fSrnLECTw0Yp9R/clx0cwoGsUAxKiGNg1mgFdo+jUvg0iwr7sQuZsPMDcTQdYuiubCpehQ7s2XDI0gbP7d+bU3h0ID20h6ygf2g4f3wgHNx6/PTQCOvY9ljA69bev2ydA1g744Bf2wXr9TAivwxTi0Ukw7l748VkYe9exrrC+VFEOu+bbpLD537YNpG0sDLrSVo31OA2CncdXygRY/DLM+R0cmAjXvAddBvk+Rl+oKLNJPDcVup8Cc56EhGGQcrpfwpHWVDc/cuRIs2KFrkzaWu3MPMK1ry0B4KOpY0npEElGXjEb0/Lsn3T7Z09W4dFz4iPDiIkIZUdmAQB9OrXjnAGdOXtAZ4YlxbS8xXM2fGGriIJD7a/5pNGQucUmi8zN9u+Dm+DIgWPntIm21RRBIXD7txBfj7UhSvLhH8Nt19dbZ/mm2iN3P+z+0Xa73fqNbfsIa29LCQOvhJ4TIaSGDgN7FsMnt9hkcvFztnqmJTEGvn7I9hq74lXodxG8dgYU58CdC3w2N5aIrDTGjPS4TxOEagn2ZBVw7atLKKtwMWPqWPrUMCYhv7iMzRn5bNify8b0PDLzSzilVwfOGdC55S6zWVFmf00ueRkSR8Iv3oaYbtUfX5htE4V7whj/K9umUF8r3oKvH4Zrp9uHdkMdyTyWEHYtgOwddnvbWOh1pk0Kvc+G0Dp0vjhyEGbeZq978k1wwV/rdr4/LZkG3zwCp/0PnO0spnlwM7x+pq0WvOU/NSfIetIEoVq0fdmFTHptCYWl5Xw4dSz9ujRshbUWJy/N/jLetxRGT7U9j3zwoKhVRTlMO9Umq3uW1D2GosOw+6djSaGyiqxNFPQ4BVLG2z+dBtbeeF5bnD/8CRY+B12GwDXvQlwzb4fcNtdWAfa90FaRud//+s9g5q0w+k648P81+kfXlCC0DUI1a2k5RVz3xhLyi8v44I4ATA4758HM223vnavehMFX+y+W4BA45w/2QbbiLdseURuXy/Y6WvYq7PgBMBDSFnqMg8G/sO0HCUOPtSc0VpxnPwndRtuR4q9NsFU2fS9ovM9oTAc32wTQeaCNs2pyHHSlnYZ98UuQNBKGXNNkoWkJQvlORTm8copt4Bxxc51PP5BXzKTXlnAov4T3p4xhaLcYZ8cGeOdSO/p34qPQpplPgeFygavM+15DlecsfBZ++LOt97/2PdsA7W/GwHuX25leH1htq4M8KcqBNdNh2etweJdtKB9+g606ShxRt/8WDZG9Cz652cZ72sNwxv82bjJqqIIseONM2ytt6g+2Q4AnFWX233zaapgyt1Eb4bWKSfnHgQ02QXQfB7d9U6dTM/NLmPTaYjJyi3n39tGM6OE2g+p3/wc/PgcYaNcFzvuT7dnSHPvAH9oOH06y9etxvaBTv2O9jDr2tw3GwaHHn1OYDZ/fBdtmw6Cr4ZIXoE07/8TvScY6mHY6nHIfnPvH4/cd3AzLXoO1M6CsALqNhTFTof+lJ95nUykrhv/+Bla9A7Eptppu+PX+7w5bXmqTbeoK2/Cf5PEZfUz+AXh1PIRFwB0/QNuYRglDE4Tyj1Xvwlf3gwTBr3dAhHfTZGcXlDL5tSXszS7k7VtHMaZn/PEHvHKaLTWc+wf4zy8hfY3t9njhX6HzgMa/j/ravdDOlCpBMPxGyNpuexpl7wTjsscEhUKHk5zE0R+ikmDenyEvHc5/GkZNaZ6J74t7Yd3HcN9yiO5mZ31dOs12TQ1uY6vCRk+FrsP8Hekxm76GRf+wbTmhkTBsso2xISWzsmLI2gYd+tatTcYY+Oo+WP1+3aoO9yyGdy6GPufZsSwNaatxaIJQTa603EXx5/fTfsN0BEPxZa8TPrz2utOcwlKue30pOzKP8K9bRnFK7w7HH5CXBs/1h7OfslUGrgr7y/C7/4PiPBhzl612qks/f19Y86FNjnEpcN1HENfz2L6yIjsJ3sFNx//J3Wv3RyXBNe/U/ovSn/LS4MUR9uFamA05eyAqEUbdDiffDJEdar+Gv6SthqWvwfqZUFFqu8+OuQv6nAtBtYyFqSiD/atgt9Pzat8yKC+2pZH+l9iSbPL42quxFr0E3/4Wxv8Gzvxt3eKv7O101u/g9F/W7VwPNEEonzhSUs6erAL2ZhWyJ7uQPVmF7MkqYE9WIem5RXwV+ji5JpJ+QXuZ7xrKX9o+TM8O7ejVKZKeHdrRs2MkvTq2IzGmLUFBQm5RGTe8sZQtGfm8fvNIJpzU8cQPXfk2/PtBuHvx8aWFwmz47vew8h1o18k2pg65pul/fRtje9As+KvtkXPNu9XX01dVkm9LGXG9/J/gvDHvL7a00+NU+0u838XNq36/NgWH7L+n5W9CfpodiT56qm0rqay+cVVAxs9OV9wfYc8iW3UG0Hmw/Y67DIKd82Hzf6A0HyI6wIDLbONy91NO/JW/dTZ8cC0MuBSufrvupQBj4NMpsOEzuOFT267TAJogVKPZk1XA07M2s3x3NlkFpcfti4sMo1tcBD3iIugVE8T9y84kfdBUyE0lLm0+T/T5gh1ZRew4eIS84vKj57UJCSKlQyQl5S5SDxcy7YYRnNW/s+cAPrzO/g/70DrPD//9K+E/v4K0VfZ/zgv/2nSjasuK4ct77DxBw2+Ai/7un+6oTcXlgvx0iE70dyQNU1EGm7+Gpa/C3sV2RPrAK2233D0Lj81k26GvHdGcMt5WaUZWqfosK4btc+z3v+UbO09U+wQYeIW9XtJIW1J881yI7wm3fmPbE+qj5Ai8cbYd33LngprHxNRCE4RqsJLyCl6bv5OXfthOaHAQFw1OILlDJD3iI+geF0H3+Aiiwt0aIfctgzfPsYOqyovh09vh9jnQbbSdfbaglJ2ZBezIPMLOzCPszCzgYH4J95/Zm3MHdvEcRHkJ/CXF9l66+Lnqg3W5YPV7dsbS4lwYe7ctUTRCfW21Cg7Bh5MhdRmc9aSt/mqObQeqZulrbSP7z59AVIJNBsnjbWJoX82/S09KjtjR4Bs+t918K0rtRIcVZbYEcMf3DU+sh7bDaxOhQx/bCaSePcN0HIRqkEXbD/G/X65nZ2YBFw1J4HcXD6BzVC2jU/evsn8njrAjWSXYFq27jUZE6NCuDR3atWF0Sh3Wd9690BbvTzqv5uOCgmy32v6X2Pl5Fr8EiSfb+mFfyNwC039hf8394m37i1G1TAlD7YR/l7zYsB8UbdrZhufBV9sfKZtn2ZJFxjqY/EHjlLo69IYrXrEdIf77CFzyfMOvWYVP14MQkfNFZIuIbBeRRz3sjxWRz0XkZxFZJiKD3PbFiMhMEdksIptEZJwvY1Unyswv4aEZq7nujaVUuAzv3Daal687ufbkALaqp32C/RXWNha6jbHdNhti27cQEm4XsfFGRBxc8g/bnXTeM7Y+ubHtnAdvnANlhXYqBE0OrUNjljbDo22PqRtmwq+2NO5kh/0vgVMfgv0rbKmlkfksQYhIMPAycAEwAJgsIlX7ID4OrDHGDAFuAl5w2/cC8I0xph8wFNjkq1jV8SpchveW7OHMZ+cxa10GD5zZm9kPjffcaFydtFXQ9eRj70861/56ykurX1DG2BJI8ul1q7cNCoIzHrO9htbNrN9nV2fVu/D+VTYJTvmuefc6Uq3XmU/Y6lsfjJXxZQliNLDdGLPTGFMKzAAuq3LMAOA7AGPMZiBZRDqLSBQwHnjT2VdqjMnxYazKsX5/Lle+sognvljP4MRo/vvQ6fzPuX3rNhV2UY7tjZPoliD6ONVC276tX2BZO+yI3Nqqlzzpd4ntcTL/GTu6uzEsetF2Y00+3c6QGtujca6rVF0Fh0Cob9ZM92WCSAT2ub1Pdba5WwtcCSAio4EeQBLQE8gE/iUiq0XkDRHxOA2niEwVkRUisqK1rxrnS/nFZfz+3xu49KWF7D9cyAuThjF9yhh6dazHr5K01fZv9wTRqb8zoKqeCaKyeqrPuXU/t7IUkb0Tfv6ofp/v7tB2O+6i38Vw/Sf+H5GrlI/4MkF46sJRtcvUM0CsiKwB7gdWA+XYxvOTgVeMMcOBAuCENgwAY8xrxpiRxpiRHTvWoQpEHXUgr5jLX/6Jtxft5voxPfjulxO5bFhi/Zfc3L/S/u0+tbSIfbjvnGd7I9XV1tnQsV/9f6n3vdAuvDL/L7YnSX0ZA//5Hzvh3EXP+W/6CKWagC8TRCrg3jk3CTiuAtoYk2eMudUYMwzbBtER2OWcm2qMWeocOhObMFQjS8sp4tpX7ZxHH0wZyx8uH0R02wY+9NJW28FeVQeInXSe7YW056e6Xa8k3w5Qqk/poZIInPFbO+J3zfT6X2fdTDudxFlPQPtqxmoo1Ur4MkEsB/qISIqIhAGTgK/cD3B6KlWOJJoCLHCSRgawT0QqJ0k5C6iyvqJqqH3ZhVzz6mKyCkp5b8oYxvWKr/0kb+xf6bmnRvLpthdSXauZdvxgZ0OtT/uDuz7n2MV2FvytfqWYohyY/ZhtfB95W8NiUaoF8FmCMMaUA/cBs7E9kD42xmwQkbtEpHIi+f7ABhHZjO3t9KDbJe4HpovIz8Aw4M++ijUQ7T5UwLWvLia/uJzpU8Zwcncvp4OoTV6aHV2b6KHAFxZhBx7Vtbvrttl22cxuYxoWmwic8Tjk7rMD6erq+z/YZTAv/nvtc/Yo1Qr4dKCcMWYWMKvKtmlurxcDfao5dw2g/QZ9YPvBI1z/xhLKKgwf3DGGgV0bsZHVfYCcJ33OtT2ZDm23A31q43LBtjnQ+8zGqe/vdaadgnrBszDsBu+Xo0xdaefsGXNn85qhVCkf8ulAOdX8bMnIZ9JrS6hwwYd3jG3c5AB2/ENQCHQZ7Hl/ZTuCt6WIjLV2hHKfBlYvVRKxs2fmp9mJ2rxRUW4Xk2/X2bZjKBUgNEEEkI1peUx+fQlBAjOmjqVvFx+sxLZ/pV0Qp7p+2bE9bG+krV4miG1zALGL1zeWlPG2PWThc3Ylr9osf8NOEHj+0y1jllWlGokmiADxc2oOk19fQnhIEB/fOY7enXywQpnLZXsweWp/cNfnXNsrqSS/9mtunW2v166RuzBPfMyWTFa8VfNxeWnw/R+h11k6jYYKOJogAsCqvYe5/vWltA8P4aM7x5HcweOYw4bL3mknJqttrpmTzrO9knb8UPNxBYdsiaSxqpfcJZ9qF4pZ+Pea57D55jE7E+dFf9PZWVXA0QThK4e2wezf2jnlG9OBDXYhe5fLq8OX7crmxjeWEtcujI/uHEe3uHrOP++NNKeBumstJYhuY2yvpNraIbbNAYydx8kXzvgtFB6C5a9X8/lzYeMXMP7Xx68Ip1SA0ATR2FwuuyTgtNPsNNPrP23c6y95xY4G3vx1jYdl5Bbz3pI93PzWMjpHh/PR1HEkxtQwX8uKt2wvnYbYv9IuttKxX83HBYfaXknb5tSc6LbNtg3DXYY2LK7qdBsNvc+Bn16wy5W6KyuCWb+E+D5w6gO++XylmjlNEI0pZy+8e6ldLzZlArTvapcibEy7nOvNe/q4h2uFy7BidzZ/nb2ZC1/4kbFPf8cTX6ynZ8dIZkwdS5foGrpzHt5j55Of8zvvGm2rs3+VnU/fm2Un+5xn2wAyfva8v6IMtn9vB7f5cqGfMx6zpbylrx6//cdn4fBuuOjZei/EolRLpwsGNQZj7PQN/30UMHDpizD8RvjyXtgyyz7IG+Mhd3i3TUI9ToU9P3Fk9UzmBp3K95sPsmBbJjmFZQQHCSO6x/LI+f04o19H+nZuX/ucSvOesQ/kilK7ru6QX9Q9tooy+7AfNcW74/ucA4gdE+FpXMG+pVCS65v2B3eJI+CkC2DxizD6DrsWceZWWPg8DLkWek7w7ecr1Yxpgmio/APw7wdh63/tOrWXvwyxyXZfygSbOA6ss7+sG8opjUyPf4DTU/dR8uVT/E/pX4iNDOfMfp04s18nTu/dkeiIOgwoO7AR1n4I4+6FjV/a1/VJEAc32qVF3Sfoq0lkB/tw3jobJvzmxP1bZ0NQKPQ6o+6x1NUZj8Gr42313cRH7WR8YRFw7h99/9lKNWOaIBpiwxfw9cNQWgDn/RnG3H18SSFlvP175/xGSRCunfPJCYrjt4vKubvj9TyS/zQ/nHeIbhNuISionj1svv8jtGkPp//SzpO08DnIz6jb+rtwbAbXuqyWddJ5tsG94JBNGO62zYEep9jYfC1hqF2Za8k/7QSDu3+0M7W26+T7z1aqGdM2iPooOgyfToFPbrYDv+760f4Cr1qNFJUAHU6CXQsa/pnGULz1B+aX9eevVw/lkYd/A50H0WPdiwSZei6luW8ZbPmPbYSNiIOhk8C4YN0ndb/W/lXQNu5Y6ckbfc4BjNNbyU3OXsjc1PDJ+epi4mNQkmfbjxJHwohbm+6zlWqmNEHURUU5bPkv/HMcbPgcJj5ul/rr2Lf6c1Im2EFh5aUN+uh1q5cQUZZNUeJpXD0iySajiY9B9o76LYJjDMx9CiI72ZIPQIc+tgSwdkbdr1c5QK4uYwW6DLW9lKp2d60cZd2Q6b3rqvNAGHglSBBc/JxvG8aVaiH0/4KauFyQ/jMsfhk+uBb+Xwp8OMmuIDZlLkx8pPYJ5FLG2zUQKscI1ENOYSlz/2N/1V925eRjjc79LrLVI/VZBGf7XLsuw4TfHL+W7dDJcGC9XT/aW6UFtg2itvEPVQUF2VLE9u+Pj3/btxCbAvFeTObXmC57Ge5e1DjtRUq1Apog3BkDmVtg2evw0Y3w117w6ukw+3E78G3QVXD1v2DqfO8bY5NPA6Te3V2NMfxm5s8MLltDSVQPIjulHNt53CI4H3h/UZcL5v7eVgedfPPx+wZeaSfbq0spIv1nWzVVl/aHSn3Os72V9jlrQ5UW2iq5k85r+pHLYRF2aVSlFKCN1LbqZ+2H9qG0+0fbNx/s+sl9Lzg2sVt01eW0vRQRBwlD7PiFiY/U+fT3l+zhu41pvBS5hbDeV514QJ9z7YN5wV/tr/+QsBOPqWrDZ7Zn1ZVvnHh8ZLx9aK/7BM7+vXdjGo42UNdj0b9eZ9jeSltn22S6+0fbG6opq5eUUh5pCSIoxA4Q2/2jTQaXvggPrIGH1sHl/7QNt/VNDpVSJtgG4ToOQtuUnscf/rOJm5JzCKs44rlP/nGL4Lxb+0XLS+3CN50H2RKRJ0Mn2US5a553gaatsgm1Pr1+2rS3vZW2OavMbZ0NoZFOyUsp5U+aIIKC4J4l8MstcNUbcPJNEJfSuNUbKRPs5HR7F3t9SmFpOfd9sIrotqH8qk+G3Zg83vPBvc46tghOWXHNF179rh1wd9aT1TfEnnQehMd4X820f6X3VW7VfV7mZhvXtm/tJHo6elkpv9MEAbY7qi/ru3uMs9Uodeju+vuvNrLzUAHPXzuMyP0/QaeB1U95XVmKyE+DVe9Uf9HSApj//+xI7D7nVH9cSBsYdCVs+rr2KbkLs+2DvT7tD5UqR0svetGWhHw1OZ9Sqk40QTSFsEhIGnVsHqVa/HttGh+t2Mc9E3txao92sHdJ7VM+pIy3I7l/fLb6qqwlr9iqo7OerD0hDp0M5UWw8auajzu6xGg92h8qdehtZ0utXJtB2x+UahY0QTSVlPGQvrbW6b/3ZhXy+GfrOLl7DA+dfRKkLreNtinVVC9VqixFVLcITmG2nbW074XQfUzt8SaNsg/ttR/WfFzaKkAgYVjt16xJn/NsT6gugyGqa8OupZRqFJogmkrPCfYBuPunag8pq3Bx/4zVIPDCpOGEBgfZUocE22qh2rgvglNacPy+hX+31UVnPuFdvCIwZBLsXgg5+6o/bv9KO1q8oUtxVlYr+XpyPqWU1zRBNJXEkXathBqqmf727RbW7svhL1cNObawz05nzIW3D+CJj9tFcJa9dmxb7n77fugk6DzA+5iHXAMYWPex5/3G2CqmhrQ/VEoeD+N/A6Nub/i1lFKNQhNEUwkJg+7jqm2oXrA1k1fn7+S6Md25cHCC3ViSb3+h12XK6e5joPfZxy+CM/8vtvQy8bG6xRyXYmNe+5FNBlXlpkLBwYa1P1QKDoEzf6vVS0o1I5ogmlLKeNudMz/juM0H84v5n4/XclLndvzuYrdf+HsWgamw3WTr4ozHbVvHslftCPDV78PI2+3EgnU1dBIc2mLnWqoqrREaqJVSzZYmiKZUWRLY9ePRTcVlFUx9dyVHSsp4cfLJhIcGHzt+1wIIbmOXxqyLykVwFr0I3zwGoW3tdN71MeByG4OnMRH7V9ruu50H1e/aSqlmTRNEU+oyxA5Ac0You1yGX32ylrWpOTx/7XD6dqmy9sHO+bbKKLSGtaSrc8ZjUJwL2+fAuPuqH0NRm7YxdsqR9TNPnBBw/yrb60gHtSnVKmmCaEpBwXYKCacd4vm5W/n653QePb8f5w+qskBPQZadL6m27q3VSRhqf/1HdrJrVTTE0MlQmGVngK3kckHaGq1eUqoV8ypBiMinInKRiGhCaaieEyFnL9/8uJh/fL+da0d2Y+r4nicet9tpzE6ZWP/PumKanUakoV1Qe58FER2OHxORtQ1K8xunB5NSqlny9oH/CnAdsE1EnhGRft6cJCLni8gWEdkuIo962B8rIp+LyM8iskxEBlXZHywiq0Xkay/jbP6cEsHC2Z8yrmc8f7h80LH1HdztnA9h7Rs2x1FoWzs7a0MFh8Lgq2HLN8cG+lXO4FrXNSCUUi2GVwnCGDPXGHM9cDKwG5gjIotE5FYR8bhijogEAy8DFwADgMkiUrUT/uPAGmPMEOAm4IUq+x8ENnl7My3BHkkkk1jODN/MtBtGEBZSzVewa4Ed+ObNdNtNYci1UFFi1+EG2/4Q1t6uQqeUapW8rjISkXjgFmAKsBr7MD8ZmFPNKaOB7caYncaYUmAGcFmVYwYA3wEYYzYDySLS2fm8JOAi4A1vY2zucovKuO2dFSxjEBNCNxHdtpqHf26qXUq0vu0PvtB1OHToe2x50/0roesw266ilGqVvG2D+Az4EYgALjHGXGqM+cgYcz/QrprTEgH3ORpSnW3u1gJXOp8xGugBJDn7ngd+A7hqiW2qiKwQkRWZmZne3I5flFW4uGf6SvZmF9LvlIsJLjoEB6spHFUOpqvr+AdfErFjIvYutqvuHVivDdRKtXLeliBeMsYMMMY8bYxJd99hjBlZzTmepgutOhz3GSBWRNYA92NLJuUicjFw0BizsrbAjDGvGWNGGmNGduxYz66cPmaM4XdfbuCn7Vk8feUQeo2+0O6obtqNnfNto3CnOkyL0RSGXAOIXWCpolTbH5Rq5bxNEP1FJKbyjdO4fE8t56QC3dzeJwFp7gcYY/KMMbcaY4Zh2yA6AruAU4FLRWQ3tmrqTBF538tYm503F+7iw2V7uWdiL64ekQQx3SE2xfO0G8bY7SmnV7+gj79EJ9m4tn5j32sPJqVaNW+fQHcYY3Iq3xhjDgN31HLOcqCPiKSISBgwCThucQERiXH2gW3bWOAkjceMMUnGmGTnvO+NMTd4GWuzMmfjAf40axMXDu7Cr87te2xHzwl2ptSK8uNPyNpuF/5pTtVL7oZOtn9HdrQJQynVanmbIILErS+m00MprIbjMcaUA/cBs7E9kT42xmwQkbtE5C7nsP7ABhHZjO3t9GBdb6A525CWy4MzVjMkMZpnfzGMoCC3WreU8VCSB+lrjj9p57xj+5uj/pdASFtbevDlKnxKKb/ztg/lbOBjEZmGbUe4C/imtpOMMbOAWVW2TXN7vRiosZ+kMWYeMM/LOJuNjNxibn97BTFtQ3n9ppG0DavS26dyfeld8yHJrRln1wKI7mYX62mO2rSHyR9C+wR/R6KU8jFvSxCPAN8DdwP3Yrum/sZXQbV0uYVl3PzWMo6UlPPGzaPoFBV+4kHtOtp1pne6NVS7XLD7R1t6aM6/znudAZ28GiuplGrBvCpBGGNc2NHUr/g2nJavuKyCO95dwc5DR3j71tEM6FrDNBc9J9jlQcuKITTczr1UdLj5tj8opQKKt+Mg+ojITBHZKCI7K//4OriWpsJleGjGGpbtzubZa4Zxau8ONZ+QMt6uN526zL6vLE001/YHpVRA8baK6V/Y0kM5cAbwLvCer4JqiexYh/V8syGD3108gEuHerEyWo9T7XrTld1ddy2w6ztHaf2+Usr/vE0QbY0x3wFijNljjHkKONN3YbU8L36/nelL93LnhJ7cdlqKdyeFR9nRyDvnQ3mpXUFOq5eUUs2Etwmi2Jnqe5uI3CciVwCdfBhXizJj2V6em7OVK09O5NHz69h4mzLezmu0az6UFWj1klKq2fA2QTyEnYfpAWAEcANws49ialHmbDzA45+vY8JJHfnLVUM8T91dk5QJdt3peU8DYhcUUkqpZqDWXkzOoLhrjDG/Bo4At/o8qhZi5Z5s7vtgFYMTo/nn9ScTGlyPqTG6jbZrPu9faVeBi4hr/ECVUqoean2iGWMqgBFS55/Grdv2g/nc9vYKEqLDeeuWUUS2qee6DaFt7brToO0PSqlmxdun2mrgSxH5BCio3GiM+cwnUTVz6blF3PTmMkKDg3j3tjHEt2vTsAumjHcm6NMEoZRqPrxNEHFAFsf3XDJAwCWI3MIybnlrOXnF5cyYOpbu8RENv+jwm6C0UBuolVLNircjqbXdAXC5DFPfs6Ok37l1NIMSoxvnwu07w9lPNs61lFKqkXiVIETkX5y42A/GmNsaPaJmbN/hQpbuyuaxC/pxSm2jpJVSqoXztorpa7fX4cAVVFn8JxCk5RQDMLBrI5UclFKqGfO2iulT9/ci8iEw1ycRNWPpuUUAJMR4mJ1VKaVamfquadkH6N6YgbQE6bm2BNE1uq2fI1FKKd/ztg0in+PbIDKwa0QElPTcImIiQk9c/EcppVohb6uY2vs6kJYgPaeYLp4W/1FKqVbI2/UgrhCRaLf3MSJyuc+iaqbScovpGqPVS0qpwOBtG8STxpjcyjfGmBwg4DruZ+QWkRCtJQilVGDwNkF4Oq6ekw+1TEWlFRwuLNMShFIqYHibIFaIyHMi0ktEeorI34GVvgysuTnaxVVLEEqpAOFtgrgfKAU+Aj4GioB7fRVUc1TZxbWLJgilVIDwthdTAfCoj2Np1nQMhFIq0Hjbi2mOiMS4vY8Vkdk+i6oZSs+xVUxaglBKBQpvq5g6OD2XADDGHCbA1qROyy0mPjKM8FAdJKeUCgzeJgiXiBydWkNEkvEwu2trlp5bpKUHpVRA8bar6m+BhSIy33k/Hpjqm5Cap4zcYpJiG2FxIKWUaiG8KkEYY74BRgJbsD2ZfontyVQjETlfRLaIyHYROaGR22nL+FxEfhaRZSIyyNneTUR+EJFNIrJBRB6s0135QFpOEV11FlelVADxdrK+KcCDQBKwBhgLLOb4JUirnhMMvAycA6QCy0XkK2PMRrfDHgfWGGOuEJF+zvFnAeXAL40xq0SkPbBSROZUObfJFJSUk1dcToL2YFJKBRBv2yAeBEYBe4wxZwDDgcxazhkNbDfG7DTGlAIzgMuqHDMA+A7AGLMZSBaRzsaYdGPMKmd7PrAJSPQy1kang+SUUoHI2wRRbIwpBhCRNs7DvG8t5yQC+9zep3LiQ34tcKVz3dFAD2wp5SinQXw4sNTTh4jIVBFZISIrMjNry1n1UzkGQhOEUiqQeJsgUp1xEF8Ac0TkS2pfclQ8bKva8+kZIFZE1mBHa6/GVi/ZC4i0Az4FHjLG5Hn6EGPMa8aYkcaYkR07dvTiVuou3VlqVOdhUkoFEm9HUl/hvHxKRH4AooFvajktFejm9j6JKknFeejfCiAiAuxy/iAiodjkMN0Y85k3cfpKmlPF1FnXglBKBZA6z8hqjJlf+1EALAf6iEgKsB+YBFznfoBTKil02iimAAuMMXlOsngT2GSMea6uMTa29JxiOrRrQ1hIfVdoVUqplsdnU3YbY8pF5D5gNhAMvGWM2SAidzn7pwH9gXdFpALYCNzunH4qcCOwzql+AnjcGDPLV/HWJD2vWLu4KqUCjk/XdHAe6LOqbJvm9nox0MfDeQvx3IbhF+k5RfTsGOnvMJRSqklpnYkX0nOLdQyEUirgaIKoRV5xGUdKyrWLq1Iq4GiCqEVG5RgI7eKqlAowmiBqkeasA9FVSxBKqQCjCaIW6VqCUEoFKE0QtUjPKUIEOrVv4+9QlFKqSWmCqEV6bjGd2rchNFj/UymlAos+9WqhXVyVUoFKE0Qt0nJ1oSClVGDSBFEDYwzpOcV0idIShFIq8GiCqEFeUTlFZRVaglBKBSRNEDVIO7qSnJYglFKBRxNEDY4uNaolCKVUANIEUYO0HF1qVCkVuDRB1CAjt5jgIKFTe00QSqnAowmiBmm5RXRu34bgoGazNIVSSjUZTRA1SM8p1jmYlFIBSxNEDdJzi+ii7Q9KqQClCaIaxhjSc4t1mm+lVMDSBFGNw4VllJS7dAyEUipgaYKoxtGFgnQMhFIqQGmCqEblQkFdtAShlApQmiCqkZGrS40qpQKbJohqpOUWExosdGinK8kppQKTJohqpOcU0TkqnCAdJKeUClCaIKphu7hq+4NSKnBpgqhGem6xDpJTSgU0TRAeuFyGjNxineZbKRXQfJogROR8EdkiIttF5FEP+2NF5HMR+VlElonIIG/P9aWsglJKK1xaxaSUCmg+SxAiEgy8DFwADAAmi8iAKoc9DqwxxgwBbgJeqMO5PpORq+tAKKWUL0sQo4HtxpidxphSYAZwWZVjBgDfARhjNgPJItLZy3N9RpcaVUop3yaIRGCf2/tUZ5u7tcCVACIyGugBJHl5Ls55U0VkhYisyMzMbJTA03N0qVGllPJlgvA0gMBUef8MECsia4D7gdVAuZfn2o3GvGaMGWmMGdmxY8cGhHtMem4xYSFBxEeGNcr1lFKqJQrx4bVTgW5u75OANPcDjDF5wK0AIiLALudPRG3n+lJ6bjEJ0eHYkJRSKjD5sgSxHOgjIikiEgZMAr5yP0BEYpx9AFOABU7SqPVcX0rPLaJLlFYvKaUCm89KEMaYchG5D5gNBANvGWM2iMhdzv5pQH/gXRGpADYCt9d0rq9irSotp5jRKXFN9XFKKdUs+bKKCWPMLGBWlW3T3F4vBvp4e25TqHAZDuQVaxdXpVTA05HUVWQdKaHcZUiI0S6uSqnApgmiirTKQXLaBqGUCnCaIKrQMRBKKWVpgqiisgSh8zAppQKdJogqMnKLCA8NIiYi1N+hKKWUX2mCqCItt5iE6LY6SE4pFfA0QVSRnlOkXVyVUgpNECdId0oQSikV6DRBuCmvcHEwv4Su2oNJKaU0QbjLPFJChcvoWtRKKYUmiOOk5WgXV6WUqqQJwk16rg6SU0qpSpog3Bxbi1pLEEoppQnCTVpOMRFhwUSF+3SSW6WUahE0QbhJzy3SleSUUsqhCcJNWm4xXXWab6WUAjRBHCcjV0dRK6VUJU0QjjJnkFwXbaBWSilAE8RRB/KKMQa6aglCKaUATRBHpVd2cdU2CKWUAjRBHJV+dKEgLUEopRRogjiqcqlRnYdJKaUsTRCO9Nxi2rcJoX24riSnlFKgCeKotJwinYNJKaXcaIJwZOTpQkFKKeVOE4QjLadYB8kppZQbTRBASXkFh46UaAlCKaXcaIIADuSWALoOhFJKufNpghCR80Vki4hsF5FHPeyPFpF/i8haEdkgIre67XvY2bZeRD4UEZ89vSsXCtKV5JRS6hifJQgRCQZeBi4ABgCTRWRAlcPuBTYaY4YCE4FnRSRMRBKBB4CRxphBQDAwyVexHhtFrSUIpZSq5MsSxGhguzFmpzGmFJgBXFblGAO0F7sAQzsgGyh39oUAbUUkBIgA0nwVaFrlUqPaSK2UUkf5MkEkAvvc3qc629y9BPTHPvzXAQ8aY1zGmP3A34C9QDqQa4z51leBpucUE902lIgwXUlOKaUq+TJBeFqWzVR5fx6wBugKDANeEpEoEYnFljZSnH2RInKDxw8RmSoiK0RkRWZmZr0CTc/VLq5KKVWVLxNEKtDN7X0SJ1YT3Qp8ZqztwC6gH3A2sMsYk2mMKQM+A07x9CHGmNeMMSONMSM7duxYr0DTc4t0JTmllKrClwliOdBHRFJEJAzbyPxVlWP2AmcBiEhnoC+w09k+VkQinPaJs4BNvgo0PbdYJ+lTSqkqfFbpbowpF5H7gNnYXkhvGWM2iMhdzv5pwB+At0VkHbZK6hFjzCHgkIjMBFZhG61XA6/5Ik6XyzDhpI6MSo71xeWVUqrFEmOqNgu0XCNHjjQrVqzwdxhKKdViiMhKY8xIT/t0JLVSSimPNEEopZTySBOEUkopjzRBKKWU8kgThFJKKY80QSillPJIE4RSSimPNEEopZTyqFUNlBORTGBPPU/vABxqxHCaC72vlqe13ltrvS9o2ffWwxjjcSK7VpUgGkJEVlQ3mrAl0/tqeVrrvbXW+4LWe29axaSUUsojTRBKKaU80gRxjE9mi20G9L5antZ6b631vqCV3pu2QSillPJISxBKKaU80gShlFLKo4BPECJyvohsEZHtIvKov+NpTCKyW0TWicgaEWmxKymJyFsiclBE1rttixOROSKyzfm7RS4JWM29PSUi+53vbY2IXOjPGOtDRLqJyA8isklENojIg872Fv291XBfLf478ySg2yBEJBjYCpwDpGLX0Z5sjNno18AaiYjsBkY6y7i2WCIyHjgCvGuMGeRs+39AtjHmGSexxxpjHvFnnPVRzb09BRwxxvzNn7E1hIgkAAnGmFUi0h5YCVwO3EIL/t5quK9raOHfmSeBXoIYDWw3xuw0xpQCM4DL/ByTqsIYswDIrrL5MuAd5/U72P9JW5xq7q3FM8akG2NWOa/zgU1AIi38e6vhvlqlQE8QicA+t/eptK4v2wDfishKEZnq72AaWWdjTDrY/2mBTn6Op7HdJyI/O1VQLaoapioRSQaGA0tpRd9blfuCVvSdVQr0BCEetrWmOrdTjTEnAxcA9zrVGar5ewXoBQwD0oFn/RpNA4hIO+BT4CFjTJ6/42ksHu6r1Xxn7gI9QaQC3dzeJwFpfoql0Rlj0py/DwKfY6vUWosDTn1wZb3wQT/H02iMMQeMMRXGGBfwOi30exORUOxDdLox5jNnc4v/3jzdV2v5zqoK9ASxHOgjIikiEgZMAr7yc0yNQkQinUY0RCQSOBdYX/NZLcpXwM3O65uBL/0YS6OqfIA6rqAFfm8iIsCbwCZjzHNuu1r091bdfbWG78yTgO7FBOB0R3seCAbeMsb8yb8RNQ4R6YktNQCEAB+01HsTkQ+BidgplQ8ATwJfAB8D3YG9wC+MMS2usbeae5uIraowwG7gzsp6+5ZCRE4DfgTWAS5n8+PY+voW+73VcF+TaeHfmScBnyCUUkp5FuhVTEoppaqhCUIppZRHmiCUUkp5pAlCKaWUR5oglFJKeaQJQqlmQEQmisjX/o5DKXeaIJRSSnmkCUKpOhCRG0RkmTPn/6siEiwiR0TkWRFZJSLfiUhH59hhIrLEmcDt88oJ3ESkt4jMFZG1zjm9nMu3E5GZIrJZRKY7o3aV8htNEEp5SUT6A9diJ0EcBlQA1wORwCpnYsT52NHQAO8CjxhjhmBH3lZunw68bIwZCpyCndwN7MygDwEDgJ7AqT6+JaVqFOLvAJRqQc4CRgDLnR/3bbGTzbmAj5xj3gc+E5FoIMYYM9/Z/g7wiTM/VqIx5nMAY0wxgHO9ZcaYVOf9GiAZWOjzu1KqGpoglPKeAO8YYx47bqPIE1WOq2n+mpqqjUrcXleg/38qP9MqJqW89x1wtYh0gqPrK/fA/n90tXPMdcBCY0wucFhETne23wjMd9YOSBWRy51rtBGRiKa8CaW8pb9QlPKSMWajiPwvdpW+IKAMuBcoAAaKyEogF9tOAXY662lOAtgJ3OpsvxF4VUT+z7nGL5rwNpTyms7mqlQDicgRY0w7f8ehVGPTKiallFIeaQlCKaWUR1qCUEop5ZEmCKWUUh5pglBKKeWRJgillFIeaYJQSinl0f8H0Ndi10dMcn4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8/fFQqAAAACXBIWXMAAAsTAAALEwEAmpwYAABPlUlEQVR4nO2dd3xUVdrHv096CEkIECAkdCnSm1hAxY5YsKJY1raWtay66+66+77b13d117WtXXTVXcvaRUGxIgiCgPTQeyCNQEIS0nPeP85MGMJkMjOZkkye7+eTz8zce8+dc5PM/O55qhhjUBRFUZTGRIV7AoqiKErrRAVCURRFcYsKhKIoiuIWFQhFURTFLSoQiqIoiltUIBRFURS3qEAoSgAQkZdF5C9eHrtDRM5s6XkUJdioQCiKoihuUYFQFEVR3KICobQbHKadX4jIahEpF5EXRaS7iHwiIqUi8oWIpLkcf6GIrBORYhGZJyLHuuwbIyI/OMb9F0ho9F7ni8hKx9hFIjLSzznfLCJbRGS/iMwSkZ6O7SIij4pIgYiUOK5puGPfVBHJdsxtj4jc59cvTGn3qEAo7Y1LgbOAQcAFwCfAb4Cu2M/DTwFEZBDwBnAPkA7MAT4SkTgRiQM+AP4NdAbedpwXx9ixwEvArUAX4DlglojE+zJRETkd+CswHcgAdgJvOnafDZziuI5OwBVAkWPfi8CtxphkYDjwlS/vqyhOVCCU9sY/jTH5xpg9wAJgiTFmhTGmCngfGOM47gpgtjHmc2NMDfAwkAicBJwAxAKPGWNqjDHvAEtd3uNm4DljzBJjTJ0x5hWgyjHOF64GXjLG/OCY36+BE0WkL1ADJANDADHGrDfG5DrG1QBDRSTFGHPAGPODj++rKIAKhNL+yHd5XuHmdUfH857YO3YAjDH1wG4g07Fvjzmy0uVOl+d9gJ87zEvFIlIM9HKM84XGcyjDrhIyjTFfAU8CTwH5IvK8iKQ4Dr0UmArsFJFvROREH99XUQAVCEVpir3YL3rA2vyxX/J7gFwg07HNSW+X57uBB4wxnVx+Ohhj3mjhHJKwJqs9AMaYJ4wx44BhWFPTLxzblxpjpgHdsKawt3x8X0UBVCAUpSneAs4TkTNEJBb4OdZMtAj4DqgFfioiMSJyCTDBZewLwG0icrzDmZwkIueJSLKPc3gduEFERjv8F/+HNYntEJHjHOePBcqBSqDO4SO5WkRSHaaxg0BdC34PSjtGBUJR3GCM2QhcA/wT2Id1aF9gjKk2xlQDlwDXAwew/or3XMYuw/ohnnTs3+I41tc5fAn8FngXu2oZAFzp2J2CFaIDWDNUEdZPAnAtsENEDgK3Oa5DUXxGtGGQoiiK4g5dQSiKoihuUYFQFEVR3KICoSiKorhFBUJRFEVxS0y4JxBIunbtavr27RvuaSiKorQZli9fvs8Yk+5uX0QJRN++fVm2bFm4p6EoitJmEJGdTe1TE5OiKIriFhUIRVEUxS0qEIqiKIpbIsoH4Y6amhpycnKorKwM91SCSkJCAllZWcTGxoZ7KoqiRAgRLxA5OTkkJyfTt29fjiy+GTkYYygqKiInJ4d+/fqFezqKokQIEW9iqqyspEuXLhErDgAiQpcuXSJ+laQoSmiJeIEAIlocnLSHa1QUJbS0C4Fo9dTXwaEi0Mq6iqK0IlQggkxxcTFPP/2054Mqi6F4F9QeNhFNnTqV4uLioM5NURTFEyoQQaYpgairc2nyVVdrH+trGzbNmTOHTp06BXl2iqIoTRPxUUzh5v7772fr1q2MHj2a2NhYOnbsSEZGBitXriQ7O5uLLrqI3Tu2UVlRzt133cUtd94DHC4bUlZWxrnnnsukSZNYtGgRmZmZfPjhhyQmJob3whRFiXiCKhAiMgV4HIgGZhpjHmziuOOAxcAVxph3fBnrC3/8aB3Zew+29DRHMLRnCr+/YFiT+x988EHWrl3LypUrmTdvHueddx5r165tCEd96aWX6MxBKg7kctwFN3LpjGvp0qXLEefYvHkzb7zxBi+88ALTp0/n3Xff5ZprtIukoijBJWgCISLRwFPAWUAOsFREZhljst0c9xAw19exbZEJEyYckavwxBNP8P47b0J9Pbv35LF58+ajBKJfv36MHj0agHHjxrFjx44QzlhRlPZKMFcQE4AtxphtACLyJjANaPwlfxe2Kftxfoz1CU93+qEiKSmp4fm8efP44osv+O7j1+kQJ0y+4na3uQzx8fENz6Ojo6moqAjJXBVFad8E00mdCex2eZ3j2NaAiGQCFwPP+jrW5Ry3iMgyEVlWWFjY4kkHmuTkZEpLS93uKykpIS0tjQ4JsWzYsp3FS1eEeHaKoihNE8wVhLvMrcaB/o8BvzLG1DVK9PJmrN1ozPPA8wDjx49vdYkEXbp0YeLEiQwfPpzExES6d+/esG/KlCk8++yzjDz9Egb378sJ40eHb6KKoiiNCKZA5AC9XF5nAXsbHTMeeNMhDl2BqSJS6+XYNsPrr7/udnt8fDyffDwL8tfaDXEdoetAgAY/Q9euXVm7dm3DmPvuuy+oc1UURXESTIFYCgwUkX7AHuBK4CrXA4wxDd5aEXkZ+NgY84GIxDQ3NmKoqzn8vL6u6eMURVFCTNAEwhhTKyJ3YqOTooGXjDHrROQ2x/7GfodmxwZrrmHFmRwXHX9EopyiKEq4CWoehDFmDjCn0Ta3wmCMub65sRFJvWMFEZMA1e6d2YqiKOFAS22EG+eqITYeTD3U14d3PoqiKA5UIMJNXS0gEB1nXxs1MymK0jpQgQg39TUQHQtRDmufOqoVRWklqECEm7paKw4OgeiYlh7mCSmKolhUIMJNvUMgJDrcM1EURTkCLfcdZH71q1/Rp08fbr/9dgD+8Ic/ICLMnz+fAwcOUFNRyl9++yumTb/WMaLVJYMritJOaV8C8cn9kLcmsOfsMQLObboS+ZVXXsk999zTIBBvvfUWn376Kffeey8pycnsW/c1J1x4Exdecb3b+iKKoijhon0JRBgYM2YMBQUF7N27l8LCQtLS0sjIyODee+9l/vxviKqrZk9uHvkFhfRQiVAUpRXRvgTCw51+MLnssst45513yMvL48orr+S1116jsLCQ5YsXElu8lb4nTqOyqgoS1A+hKErroX0JRJi48sorufnmm9m3bx/ffPMNb731Ft26dSM2Cr5euJSduxyVzaP0z6EoSutBv5FCwLBhwygtLSUzM5OMjAyuvvpqLrjgAsafdAqjh/RjyODB9sCoaPVRK4rSalCBCBFr1hx2jnft2pXvvvsOygrg4B7r6I6KgaJtlG1fHsZZKoqiHEbzIMJJfQ0gh3MgoqK1oquiKK0GFYhw4syidnbTi4oBo6U2FEVpHbQLgTCmlRr2nXWYnERF+13RtdVeo6IobZaIF4iEhASKiopa5xeocwXhJMphavJxFWGMoaioiISEhABOTlGU9k7EO6mzsrLIycmhsLAw3FM5moN7baOggmr7uvoQHNoH+9cfubLwgoSEBLKysoIwSUVR2isRLxCxsbH069ev+QNDTX09/GUSnPRTGPd7u23r1/D+dLh+DvSdGN75KYrS7ol4E1OrpbLYRix17HZ4W4fO9rHiQFimpCiK4ooKRLgoK7CPSS79HxLT7KMKhKIorYCgCoSITBGRjSKyRUTud7N/moisFpGVIrJMRCa57NshImuc+4I5z7BQ7hAI1xVEg0DsD/18FEVRGhE0H4SIRANPAWcBOcBSEZlljMl2OexLYJYxxojISOAtYIjL/tOMMfuCNcew0rCCcBGIuI42qklXEIqitAKCuYKYAGwxxmwzxlQDbwLTXA8wxpSZw/GnSbSnSkTljqgq1xWECCR2VoFQFKVVEEyByAR2u7zOcWw7AhG5WEQ2ALOBG112GeAzEVkuIrc09SYicovDPLWsVYayNkVZgV0tJHQ6cntimgqEoiitgmAKhLvuN0etEIwx7xtjhgAXAX922TXRGDMWOBe4Q0ROcfcmxpjnjTHjjTHj09PT3R3SOikvsA7qqEZ/gsQ0OKQ+CEVRwk8wBSIH6OXyOgvY29TBxpj5wAAR6ep4vdfxWAC8jzVZRQ5lhUdGMDnp0BkqikM+HUVRlMYEUyCWAgNFpJ+IxAFXArNcDxCRY0RspToRGQvEAUUikiQiyY7tScDZwNpgTLKmrp6f/Gc5by3b3fzBgaS84Ej/gxM1MSmK0koIWhSTMaZWRO4E5gLRwEvGmHUicptj/7PApcCPRKQGqACucEQ0dQfed2hHDPC6MebTYMwzNjqKlbuLiY+JYvr4Xs0PCBRlhZB+7NHbE9M0zFVRlFZBUEttGGPmAHMabXvW5flDwENuxm0DRgVzbq4MzUghO/dgqN4OjHGsINyYmBI7Qc0hqKmEWC2+pyhK+NBMamBozxS2FpZTWROiXgyVJVBXfWQOhJNER7mNyuLQzEVRFKUJVCCwK4i6esOm/NLQvKG7HAgnWm5DUZRWggoEdgUBkL03RGYmd3WYnDgFQkNdFUUJMyoQQK+0DnSMjwmdH8JdHSYnuoJQFKWVoAIBREUJx2Ykh3AF4TAxufNBaMlvRVFaCSoQDoZmpLA+9yD19SEoB1VeABJ1WAxc0RWEoiitBBUIB0N7plBeXceu/YeC/2ZlBdCh6+Ee1K40VHRVH4SiKOFFBcLB0IxUgND4IcoL3fsfwFHRVbOpFUUJPyoQDgZ270h0lITGD1FW4D6CyYmW/FYUpRWgAuEgITaaY9I7hmgF0UQdJie6glCU8FJ50FY8aOeoQLgwtGdK8FcQxjRdydVJYhocUoFQlLBQcQAeORZWvRnumYQdFQgXhmakkHewkqKyquC9SXUZ1FZ4XkF0UBOTooSN3NX2c7ptXrhnEnZUIFxwZlSvzw1iyQ13vagb015NTKX5uqxXwk/+Ovu4Z3l459EKUIFw4dgMR8mN3JLgvUlDHSZPJqZOUFMOtUFcybQ2SvbAo8Ng/UfhnonS3sl3tJ4p2tzum3epQLjQOSmOjNSE4PohvF1BQPtaReSugvoaKNwQ7pko7Z28NTYfCSB3ZVinEm5UIBoR9N4QZfn20WMUUzsst1HgWNaX5IR3Hkr7ps5xkzL8Evu6nZuZVCAaMSzYvSHKCwGxmdRN0R5XEAXr7ePBPeGdh9K+2bfZ9mrpezJ0HgB7fgj3jMKKCkQjhvYMcm+IsgIbpRTtoZlfeyz5nZ9tH0tUIJQw4nRQdx8GmeN0BRHuCbQ2GkpuBMsPUV7o2f8A7W8FUVttHYKgJiYlvOSvgeg46DrICkRpLhzcG+5ZeaaqFOqDY/FQgWhEVloiycHsDVHWRC9qV9pbye+izVBfCz1GQnWpbcmqKOEgby2kD4boWCsQ0PpXEfMehMdGQl1twE8dVIEQkSkislFEtojI/W72TxOR1SKyUkSWicgkb8cGC9sbIogZ1eUFza8gGiq6thOBcPofBp5tH9XMpISL/LXQfYR93mOE/Ry2ZoEwBjbMhu5DPZut/SRoAiEi0cBTwLnAUGCGiAxtdNiXwChjzGjgRmCmD2ODxtCeQewNUeahkquThoqu7cQHkb/OfhD7T7av1VGthIOyQhtl2GO4fR2bAN2Ht26BKFgPB7bD4KlBOX0wVxATgC3GmG3GmGrgTWCa6wHGmDJjGlJnkwDj7dhgMjQjSL0hqsttApynOkxO2lNF14L10GUgpPW1r9UPoYQDZ4Jc9+GHt2WOgz0roL4+PHNqjg2zAWmTApEJ7HZ5nePYdgQicrGIbABmY1cRXo91jL/FYZ5aVlhYGJCJO0tuBNwPUeahF3Vj2lO5jYJ1domcnGE77ekKQgkHTQlEdenhIIrWxsbZkDUekrsH5fTBFAhxs+0om40x5n1jzBDgIuDPvox1jH/eGDPeGDM+Pd2LO3MvOKZbR2KC0Rui3EMv6sa0F4GoKoXiXdDtWGtD7dhDfRBKeMhba29Skroc3taaHdUlObB3BQw5L2hvEUyByAF6ubzOApqMFzPGzAcGiEhXX8cGmoTYaI7pFoTeEA0rCG9MTO2k5HeBo7RGt2H2MTUTDqqJSQkD+WuPXD0AdB0IccmtUyA2fmIfB7dNgVgKDBSRfiISB1wJzHI9QESOERFxPB8LxAFF3owNNkODEclU7kUdJiftpeR3gSNBrtux9jElU1cQSuiprYbCjYcd1E6ioqHn6NYpEBtmW99d+qCgvUXQBMIYUwvcCcwF1gNvGWPWichtInKb47BLgbUishIbtXSFsbgdG6y5umNozyD0hihzmpi8WUF0ah8VXQuyITYJOvWxr1OzrA9Cy34roWTfRlsssvEKAqyZKW8t1FSGfl5NUVEMOxYE1bwEEPjAWReMMXOAOY22Pevy/CHgIW/HhpKhGYd7Q0waGB+Yk5YXQEIniIlr/tiGbOrioDmgWgUF2dBtCEQ57lVSs6C20pYZcbUFh4Lt82Hte3D+ozbUWGk/NJTYaEIg6musCSprfGjn1RRbvrDJpUEWCM2kboKg9IYoa6YXtSsNAhHhuRD52dDNJcUlxRGsFg4/xIrXYPm/DifuKe2HvDUQHQ9djjl6X4OjuhUV7tvwsTVVZwZXsFQgmiAtKY6ege4N4U0dJiftoeR3WSEc2nekQKQ6BCIcfojcVfZx06ehf28lvOSvPRxJ15iUnja6rrX4IWqrYPPnMPjcwyvvIKEC4YGhPQPcG8KbOkxO2kPBPmcPiO6uK4gs+xjqXIjqQ9YODbBpbmjfWwkvxlgfQ2MHtROR1lXZdfsC2zN7yPlBfysVCA8MzQhwbwifVhDtQSAcphzXFURSOkTFhj6bOn8dmHpbfyfneygvCu37K+GjLN+uZJ01mNyRObb1tCDd8LEN7Oh3StDfSgXCAwHtDVFTCVUHfV9BRHJPiPx1tnGSq18mKsou6UMtEM7Wkqf+ygrFli9C+/5K+GjIoB7W9DGZY+3j3hXBn48n6utt/sPAM22tqCCjAuGBgPaG8CUHAiA+OfIruhasP9K85MQZ6hpKcldZv8/g8+zfSP0Q7Yc8h0A0ZWIC6DnGPobbzLT3ByjLC4l5CVQgPBLQ3hDOHAhvo5gaKrpGqEDU11uB6OZGIMKRLJe7CjJG2RXMoLNhy5e2P7ES+eSvtb4v56rdHYlpNsIp3JFMGz4GiYaBZ4Xk7VQgPBDQ3hDlPhTqcxLJJb+Ld9pEQHcCkZoJpXuD1iXrKGqrrFhljLKvB02BqhLYvSQ076+EF08Oalcyx8GeZeFN4twwB/pO8ixmAUQFohkC1huizEcTE0R2yW93DmonKZk2Ccj5OwvFXOprDgtE/8m27aSamSKfmkrYt8l9glxjMsdZh3a4WpDu22wj7YKcHOeKCkQzBKw3RIMPwoeKs5FsYnKGuHYbcvS+1BCHuuatto9OgYhPtndpGu4a+ezbCKbO+xUEhM8PsWG2fQxS7wd3qEA0Q8B6Q5QVQnyqb5EHiWmtI6wuGBSsh0697ZdxY5zZ1KGKZMpdBfEpkNbv8LZBU+ydZdHW0MxBCQ9OB7U3K4juw20IdrgEYuMc27e9U6/mjw0QKhDNELDeEOU+JMk5SUyL3DDX/OzDJb4bE+oVRO4q+8FzzUp19sfe/Flo5qCEh/y1EJMInfs3f2xsgl1phEMgygpg9/chi15yogLRDAHrDVHmQ5Kckw5pkVnRtbbaJh05S3w3JjHNfmhDEclUV2vvIp3mJSed+0HXweqHiHTy1thQ66ho747PHAd7V4YugMLJxk8AE1L/A6hAeEVAekP4u4KAyDMzFW22TuimEpNEQtc4qGgz1FYcLRAAg86BHQtt1zsl8jDGfZMgTzhbkO4LcQvSDbOtSdZTMl8QUIHwgoD0higr8H0FEanlNhoimJpYQYA1M4ViBeEs0Jcx8uh9g6bY6KatXwd/HkroKc21n60eHkpsNKaljmp/cmuqymDbPGteCnEZehUIL3DtDeEXtdVQWexbDgS03pLfs++DpS/6Pz5/nc0S7zKw6WNSQpRNnbvKmrPczaXX8ZCQqtFMkUqeFyU2GtOlBS1IV78Nf82Cla/7Nm7rl1BXFdLoJScqEF7Q4t4Q5T50knOlNZb8rjwIy16EBf+w2dD+ULDeftA8NU5KzYTSvOBnM+euso5Hd2Weo2PgmLNg81z/r1VpveSvsY++CERUFGSOsSUvfOHQfvjkl9Z38cFP4LunvR+7Yba9Wex9om/vGQBUILygxb0h/MmihtZpYtr9vS1md3AP7PrOv3MUrHNfg8mVlEzAWDNAsKivh9zV7v0PTgZNsQIf7iJtSuDJW2vt+gmpvo3zpwXp57+1xTpv+gyOvRDm/hq++kvzWdl1NTZQYtC57m9igowKhJe0qDdEQy/qCBCInQuteSgmEda+4/v4qlIo3uXZ/wChaRx0YLt1OHoSiGPOAInSaKZIJH+t5xLfTeHagtQbdi6CFf+BE++wVWEvfxnGXAvz/w6zf+55dbpzEVSWwJDQm5dABcJrWtQbomEF4aOJyVnRtTXlQuxcZCtbDj4X1n3guwmoYIN9bCoHwomzcVAwk+UaHNQeBKJDZ+uLUIGILGoqoGiLdxnUjfHFUV1bDR/fC6m9bSl5sCG1F/4TJt5tzbXv/dge544NsyEmAQac7vs8A0BQBUJEpojIRhHZIiL3u9l/tYisdvwsEpFRLvt2iMgaEVkpIsuCOU9vaFFvCH/qMEHLK7queQfm/NK/se6oqbAfij4nwYjLrfPc1wifgmz76O0KIpihrrmrbGZsejNzGXSOLccRrho8SuApWG9Npb6EuDpJ6QnJGd4JxHf/hMINcN7DEJd0eLsInPUnOPOPsPZdeHMGVJcfOdYYmz3d/7Qjx4aQoAmEiEQDTwHnAkOBGSLS2PC8HTjVGDMS+DPwfKP9pxljRhtjgtuZ2wta1BuivBDiOkJcB9/HtkQgVr4G3z932MTVUnKW2aV1n4nW9JKQ6ruZqSDbdsPq1MfzcfHJtjRJME1MuausL8STsxysHwI0qzqS8KZJkCe8aUG6fzt88zc49gJ7k+GOSffABU/A1q/g3xcf+VnPWw0lu0OeHOdKMFcQE4Atxphtxphq4E1gmusBxphFxhjnb2QxkBXE+bQIZ2+Idf4IRFmB7xFMTloiEPmOu/Vt8/wb35idiwCB3idATLx1tm2Ybfs5e0tBti3Q502z9dTM4IW6GnO4xEZzpA+xzkwNd40c8tbamzbX+lu+kDnWmqia+mwaA3PusybiKQ95Pte466xfYu8K+Nd5NnoPHMX55PANShjwSiBE5G4RSRHLiyLyg4ic3cywTGC3y+scx7amuAn4xOW1AT4TkeUicouHud0iIstEZFlhYYDulN3Q0BvCC0d1fb0he+9BPly5h+JD1Y4sah/NS04SO/uXB3Fov+08BfbuJBDsXGiTipxRHyMut83TfbHP52e7L/HtjpTM4PkgSnLs79WT/8GJOD6k2+ZZM5vS9slfa/8PvblRcUfPZlqQZn9g29ae/r+HzaWeGDoNrnoLDuyAl86xq48Nc+zNmK++ywDi7W/nRmPMQeBsIB24AXiwmTHuUv7cxnSJyGlYgfiVy+aJxpixWBPVHSLitkO3MeZ5Y8x4Y8z49PTg/iKb6g1RW1fPqt3FvDB/Gz9+ZSmj//QZU59YwN1vruTkh76mKD+H2sSu/r2pvxVd8x3ltJO6WYFoaZOT2mob4tpn4uFtfSdBxx7WhuoNZYW2Oby3AhHMFUSDg3q0d8cPOgdqDsGOb4MzHyV0GON9k6Cm8NSCtLIEPrnf3nwcd7P35xxwGlw3y46feabN0wijeQnA28Ba55f9VOBfxphVIs3mfOcArnVps4CjvHwiMhKYCZxrjClybjfG7HU8FojI+1iT1Xwv5xsUhmakcKi6ji2FZZRW1rB4236+376f5TsPUFZVC0C/rklMHZHB8f07k9mpAzMXbIOthby/+RiK52/j2hP7kBDrZWEw8N/E5HQGT7gFvv5L0/2fvSV3la1Z1Oekw9uiomHYxTYSo6IYEjs1MyeHaHk7j5QsOFRk79pjE/2ZddPkrrLhq97aoPtMsr6TTZ+GrN2jEiRKcmzHQH8c1E4SO9lkT3ctSL/6i20sNON133MXssbDDZ9YfwSEJXvaFW9nv1xEPgP6Ab8WkWSgudTSpcBAEekH7AGuBK5yPUBEegPvAdcaYza5bE8CoowxpY7nZwN/8nKuQcPZG2LKY/NxLiIGde/IRWN6cny/LhzfrzPdUo7s9zChdwrmz2VEJ3fjgTnrmfntNu48fSBXjO9FXIwXC7jENGvGqa1u3pnqSv5aa54adaUViK1ftUwgdi60j64CATDiMljyDKz/CMZe6/kcDTWYvPxSbohk2gtdBng/V2/IW22rtXobOBCbYDvNbZoLUx8OeU0cJYA0OKhbIBBgHdXO1bnz/2HPD/D9CzDh5sPhsL7S7Vi4+StrBQj0/72PeCsQNwGjgW3GmEMi0hlrZmoSY0ytiNwJzAWigZeMMetE5DbH/meB3wFdgKcdC5JaR8RSd+B9x7YY4HVjTNgD0Qf3SOaycVkkJ8RwfL/OHNe3M106xnsedGgfguGSk8fQs8sJPDx3I7/9YC3Pz9/K3WcM4uIxmURHefiy6eCSLJfc3fvJ5mfbu+NOvaDrINj2NZx0p/fjG7Nzkf1CTWpkKsscB2l9bTRTcwKRvw46dPXepuraOCjQH5TcVdDvVN/GDDoHNs5u+WpMCS8NNZha+DfMHAer37Rm0NQsWzr+43ugY3fre2gJKT3tT5jxViBOBFYaY8pF5BpgLPB4c4OMMXOAOY22Pevy/MfAj92M2wZ44T0MLbHRUTx8uY/TcsmBOKF/F96+7UTmbSrkH59t5L63V/HMvC387KzBnDu8B1HuhCLRD4Gor7dfYs4v7AGnw/JXbGkAXzraNZyvDnYthuGXHL1PBIZfBt8+AqX5nufo6xdrsBoHlebbEh7eOKhdcTYR2vSpCkSwMAY++qntgXJJ46j3AJG/xkYvuetm6AuuCXOpWbD0BXvjcdm/fC/f0Urx1kn9DHDIkcj2S2An8GrQZhVJNKrDJCKcNrgbH905iWeuHkuUCHe8/gMXPb2Qkgo3Wcn+lNso3mkbDTmdwQNOt/6D3Yv9u4b8ddZm6+qgdmXEZTbpKPuDps/hFC1vHdTgsoIIsEA07kHt9XwyrFNbw12Dx/cvwA+v2sCHYPXhaKmD2kkPlxakJXus7+GYM61fLkLwViBqjTEGm8fwuDHmcaCF8ttOaKjDdKRZRUQ4d0QGn95zCn+7bCSrc0p4ccG2o8f7U/Lb6aB2OmD7TLT/yP6Gu+5c5DhPE9Ukux1r7blr3m76HCW7jhQtb4hNsCapQGdT5660j770AXAyaArkfA/lRc0fq/jG7qUw9zfQeYBtKOX8vwsk1eWwf5t/NZgaExNv/4f2/ACf3m/nHGH+KW8FolREfg1cC8x2ZEnHBm9aEUQzlVyjo4Tp43sxdUQPXlq4gwPljWqy+FPy25kglz7EPsZ3tPHUfgvEQpv5nOohj3H4pZCz1MZxe5qTLwIB1lEd6BVE7irbgzghxfexg86xq6UtXwR2Tu2d8iJ4+3prd79hjq0/FKgET1cK1gMmcJ3ZMsdZIVs/C079pW1VG0F4KxBXAFXYfIg8bMLb34M2q0iirMBWPo3r6PGwe84cRHl1Lc/Nb7SK8MfEVLDOOo7jXd5zwGm2/67TJ+ItxtgPQFPmJSfDL7WPTeVENNRgGuLb+wejcVDuKt/NS04yRtvcEi3eFzjq6+C9m21JmumvQnIP2/sgGAKR5+gBEQgTE1iBMHX2ZuzEuwJzzlaEVwLhEIXXgFQROR+oNMaoD8Ibygtt1E4zy85B3ZO5cFRPXlm0g8JSl9amzoquvq4gGoeSOqtB+vqh27fJJrc1Dm9tTFofW/V0TRO1mQqybbkKXx2DgV5BHNpvy437KxBRUTDobNjyZfCbGbUX5j9su6ad+xD0HG239Z9s/2dK8wP7XvlrIT6l+Vpg3tL/VCsOFzzhWxh6G8HbUhvTge+By4HpwBIRuSyYE4sYfOhFffcZA6mqrePZb7Ye3uis6Optye+aSlsjpnGUTY9R1lzlq5mpqfwHdwy/zH6oneYkV9yJljekZFoHeaWfvTga46+D2pVBU+ycdi8JzJzaM1u+hHl/hVEzYNz1h7f3n2wft38T2PfLW2vNS4HyE6T0hDuWQO/jA3O+Voa3Jqb/AY4zxlxnjPkRNqv5t8GbVgRRXuh1Hab+6R25ZGwW/1m8k7wSl25VvmRT79tol7yNbf1RUfZD52vZjZ2LbDmNzv2bP3bYRTY7uXGF19pqKNrcfIlvdwQ61DXXIRA9WiAQ/SdDdJyamVpKSQ68+2P7f3HeI0d+afcYaf/vA2lmMsZG5LU0Qa4d4a1ARBljXI3XRT6Mbd/4WMn17jMGUldveOrrLYc3+iIQ+Y0imFwZcLotAVDg5g7fHcbAjoV29eDNHVfHbvbLc807R4pQ0WYb4eGPYzDQoa65qyC1FyR18f8c8cm2DtX6j6z9XPGd2mrrlK6rgen/PjqjPSrKJjJum9fyOmJOinfaDoKB8j+0A7z9kv9UROaKyPUicj0wm0YJcIob6uus/d6HSq69Onfg8vG9eHPpLnIOOMpo+yIQBesgOt6GCjZmwGn20VszU/FOKN0LfZtxULsy/DI7Lselx1NDiQ1/VhABbhzUEge1K2OutRFbGz9p9lDFDZ//zka9TXsSuh7j/pj+k+3KsWiL+/2+khegEhvtCG+d1L/ANvMZic1wft4Y8yvPoxQOFdmQSB87yd11+jEIwpNfOT4YiZ19W0GkD3ZfJCw1y5bL8FYgGvIffBCIY8+3AuVqZspfZx3tXQZ6fx4nyRmABGYFUVVqv2wCIRDHXmgdnYueaPm52htr37P1u0643Zolm8LphwiUmSl/LSD+3ai0U7w2Exlj3jXG/MwYc68x5v1gTipicIaU+ljPvWenRGZM6MXby3PYWVTu4woi27MpZ8Dp9ovfm74GOxdaceo62Lv3BltiYOBZ9kvAaX4pWG/FwZ8oj+hYG/YYCB9E3lrAeNckqDmiY2wT+t1LYJc6q71m32aYdRdkTbDtNj3RuZ8V4UAJRN4aW9MrTO072yIeBUJESkXkoJufUhEJUFhJBONMkvO1FzVwx2nHEBMlPP7l5iMrunri0H5bY8hTMtqA06G2EnZ91/wkdi6y/gdfm6qMuNxe+3ZHdfaCdS2rXRSoxkENPSACVOZr9NWQ0Mn2HVaap7oc/nutzUC+/GXvbhj6T4btC2whvJaSv1bNSz7i8ZNvjEk2xqS4+Uk2xviRhtrOcJbZ8KObXLeUBK49oQ8frNhDQZ2jF0JlsedBDSU2PHwZ9/Wy7MbBXFuSwJvw1sYMOgfikq2ZqarU5h20ZFkfqMZBuausWCf3aPm5wCYiHvdjWP8xFG1t/vj2jDHw8b1QuAEuneldlzWwAlFVcrg8ir/kr7M+I3/+n9sxGokUTBpWEP51urtt8gASYqOZvdmRONdcLkRDOQsPJqa4JEfZja89n2uX0//gxwcqNtF2wsr+6PBduz85EE5SsqwPoqXRLE4HdSBr5Uy4xZrBvnsqcOeMNPZvt+Kw+r9w2m8OJ216g7Mk+7Zm/l+bY8Vr9sZouKZv+YIKRDApK7Dx8n6W/u3aMZ7rTurLVzsdGbvN+SEK1llzVHN3yANOt8ttT1mqOxfZVYC/Rc1GXG7v/BY6nLgtWkFk2Wq0/nTWc1JTYe9eA2VecpLcHUZeAStfg/J9gT13W8YZIv3m1fDEGFjxbxh/I5x8n2/nSepifUbbWpAwV1djxWnwuS0Lb26HqEAEk/JCa9JowR3rraf0pzrWITDNfUHmr7N36s29nzdlN3YustmhvrZMdNL/VOjQBTbPta06W1LawGmOaIkfoiDbJhAGWiAATrrL+nWWzgz8udsatVWw8g147hR4eaoNdDj5Z3DPWjj/Ud/9WWDNTLuXWB+GP2yaa8PNx1zj3/h2jApEMCkr8DmCqTGdOsRx9nhb4C5nrwc7vLPfgjfO4B4j7Zd3U36IQ/vtF2pL7LXRsTD0Ivu82xD/vhicpAQgmzrQDmpX0gfb8hvfP+9ddFgkUlYI3/wNHhsBH9xmheL8x+DebDjjd7aXhr/0nwx11d4FVrhj5Wu2GsCAM/yfQztFBSKYlHtfh8kTl59iv9QWrtnc9EElu2ykkzfltJsru+H8IPqS/+COEZfbR19LfDcmECuI3FU24qhT75bNpSlOusvmvax6Izjnb63kr4MP74BHh8HXD9ibj2ves/WJxt/gfc9vT/Q+0Zpq/Ql3LSuwK4hRV/i/Gm7HqEAEk7LCFq8gAFJS0qgnmsLCPFbsasLM1FBiw8swvgGnWwHLX3f0vh0LbT3+nmP8m7CTXsfbUFCnUPhLUjfrYGzpCiLQDmpX+kyEnmNh0ZP+l98wBhY80nwAQWth6YvwzEmw5l0YczXc8T1c8w4cc0Zgf89xHez/kj8Csfq/1rQ4Ws1L/qACESzq6w/7IFqKCNIhje6xFTzy+Sb3xxQ4vui97bfQ30PZjZ0LIes4G6/eEqKi4KKnrT+ipedJyfB/BVFXY4UwIwAJck0hYlcR+7f6X37j6wfgyz/CWz+ykT+tmX2bbfe3AafDz7KtfyHdh4RKX+k/2Sa6+RIIYIyNXso6DtIHBW1qkUxQBUJEpojIRhHZIiL3u9l/tYisdvwscvS89mpsq6figL1z8SMHwh3SoTNj0w0LNu/j++1uwl3zs60j2Nt+C6mZto59Y4GoPGhLYre2eHFnqKs/FG6wNuyM0QGd0lEce6E1YS3yI3Huh3/D/L87/DZiq5y21n4TdbXw/q02nPmiZ6BD5+C/p/OGxpfy33t/gML1dhWr+EXQBMLRlvQp4FxgKDBDRBobo7cDpxpjRgJ/xtZ78nZs66aFORBHkZhG3w7VpCfH8/DcjdTW1R+5v7kSG+5wV3Zj9/e2flRrE4jUTP8L9gXTQe1KdAyceCfsXmx/j96y5Uv46G7rRL10Jlz4OOxZZvsktEYWPgp7lsN5/whc0mFz9BwN8am+mZlWvGa7OQ6/JFiziniCuYKYAGwxxmwzxlQDbwLTXA8wxiwyxjiN6ouBLG/HtnrWOcpV9QhAc3SAxDSiKw/ws7MG8f2O/fz41WWUVTnKD9RW2SW/r87gAadDXdWRzeF3LrSF9bKOC8y8A0VKps3urq9v/tjG5K6yLV/dVbgNNM7yG94W8ctbA29dZ/NELn/ZRn8Nu9hWi13wSMvi/4NB7mqY9xAMu+Rwm9lQEBUN/U6GrfO8S5isqbSZ/Mde4HcekhJcgcgEdru8znFsa4qbAKfx1uuxInKLiCwTkWWFhYUtmG4AObQfvnsahk4LnF02MQ0qipkxoTcPXDycBZv3cdkzi9hbXGHbgpo63+sd9TnJRoe4mpl2LrLO6dZW0Cw1C+prDq/MfCF3lRXqloTaekt8RzjuJu/Kb5TsgdemW7PgVW9Bgkv1mnMfgi7HWFNOeVFw5+wttVV2Ph0629VDqOk/2UbrHfDCP7PhY6gssc5zxW+C+YlxF8bgVvpF5DSsQDhLiHs91hjzvDFmvDFmfHp6gMw5LWXRP23I6akBdJ0kdoYK63u4+vg+/Ov649hzoIJpTy1k1/ql9hhfy1k0LrtRU2FNBy0Nbw0G/jYOqq+zVVyDbV5yZcKtzZffqDwIr0+3taqufvvo2kRxSXDZSzZ09sM7WlZmpKLY+jeKd/l/DrAmr4JsuPDJ0PgdGuP0Q3hjZlr5GqT2hr6nBHVKkU4wBSIH6OXyOgvY2/ggERkJzASmGWOKfBnbKinfB0ues3bPllQwbUyjiq6nDErnnZ+cRFx0FJ99/RX1UbG2lLGvDDjdRkCV5tkmP/U1rVMg/G0cVLQVaspDKxDNld+oq7Hd1ArWw/SXm+5wljESzvoTbPrE/yztfZth5hnw1V9g5pmwd4V/59m1BBY+DmN/BIPO9u8cLaXLABus0JxAlOTYm57RM0KzaoxggvnbWwoMFJF+IhIHXAnMcj1ARHoD7wHXGmM2+TK21bLwMVs3KJCrB4DETvbRpaLr4B7JfHDHRMbE72VDbU+eX7gL4+udprPsxtavHb4IaZ0N2J3Z1L6uIHIczuJA9IDwhYbyGy8eud0YmP0z2PolXPAYHHOm5/McfxsMPBvm/s/hjmjesvlzeOEMu4K46BnbyOlfU2Gjj720q8ttdnRKFpz9gG9jA4mIo/z3fM+5JqveAAyMvipUM4tYgiYQxpha4E5gLrAeeMsYs05EbhOR2xyH/Q7oAjwtIitFZJmnscGaa8AozYfvZ8KI6YGPu05Ms4+N6jGlJ8czNmEvpamD+L85G/ifD9ZS0zjCyRPdR0CHrtYPsXOhtdW3Rqdeh842ec+XZDlj7J135wEtz+b2labKbyz4B/zwqi1aN/ZHzZ9HBKY9bW8Q3r0Jqg81P8YYa+Z8fboNu73la/tl+eMvoOsgeHMGfP+C99fy+e9t6feLnj7STxIO+k+2n4G81e73O3Mf+p4MaX1DObOIJKjrL2PMHGPMIGPMAGPMA45tzxpjnnU8/7ExJs0YM9rxM97T2FbPwsdsvP2pvwz8uZ0238YlvysOIKW5HHf8ydw+eQCvL9nFjS8v5WCllzH0rmU3cpa2TvMS2C/K1CzfkuV2LbYmlRNvD4+p4aS7bJE4Z/mN1W/DV3+2NxCn/6/35+mYDhc/a/M55v7G87E1lfDBT+Cz/7URPDfNPVxeJLk73DAHBp4Dc+6zxzQXFbb1a1j6gm0P2u9k7+ccLJxJl02ZmXYusk5szX0ICGqgCxQHc605YdQM/3wBzdHECsJZYiOq+zB+OWUIf7tsJN9tLeLSpxexe78Xd5tgzUyH9kHNodaX/+BKio+NgxY/ZUNOR80I2pQ80meijQhb9KQ1i3x4u72znfak76UoBpwOE++G5f+C7CasrQdz4eXzrCCd9j9w+StHR6PFJcGVr8FxN9tVxjvXN11gsKLYOsi7DrIF91oDHbvZYIymBGLla7ZM/dALQzqtSEUFIlAs+IcNNT31F8E5f1MC0aiL3PTxvXj1pgnkH6zk4qcX8kNTtZtcGXDa4eetWSBSfcim3r8dNsy2PQjCFbLrWn7j35dAWj+44t/+lzA57X+t4My66+iV1J7l8MJp1vF9xX/sKrYpEYqKhql/h7P/AtkfwqvT3IfSfnq/DV64+FmbNd1a6D8Zdn53tLBVlcG6D2DYRa0vTLuNogIRCIp3ww+v2HrzwbJ7JjpMTEetINbZu+Tkw+WUTxrQlffvmEiHuBiunbmELQWlns+d0hPSj4WugyGpa2DnHUhSMqEsz7v+xEueA4mCCTcHf16eOHaa/Z9ITLPhrE6h94eYOLj0RaivhXdvPuyoXf0WvHSuDa296TNrWmoOp3hd/grsXQkvnnlk3sb6j+1K5OSfQ+Y4/+ccDPpPtgmeu5ccuT37Axuxpn0fAoYKRCBY8A/rHPO1W5YvxCeDRDfkQjTgLLHR6G5xQHpH/nvrCSTGRXPLq8ub90lc8pz9ac2kZtoyIKW5no+rLLEdzIZfasUvnETHwA2fwG3fQloLmiY56TLAJqntWgTfPGQdyO/dbDPfb57XdMhsUwy7CK77yJqTZp5pw1nLCm3pjx4j4ZQgrYhbQp+TbLZ/YzPTitdscmGvVhiF10ZRgWgpB3bYL6Nx10GnXs0e7jcijmxqlxWEMdYH0USETkZqIk9dNZZd+w9x75srqa/3EAKbMarl5b1bQG1dPS9+u52isqqmD/K2cdAPr9qckRNuD9wEW0JKT+sgDhSjrrR5Ft88ZAMjxt8IP/rA/3aavY+3EU6JneCVCxwJfAfh4ufsqqW1Ed8RsiYcKRBFW61ojr4qeCXd2yEqEC1l/t/tnf3JPw/+ezUWiOJdUF3qMSHv+P5d+O35Q/lyQwGPfemh4VCYeW/FHv78cTYvL9rR9EHeNA6qq7XmpT6TbIG3SGXqwzB4qi2zff6j1rzUEroMgJu+sL+zvT/YKKtAJnoGmv6TrWnMGdW38nVrUgxXQEKEogLREoq22v67428MjSmjQ+cjw1ydDupmSmz86MQ+XDYuiye+3MzcdXlBnKB/VNXW8fgXVrxmr8ltOtkvxQuBWD8LSnbb0NZIJiEFZjj+9wJFUhf40SzbEe7EuwJ33mDQfzJgYMcC64tZ9YaN9Aq3STHCUIFoCfP/bovdTbo3NO/XeAXh7AbX7ViPw0SEv1w0nFFZqfzsvyubd1qHmDeW7GJPcQVTR/RgW2E5m/LL3B+YkALxKZ5NTIufhs79bZKa4juxCbYjXGsvUZE51oazbptnfw7u0dyHINDK/wtaMYWbbDvD424KrH3ZE46Krg0UZNskKC+yWxNio3n22nHeO61DxKHqWp78egsn9O/MHy8cTpTYVUSTpGQ2Heq6+3ub7Hf8T2wopxK5RMdC30lWHFa+ZiP5Bk8N96wiDhUIf/nmIduMZOI9oXvPo1YQ2T5VcM1ITeTpq8d557QOEf9auIN9ZdX84pzBpCfHM6FfZ+Z4EghPjYO+e8qWCdEaPO2D/pNtCZDsD23f89iEcM8o4lCB8IeC9bD2XTj+FlsGIVQkdrZO6dpq+1O02WdH4oR+nfndBQ6n9RdN9LcOESUVNTz3zVbOGNKNcX1snsfUERlsKShjc34TZrCmVhAHdlr/w7jrbZSLEvn0n2wf62u170OQUIHwh3kP2kzNk34a2vd1rei6b5P9YPhRhO7aE/pw+bgsnvhqS1id1i/M38bBylp+fvbhpkpThvVAPJmZUrMcZUEqj9z+/fOOxLhbgzhjpVWRPtgmiHYbFvx+4+0UFQhfyVtjMzZP+Enom6a4lttoKLHhY5MgrNP6z2F2WheWVvHSwu1cMKonQ3se9qF0S0nguD6d+WRNE8LljGRydVRXHrS5D0MvOrrxjhK5iMD0f9s+3pr7EBRUIHxl3oO2efqJd4T+vV0FIn8tRMXazFE/CLfT+ul5W6iqrefeMwcetW/qiB5szC9lS4GbaKZUN8lyK/5jE7vC8TdRwkuv41p3vkYbRwXCW+pq4dPf2F63J97Rspo6/uJa8js/2y6xW5AgFS6n9Z7iCl5bvIvLxmbRP/1of8G5I2xdKbfOaqdAOP0Q9XWw5BnofaINfVQUJWCoQHhDeRH852JbPnrCLXDyz8Izj8YmpgA0wXF1Wv/1k/W+d6TzgyccSXE/dbN6AOieksD4PmnuBcKZCOWMZNrwsc0o19WDogQcFYjmyFsDL0y2zWemPWXLJLe0rIG/OAXiwHZrYgnQ0vraE/pwzQm9eWHBdu56YwUV1R7aObaQbYVlvPNDDlef0JvMTk2XkJ46IoMNeaVsK2xkZopNhA5dDq8gvnvaVkvVGHhFCTgqEJ5Y8w7MPMual274NPxlhONTbN2nnYvs6+4+Vu5sAhHhz9OG86spQ5i9Jpfpz31HXkll8wP94NEvNhMfE8Xtkz37Ts4d0QOAT9a6cVY7GwflLIfdizUxTlGChAqEO+rr4LPf2h7APUfDLfMgqxXUxHdWdM1Zal8HsM+yiPCTyQN44drxbCss48Inv2Xl7uKAnR8ge+9BPlq1lxsn9iM92XPTnIzURMb27sTs1U34IUr2WJNffIrGwCtKkFCBaMyh/fDaZbDoCRh/ky1eFqpSGt6QmGb7XiekBqUw2ZlDu/Pe7ROJi4li+nPf8eFKH1p8NsM/PttISkIMN5/S36vjp47IIDv3IDv2lR+5IyXTdmlb9wGM/ZHtlaEoSsAJqkCIyBQR2SgiW0Tkfjf7h4jIdyJSJSL3Ndq3Q0TWiMhKEVkWzHk2kL/Otm3cvgAueALOf6T11cN3+iG6Hd0kKFAM7pHMh3dMZHSvTtz95kr+PndDiyOclu/cz5cbCrj11AGkJnrnw2mIZlrbaBWRmgm1DhPY8ZoYpyjBImgCISLRwFPAucBQYIaINLaJ7Ad+CjzcxGlOM8aMNsaMD9Y8G1j3gfU31FTCDXNsA6DWiDPUNcix3106xvOfm47nyuN68dTXW7n1P8spr/Ki1acbjDH8fe5GunaM44aJfb0el9kpkdG9Oh0dzeRsHDT0QlusUFGUoBDMFcQEYIsxZpsxphp4E5jmeoAxpsAYsxQIX2nR+jr44o/w9nU2K/mWedBrQtim0ywNK4jgJwfFxUTx10tG8PsLhvLl+nwufWYROQcO+XyehVuKWLxtP3eedgwd4mJ8Gjt1RA/W7jnIriKX9+052kYyhbJQoqK0Q4IpEJnAbpfXOY5t3mKAz0RkuYjc0tRBInKLiCwTkWWFhYW+z7Kq1BbeG3sdXP8xpGT4fo5Q4hQIP0ps+IOIcMPEfrx8wwT2FFcw7cmFLN2xv/mBDuzqYQOZnRKZcbzvd/vnDndjZuo6EH65LbI7xilKK8C32znfcGcg98WQPdEYs1dEugGfi8gGY8z8o05ozPPA8wDjx4/33VCe2MmuGkJdV8lfknvYEhvNNAkKNKcMSueDOyZy8yvLuOqFxVwwsifpKfF0SYqjS1I8nTvG0dXx2CUpjoRYG3b6WXY+q3JK+NtlI4mP8T0UtVfnDozMSmXOmlxuO3VAoC9LURQPBFMgcoBeLq+zgL3eDjbG7HU8FojI+1iT1VECERDaijiAbTHZ71QbxRRiBqR35P3bJ/KbD9aweFsR+8qrqa6td3tsUlw0XTrGU1JRQ//0JC4Z438RvakjMnjwkw3s3n+IXp07+H0eRVF8I5gCsRQYKCL9gD3AlYBXnVxEJAmIMsaUOp6fDfwpaDNtS8Qnh9W0ktohlqeusjWPjDGUV9dRVFZFUXk1RWXVRzzfX17FgUM13HJKf2Ki/bdmTh1uBeKTtbnccopvq4i6esPu/Yfo2zXJ7/dXlPZK0ATCGFMrIncCc4Fo4CVjzDoRuc2x/1kR6QEsA1KAehG5Bxvx1BV4X2wYZwzwujHm02DNVfEPEaFjfAwd42Po0yV4X8C9u3RgeGYKc9bk+SQQ1bX13Pn6D3yWnc/1J/XlN1OPJS5GU38UxVuCuYLAGDMHmNNo27Muz/OwpqfGHARGBXNuStti6ogM/vbpRvYUV3is4eSksqaO21/7ga82FHDqoHReXrSDFbuLeeqqMWSlqZlKUbxBb6eUNsFURzTTJ576VTuoqK7j5leX8dWGAh64eDiv3DiBZ64ey7aCMs574lu+2pAf7OkqSkSgAqG0Cfp2TWJoRor7EuAuHKqu5caXl/Ltln387bKRXH18H8BmZX901yR6dkrkxpeX8bdPN1Bb597BriiKRQVCaTNMHdGDH3YVs7e4wu3+0soarnvpe5ZsL+KR6aOYPr7XEfv7dk3i/dtPYsaEXjw9bytXz1xCwcHgVK1VlEhABUJpM0x11GZyVwK8pKKGH730PT/sKuaJGWO4eIw715ZttfrXS0byyPRRrM4pYeoT3/Ld1qKgzltR2ioqEEqboX96R4b0SD7KD1F8qJprZi5h7Z4SnrpqLOePbL7K7SVjs/jwzomkJsZw9czFPPX1lpC1XFWUtoIKhNKmmDoig2U7DzQ0NCoqq2LGC0vYmFfKs9eMY8rwHl6fa1D3ZGbdOYnzR/bk73M3ctMrSzlQXh2sqStKm0MFQmlTOM1Mn67NpaC0khkvLGZbYRkzrxvPGcf63rcjKT6Gx68czZ8vGs7CLUWc+/gCPlmTG5Le3IrS2pFI+iCMHz/eLFsWmtYRSvg4+9FviImKorK2jtziSl68bjwnHdO1xeddk1PCL95ZxYa8Uk4dlM4fLxymGdhKxCMiy5tqqaArCKXN4ew0l19SySs3TgiIOACMyErl47sm8bvzh7J85wHOfmw+j3y+icqauoCcX1HaGioQSptj+vhenDoonX//+Hgm9AtsocWY6ChunNSPL39+KlOG9eCJLzdz9qPz+XpDQUDfR1HaAmpiUhQPLNqyj99+uJatheWcPbQ7v7tgqJbqUCIKNTEpip+cdExXPrn7FH45ZTDzNxdy5iPf8PS8LU2WOVeUSEJXEIriJTkHDvGnj7L5LDufAelJ3HrqADp3iCPJUdE2KT7a8RhDh7hoHNWIFaVV42kFoQKhKD7y1YZ8fj9rHbv3uy/5ASACSXFWNJLiY+iSFEeP1EQyUhMafnqkJtIzNYEuHeOJjlIxUcKDJ4EIarlvRYlETh/SnUnHpLNr/yHKq2opr6qlrKqW8upayqrqjtxWVUt5VR1F5VWszilm7rrKo8xTMVFC9xSnaCRw7Ql9OL5/lzBdnaIcRgVCUfwgLiaKY7p19HmcMYb95dXkllSSW1JJXklFw/Pckgq+21rEVxsKeOvWExmeGfq2soriigqEooQQEaFLx3i6dIx3KwAFByu56KmF3PjyUj64YyI9vWiOpCjBQqOYFKUV0S0lgZduOI5D1XXc+PJSSitrwj0lpR2jAqEorYwhPVJ4+uqxbC4o447XV1CjjY2UMKECoSitkFMGpfPARcOZv6mQ3324TosHRgDfbS3iTx9lt6lVYVB9ECIyBXgciAZmGmMebLR/CPAvYCzwP8aYh70dqyiRzpUTerNz/yGembeVPl06cNupA/w+197iCgpKqyitrKG0spbSyhoOVjgeK2sprazlYGUNpZU1VFTXkRQfQ0pCLCmJMSQnxDZ6HkNKot3WLSWerh3jA3jVkcfOonL+b8565q6zvdD3FlfwzDVj20SeTNAEQkSigaeAs4AcYKmIzDLGZLscth/4KXCRH2MVJeL5xdmD2b3/EA9+soFeaR04b2SGT+NzDtixH69238tbBDo6xCA5wT6mdojjUFUt2/aVNYhIeXXTBQvH9UnjwlE9mToig/RkFQsnBytrePKrLfxr4XZio6P4xTmDMcbw8GebePHb7fz45P7hnmKzBHMFMQHYYozZBiAibwLTgIYveWNMAVAgIuf5OlZR2gNRUcLDl48it6SSe99aSY/UBMb1SWt2XGllDc/M28rMb7cTJXDHaQMY1yeNZIcQOFcCSXExRHmRpFdbV++yyqjlYEUNBytr2FJQxkercvn9rHX88aN1TDymKxeO6sk5w3uQkhAbiF9Bm6O2rp43l+7m0c83sf9QNZePy+K+swfTLSUBYwyrc0p48JMNjO7VifF9A1tsMtAELZNaRC4Dphhjfux4fS1wvDHmTjfH/gEoc5qYfBx7C3ALQO/evcft3LkzKNejKOFkf3k1Fz+9kLLKWt6/fSK9u7gvGFhXb3hr2W7+8dlG9pVVc/GYTH5xzuCgh8tuzCtl1qo9zFq1l937K4iLieK0welcOCqTM47tRkJsdFDfv7WwYHMhf/l4PRvzS5nQrzO/O3/oUeHMJRU1XPjkt1TW1DH7pyeH3UQXrkxqd7cl3qqR12ONMc8Dz4MtteHl+RWlTdE5KY5/XX8clzyziOtf/p73fnISnTrEHXHMt5v38ZfZ2WzIK2V8nzRmXncco3t1Csn8BvdI5hc9hnDf2YNZubuYWav28vHqXOauyycpLpqzh/Xg8vFZnDQgML07WhtbC8v4v9nr+XJDAb06J/LsNWM5Z1gPt36G1MRYnr56LJc8vYi731zBqzce32pLrQRTIHKAXi6vs4C9IRirKBFJ//SOPH/teK6ZuYRb/72cV2+aQHxM9BFfTllpiTx11VimjnD/5RRsRIQxvdMY0zuN/z1vKEu2FTFr1V7mrMnl/RV7OH9kBr+/YFjE+Cqqa+t58JMNvPrdDhJio/n1uUO4fmJf4mM8r5iG9Uzlz9OG88t3V/Po55u475zBIZqxbwTTxBQDbALOAPYAS4GrjDHr3Bz7B440MXk91hUt1qe0Bz5cuYe731zJhaN60jkpjv8s3klCbDR3nHYMN0zs2yrNOVW1dTz/zTb++dUWEuOi+Z+px3L5+Kw2EcnjiT9/nM2L325nxoRe/OyswT4L3y/fWcVby3J46frxnD7E957qgSBs1VxFZCrwGDZU9SVjzAMichuAMeZZEekBLANSgHqgDBhqjDnobmxz76cCobQX/vnlZv7x+SaixIbD/uysQWG3ZXvD1sIyfv3eGr7fvp8T+nfmr5eMpF8b7fv9eXY+N7+6jOtP6ssfLhzm1zkqa+q4+OlF7C2u4OO7JtGrc+ibUWm5b0WJMIwxvL9iD8N6pjK4R3K4p+MT9fWG/y7bzf/NWU9VbT13nzGQW07pT2x028nb3VNcwdTHF9CrcyLv/uSkZk1Kntixr5wL/vkt/dKTePu2E1t0Ln9QgVAUpdVRcLCSP3y0jjlr8hjSI5m/XjKCMb2bD+EtOVTDypxiVu4qZs2eYo7NSOGu0wcSFxMagampq+fK5xezMa+Uj++aRN8ArIDmrsvj1n8v55oTevOXi0YEYJbeo/0gFEVpdXRLSeDpq8fxeXY+v/1gLZc8s4jrTuzLfecMpmO8/WqqqatnY14pK3ZbQVix+wDbCssBm+TXp3MHvlhfwDebCvnnjDH06RJ8c9Wjn29i+c4DPDFjTEDEAeCcYT249ZT+PDd/G+P7dOaiMZkBOW9L0RWEoihhp7Syhr/P3ci/F+8kIyWBc4b3YO2eEtbsKaGyxhYr7NoxjtG90hjTuxNjenViRFYqyQmxzF2Xxy/fWU1dveGBi4czbXTwvly/2VTIdS99z4wJvfjrJSMDeu7aunquemEJa/aU8OGdExnUPTSmQzUxKYrSJli+cz+/fm8NO/YdYlhmCmN6pTHaIQhZaYlNRj3tKa7g7jdWsGznAaaPz+IPFw6jQ1xgDSQFBys59/EFdO0Yzwd3TCQxLvC+goKDlUx94ltSEmOYdeekhpVUMFGBUBSlzWCMoa7eEOOj07q2rp7Hv9zMk19voX/XJJ68aizHZqQEZE519YZrZi5h5e5iPrprIsd0C97d/Xdbi7h65mJOGZTO5eN60T89ib5dkoIiSKACoShKO2LRln3c/d+VlFTU8NvzjuWaE/q0ON/isS828dgXm3n48lFcNi4rQDNtmhe/3c4Ds7Opd/l6zuyUSP/0JPp1TaJ/1yT6p3ekX9ckMjslelVPqylUIBRFaVfsK6vivrdXMW9jIecM687fLh1Fagf/igcu2rqPq2cu4eIxmTwyfXRgJ+qBQ9W1bN9XzrbCcsdjGdscr8uqahuOi4+JYkiPZD64Y6JfQqgCoShKu6O+3vDit9t56NMNdE9J4PErR/tcPXVfWRVTH19Ax4QYPrpzEkkh8Ak0hzGGwrKqI4SjqraeP00b7tf5VCAURWm3rNpdzF1vrGBPcQXnjcjglEHpnDywK91TEjyOq683XP/yUhZvK+LDOyYGzJ/R2tA8CEVR2i2jenXi459O4m+fbuDTtXnMWmXrfg7unszJA7ty8qB0JvTtfJQT+Ln525i/qZAHLh4eseLQHLqCUBSl3VBfb1ifd5AFm/exYHMhS3ccoLq2nriYKCb07WwFY2A6h6prueL5xUwZ3oMnZ4xp80UFPaEmJkVRFDdUVNfx/Y79LNhUyILN+9iYX9qwr3fnDnz800kR3xlPTUyKoihuSIyL5tRB6Zw6KB2AvJJKvt2yj2U79nPdSX0jXhyaQwVCURTFQY/UBC4blxWSXIe2QNupr6soiqKEFBUIRVEUxS0qEIqiKIpbVCAURVEUt6hAKIqiKG5RgVAURVHcogKhKIqiuEUFQlEURXFLRJXaEJFCYKefw7sC+wI4ndaCXlfbI1KvLVKvC9r2tfUxxqS72xFRAtESRGRZU/VI2jJ6XW2PSL22SL0uiNxrUxOToiiK4hYVCEVRFMUtKhCHeT7cEwgSel1tj0i9tki9LojQa1MfhKIoiuIWXUEoiqIoblGBUBRFUdzS7gVCRKaIyEYR2SIi94d7PoFERHaIyBoRWSkibbYXq4i8JCIFIrLWZVtnEflcRDY7HtPCOUd/aeLa/iAiexx/t5UiMjWcc/QHEeklIl+LyHoRWScidzu2t+m/m4fravN/M3e0ax+EiEQDm4CzgBxgKTDDGJMd1okFCBHZAYw3xrTVBB4AROQUoAx41Rgz3LHtb8B+Y8yDDmFPM8b8Kpzz9Icmru0PQJkx5uFwzq0liEgGkGGM+UFEkoHlwEXA9bThv5uH65pOG/+buaO9ryAmAFuMMduMMdXAm8C0MM9JaYQxZj6wv9HmacArjuevYD+kbY4mrq3NY4zJNcb84HheCqwHMmnjfzcP1xWRtHeByAR2u7zOIbL+2Ab4TESWi8gt4Z5MgOlujMkF+6EFuoV5PoHmThFZ7TBBtSkzTGNEpC8wBlhCBP3dGl0XRNDfzEl7Fwhxsy2SbG4TjTFjgXOBOxzmDKX18wwwABgN5AL/COtsWoCIdATeBe4xxhwM93wChZvripi/mSvtXSBygF4ur7OAvWGaS8Axxux1PBYA72NNapFCvsMe7LQLF4R5PgHDGJNvjKkzxtQDL9BG/24iEov9En3NGPOeY3Ob/7u5u65I+Zs1pr0LxFJgoIj0E5E44EpgVpjnFBBEJMnhRENEkoCzgbWeR7UpZgHXOZ5fB3wYxrkEFOcXqIOLaYN/NxER4EVgvTHmEZddbfrv1tR1RcLfzB3tOooJwBGO9hgQDbxkjHkgvDMKDCLSH7tqAIgBXm+r1yYibwCTsSWV84HfAx8AbwG9gV3A5caYNufsbeLaJmNNFQbYAdzqtNu3FURkErAAWAPUOzb/Bmuvb7N/Nw/XNYM2/jdzR7sXCEVRFMU97d3EpCiKojSBCoSiKIriFhUIRVEUxS0qEIqiKIpbVCAURVEUt6hAKEorQEQmi8jH4Z6HoriiAqEoiqK4RQVCUXxARK4Rke8dNf+fE5FoESkTkX+IyA8i8qWIpDuOHS0iix0F3N53FnATkWNE5AsRWeUYM8Bx+o4i8o6IbBCR1xxZu4oSNlQgFMVLRORY4ApsEcTRQB1wNZAE/OAojPgNNhsa4FXgV8aYkdjMW+f214CnjDGjgJOwxd3AVga9BxgK9AcmBvmSFMUjMeGegKK0Ic4AxgFLHTf3idhic/XAfx3H/Ad4T0RSgU7GmG8c218B3nbUx8o0xrwPYIypBHCc73tjTI7j9UqgL/Bt0K9KUZpABUJRvEeAV4wxvz5io8hvGx3nqX6NJ7NRlcvzOvTzqYQZNTEpivd8CVwmIt2gob9yH+zn6DLHMVcB3xpjSoADInKyY/u1wDeO3gE5InKR4xzxItIhlBehKN6idyiK4iXGmGwR+V9sl74ooAa4AygHhonIcqAE66cAW876WYcAbANucGy/FnhORP7kOMflIbwMRfEareaqKC1ERMqMMR3DPQ9FCTRqYlIURVHcoisIRVEUxS26glAURVHcogKhKIqiuEUFQlEURXGLCoSiKIriFhUIRVEUxS3/D0UTOcYwEGUAAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_nn_training(results_TL_3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d76e6b83",
   "metadata": {},
   "source": [
    "### TF Model 4 - Reduced learning rating and Convolution Kernel Size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "a0afb2ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_TL_model_4(input_shape, num_classes):\n",
    "    \n",
    "    '''TL model, Conv kernel=3 '''\n",
    "    \n",
    "    weights = 'imagenet'\n",
    "    inputs = Input(input_shape)\n",
    "\n",
    "    #what does include_top=False remove?\n",
    "    base_model = ResNet50(include_top=False, weights=weights, input_shape=(128, 128, 3))\n",
    "\n",
    "    #freezing the resnet layers\n",
    "    for layer in base_model.layers:\n",
    "        layer.trainable = False\n",
    "\n",
    "    #these convolution layers come before the resnet layers\n",
    "    x = Conv2D(32, kernel_size=3, strides=1, padding='same', activation='relu', input_shape=(128, 128, 3))(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Conv2D(64, (3, 3), strides=2, padding='same', activation='relu')(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "\n",
    "    x = Flatten()(x)\n",
    "\n",
    "    base_resnet = base_model(inputs)\n",
    "    base_resnet = Flatten()(base_resnet)\n",
    "\n",
    "    concated_layers = Concatenate()([x, base_resnet])\n",
    "\n",
    "    # dense layers come after the resnet layers\n",
    "    # dropout to reduce overfit\n",
    "    concated_layers = Dense(1024, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.3)(concated_layers)\n",
    "    concated_layers = Dense(512, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.2)(concated_layers)\n",
    "    concated_layers = Dense(256, activation='relu')(concated_layers)\n",
    "    concated_layers = Dropout(0.1)(concated_layers)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(num_classes, activation='softmax')(concated_layers)\n",
    "\n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "235d8ec1",
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 100\n",
    "\n",
    "#Adds adam optimizer\n",
    "adam = tf.keras.optimizers.Adam(lr=0.0001,\n",
    "                                beta_1=0.9,\n",
    "                                beta_2=0.999,\n",
    "                                decay=0.0,\n",
    "                                amsgrad=False)\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model5/best_model_tl_smallkernel.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "TL_model_4 = generate_TL_model_4(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "TL_model_4.compile(optimizer=adam, loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "37722ce3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x2b1a56820> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x2b1a56820> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-12-09 20:50:52.848089: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
      "2021-12-09 20:50:52.848285: W tensorflow/core/platform/profile_utils/cpu_utils.cc:126] Failed to get CPU frequency: 0 Hz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x2b1188550> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING: AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x2b1188550> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.95100, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.95100\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.95100 to 0.95650, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.95650 to 0.96350, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy did not improve from 0.96350\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.96350 to 0.96500, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy did not improve from 0.96500\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.96500\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy improved from 0.96500 to 0.96800, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy improved from 0.96800 to 0.97100, saving model to models/Model5/best_model_tl_smallkernel.h5\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.97100\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.97100\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.97100\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.97100\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.97100\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.97100\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy did not improve from 0.97100\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.97100\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.97100\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.97100\n",
      "Epoch 00020: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results_TL_4 = TL_model_4.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "ac140a1b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results_TL_4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ebf556c9",
   "metadata": {},
   "source": [
    "## Non-TL Model - with decreased learning rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "935a68a6",
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_model3_lr(input_shape, num_classes):\n",
    "    \n",
    "    '''CNN model architecture with max pooling + dropout layers'''\n",
    "    \n",
    "    inputs = Input(shape=input_shape)\n",
    "    \n",
    "    #x = Rescaling(1.0 / 255)\n",
    "    \n",
    "    # convolution (with max pooling)\n",
    "    x = Conv2D(32, kernel_size=3, strides=1, padding=\"same\")(inputs)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x) \n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "\n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    x = Conv2D(64, kernel_size=3, strides=2, padding=\"same\")(x)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), strides=2, padding='valid', data_format=None)(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Activation(\"relu\")(x)\n",
    "    \n",
    "    #flatten\n",
    "    x = Flatten()(x)\n",
    "    \n",
    "    #dense layers (with dropout to avoid overfitting)\n",
    "    x = Dense(512, activation='relu')(x)\n",
    "    x = Dropout(0.3)(x)\n",
    "    x = Dense(256, activation='relu')(x)\n",
    "    x = Dropout(0.2)(x)\n",
    "    x = Dense(128, activation='relu')(x)\n",
    "    x = Dropout(0.1)(x)\n",
    "    \n",
    "    # output layer\n",
    "    output = Dense(2, activation='softmax')(x)\n",
    "    \n",
    "    mod = Model(inputs=inputs, outputs=output)\n",
    "    \n",
    "    return mod\n",
    "\n",
    "epochs = 100\n",
    "\n",
    "# patient early stopping\n",
    "es = EarlyStopping(monitor='val_accuracy', mode='max', verbose=1, patience=10)\n",
    "\n",
    "# save best model\n",
    "mc = ModelCheckpoint('models/Model3/best_model_test.h5', monitor='val_accuracy', mode='max', verbose=1, save_best_only=True)\n",
    "\n",
    "# Instantiate model\n",
    "model3_lr = generate_model3_lr(input_shape=(128,128,3), num_classes=2)\n",
    "\n",
    "# Compile model\n",
    "model3_lr.compile(optimizer=adam, loss='sparse_categorical_crossentropy', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "f1327860",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_train_function.<locals>.train_function at 0x8f4c42dc0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "WARNING:tensorflow:AutoGraph could not transform <function Model.make_test_function.<locals>.test_function at 0x907f3f5e0> and will run it as-is.\n",
      "Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.\n",
      "Cause: unsupported operand type(s) for -: 'NoneType' and 'int'\n",
      "To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.50000, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 2/50\n",
      "\n",
      "Epoch 00002: val_accuracy did not improve from 0.50000\n",
      "Epoch 3/50\n",
      "\n",
      "Epoch 00003: val_accuracy did not improve from 0.50000\n",
      "Epoch 4/50\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.50000 to 0.50400, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 5/50\n",
      "\n",
      "Epoch 00005: val_accuracy improved from 0.50400 to 0.78300, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 6/50\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.78300 to 0.87100, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 7/50\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.87100 to 0.95950, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 8/50\n",
      "\n",
      "Epoch 00008: val_accuracy did not improve from 0.95950\n",
      "Epoch 9/50\n",
      "\n",
      "Epoch 00009: val_accuracy did not improve from 0.95950\n",
      "Epoch 10/50\n",
      "\n",
      "Epoch 00010: val_accuracy did not improve from 0.95950\n",
      "Epoch 11/50\n",
      "\n",
      "Epoch 00011: val_accuracy did not improve from 0.95950\n",
      "Epoch 12/50\n",
      "\n",
      "Epoch 00012: val_accuracy did not improve from 0.95950\n",
      "Epoch 13/50\n",
      "\n",
      "Epoch 00013: val_accuracy did not improve from 0.95950\n",
      "Epoch 14/50\n",
      "\n",
      "Epoch 00014: val_accuracy did not improve from 0.95950\n",
      "Epoch 15/50\n",
      "\n",
      "Epoch 00015: val_accuracy did not improve from 0.95950\n",
      "Epoch 16/50\n",
      "\n",
      "Epoch 00016: val_accuracy did not improve from 0.95950\n",
      "Epoch 17/50\n",
      "\n",
      "Epoch 00017: val_accuracy improved from 0.95950 to 0.96650, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 18/50\n",
      "\n",
      "Epoch 00018: val_accuracy did not improve from 0.96650\n",
      "Epoch 19/50\n",
      "\n",
      "Epoch 00019: val_accuracy did not improve from 0.96650\n",
      "Epoch 20/50\n",
      "\n",
      "Epoch 00020: val_accuracy did not improve from 0.96650\n",
      "Epoch 21/50\n",
      "\n",
      "Epoch 00021: val_accuracy did not improve from 0.96650\n",
      "Epoch 22/50\n",
      "\n",
      "Epoch 00022: val_accuracy did not improve from 0.96650\n",
      "Epoch 23/50\n",
      "\n",
      "Epoch 00023: val_accuracy did not improve from 0.96650\n",
      "Epoch 24/50\n",
      "\n",
      "Epoch 00024: val_accuracy did not improve from 0.96650\n",
      "Epoch 25/50\n",
      "\n",
      "Epoch 00025: val_accuracy did not improve from 0.96650\n",
      "Epoch 26/50\n",
      "\n",
      "Epoch 00026: val_accuracy improved from 0.96650 to 0.96800, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 27/50\n",
      "\n",
      "Epoch 00027: val_accuracy did not improve from 0.96800\n",
      "Epoch 28/50\n",
      "\n",
      "Epoch 00028: val_accuracy did not improve from 0.96800\n",
      "Epoch 29/50\n",
      "\n",
      "Epoch 00029: val_accuracy did not improve from 0.96800\n",
      "Epoch 30/50\n",
      "\n",
      "Epoch 00030: val_accuracy did not improve from 0.96800\n",
      "Epoch 31/50\n",
      "\n",
      "Epoch 00031: val_accuracy improved from 0.96800 to 0.97350, saving model to models/Model3/best_model_test.h5\n",
      "Epoch 32/50\n",
      "\n",
      "Epoch 00032: val_accuracy did not improve from 0.97350\n",
      "Epoch 33/50\n",
      "\n",
      "Epoch 00033: val_accuracy did not improve from 0.97350\n",
      "Epoch 34/50\n",
      "\n",
      "Epoch 00034: val_accuracy did not improve from 0.97350\n",
      "Epoch 35/50\n",
      "\n",
      "Epoch 00035: val_accuracy did not improve from 0.97350\n",
      "Epoch 36/50\n",
      "\n",
      "Epoch 00036: val_accuracy did not improve from 0.97350\n",
      "Epoch 37/50\n",
      "\n",
      "Epoch 00037: val_accuracy did not improve from 0.97350\n",
      "Epoch 38/50\n",
      "\n",
      "Epoch 00038: val_accuracy did not improve from 0.97350\n",
      "Epoch 39/50\n",
      "\n",
      "Epoch 00039: val_accuracy did not improve from 0.97350\n",
      "Epoch 40/50\n",
      "\n",
      "Epoch 00040: val_accuracy did not improve from 0.97350\n",
      "Epoch 41/50\n",
      "\n",
      "Epoch 00041: val_accuracy did not improve from 0.97350\n",
      "Epoch 00041: early stopping\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "results3_lr = model3_lr.fit(\n",
    "    train_generator,\n",
    "    steps_per_epoch=100, # num_samples//batch_size\n",
    "    epochs=50,\n",
    "    callbacks= [es, mc],\n",
    "    validation_data=validation_generator,\n",
    "    verbose=4,\n",
    "    validation_steps=20) # num_samples//batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "db57b19f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plot_nn_training(results3_lr)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf_env_2",
   "language": "python",
   "name": "tf_env_2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}