Copy_of_StudentCopy_DoNotEdit_EmotionDetection_section3.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TiM6gYg0nhkY"
      },
      "source": [
        "<font color=\"#de3023\"><h1><b>REMINDER MAKE A COPY OF THIS NOTEBOOK, DO NOT EDIT</b></h1></font>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Qhr-G_92H2Bc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8afce65e-21c7-4824-8e05-167978174a69"
      },
      "source": [
        "#@title Run this to download data and prepare our environment!  { display-mode: \"form\" }\n",
        "\n",
        "import cv2\n",
        "import dlib\n",
        "import pickle\n",
        "import warnings\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import itertools\n",
        "\n",
        "import urllib.request\n",
        "\n",
        "from sklearn import metrics\n",
        "from scipy.spatial import distance\n",
        "from sklearn.metrics import accuracy_score\n",
        "from matplotlib import pyplot as plt\n",
        "from tqdm import tqdm,tqdm_pandas\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import re\n",
        "import keras\n",
        "\n",
        "from keras.models import Sequential\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "from keras.optimizers import Adam\n",
        "from keras.layers import Dense, Dropout, Activation, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D, BatchNormalization\n",
        "from keras.losses import categorical_crossentropy\n",
        "from keras.optimizers import Adam, SGD\n",
        "from keras.regularizers import l2\n",
        "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
        "from keras.models import load_model\n",
        "\n",
        "# grab tools from our tensorflow and keras toolboxes!\n",
        "import tensorflow as tf\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Activation, Dropout, Flatten, Dense\n",
        "from keras import optimizers\n",
        "\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "def model_to_string(model):\n",
        "    import re\n",
        "    stringlist = []\n",
        "    model.summary(print_fn=lambda x: stringlist.append(x))\n",
        "    sms = \"\\n\".join(stringlist)\n",
        "    sms = re.sub('_\\d\\d\\d','', sms)\n",
        "    sms = re.sub('_\\d\\d','', sms)\n",
        "    sms = re.sub('_\\d','', sms)  \n",
        "    return sms\n",
        "\n",
        "###Getting the csv data loaded\n",
        "!wget -O ./ferdata.csv \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/fer2013_5.csv\"\n",
        "\n",
        "###Getting the Dlib Shape predictor!\n",
        "!wget -O ./shape_predictor_68_face_landmarks.dat \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/shape_predictor_68_face_landmarks.dat\"\n",
        "\n",
        "###Getting the Xpure loaded\n",
        "!wget -O ./pureX.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/pureX.npy\"\n",
        "\n",
        "###Getting the Xdata loaded\n",
        "!wget -O ./dataX.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataX_edited.npy\"\n",
        "\n",
        "###Getting the Ydata loaded\n",
        "!wget -O ./dataY.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataY.npy\"\n",
        "\n",
        "print (\"Data Downloaded!\")\n",
        "\n",
        "'''\n",
        "Plots the confusion Matrix and saves it\n",
        "'''\n",
        "def plot_confusion_matrix(y_true,y_predicted):\n",
        "  cm = metrics.confusion_matrix(y_true, y_predicted)\n",
        "  print (\"Plotting the Confusion Matrix\")\n",
        "  labels = list(label_map.values())\n",
        "  df_cm = pd.DataFrame(cm,index = labels,columns = labels)\n",
        "  fig = plt.figure()\n",
        "  res = sns.heatmap(df_cm, annot=True,cmap='Blues', fmt='g')\n",
        "  plt.yticks([0.5,1.5,2.5,3.5,4.5], labels,va='center')\n",
        "  plt.title('Confusion Matrix - TestData')\n",
        "  plt.ylabel('True label')\n",
        "  plt.xlabel('Predicted label')\n",
        " \n",
        "  plt.show()\n",
        "  plt.close()\n",
        "\n",
        "def plot_graphs(history, best):\n",
        "  \n",
        "  plt.figure(figsize=[10,4])\n",
        "  # summarize history for accuracy\n",
        "  plt.subplot(121)\n",
        "  plt.plot(history.history['accuracy'])\n",
        "  plt.plot(history.history['val_accuracy'])\n",
        "  plt.title('model accuracy across training\\n best accuracy of %.02f'%best[1])\n",
        "  plt.ylabel('accuracy')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train', 'test'], loc='upper left')\n",
        "  \n",
        "  # summarize history for loss\n",
        "  plt.subplot(122)\n",
        "  plt.plot(history.history['loss'])\n",
        "  plt.plot(history.history['val_loss'])\n",
        "  plt.title('model loss across training\\n best loss of %.02f'%best[0])\n",
        "  plt.ylabel('loss')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train', 'test'], loc='upper left')\n",
        "  plt.show()\n",
        "\n",
        "#Integer to Label Mapping\n",
        "label_map = {\"0\":\"ANGRY\",\"1\":\"HAPPY\",\"2\":\"SAD\",\"3\":\"SURPRISE\",\"4\":\"NEUTRAL\"}\n",
        "\n",
        "\n",
        "#Load the 68 face Landmark file\n",
        "predictor = dlib.shape_predictor('./shape_predictor_68_face_landmarks.dat')\n",
        "\"\"\"\n",
        "Returns facial landmarks for the given input image path\n",
        "\"\"\"\n",
        "def get_landmarks(image):\n",
        "  \n",
        "  \n",
        "  #:type image : cv2 object\n",
        "  #:rtype landmarks : list of tuples where each tuple represents \n",
        "  #                  the x and y co-ordinates of facial keypoints\n",
        "  \n",
        "  #Bounding Box co-ordinates around the face(Training data is 48*48(cropped faces))\n",
        "  rects = [dlib.rectangle(left=1, top=1, right=47, bottom=47)]\n",
        "\n",
        "  #Read Image using OpenCV\n",
        "  #image = cv2.imread(image_path)\n",
        "  #Detect the Faces within the image\n",
        "  landmarks = [(p.x, p.y) for p in predictor(image, rects[0]).parts()]\n",
        "  return image,landmarks\n",
        "\n",
        "\"\"\"\n",
        "Display image with its Facial Landmarks\n",
        "\"\"\"\n",
        "def image_landmarks(image,face_landmarks):\n",
        "  \"\"\"\n",
        "  :type image_path : str\n",
        "  :type face_landmarks : list of tuples where each tuple represents \n",
        "                     the x and y co-ordinates of facial keypoints\n",
        "  :rtype : None\n",
        "  \"\"\"\n",
        "  radius = -4\n",
        "  circle_thickness = 1\n",
        "  image_copy = image.copy()\n",
        "  for (x, y) in face_landmarks:\n",
        "    cv2.circle(image_copy, (x, y), circle_thickness, (255,0,0), radius)\n",
        "    \n",
        "  plt.imshow(image_copy, interpolation='nearest')\n",
        "  plt.show()\n",
        "  \n",
        "\"\"\"\n",
        "Computes euclidean distance between 68 Landmark Points for our features\n",
        "e_dist is a list of features that will go into our model.\n",
        "Each feature is a distance between two landmark points, and every pair of points\n",
        "must have a feature.\n",
        "\"\"\"\n",
        "  \n",
        "\n",
        "def landmarks_edist(face_landmarks):\n",
        "  \n",
        "    e_dist = []\n",
        "    for i, j in itertools.combinations(range(68), 2):\n",
        "      e_dist.append(distance.euclidean(face_landmarks[i], face_landmarks[j]))\n",
        "    \n",
        "    return e_dist\n",
        "  \n",
        "def compare_learning(mlp, lm, cnn, vgg): # there's one model missing: MLP from pixels\n",
        "  \n",
        "  # summarize history for accuracy\n",
        "  plt.plot(vgg.history['val_accuracy'],)\n",
        "  plt.plot(cnn.history['val_accuracy'])\n",
        "  plt.plot(mlp.history['val_accuracy'],)\n",
        "  plt.plot(lm.history['val_accuracy'])\n",
        "  plt.ylabel('validitation accuracy')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['cnn_transfer', 'cnn_scratch', 'mlp_pixels', 'mlp_landmarks'], bbox_to_anchor=[1,1])\n",
        "  plt.xticks(range(0, epochs+1, 5), range(0, epochs+1, 5))\n",
        "  plt.show()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2021-07-05 05:03:26--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/fer2013_5.csv\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 167738172 (160M) [text/csv]\n",
            "Saving to: ‘./ferdata.csv’\n",
            "\n",
            "./ferdata.csv       100%[===================>] 159.97M   229MB/s    in 0.7s    \n",
            "\n",
            "2021-07-05 05:03:27 (229 MB/s) - ‘./ferdata.csv’ saved [167738172/167738172]\n",
            "\n",
            "--2021-07-05 05:03:27--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/shape_predictor_68_face_landmarks.dat\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.251.2.128, 142.250.141.128, 2607:f8b0:4023:c0d::80, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.251.2.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 99693937 (95M) [application/octet-stream]\n",
            "Saving to: ‘./shape_predictor_68_face_landmarks.dat’\n",
            "\n",
            "./shape_predictor_6 100%[===================>]  95.08M   150MB/s    in 0.6s    \n",
            "\n",
            "2021-07-05 05:03:28 (150 MB/s) - ‘./shape_predictor_68_face_landmarks.dat’ saved [99693937/99693937]\n",
            "\n",
            "--2021-07-05 05:03:28--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/pureX.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 46080128 (44M) [application/octet-stream]\n",
            "Saving to: ‘./pureX.npy’\n",
            "\n",
            "./pureX.npy         100%[===================>]  43.95M  35.0MB/s    in 1.3s    \n",
            "\n",
            "2021-07-05 05:03:29 (35.0 MB/s) - ‘./pureX.npy’ saved [46080128/46080128]\n",
            "\n",
            "--2021-07-05 05:03:29--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataX_edited.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 364480128 (348M) [application/octet-stream]\n",
            "Saving to: ‘./dataX.npy’\n",
            "\n",
            "./dataX.npy         100%[===================>] 347.59M   158MB/s    in 2.2s    \n",
            "\n",
            "2021-07-05 05:03:32 (158 MB/s) - ‘./dataX.npy’ saved [364480128/364480128]\n",
            "\n",
            "--2021-07-05 05:03:32--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataY.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 160128 (156K) [application/octet-stream]\n",
            "Saving to: ‘./dataY.npy’\n",
            "\n",
            "./dataY.npy         100%[===================>] 156.38K  --.-KB/s    in 0.001s  \n",
            "\n",
            "2021-07-05 05:03:32 (118 MB/s) - ‘./dataY.npy’ saved [160128/160128]\n",
            "\n",
            "Data Downloaded!\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jcpypKzI-ol0",
        "cellView": "form",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "73ca7610-553a-49b9-af40-cc19cba2b644"
      },
      "source": [
        "#@title If the previous code cell fails to download the data properly, please run this.\n",
        "import cv2\n",
        "import dlib\n",
        "import pickle\n",
        "import warnings\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import urllib.request\n",
        "from sklearn import metrics\n",
        "from scipy.spatial import distance\n",
        "from matplotlib import pyplot as plt\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.tree import DecisionTreeClassifier \n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.model_selection import train_test_split\n",
        "import io\n",
        "\n",
        "###Getting the csv data loaded\n",
        "!wget -O ./ferdata.csv \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/fer2013_5.csv\"\n",
        "\n",
        "###Getting the Dlib Shape predictor!\n",
        "!wget -O ./shape_predictor_68_face_landmarks.dat \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/shape_predictor_68_face_landmarks.dat\"\n",
        "\n",
        "###Getting the Xpure loaded\n",
        "!wget -O ./pureX.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/pureX.npy\"\n",
        "\n",
        "###Getting the Xdata loaded\n",
        "!wget -O ./dataX.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataX_edited.npy\"\n",
        "\n",
        "###Getting the Ydata loaded\n",
        "!wget -O ./dataY.npy \"https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataY.npy\"\n",
        "\n",
        "print (\"Data Downloaded!\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2021-07-05 05:03:33--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/fer2013_5.csv\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 167738172 (160M) [text/csv]\n",
            "Saving to: ‘./ferdata.csv’\n",
            "\n",
            "./ferdata.csv       100%[===================>] 159.97M   261MB/s    in 0.6s    \n",
            "\n",
            "2021-07-05 05:03:34 (261 MB/s) - ‘./ferdata.csv’ saved [167738172/167738172]\n",
            "\n",
            "--2021-07-05 05:03:34--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/shape_predictor_68_face_landmarks.dat\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.251.2.128, 142.250.141.128, 2607:f8b0:4023:c06::80, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.251.2.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 99693937 (95M) [application/octet-stream]\n",
            "Saving to: ‘./shape_predictor_68_face_landmarks.dat’\n",
            "\n",
            "./shape_predictor_6 100%[===================>]  95.08M   267MB/s    in 0.4s    \n",
            "\n",
            "2021-07-05 05:03:34 (267 MB/s) - ‘./shape_predictor_68_face_landmarks.dat’ saved [99693937/99693937]\n",
            "\n",
            "--2021-07-05 05:03:34--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/pureX.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.250.101.128, 142.251.2.128, 142.250.141.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.250.101.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 46080128 (44M) [application/octet-stream]\n",
            "Saving to: ‘./pureX.npy’\n",
            "\n",
            "./pureX.npy         100%[===================>]  43.95M   247MB/s    in 0.2s    \n",
            "\n",
            "2021-07-05 05:03:35 (247 MB/s) - ‘./pureX.npy’ saved [46080128/46080128]\n",
            "\n",
            "--2021-07-05 05:03:35--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataX_edited.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.251.2.128, 142.250.141.128, 2607:f8b0:4023:c0d::80, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.251.2.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 364480128 (348M) [application/octet-stream]\n",
            "Saving to: ‘./dataX.npy’\n",
            "\n",
            "./dataX.npy         100%[===================>] 347.59M   263MB/s    in 1.3s    \n",
            "\n",
            "2021-07-05 05:03:36 (263 MB/s) - ‘./dataX.npy’ saved [364480128/364480128]\n",
            "\n",
            "--2021-07-05 05:03:36--  https://storage.googleapis.com/inspirit-ai-data-bucket-1/Data/AI%20Scholars/Sessions%206%20-%2010%20(Projects)/Project%20-%20Emotion%20Detection/dataY.npy\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 142.251.2.128, 142.250.141.128, 2607:f8b0:4023:c0d::80, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|142.251.2.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 160128 (156K) [application/octet-stream]\n",
            "Saving to: ‘./dataY.npy’\n",
            "\n",
            "./dataY.npy         100%[===================>] 156.38K  --.-KB/s    in 0.001s  \n",
            "\n",
            "2021-07-05 05:03:36 (109 MB/s) - ‘./dataY.npy’ saved [160128/160128]\n",
            "\n",
            "Data Downloaded!\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EIrowe2Atd1W"
      },
      "source": [
        "In this notebook we'll be:\n",
        "1.   Understanding and Building Neural Networks\n",
        "2.   Applying Neural Networks to Predict Emotions\n",
        "3.   Implementing CNNs for Emotion Detection\n",
        "4.   Implementing Transfer Learning\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xq10yh_OuVlg"
      },
      "source": [
        "#Understanding and building Neural Networks"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OOfhabJ_4hvE"
      },
      "source": [
        "A neural network is a series of algorithms that endeavors to recognize underlying relationships in a set of data through a process that mimics the way the human brain operates. Basically, neural network algorithm helps computers think and learn like humans.\n",
        "\n",
        "The whole idea of artificial neural network is based on the concept of the structure and functions of a human brain. A human brain consists of neurons that process and transmit information between themselves."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4DcLenvXu7th"
      },
      "source": [
        "### Why try neural networks?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s77Fm2CbvZi7"
      },
      "source": [
        "Our Baseline model performed at ~50 %\n",
        "\n",
        "Human accuracy for fer2013 is around 65 %, suggesting that we could do better!\n",
        "\n",
        "What we require now is a model that can capture more complicated patterns. One class of models that is able to do this and has been very effective for doing this for images are neural networks. \n",
        "\n",
        "\n",
        "### What are neural networks?\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZB_uPibnv2UG"
      },
      "source": [
        "## Instructor-Led Discussion: Building networks\n",
        "\n",
        "To build neural networks in Python, we use the packages known as `tensorflow` and `keras`. Let's learn how to build and use these networks!\n",
        "\n",
        "Tensorflow calls the various machine learning algorithms that it uses 'models'.  These 'models' are 'learning machines.''\n",
        "\n",
        "1. We **teach** models by **training** them on **data**. \n",
        "2. We **use** models to **predict** things. \n",
        "\n",
        "Here is example code for building a Tensorflow model with Keras:"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RUtJ7_FkkPr-"
      },
      "source": [
        "The things you'll want to pay most attention to as we go over how to build networks are: \n",
        "1. The number of neurons\n",
        "2. The activation of the neurons\n",
        "3. The losses and metrics\n",
        "\n",
        "Everything else will work with the default settings!\n",
        "\n",
        "Let's walk though what each of these lines of code means!\n",
        "\n",
        "**1. Specify model**\n",
        "\n",
        "```\n",
        "model = Sequential()\n",
        "```\n",
        "In this line of code, we build our network where the information flows from LEFT to RIGHT through the network in ONE DIRECTION as opposed to multiple directions. Neurons on the right never pass informations to neurons on the left of it. \n",
        "\n",
        "\n",
        "**2. Add layers to the network**\n",
        "```\n",
        "model.add(Dense(4,input_shape = (3,), activation = 'sigmoid'))\n",
        "```\n",
        "In this code, we `add` a `layer` of neurons to our network. \n",
        "\n",
        "This layers consists of 4 neurons. Each neuron is DENSE and connects to all of the previous layer's inputs and all of the subsequent layers outputs. We specify that there are 3 inputs here.\n",
        "\n",
        "We also specify what kind of output the neuron will give. If you want the neuron to output a number between 0 and 1 (like a probability!) you would use 'softmax' or 'sigmoid'. If you want the neuron to output any number, you can use 'linear'! You'll also often see 'relu', which is when a neuron will only output positive numbers. \n",
        "\n",
        "```\n",
        "model.add(Dense(1, activation = 'linear'))\n",
        "```\n",
        "This code adds ANOTHER layer to the network that has 1 neuron. This one neuron is used to predict a continuous value!\n",
        "\n",
        "**3. Turn the model on by compiling it** \n",
        "\n",
        "After having built the network, we want to train and use it, so we have to 'turn it on' and 'compile' it. To turn it on, we have to specify at the very least, a loss, an optimizer, and some ways of evaluating the model (metrics). Don't worry too much about what this means! Just know that this is necessary. \n",
        "\n",
        "```\n",
        "model.compile(loss='mean_squared_error',\n",
        "               optimizer = 'adam',\n",
        "               metrics = ['mean_squared_error'])\n",
        "  ```\n",
        "  \n",
        "  Once we've created our network, we can use it very simply! Just like we did with sklearn, we define our input data (x), the true predictions from that data (y), and then train our model with `fit`. \n",
        "\n",
        "```\n",
        "model.fit(x, y)\n",
        "```\n",
        "\n",
        "To use the model, you can use it to predict something with:\n",
        "```\n",
        "y = model.predict_classes(x)\n",
        "```\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ACtfnujXvzaT"
      },
      "source": [
        "### For reference, this is a lot like the kind of models we've been training with scikit learn.\n",
        "\n",
        "The big difference is that defining the model is more involved. We can look at an example from KNN: \n",
        "\n",
        "```\n",
        "# define the model instance \n",
        "knn = KNeighborsClassifier()\n",
        "\n",
        "# train model with X_train\n",
        "knn.fit(X_train, y_train) \n",
        "\n",
        "# predict testing data\n",
        "y_predict = knn.predict(X_test)\n",
        "\n",
        "```"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YJXxWYXcwoGn"
      },
      "source": [
        "##Exercise: Coding a 2 hidden layer MLP\n",
        "\n",
        "Just as we went over last week, neural networks look something like this: \n",
        "\n",
        "![A 2 layer neural network](https://cdn-images-1.medium.com/max/1600/1*DW0Ccmj1hZ0OvSXi7Kz5MQ.jpeg)\n",
        "\n",
        "Each orange and blue node is a neuron. The network itself is composed of a bunch of neurons that talk to each other and eventually give us a prediction. Let's get a bit more concrete with this..."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2MamYyRf16Ii"
      },
      "source": [
        "### How can we build this network above? Fill in the ___ 's below "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g1FadMwUkOwV"
      },
      "source": [
        "# First, we initialize our model\n",
        "tmp_model = Sequential()\n",
        "# then we add a \"Dense\" (i.e. fully connected) layer\n",
        "tmp_model.add(Dense(7, input_shape=(5,), activation = 'relu')) # for the first layer we specify the input dimensions \n",
        "# then we have to add another layer \n",
        "tmp_model.add(Dense(7, activation = 'relu'))\n",
        "# we end by defining the output layer, which has the number of dimensions of the predictions we're making\n",
        "tmp_model.add(Dense(4, activation = 'linear'))\n",
        "# we finalize the model by \"compiling\" it and defining some other hyperparameters \n",
        "tmp_model.compile(loss='mean_squared_error', optimizer='adam', metrics=['mean_squared_error'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-rrqPkJWVLOY"
      },
      "source": [
        "## Exercise: Building our custom neural network"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "25DoFvA0VWnq"
      },
      "source": [
        "###Keypoints"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tchJVYOTVZAh"
      },
      "source": [
        "We will build a simple 3-layer network for our first model!\n",
        "\n",
        "\n",
        "For our model, we have as our layers: \n",
        "* Input Layer:  However many inputs there are!\n",
        "* Layer 1 (Hidden): 1024 neurons that are activated by `'relu'` and weights are intialized using `kernel_initializer='glorot_normal'`. Specify the input shape as `(2278,)`.\n",
        "* Layer 2 (Hidden): 512 neurons that are activated by `'relu'` and weights are intialized using `kernel_initializer='glorot_normal'`\n",
        "* Layer 3 (Output): 5 neuron that should have an appropriate activation. \n",
        "* We will compile with the `optimizers.SGD(lr=0.001)` optimizer\n",
        "\n",
        "As a hint for the output activation and the compilation loss, we know that:\n",
        "* Multi-class classification problems require an output activation of `'softmax'` and a loss of `'categorical_crossentropy'`\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sv3U4FxbVyfs"
      },
      "source": [
        "###Build your Model (call it `perceptron`)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WJh4klMtV3Q_"
      },
      "source": [
        "#YOUR CODE HERE\n",
        "\n",
        "perceptron = Sequential()\n",
        "\n",
        "perceptron.add(Dense(1024, input_shape = (2278,), activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "perceptron.add(Dense(512, activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "perceptron.add(Dense(5, activation = 'softmax'))\n",
        "\n",
        "perceptron.compile(loss='categorical_crossentropy', optimizer=optimizers.SGD(lr=0.001), metrics=['accuracy'])\n",
        "\n",
        "#END CODE"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rWDzNTh9Wm0e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "cebd4b2b-dfd9-4413-8476-fa74977e2cc0"
      },
      "source": [
        "#@title Run this to test if your model is right! { display-mode: \"form\" }\n",
        "perceptron_answer = Sequential()\n",
        "perceptron_answer.add(Dense(units = 1024, input_shape = (2278,),kernel_initializer='glorot_normal',activation = 'relu'))\n",
        "perceptron_answer.add(Dense(units = 512,kernel_initializer='glorot_normal' , activation = 'relu'))\n",
        "perceptron_answer.add(Dense(units = 5, activation = 'softmax'))\n",
        "    \n",
        "perceptron_answer.compile(\n",
        "    loss='categorical_crossentropy',\n",
        "    optimizer=SGD(lr=0.001),\n",
        "    metrics=['accuracy'])\n",
        "\n",
        "if model_to_string(perceptron) == model_to_string(perceptron_answer):\n",
        "  print('Good job, you specified it correctly!')\n",
        "else: \n",
        "  print('Please check your code again!')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Good job, you specified it correctly!\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AgQEGSXJueyV"
      },
      "source": [
        "###Keras Neural Network Model Hyperparameters\n",
        "\n",
        "Neural Network performance depends a lot on *how much* they train. As we'll see, they usually get better with more training BUT actually can get worse with too much training. With too much training, our model can get overconfident in its abilities with the training manual (overfitting), and so doesn't actually think (generalize) when it is tested. \n",
        "\n",
        "The  `fit()` function pertain to how the neural networks train. Don't worry too much about the extra options, what really matters for us is that the right data is specified.\n",
        "\n",
        "\n",
        "What are all these options?\n",
        "* `epochs`: how many times the model trains on the entire data set\n",
        "* `batch_size`: number of samples processed at a time\n",
        "* `shuffle`: mixes the training dataset so the model pays better attention to the data and learns better while training\n",
        "* `validation_data`: we request that our model tests itself on the `test_data` after every epoch. **It is essential to use validation data to test for overfitting!**\n",
        "* `callbacks`: With a custom command, we tell our model to save the best version of itself to a model file called `best_dnn_model.h5`. \n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oeb1iFnZw3W1"
      },
      "source": [
        "#Applying Neural Networks (MLPs) to predicting emotions"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kGdpkyMmqs-G"
      },
      "source": [
        "### Loading and formatting data for all models in notebook "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pf4GspMjAGoE"
      },
      "source": [
        "First, let's try to build a 4-Layer Neural Network  for our training data to achieve accuracy closer to the human accuracy on the dataset (around 65%)\n",
        "\n",
        "We want to identify the key things that we need to design our network. \n",
        "\n",
        "In your group, discuss: \n",
        "\n",
        "* What are our inputs?\n",
        "* What is/are our outputs?\n",
        "\n",
        "How could this look in a neural network diagram?\n",
        "\n",
        "**Show your instructor your diagram once you're done!!**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G-hJkIeQthKl"
      },
      "source": [
        "##Activity: Train Neural Network on Emotion Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LDZqO4EMHeSZ"
      },
      "source": [
        "###Set some hyper parameters for all models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "284_6dNsHiS5"
      },
      "source": [
        "# the number of times we pass all the training data through the model \n",
        "epochs = 20\n",
        "# the number of examples we pass to the model at each time\n",
        "batch_size = 64\n",
        "# the proportion of testing data we set aside (e.g. 10%)\n",
        "test_ratio = .1\n",
        "# the number of emotion categories we have to predict\n",
        "n_labels = 5 "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gNJx9QCE8w8p"
      },
      "source": [
        "###Load the original (unprocessed) data saved on Day 2!"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oG7AIjf3xWBa"
      },
      "source": [
        "# load data \n",
        "dataX_pixels = np.load('pureX.npy')\n",
        "dataY_pixels = np.load('dataY.npy')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7y6HwMAg85jx"
      },
      "source": [
        "### Convert labels to one-hot encoded labels\n",
        "\n",
        "One hot encoding is a representation of categorical variables as binary vectors"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qA_0acGq9AJN",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bf2a31df-2179-4aae-f828-af4d7533af72"
      },
      "source": [
        "# convert labels to one hot encoding\n",
        "y_onehot = to_categorical(dataY_pixels, len(set(dataY_pixels)))\n",
        "\n",
        "# what does this data type look like? \n",
        "\n",
        "y_onehot"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[1., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 0., 1.],\n",
              "       [0., 0., 0., 0., 1.],\n",
              "       ...,\n",
              "       [1., 0., 0., 0., 0.],\n",
              "       [0., 0., 1., 0., 0.],\n",
              "       [1., 0., 0., 0., 0.]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a-LTjtaknAqa"
      },
      "source": [
        "#@title Is it clear how these are two equivalent ways to represent the target?\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1RodV9eQ71Yz"
      },
      "source": [
        "###Split and Standardize your Data\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cVO4wnTY7rRk"
      },
      "source": [
        "# split Data into Train, Test (90-10)\n",
        "X_train, X_test, y_train, y_test = train_test_split(dataX_pixels, y_onehot, test_size=test_ratio, random_state=42)\n",
        "\n",
        "#### Standardize the data ##########\n",
        "pixel_scaler = StandardScaler()\n",
        "pixel_scaler.fit(X_train)\n",
        "X_train = pixel_scaler.transform(X_train)\n",
        "X_test = pixel_scaler.transform(X_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iaQQkwNsHGSM"
      },
      "source": [
        "## Exercise: Build a simple MLP for emotion detection (call it `mlp_model`)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QBk3kFJSgQJ0",
        "cellView": "both"
      },
      "source": [
        "#YOUR CODE HERE\n",
        "\n",
        "mlp_model = Sequential()\n",
        "\n",
        "mlp_model.add(Dense(1024, input_shape = (X_train.shape[1],), activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "mlp_model.add(Dropout(rate = 0.5))\n",
        "mlp_model.add(Dense(512, activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "mlp_model.add(Dense(5, activation = 'softmax'))\n",
        "\n",
        "mlp_model.compile(loss='categorical_crossentropy', optimizer=optimizers.SGD(lr=0.001), metrics=['accuracy'])\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "#END CODE"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IwYofcaXzWVa",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b2e59823-d334-49df-a8f5-8d92487e16e8"
      },
      "source": [
        "# Compiling the model with SGD optimixer and categorical crossentropy loss\n",
        "mlp_model.compile(loss=categorical_crossentropy, optimizer=SGD(lr=0.001), metrics=['accuracy'])\n",
        "              \n",
        "#Saves the Best Model Based on Val Loss\n",
        "checkpoint = ModelCheckpoint('best_mlp_model.h5', verbose=1, monitor='val_accuracy', save_best_only=True,  mode='auto')  \n",
        "\n",
        "#training the model\n",
        "mlp_history = mlp_model.fit(X_train, y_train, batch_size=batch_size, epochs=30, verbose=1, \n",
        "                            callbacks=[checkpoint], validation_data=(X_test, y_test), shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "282/282 [==============================] - 15s 5ms/step - loss: 1.8875 - accuracy: 0.2358 - val_loss: 1.5581 - val_accuracy: 0.3095\n",
            "\n",
            "Epoch 00001: val_accuracy improved from -inf to 0.30950, saving model to best_mlp_model.h5\n",
            "Epoch 2/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.6920 - accuracy: 0.2885 - val_loss: 1.5046 - val_accuracy: 0.3515\n",
            "\n",
            "Epoch 00002: val_accuracy improved from 0.30950 to 0.35150, saving model to best_mlp_model.h5\n",
            "Epoch 3/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.6216 - accuracy: 0.3105 - val_loss: 1.4775 - val_accuracy: 0.3585\n",
            "\n",
            "Epoch 00003: val_accuracy improved from 0.35150 to 0.35850, saving model to best_mlp_model.h5\n",
            "Epoch 4/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.5904 - accuracy: 0.3354 - val_loss: 1.4558 - val_accuracy: 0.3790\n",
            "\n",
            "Epoch 00004: val_accuracy improved from 0.35850 to 0.37900, saving model to best_mlp_model.h5\n",
            "Epoch 5/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.5472 - accuracy: 0.3522 - val_loss: 1.4404 - val_accuracy: 0.3825\n",
            "\n",
            "Epoch 00005: val_accuracy improved from 0.37900 to 0.38250, saving model to best_mlp_model.h5\n",
            "Epoch 6/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.5189 - accuracy: 0.3580 - val_loss: 1.4252 - val_accuracy: 0.3940\n",
            "\n",
            "Epoch 00006: val_accuracy improved from 0.38250 to 0.39400, saving model to best_mlp_model.h5\n",
            "Epoch 7/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.5019 - accuracy: 0.3665 - val_loss: 1.4156 - val_accuracy: 0.4040\n",
            "\n",
            "Epoch 00007: val_accuracy improved from 0.39400 to 0.40400, saving model to best_mlp_model.h5\n",
            "Epoch 8/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.4942 - accuracy: 0.3681 - val_loss: 1.4085 - val_accuracy: 0.4090\n",
            "\n",
            "Epoch 00008: val_accuracy improved from 0.40400 to 0.40900, saving model to best_mlp_model.h5\n",
            "Epoch 9/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.4731 - accuracy: 0.3850 - val_loss: 1.4044 - val_accuracy: 0.4110\n",
            "\n",
            "Epoch 00009: val_accuracy improved from 0.40900 to 0.41100, saving model to best_mlp_model.h5\n",
            "Epoch 10/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.4689 - accuracy: 0.3754 - val_loss: 1.3938 - val_accuracy: 0.4185\n",
            "\n",
            "Epoch 00010: val_accuracy improved from 0.41100 to 0.41850, saving model to best_mlp_model.h5\n",
            "Epoch 11/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.4445 - accuracy: 0.3949 - val_loss: 1.3883 - val_accuracy: 0.4205\n",
            "\n",
            "Epoch 00011: val_accuracy improved from 0.41850 to 0.42050, saving model to best_mlp_model.h5\n",
            "Epoch 12/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.4359 - accuracy: 0.3995 - val_loss: 1.3825 - val_accuracy: 0.4235\n",
            "\n",
            "Epoch 00012: val_accuracy improved from 0.42050 to 0.42350, saving model to best_mlp_model.h5\n",
            "Epoch 13/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.4249 - accuracy: 0.4049 - val_loss: 1.3815 - val_accuracy: 0.4230\n",
            "\n",
            "Epoch 00013: val_accuracy did not improve from 0.42350\n",
            "Epoch 14/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.4131 - accuracy: 0.4052 - val_loss: 1.3773 - val_accuracy: 0.4315\n",
            "\n",
            "Epoch 00014: val_accuracy improved from 0.42350 to 0.43150, saving model to best_mlp_model.h5\n",
            "Epoch 15/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.4113 - accuracy: 0.4057 - val_loss: 1.3729 - val_accuracy: 0.4325\n",
            "\n",
            "Epoch 00015: val_accuracy improved from 0.43150 to 0.43250, saving model to best_mlp_model.h5\n",
            "Epoch 16/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3976 - accuracy: 0.4184 - val_loss: 1.3696 - val_accuracy: 0.4330\n",
            "\n",
            "Epoch 00016: val_accuracy improved from 0.43250 to 0.43300, saving model to best_mlp_model.h5\n",
            "Epoch 17/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3912 - accuracy: 0.4173 - val_loss: 1.3662 - val_accuracy: 0.4320\n",
            "\n",
            "Epoch 00017: val_accuracy did not improve from 0.43300\n",
            "Epoch 18/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3786 - accuracy: 0.4309 - val_loss: 1.3614 - val_accuracy: 0.4375\n",
            "\n",
            "Epoch 00018: val_accuracy improved from 0.43300 to 0.43750, saving model to best_mlp_model.h5\n",
            "Epoch 19/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3847 - accuracy: 0.4214 - val_loss: 1.3600 - val_accuracy: 0.4395\n",
            "\n",
            "Epoch 00019: val_accuracy improved from 0.43750 to 0.43950, saving model to best_mlp_model.h5\n",
            "Epoch 20/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3685 - accuracy: 0.4255 - val_loss: 1.3555 - val_accuracy: 0.4425\n",
            "\n",
            "Epoch 00020: val_accuracy improved from 0.43950 to 0.44250, saving model to best_mlp_model.h5\n",
            "Epoch 21/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3793 - accuracy: 0.4240 - val_loss: 1.3558 - val_accuracy: 0.4420\n",
            "\n",
            "Epoch 00021: val_accuracy did not improve from 0.44250\n",
            "Epoch 22/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3629 - accuracy: 0.4341 - val_loss: 1.3516 - val_accuracy: 0.4480\n",
            "\n",
            "Epoch 00022: val_accuracy improved from 0.44250 to 0.44800, saving model to best_mlp_model.h5\n",
            "Epoch 23/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3587 - accuracy: 0.4334 - val_loss: 1.3497 - val_accuracy: 0.4525\n",
            "\n",
            "Epoch 00023: val_accuracy improved from 0.44800 to 0.45250, saving model to best_mlp_model.h5\n",
            "Epoch 24/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3526 - accuracy: 0.4325 - val_loss: 1.3454 - val_accuracy: 0.4560\n",
            "\n",
            "Epoch 00024: val_accuracy improved from 0.45250 to 0.45600, saving model to best_mlp_model.h5\n",
            "Epoch 25/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3519 - accuracy: 0.4410 - val_loss: 1.3435 - val_accuracy: 0.4500\n",
            "\n",
            "Epoch 00025: val_accuracy did not improve from 0.45600\n",
            "Epoch 26/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3495 - accuracy: 0.4434 - val_loss: 1.3432 - val_accuracy: 0.4500\n",
            "\n",
            "Epoch 00026: val_accuracy did not improve from 0.45600\n",
            "Epoch 27/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3368 - accuracy: 0.4485 - val_loss: 1.3394 - val_accuracy: 0.4550\n",
            "\n",
            "Epoch 00027: val_accuracy did not improve from 0.45600\n",
            "Epoch 28/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3304 - accuracy: 0.4497 - val_loss: 1.3395 - val_accuracy: 0.4605\n",
            "\n",
            "Epoch 00028: val_accuracy improved from 0.45600 to 0.46050, saving model to best_mlp_model.h5\n",
            "Epoch 29/30\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3248 - accuracy: 0.4511 - val_loss: 1.3360 - val_accuracy: 0.4540\n",
            "\n",
            "Epoch 00029: val_accuracy did not improve from 0.46050\n",
            "Epoch 30/30\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3288 - accuracy: 0.4481 - val_loss: 1.3322 - val_accuracy: 0.4565\n",
            "\n",
            "Epoch 00030: val_accuracy did not improve from 0.46050\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qF0MmGV8Lm7y"
      },
      "source": [
        "##Neural Network Model Evaluation on pixel inputs"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ORXi0y71t3o8"
      },
      "source": [
        "###Evaluate best model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DnjfotXoth4e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bce1ea16-db94-4624-c590-4fd14c3ea6a7"
      },
      "source": [
        "mlp_performance = mlp_model.evaluate(X_test, y_test, batch_size=64)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "32/32 [==============================] - 0s 2ms/step - loss: 1.3322 - accuracy: 0.4565\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3vgEP7OnGlR9"
      },
      "source": [
        "###Visualize accuracy and loss over training + display best model's performance"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ktHNwzn0G73y",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 310
        },
        "outputId": "f99a9604-6e4a-4b0f-b581-a627905979ac"
      },
      "source": [
        "plot_graphs(mlp_history, mlp_performance); "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oPeyL20OFzjN"
      },
      "source": [
        "###Plot the Confusion Matrix"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MgkxPwg2F7eL",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "outputId": "b0e4dd31-f83a-4cb7-9103-4a3418090515"
      },
      "source": [
        "y_pred_mlp = mlp_model.predict_classes(X_test)\n",
        "y_true = np.argmax(y_test,axis=1)\n",
        "plot_confusion_matrix(y_true, y_pred_mlp)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Plotting the Confusion Matrix\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAAEWCAYAAACufwpNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3gUVffA8e9JQg0lhAREWkCk2VBQigUQG1iwIBoV7KiAClgRVERRUVRQioKiKNJeUUF9FRSp0kGaguUVFEIJhJoQSDu/P2YSl5CySTbZ3fzO53nmSeZOO7Nlzt57p4iqYowxxvhCiL8DMMYYU3pYUjHGGOMzllSMMcb4jCUVY4wxPmNJxRhjjM9YUjHGGOMzllRMsRKRCiLylYgcEpH/FGE9t4vIXF/G5g8i8q2I3OnvOIwpLpZUDAAicpuIrBaRRBHZ5R78LvLBqrsBNYHqqnpzYVeiqp+q6hU+iOcEItJBRFREvshWfo5bvsDL9QwRkcn5zaeqnVV1UiHDzW3bF7vvW6KIJLlxJ3oM9QqxThWRRh7jHUQkw2OdO0RkhoicX4B1evUameBmScUgIgOAkcDLOAmgHjAW6OqD1dcHflfVNB+sq7jsBdqKSHWPsjuB3321AXEUy/dNVReraiVVrQSc4RZHZJap6j8+2tROdxuVgTbAFmCxiHTy0fpNaaCqNvw/HoCqQCJwcx7zlMNJOjvdYSRQzp3WAdgBPAbEA7uAu91pLwApQKq7jXuBIcBkj3XHAAqEueN3AX8BR4CtwO0e5Us8lmsHrAIOuX/beUxbALwI/OSuZy4Qlcu+Zcb/LtDHLQsF4oDngAUe844CtgOHgTXAxW75Vdn2c71HHMPcOJKBRm7Zfe70ccBMj/UPB+YBUoT3M/vrWRX4wH1f4oCXgFB3WiNgofsa7gOmu+WL3HUkuftzS+brlMP2RgOri/Aa3Q1sdt+nv4AH/P2dsKFog98DsMHPHwDny56WeRDKZZ6hwHKgBhANLAVedKd1cJcfCpQBugBHgWru9CGcmESyj2cdBIFw92DUxJ1WCzjD/f8u3KQCRAIHgB7ucrHueHV3+gLgf0BjoII7/mou+9YBJ6m0A1a4ZV2AOcB9nJhU7gCqu9t8DNgNlM9pvzzi+Aen9hDmvj4L+DepVMSpDd0FXIxzYK9TxPcz6/V0x78A3nNf2xrAyswDNzAVGITTYlEeuMhjPQo0yv465bC9S4EMILyQr9HVwGmAAO3dz855/v5e2FD4wZq/THVgn+bdPHU7MFRV41V1L04NpIfH9FR3eqqq/hfnl2iTQsaTAZwpIhVUdZeq/pLDPFcDf6jqJ6qapqpTcZpirvWY50NV/V1Vk4EZQIu8NqqqS4FIEWkC9AQ+zmGeyaqa4G7zDZwaXH77+ZGq/uIuk5ptfUdxXsc3gcnAw6q6I5/1eU1EauIkyH6qmqSq8cBbwK3uLKk4zZOnquoxVV1SiM3sxEkIEVDw10hVv1HV/6ljIU6t8uJCxGEChCUVkwBEiUhYHvOcCvztMf63W5a1jmxJ6ShQqaCBqGoSTlPLg8AuEflGRJp6EU9mTLU9xncXIp5PgL5AR5xf+CcQkcdFZLN7JttBnKalqHzWuT2viaq6AqfZR3CSX45E5BePTnJvD7r1cWpHu0TkoBvzezg1FoAn3e2udNd/j5fr9VQbp1Zz0I2zQK+RiHQWkeUist+dv0te85vAZ0nFLAOOA9fnMc9OnANUpnpuWWEk4TT7ZDrFc6KqzlHVy3GavrYAE7yIJzOmuELGlOkToDfwX7cWkcU9kD8JdMdp2ovA6YuQzNBzWWeetwEXkT44v+Z3uuvPeSWqZ+i/He+LvdkZnIR2HKc/KcIdqqjqGe46d6vq/ap6KvAAMNbzjC8v3QCsVdWkgr5GIlIOmAmMAGq68//XY34ThCyp/D+nqodwOqTHiMj1IlJRRMq4vyBfc2ebCgwWkWgRiXLnL+ypoeuAS0SknohUBQZmThCRmiLSVUTCcQ6GiTjNYdn9F2jsngYdJiK3AM2BrwsZEwCquhWnXX9QDpMr4/Qd7QXCROQ5oIrH9D1ATEHO8BKRxjgd53fgNIM9KSJ5NtMVhKruwmlOekNEqohIiIicJiLt3e3fLCJ13NkP4Bz0M1/vPUDDXOIWEaktIs/j9Ds9404q6GtUFieh7gXSRKQz4PPTxk3JsqRicNu+BwCDcb7g23Gagb50Z3kJWA1sADYCa92ywmzre2C6u641nJgIQtw4dgL7cQ7wD+WwjgTgGpyO4AScX8fXqOq+wsSUbd1LVDWnWtgc4DucjvW/gWOc2LSVeWFngoiszW87bnPjZGC4qq5X1T9wDs6fuL/gfaUnzsH7V5zE8RlOLRDgfGCFiCQCs4FHVfUvd9oQYJLbbNbdLTvVnTcR54y7s4AOqpp5UWqBXiNVPQI8gtPsdwC4zY3DBDFRtYd0GWOM8Q2rqRhjjPEZSyrGGGN8xpKKMcYYn7GkYowxxmfyuuDN5OKNhX+VurMbrm1ySv4zBaH9SSn+DsHnIsPL+juEYnEoOTX/mYLQ+Q2qFvm6mwrn9vXqmJP882i/X+NjNRVjjDE+YzUVY4wJdMXz1IRiYUnFGGMCXUiovyPwmiUVY4wJdOL3rhKvWVIxxphAZ81fxhhjfMZqKsYYY3zGairGGGN8xmoqxhhjfMbO/jLGGOMz1vxljDHGZ6z5yxhjjM9YTcUYY4zPWFIxxhjjM6HWUW+MMcZXrE/FGGOMz1jzlzHGGJ+xmkrRicj1wBdAM1XdIiIxwFbgEVV9x51nNLBaVT9yxwcAvYBUIAOYBzylqqkisg04AihwAOgJXAFcpqq3uMtXAX4GLlfVv4p7Hxd89Cb/bFxJhcoR3DzkXQD+Wr2YNV9N5sDu7dwwcCTRMY0B+GPFj2yYMzNr2YS4rdw4+B2i6p5W3GEW2b23dKFChXBCQkMIDQ3lrfFTGD7kKeK2bwMgKfEI4ZUq8/YH0/0bqJd27fibsa8OyhqP3xXHjT168efmTeyO+xuAo4mJVKxUiRdHT/ZXmIVS2t4rgJ3b/2b0K89kjcfv3km3Hr246oZYAP4781OmTBjFuOlzqVw1wl9h5s1qKj4RCyxx/z7vlsUDj4rIe6p6wnNiReRBnCTRRlUPikhZYABQASfJAHRU1X0i8gIwGCcB3S0il6nqD8BQYGJJJBSAJu0u58yO1zH/wxFZZdVq1+fyh55l8eS3T5j39NaXcnrrSwHYv2Mrc8YODYqEkmnYyPFUjaiWNf7UkOFZ/38w5g0qhlfyR1iFUqtO/axkkZGeTr+e19CybQeuvD42a56pE0ZRITzcXyEWSWl6rwBOrVufl8d+Cjjv18N3XE2rdh0ASNi7h41rllO9RoA/TjuIaioBmf5EpBJwEXAvcKvHpL04tY87c1hsEPCQqh4EUNUUVX1VVQ/nMO8yoLaqKvAgMFJEWgGdgNd9tyd5q9X4LMqFVz6hrFqtekScUifP5f5ctZDTzm9fnKGVGFVlyfzvaX/ZVf4OpVB+Wb+K6FPqEFWzVlaZqrJy8Q+0aX+FHyPzvWB/rwB+WbeKGrX+fb8mv/cWt973MEKAH7RDQr0bAkBAJhWgK/Cdqv4OJIhIS49pw4HHRSTrFXSbrSqp6lYv138V8CWAqm4A5uAkq4ez14AC0f9WLaTRBR38HUYBCM893pt+99/Gd7NnnjDllw1riYiM5NQ69f0UW9GsWPg9bTqcmDx+27SOKhGRnFK7np+iKorS+14BLFv4PW3d92vNsoVUqx5N/YaN/RyVFyTEuyEABEYUJ4sFprn/T3PHAXCbplYAt+W2sIhcKSLrRGSbiLTzmDRfROKAzsBUj/IxQJyqLshjnb1EZLWIrF7+1dTcZit28X9tIaxseSJrx/gthoJ6bfSHjHp/KkNeG803X05n0/o1WdMW/fAdl3QKzl++aamp/LxiMRdcdOkJ5csXzj0p0QSL0vpegfN+rV2+iNYXd+L4sWPMnvYR3Xo+4O+wvCPi3RAAAi6piEgkcCnwvtu5/gTQHU6on74MPJVZ5jZxJYpIA3d8jqq2ADYBZT2W6wjUB9YBL3iUZ7hDrlR1vKq2UtVWba6NzWvWYvXnqoU0uiC4mr6qR9cAIKJaJG0vvpTfN/8CQHpaGssW/8jFHa/0Z3iFtmH1Uuqf1oSq1apnlaWnp7Fm6XxaX3KZHyMrvNL6XgGsX72UmEZNqVqtOvG7drB3906eeeh2+vXsyv598Qzu24OD+/f5O8ycWU2lSLoBn6hqfVWNUdW6OGd91c2cQVW3AL8C13os9wowTkQiAEREgPLZV66qaUA/oKebwIKGZmTw15rFQdWfciw5maNHk7L+/3nVMuo3cE4wWLdmBbXrxRBVo6Y/Qyy05QvnntRv8svPq6hVJ4bIqODbp9L8XgEsWzA3q+mrboNGjJ0+h5Efz2Lkx7OIjKrBS6M/ISIyys9R5iKIkkognv0Vi9Nv4mkmMDBb2TCc038zjQPCgRUichxIBH7KNg8AqrpLRKYCfYAXfRR3gc2b8Co7f9vAscTDfPrkHbS8rgflwiuxdOo4khMP8d07z1O9bkO69BsGwK4/NlGpWhRVomvls+bAcfBAAsMGDwAgPT2d9pd1pmXrCwFY9OMc2gdpc8rxY8ls+nkldz184sdyxaLvg7aDvrS+VwDHjiWzae0K7nkk+2EkSARIJ7w3xDkByhTEGwv/KnUv2rVNAvyUykLanxTw510UWGR42fxnCkKHklPznykInd+gapE7OypcP96rY07yl7383rESiDUVY4wxngKkacsbllSMMSbQBciZXd6wpGKMMQFOLKkYY4zxFUsqxhhjfEZCLKkYY4zxEaupGGOM8RlLKsYYY3zGkooxxhjfCZ6cYknFGGMCndVUjDHG+ExIiF1Rb4wxxkespmKMMcZ3gienWFIxxphAZzUVY4wxPmNJxRhjjM/YbVpKudvPrZv/TEGmwe0T/B1Csfh90r3+DsHnqlQo4+8QikVqeql79p3P+KqmIiJ1gY+BmoAC41V1lPto9elADLAN6K6qB9zHso8CugBHgbtUdW1e2wie89SMMeb/KRHxavBCGvCYqjYH2gB9RKQ58DQwT1VPB+a54wCdgdPdoRfOY9vzZEnFGGMCnK+SiqruyqxpqOoRYDNQG+gKTHJnmwRc7/7fFfhYHcuBCBGpldc2rPnLGGMCXHF01ItIDHAusAKoqaq73Em7cZrHwEk42z0W2+GW7SIXVlMxxphAJ94NItJLRFZ7DL1yXJ1IJWAm0E9VD3tOU1XF6W8pFKupGGNMgPP2Ni2qOh4Yn9c8IlIGJ6F8qqqfu8V7RKSWqu5ym7fi3fI4wPPMpDpuWe6xehWpMcYYv/FVn4p7NtcHwGZVfdNj0mzgTvf/O4FZHuU9xdEGOOTRTJYjq6kYY0yg812XyoVAD2CjiKxzy54BXgVmiMi9wN9Ad3faf3FOJ/4T55Tiu/PbgCUVY4wJcL7qqFfVJeSeojrlML8CfQqyDUsqxhgT4Ow2LcYYY3zGkooxxhifsXt/GWOM8RmrqRhjjPEZSyrGGGN8JohyiiUVY4wJdFZTMcYY4zMh1lFvjDHGV4KoohLYSUVEElW1ksf4XUArVe3rUbYO2KKqt3qUfQS0Bw4BGUAfVV2WUznQF1iiquPcZVsDE4CWqpparDuYg/T0dHr1vIXoGjV49a2x7IrbwQuDnuDwoYM0btqcQUNfpUyZwH7yX52oSrzfvxM1IiqgwMTvfmXMVxt4+e62dLkghpTUDLbuPkSvUT9yKCkla7m60ZVYOyaWYVNXMfKLdblvIECkp6fT++5YoqJrMOyN0agqH773Dgt//J7QkBCuvbE7N3S/3d9heu348eM8cE8PUlJTSE9Lo9NlV9Kr98PExe1g8FOPcejQQZo2a84Lw4ZTpkxZf4frtaTEI4wZMZTtW/8HAn2feJ7li39k9bLFhJUJo2atujz81BDCK1X2d6i5CqaaSlDfUFJEmgGhwMUiEp5t8hOq2gLnCWbv5VE+AHhCRKJFJAQYDfT2R0IB+GzaZOo3aJg1/u7ot7j5th5M+eJbKlepwjezZvojrAJJS8/g6Yk/cV6fabR/fCYPXH0mTetWY966HbTsM40LHpnOH3EHeaLbeScsN/zeC5m75m8/RV1wX8z4lHoxDbLG53wzi/g9u/lw2iwmTptFh8uu8mN0BVe2bFnGTviQKTO+5NPpX7Bs6RI2bljH6JFvEHtHTz7/ag6Vq1Rl1heB/xn09MHo1zn3/Ha8M+lz3pwwnTr1G3JOyzaMnDiDt96fwal16zFzykR/h5knEe+GQBDUSQWIBT4B5uI8oSwni4BGuZWr6h5gBPAa8CCwwb0/TomL37Ob5UsWcU3XmwBQVX5etYL2l14BwJVXd2XJwh/9EVqB7D5wlHX/2wdAYnIqW7Yf4NTq4cz7eTvpGc5jGlb+tofaUVmVUK5t04Btew7z6z8H/BJzQe2N382KnxbR5bobs8q++nwGPe55MOs25dUiq/srvEIRESpWdH6bpaWlkZaWioiwetVyLr3sSgCuvrYrC+fP82eYBZKUeIRfN6zlsi7OgwzLlClDeKXKtDi/LaGhTkNN42ZnkbA3Pq/V+J0PHydc7AI9qVQQkXWZAzA02/RbgGnAVJwEk5NrgY35lL8LNAeeAJ4sctSFNPrN4Tz4yICsq2cPHTpIpcqVCQtzPvw1atRkX3xgf/izq1ejMi1Oi2LVb3tOKO95eTPmrPkHgPDyYTx207kMm7rKHyEWytiRr3F/3wGIx3MudsZtZ8G87+h9960M7P8QO7YHT60rU3p6Ord3v4ErL72IC9q0o06delSuXCXrM1iz5insjd+Tz1oCR/zunVSpWo3Rrw3hsV6xjBkxlGPJySfM8+O3szjvgnZ+itA7VlPxnWRVbZE5AM9lThCRVsA+Vf0HmAecKyKRHsu+7iaiXsC9eZWragZOU9i3qppQvLuUs6WLFxBRLZImzc7wx+aLRXj5MKYOvJInJvzEkeR/WxOf7N6S9PQMpi34HYDBt13AO7PWk3QszV+hFsjyJQuJqBZJ46bNTyhPTU2hbNlyjP1wGl263sSIYc/lsobAFRoayqczvuDrOfP5ddNGtm37y98hFUl6ejp//bGFK6/rxhvjp1K+fAU+n/ph1vTPJr9PSGgYl1zWxY9R5i8kJMSrIRAEdEd9PmKBpiKyzR2vAtyE08kOTt/JZzksl1t5hjvkyH0sZy+A10aOpcfd9xU27hxtWv8zSxcvYMXSxaQcP05SUhLvjHiVxCNHSEtLIywsjPj4PUTVqOHT7RaXsNAQpg68iukL/mDWsn8PTHd0akKX8+vTefDsrLLzG9fghnYNGXZXW6qGlyNDlWMpabz7zSZ/hJ6vTRvWsWzxAlYuXUJKynGOJiXxypCBREfX5KIOzt3DL2rfiddfCr6kkqlylSq0PP8CNq5fx5Ejh7M+g3v27Ca6Rs38VxAgqkfXoHp0DRo3OwuAtpd04vOpHwHw43ezWb18MS+MeDdgmo5yE+DhnSAok4rbod4dOEtVd7plHYFn+Tep+JTnYzp3H04t9PObc9Orb3969e0PwM9rVjJ98kc8+9Jwnnt6AAt/nEunK7ow55tZXHjJpb7edLF495GO/Lb9AG/PWp9Vdvl5dRlw47lcMfBLko//Wyu57Okvs/4fFHs+ScdSAzahANzX+1Hu6/0oAOvWruI/n05i4JBXmDB2JOvWrKLWqXVY//Nq6tSr7+dIC+bA/v2EhYVRuUoVjh07xorly+h59720bNWaH3+YwxVXXc03X82ifYfg+AwCVIuMIqpGTeL+2UbtejFsWLuSuvUbsHblT3w5fRIvvvU+5cpX8HeY+Qr0pOcpKJMKcDEQl5lQXIuA5u7zlUuNB/v254VBT/DBuHdo1KQZV3e9Mf+F/Kxd81O4/dImbNyawPJRzgPknv94OW/0uphyZUL5+sXrAKez/pGxC/0Zqk/F9riHl4cMZOa0T6hQsSKPDRzi75AKZN++vbzw7EAyMtLJyMjgsiuu4uJLOtKwYSMGPfUY7455m8ZNmnHdDd38HWqB3PfwU4x8eRBpaanUrFWHvk8O4cmH7iA1NZUXnngIgMbNz+LB/oP8HGnugiinIM6DvUxBFEdNxd8a3F4sFTy/+33SvfnPFGSqVAjs65QKa8f+5PxnCkJn1A4vckpo+eJ8r445a57t6Pf0E6w1FWOM+X8jmGoqllSMMSbABdMV9ZZUjDEmwFlHvTHGGJ8JopxiScUYYwKd1VSMMcb4TBDlFEsqxhgT6Kyj3hhjjM9Y85cxxhifsaRijDHGZ4Iop1hSMcaYQGc1FWOMMT4TRDnFkooxxgQ6O/vLGGOMz4QEUVXFkooxxgS4IMopllSMMSbQWUe9McYYnwmiLpXck4qIvAPk+rQxVX2kWCIKAonH0vKfKcgseecOf4dQLPp9GbjPui+skdef6e8QisXvCUf8HUKxOKN2eJHXUVo66leXWBTGGGNyJZSCpKKqkzzHRaSiqh4t/pCMMcZ4CqKKCiH5zSAibUXkV2CLO36OiIwt9siMMcYATke9N0MgyDepACOBK4EEAFVdD1xSnEEZY4z5l4h3g3frkokiEi8imzzKhohInIisc4cuHtMGisifIvKbiFyZ3/q9OvtLVbdny4Lp3oVvjDGmqHx88eNHwGjg42zlb6nqCM8CEWkO3AqcAZwK/CAijVU11xzgTU1lu4i0A1REyojI48DmAuyAMcaYIggJEa8Gb6jqImC/l5vuCkxT1eOquhX4E7ggz1i9WOmDQB+gNrATaOGOG2OMKQG+bP7KQ18R2eA2j1Vzy2oD2z3m2eGW5SrfpKKq+1T1dlWtqarRqnqHqiYUPm5jjDEFESLi1SAivURktcfQy8tNjANOw6k07ALeKGys+fapiEhDYBTQBudiyGVAf1X9q7AbNcYY4z1vKyGqOh4YX9D1q+qerG2JTAC+dkfjgLoes9Zxy3LlTfPXFGAGUAuno+Y/wNQCxGuMMaYIivuUYhGp5TF6A5B5Zths4FYRKSciDYDTgZV5rcubs78qquonHuOTReSJggRsjDGm8Hx58aOITAU6AFEisgN4HuggIi1wWqO2AQ8AqOovIjID+BVIA/rkdeYX5H3vr0j3329F5GlgmrvBW4D/FmGfjDHGFIAv7/2lqrE5FH+Qx/zDgGHerj+vmsoanCSSuTcPeG4HGOjtRowxxhReoFwt74287v3VoCQDMcYYk7NguveXV1fUi8iZQHOgfGaZqma/GtMYY0wxKBU1lUwi8jxOp05znL6UzsASTr7E3xhjTDEInpTi3SnF3YBOwG5VvRs4B6harFEZY4zJEhoiXg2BwJvmr2RVzRCRNBGpAsRz4sUwAUNEBgG34dzwMgN4QFVXiEgYzlWiH6jq0x7zL8C5/uY4UBb4ARisqgdLOnaAu2/uTIWK4YSEhBAaGsao96dkTft82sd8MOZNpnw1n6oR1fJYS2DZuX0b77z8TNZ4/O6ddOvRi/0Je1m7fDFhZcpQs1YdHnjsOcIrVfZjpHnrfVF9WtWtyqFjafT/4lcAup9bi8saR3HYfRLolDVxrN1xmLNPrcwdrWoTFhJCWkYGH6+KY9Ou4HmqYXp6Or3vjiUqugbD3hiNqvLhe++w8MfvCQ0J4dobu3ND99v9HabXlnzzGavmfY2qcsFl13DR1TcD8NO3M1n+3ZdISAhNz2tDlx4P+TnS3JWq5i9gtYhEABNwzghLxLmqPqCISFvgGuA8VT0uIlE4iQLgcuB34GYRGaiqno9Jvl1VV4tIWeAVYBbQviRj9/TKqAknJY29e3bz88plRNeslctSgevUujG8Ms5Jjhnp6fS5vQutLuzIrh1/c+s9fQgNDWPq++8we9pHxN73sJ+jzd2CPxL4dnM8j1xy4vkrX/8Sz+xNe04oO3IsjVe+/x8HklOpG1GeZ688nV7TN5ZkuEXyxYxPqRfTgKNJSQDM+WYW8Xt28+G0WYSEhHBgf/DcpWn3P3+xat7X9HnlXULDwvhw2JM0Pa8thxLi2bzqJx4d8QFhZcqSeOiAv0PNUxDlFK/u/dVbVQ+q6rs4B+c73WawQFML2KeqxyHrnmU73WmxOLea+Qdom9PCqpoCPAnUE5FzSiBer014ZwR39+4XVB+snGxat4qateoQXbMWZ7dsQ2io85umUbMzSdi3J5+l/evXPYkkHvfuiQ9b9ydzIDkVgO0Hj1E2LISwAGmayM/e+N2s+GkRXa67Mavsq89n0OOeBwkJcQ4X1SKr+yu8AouP+5u6jZpRtlx5QkPDaND8HH5ZuYjlc2fR/vrbCCvj/O6sVDWwa//e3vsrEOR18eN5eU1T1bXFE1KhzQWeE5HfcZqxpqvqQhEpD1yGc51NBE6CWZrTClQ1XUTWA02B9SUT9r9EhGcHPAQidO56E52v68ayxfOpHh1Nw0ZNSjocn1u2YC5tO5z8jJ8Fc2bTtv3lfoio6Do3i6ZDo0j+3HeUSSt3kJRyYuJpExPB1oSjpGVoLmsILGNHvsb9fQdw9GhSVtnOuO0smPcdPy38kaoR1egz4Gnq1K3vxyi9d0rdBsyd+j5JRw5Rpmw5flu7nNqnNWHfzh1s27yBuVPfJ6xMWbr0fIi6jZr5O9xcBUi+8EpezV953aVSgUt9HEuRqGqiiLQELgY6AtPdOwEkAvNVNVlEZgLPiki/PG41kOPb597tsxfAi6+/w6097/X5Prw25kOiomty8MB+Bvd/kLr1GjDjkw946c1xPt9WSUtLTWXN8kXces+JT034cspEQkPDuPDSzn6KrPDmbN7LZ+t2oQqxLU/lzgvqMHbJ31nT60aUp0erOgyd87sfo/Te8iULiagWSeOmzVm3dlVWeWpqCmXLlmPsh9NYvOAHRgx7jpHvTvJjpN6rUSeG9l1vY+KLj1OmfHlqxTQiJCSUjIx0jiYepvfL49jx5xamvDmEJ8dMC9i+i0CNKyd5XfzYsSQD8QU3USwAFojIRuBOIAW4SES2ubNVx0mI32dfXkRCgbPI4SFknnf//DM+uVh+dkZF1wQgolokbS/pyMZ1a9izK46+d3cHYN/eeKf/h44AACAASURBVB69N5Y3x08msnpUcYRQbNatWkqDRk2pWu3fppOFc79i7colDHp1bFB9aTIdcjvoAb7/bR/PXN4oazyyYhme7HQaby/ayp4jKf4Ir8A2bVjHssULWLl0CSkpxzmalMQrQwYSHV2Tizp0AuCi9p14/aXn/BxpwZzf6WrO73Q1AN9NGU/V6tHsjfuHM1tfgohQ9/RmSEgISYcPUalqhJ+jzVloEH0/vLr4MRiISBMgQ1X/cItaAHtxOu/rZva1iMjdOE1g32dbvgzO/W22q+qGEgvcdSw5mQzNoGLFcI4lJ7N21TJi73qAKV/Nz5rn7ps7M3LClKA6+yvT0gVzaNvhiqzx9auW8vV/PuHZ19+jXPnyeSwZuCIqhHEw2UksretH8M+BZAAqlg1l0BWNmLw6jt/ik/JaRUC5r/ej3Nf7UQDWrV3Ffz6dxMAhrzBh7EjWrVlFrVPrsP7n1dSpFxxNX5kSDx2gUtVqHNy7h19WLKb3y2MRCeF/m37mtDPPY+/O7aSnpRJeJXCvlAiSLjmgFCUVoBLwjnumWhrOYy9n4dxl+bjHfLOA10SknDv+qYgcB8rh9MV0LcGYsxw4kMCwZwYAkJ6eRvvLO9Oq9YX+CMXnjh1LZtPaldz36L+nFn805nVSU1N4ZaDTHNao6Vnc+2jg3k6uf4cGnHFKZSqXD2P8LWcxfe1OzqhVmZjIioASn5jCuz85TV+dm0VzSuVy3NyiFje3cM7YGzrnj6xTj4NNbI97eHnIQGZO+4QKFSvy2MAh/g6pQCaPeJajRw4TEhZG1/v6USG8Mq06duGzccN5a8BdhIaFcXOfZwK6thxMSUVOPLvWeKO4mr/86dDRVH+HUCxe/vGP/GcKMiOvP9PfIRSL1TsC+7Tewrrh7FOKnBIe++o3r445b1zbxO/pJ99TisVxh4g8547XE5E8H3xvjDHGd0LEuyEQeHOblrE413Zk3oP/CDCm2CIyxhhzAhHvhkDgTZ9Ka1U9T0R+BlDVA+7V58YYY0pAWKBkDC94k1RS3VNtFUBEonHuq2WMMaYEBFFO8SqpvA18AdQQkWE4dy0eXKxRGWOMyRIot2DxRr5JRVU/FZE1OLe/F+B6VT3p4kBjjDHFI4hyilcP6aoHHAW+8ixT1X+KMzBjjDGOQDmzyxveNH99g9OfIjiPE24A/AacUYxxGWOMcQXKA7i84U3z11me4+7di3sXW0TGGGNOEEQ5peC3aVHVtSLSujiCMcYYczIJoqfUe9OnMsBjNAQ4D9iZy+zGGGN8rLTVVDwfHJ6G08cys3jCMcYYk12pSSruRY+VVfXxEorHGGNMNoF8B+Xs8nqccJiqpolI6bj/ujHGBKlQb+7SGCDyqqmsxOk/WScis4H/AFlPHFLVz4s5NmOMMZSyK+pxrk1JwHkEb+b1KgpYUjHGmBJQWvpUarhnfm3i32SSqdQ9pKog6kRW8HcIPlcmmOrXBTC229n+DsHnYtr393cIxWLTnNf9HULACqKKSp5JJRTnEb057c7/66RijDElKaSUXKeyS1WHllgkxhhjclRaaipBtBvGGFN6hQVRp0peSaVTiUVhjDEmV6WipqKq+0syEGOMMTkrbacUG2OM8aMgyimWVIwxJtAF0wn/llSMMSbAWfOXMcYYnwmmpBJMtSpjjPl/SbwcvFqXyEQRiReRTR5lkSLyvYj84f6t5paLiLwtIn+KyAb3yb95sqRijDEBTsS7wUsfAVdlK3samKeqpwPz3HGAzsDp7tALGJffyi2pGGNMgBMRrwZvqOoiIPslI12BSe7/k4DrPco/VsdyIEJEauW1fksqxhgT4EK8HESkl4is9hh6ebmJmqq6y/1/N1DT/b82sN1jvh1uWa6so94YYwKctx31qjoeGF+UbamqikihbxpsNRVjjAlwvmz+ysWezGYt92+8Wx4H1PWYr45blitLKsYYE+C8bf4qgtnAne7/dwKzPMp7umeBtQEOeTST5ciav4wxJsAVsRaSfV1TgQ5AlIjsAJ4HXgVmiMi9wN9Ad3f2/wJdgD+Bo8Dd+a2/WJOKiAwCbgPSgQzgAWA60EpV97nzdAAeV9VrROQu4HWc6lV54D1VfcudbwhwP7DXjfsZVZ2drbws8KKqTnWX+Qj4WlU/E5FrgBdxEnoZYJSqvpdt+UwdVPVgMbwkudq9axeDBj7J/oQEEKHbzd25vcedvDliOAsXzKdMmTLUqVuPoS+9QpUqVUoytCJLT0+n7z2xREXX4MURo/l59XImjH6TDFUqVKjI44NfpHadev4O02vHjx/n4V53kpqaQnpaOh06Xc49D/RlzaoVjB01grTUVBo3a85Tg4cSFhbYv9vq1Izg/Rd7UqN6ZVRh4syfGDN1Qdb0R3tcyqsDbqROx6dIOJhElUrlmfjSndStVY2w0FBGfjyPT2Yv998OeOmumztToWI4oSEhhISG8fb7U/j4/TEsX7yAkBCharVIBjwzlOpRNfwdao58eemjqsbmMumkO9OrqgJ9CrL+YvvEi0hb4BrgPFU9LiJROAf9/ExX1b4iUh34TUQ+U9XMsw/eUtURItIMWCwiNbKVnw6scZdJ9YilDE7n1QWqukNEygExHtt8S1VHFG2PiyY0LJTHn3yaZs3PICkpkVtvvok2bS+kTdsLeaTfY4SFhfHWG6/zwYT36P/YE/4MtcC+mPEp9WIacDQpCYC3Xx/GC8NHUS+mIbNnTmPKR+N5YvBLfo7Se2XLlmXkuIlUrFiRtLRU+tzXk/PbXMjLQ55h5NgPqFs/hg/eHc1338zimq43+TvcPKWlZ/D0m5+zbssOKlUsx9IpTzFvxRa2/LWbOjUj6NSmGf/s+vfs0we6X8KWv3bTrd97RFWrxPovnmXaf1eRmpbux73wzqujJlA1olrWeLfYO+l5n3O8nPXZFKZ8NJ6HHx/sr/DyFGpX1ANQC9inqscBVHWfqu70dmFVTcCpcp10TrSqbgbSgKhs5X/gVNGqZVukMk4CTXDnO66qv3m/K8UvOroGzZqfAUB4eCUaNmxIfPwe2l14Udav3bPPaUH8nt3+DLPA9sbvZuXSRVx17Y1ZZSKQlJQIOH+rR0X7K7xCEREqVqwIQFpaGmlpaYSGhlKmTBnq1o8BoFXrtiz88Qc/Rumd3fsOs27LDgASjx5ny9bdnBodAcBrj9/EoFFf4vxYdShQKbwcAOEVynHg0FHS0jNKPG5fqBheKev/Y8nJSAA/l9DHFz8Wq+Ksm88FnhOR34EfcGogC71dWETq4TSBbchhWmuc5rS92crPA/5Q1XjPclXdLyKzgb9FZB7wNTBVVTO/Df1F5A73/wOq2tHbOItDXNwOtmzezFlnn3NC+Zefz+TKzp39FFXhjBv5Gvf1GUDy0aSssv5PD2HwY30oV64cFcMrMWrCZD9GWDjp6enc36M7cTv+4fqbY2l2xlmkp6ez5ddNNG1+JgvmzQ26HwD1akXSokkdVm3axjUdzmJn/EE2/n7iiT7vTlvIZyMf4K+5w6gcXp4eT008IekEKhFh8ICHEBE6d72Jztd1A2DS+HeYN+drwsMr8eqoCX6OMneBnPCyK7aaiqomAi1xLu3fC0x3+0xy+gR6lt0iIhtwailjVfWYx7T+IrIOGAHcov9+mvuLyC/ACmBYLvHch9NmuBJ4HJjoMfktVW3hDn5NKEeTknis3yM88fQzVKr07y+pCe+NIzQslKuvuc6P0RXM8p8WElEtksZNm59Q/vn0ybz0xhimzPqBK67uyntvv+6nCAsvNDSUiVNm8tk389jyy0a2/u9Pnh/2OqPfeo1ed95KxXCn/T5YhFcoy9QR9/HEiJmkpafz5D1XMnTcNyfNd3m7Zmz4bQcNrxhE61tf4a2nb6ZyeHk/RFwwr4/5kHcmTmPoiDF8/fkMNq5bA8CdvR7m45lz6HB5F776fJqfo8xdMNVUivVTr6rpqrpAVZ8H+gI34TRBeTZPRQL7PManq+rZQDvgVRE5xWNa5sH/YlVdnK38DHf9H4hIjp9yVd3odvxf7s7rNc8rVT+YUKRri3KVmprKgH6P0OXqa7ns8iuyymd98TmLFi7gleEjfHoWSHH7ZcM6li9ZQI8br+Ll555k3ZqVDH6sD3/98RvNzjgbgA6druLXjev9HGnhVa5chXNbXsCKZUs48+wWjJ7wMeMnTeOcc1tmNYUFurCwEKaOuJ/p365m1o/raVgnmvq1q7Ny+kC2fPMCtWtEsGzKU9SsXpke17Vh1o/O+/XX9n1si0ugSUzNfLbgf1HRTowR1SJpe0lHft+86YTpHa/owk8L5/kjNK+EIF4NgaDYkoqINHE7zjO1wDlVbQHQw50nFLgDmJ99eVVdDXwCPOrtNlV1NrCaf8+3zoylknuWWfZYvKaq41W1laq2uvd+b+98UKD1M+S5QTRs2JCed/171t5Pixfx0cT3GTV6HBUqVPD5dovTvQ89ypRZP/DJ59/xzNDXaNHyAl4YPoqkpER2/LMNgDWrllEvpoF/Ay2ggwf2c+TIYQCOHzvG6pXLqB/TgAP7EwBISUlhyqSJXHdj97xWEzDeff52ftu6m7cn/wjAL3/upH6ngTS9+nmaXv08cfEHaXvbcPYkHGH77gN0uKAJADUiK9M4piZb4/bltXq/O5aczFG3+fVYcjI/r1pG/YaNiNv+7yFg+eIF1KkXuJ/DYKqpFGefSiXgHRGJwOlU/xOnKSwVGCci63HOlPsOyK1RfTiwVkReLsB2hwJTRMSzgVSAJ0XkPSAZSALu8pju2acCcL2qbivANovs57Vr+Hr2LE5v3JjuN3YF4OF+Axj+8kukpKbw4H1OojnrnHN49vmhJRmaT4WGhdHv6ecZ+swAQkJCqFS5Co89E1z7k7BvLy8PGUR6RjqaoXS87EraXdyBsaNGsHTJQjRD6XrTLbQ8v7W/Q81XuxYNuf2a1mz8PY7l05wb0z4/ejZzlvya4/yvTviO8S/cwaoZzyACg0bNIuFgUo7zBooDBxJ46ZkBAKSnp9Hh8s60an0hLw1+jLh/tiESQo1TatH38UF+jjR3wfQ8FQmGTrZAcywtx36hoLbn0HF/h1AsypcJnn4Nb8W07+/vEIrFpjnB17fmjdNqVChyRpi3ZZ9Xx5xOTaP8nn0C+8osY4wxQXX2lyUVY4wJcEHU+mVJxRhjAp3VVIwxxvhMSPDkFEsqxhgT6ILp7C9LKsYYE+CCJ6VYUjHGmIBnNRVjjDE+EzwpxZKKMcYEviDKKpZUjDEmwFnzlzHGGJ8JnpRiScUYYwJfEGUVSyrGGBPg7Ip6Y4wxPhNEXSqWVIwxJtAFUU6xpGKMMYEumB4jbknFGGMCXBDlFEsqhXEsNd3fIfhcanqGv0MoFjv2H/V3CD63+YcR/g6hWMS+v9LfIRSLZU9dUuR1BFFOsaRijDEBL4iyiiUVY4wJcHZKsTHGGJ+xPhVjjDE+Y0nFGGOMz1jzlzHGGJ+xmooxxhifCaKcYknFGGMCXhBlFUsqxhgT4OwhXcYYY3wmeFKKJRVjjAl8QZRVLKkYY0yAs1OKjTHG+EwQdalYUjHGmEDny5wiItuAI0A6kKaqrUQkEpgOxADbgO6qeqAw6w/xTZjGGGOKi4h4NRRAR1Vtoaqt3PGngXmqejowzx0vFEsqxhgT4ES8G4qgKzDJ/X8ScH1hV2RJxRhjApx4O4j0EpHVHkOvHFanwFwRWeMxvaaq7nL/3w3ULGysJdKnIiIKvKmqj7njjwOVVHWIiAwB7gf2eizSASdTtlLVvh7rWQA8DowBygGRQAUgzp3lemABTnuhAgeAnqr6t8c6vgROUdU2HmVDgERV9dsj9Y4fP86D9/QkJTWF9LQ0Lr3sCnr1fpj/TPuUaZ9+zI7t25kz/yciqlXzV4iFdvfNnalQMZyQkBBCQ8MY9f6UrGmfT/uYD8a8yZSv5lM1Inj2be6XU1k0dzaCUDvmNO7tN5iD+xN497XBJB05TP1GTbh/wBDCypTxd6gFlp6ezsP3xFI9ugYvjhjNutUrmDD6TVJTUzm9aXMGDBxCaFhgd8cO6tyYdqdFcuBoKndMXAPAi9c1pV5kRQAqlw/jyLE07vxoLefHRNC7fQPKhIaQmp7B6PlbWfPPQX+GfzIvayGqOh4Yn89sF6lqnIjUAL4XkS3Z1qHuMbtQSuqTcRy4UUReUdV9OUx/K/sBPa/2QVVt7c5zFycnHnDaC/eJyAvAYJykhYhEAC2BRBFpqKp/FWmvfKhs2bKMmTCRihXDSUtNpdfdd9D2oks4u8W5XHhxB3rfd6e/QyySV0ZNOClp7N2zm59XLiO6Zi0/RVU4B/bF88NXM3hp7FTKlivP2FcHsWLR92xYvYwrusbSuv3lfDx6OIu/n03HLjf5O9wC+3LGp9SNacjRpEQyMjJ4/aVnGf72eOrUi2HShDF8/+1srrr2Rn+HmadvNu7hP2t38tzVTbLKnp3977Hz4Y4NSTqeBsCho6k8MfMX9iWm0DCqIiO7n8V1Y1eUeMx58eUpxaoa5/6NF5EvgAuAPSJSS1V3iUgtIL6w6y+p5q80nOzZv4S2l2kZUNtj/EbgK2AacGsJx5InEaFixXAA0tLSSEtLQwSaNG3OqbVr57N0cJrwzgju7t0vqE6XzJSenk5KynHS09NIOX6MiGpRbNmwmlYXdQSgXacurF22yM9RFtze+D2sXLqYztfeAMDhQwcpE1aGOvViADjv/LYsWTDPjxF6Z92OQxxOTs11eqem0czd7Bw3f49PYl9iCgB/7TtKubAQyoQG1ofSV30qIhIuIpUz/weuADYBs4HMX653ArMKG2tJ1mHHABtE5LUcpvUXkTvc/w+oakcfbfMq4EuP8VhgKLAHmAm87KPt+ER6ejp3xnZjx/Z/6HbLbZx51jn+DsknRIRnBzwEInTuehOdr+vGssXzqR4dTcNGTfJfQYCpFlWDq264nSfuvp4yZctx5rkXUL9RUyqGVyY01PlKRUbV4GDC3nzWFHjeHfka9/Xpz9GjSQBUjahGeno6v2/+hcbNzmDJ/O/Zu2e3n6MsmhZ1qrI/KYUdB46dNK1jkyh+25NIanqhW3+KRYjvclxN4Au3RScMmKKq34nIKmCGiNwL/A10L+wGSiypqOphEfkYeARIzjb5pOYvnD6RHFflxebmu+ddJwLPAohITeB0YInbZpgqImeq6iZv4nc7tHoBvPXOOO66935vFiuQ0NBQJs/4giOHD/PkgEf4359/cFqj032+nZL22pgPiYquycED+xnc/0Hq1mvAjE8+4KU3x/k7tEJJSjzMzysWMfyDz6kYXplxrz7DprXL/B1WkS3/aSER1SI5vWlz1q9dBTg/CAYOHc67b79OakoKLS9oR0hoqJ8jLZrLm0fz/eaTW3caRFWkd/sG9Jux0Q9R5cc3WcVt8j/p16qqJgCdfLGNku5tGwmsBT70Yt4EIHvPbSSQU59Mdh2Bg8CnwAvAAJzMWw3Y6mbpKjg1l0HeBO7ZAXYwuXh/xlSuUoWW51/Asp8Wl4qkEhXtnEgSUS2Stpd0ZOO6NezZFUffu50fQ/v2xvPovbG8OX4ykdWj/BmqV35dt4qomqdSparz8TyvbQf++HUDR5OOkJ6eRmhoGPv3xRNRPdrPkRbMrxvWsXzJAlYtW0JKynGOJiUxfMhAnhryCm+O+wiANSuWsmP733mvKICFCnRoHMVdk9aeUB5duSyv3tCcF7/5jbiDJ9dg/C2YmohL9JRiVd0PzADu9WL2VcCFInIKgIi0wjnja7uX20oD+gE93VpLLHCVqsaoagxOh33A9Ksc2L+fI4cPA3Ds2DFWLl9KTIOGfo6q6I4lJ2c1pRxLTmbtqmWc3uwMpnw1nw//8y0f/udboqJrMOqDqUGRUAAio2vy12+bOH7sGKrK5vWrObVeA5qe1ZLVS+YDsHTefzm3zcV+jrRg7nnoUT6d9T0ff/4tA4cO55yW5/PUkFc4uD8BgJSUFGZM/pBrru/m50gL7/yYavydcJS9R1KyyiqVC+WNbmcyduFWNsQd9mN0ufP2lOJA4I/zAt8A+mYr8+xTAbheVbeJyKPAf0UkBKcpK1ZVM7zdkHsmw1SgD1AfWO4xbauIHBKR1m7RYBHp5zG9TsF2q2j27dvL0GcHkpGRQUZGBp2uuIqLLunA9Cmf8MlHE9mfsI/bu19Pu4suYdDzL5ZkaEVy4EACw54ZAEB6ehrtL+9Mq9YX+jmqojmtyZm0uvBSXuh3J6EhodQ7rTHtr7qes8+/kPeGP8sXk9+jXsPGXHzFdf4O1Sf+M2USK35ahGoGV9/QnRatWue/kJ+9cG1TzqtXlYgKZZjVuzXvL/mbrzbs5rJm0Xy/+cS+rm7n1aZORAXuaVefe9rVB6DfjI0cOJp7R39JC6aaiqgGVodUMCju5i9/2Ofxy6002XMo8Joyiqp2ZAV/h1AsYt9f6e8QisWypy4pckrYfTjVq2POKVXK+D39BPYVTMYYYwKmacsbllSMMSbABVPzlyUVY4wJcPaQLmOMMb4TPDnFkooxxgS6IMopllSMMSbQhQRRp4olFWOMCXBBlFPsIV3GGGN8x2oqxhgT4IKppmJJxRhjApydUmyMMcZnrKZijDHGZyypGGOM8Rlr/jLGGOMzVlMxxhjjM0GUUyypGGNMwAuirGJJxRhjAlww3abFnvwY4ESkl6qO93ccvlQa9wlK536Vxn2C0rtfgcBu0xL4evk7gGJQGvcJSud+lcZ9gtK7X35nScUYY4zPWFIxxhjjM5ZUAl9pbPctjfsEpXO/SuM+QendL7+zjnpjjDE+YzUVY4wxPmNJxRhjjM9YUikBInK9iKiINHXHY9zxhz3mGS0id3mMDxCRLSKyUUTWi8ibIlLGnbbNLd8gIgtFpL6I3C8i0z2WryIi/xORhsW0T4nZxu8SkdHZytaJyLRsZR+JyFZ32loRaZtbuYh8KiIPeSzb2t3nMsWxT3kRkUEi8ou7/XUi0totDxORvSLyarb5F4jIb+78W9z3N6Ik43M/J1Ee83QQka/d/+9y417nxtffY74hIhLnTtskItflUP6riMR6LPORiHRz/79GRH52P7e/isgDOSyfOeT5mrjfkzc8xh8XkSF5rS+Xz+ICEWklIivc+f7x2P917nfypO9VtnV8KSLLs5UNEZHHvXyb/l+wpFIyYoEl7t9M8cCjIlI2+8wi8iBwBdBGVc8Cznfnr+AxW0dVPRtYAAwG3gfqishl7vShwERV/cvH++IVEWkGhAIXi0h4tslPqGoL4GngvTzKBwBPiEi0iIQAo4Heqppa/HvwLzfxXQOc577mlwHb3cmXA78DN4ucdNnz7e78ZwPHgVl+iC8v093X+0JgkIjU9Zj2ljvtZmCi+/p7lncF3sue4N3x8cC1qnoOcC7OZ/SE9XoMB/OJ8Thwo2dyzKZA61PV1m78z2Xuvztsc2fJ/r3K3K8IoCVQtbh+qJUWllSKmYhUAi4C7gVu9Zi0F5gH3JnDYoOAhzK/IKqaoqqvqurhHOZdBtRW54yLB4GRItIK6AS87rs9KbBY4BNgLs4BKCeLgEa5lavqHmAE8BrOvm1Q1SXFEGt+agH7VPU4gKruU9Wd7rRYYBTwD9A2p4VVNQV4EqgnIueUcHz5UtUE4E93PdmnbQbSgKhs5X8AR4Fq2RapjHP7pwR3vuOq+pv3u3KSNJwk1T+/GX1sGVDbY/xG4CtgGid+j002llSKX1fgO1X9HUgQkZYe04YDj4tIaGaBiFQBKqnqVi/XfxXwJYCqbgDm4CSrh92DWXGp4NnsgFMz8nQLzhdwKifW0DxdC2zMp/xdoDnwBM6B2R/m4tQCfxeRsSLSHkBEyuPUCr4i7/1EVdOB9UDTkorPWyJSDygPbMhhWmsgA+dHkGf5ecAfqhrvWa6q+4HZwN8iMlVEbveo5QD09/jczPcyxDHA7SJSNYdphVmfN7K+V65YnPc4z/fZWFIpCbE4B1fcv1kfSLdpagVwW24Li8iV7hdmm4i085g0X0TigM44H/RMY4A4VV3go/hzk+zZ7IDTnJAZcyucX87/4CS4c0Uk0mPZ191E1AunBpdruapm4DSFfev+oi5xqpqI0/TRC+fgOl2c/q9rgPmqmgzMBK73/IGQg2K5K2Ae8eV0vYBn2S0isgGnljJWVY95TOvvvhcjgFv032sP+ovILzif22G5xHMfTk15JfA4MNFjsmdzVUcv9+8w8DHwSA6Tc1pfbtdJeHP9xEnfKxGpCZwOLHF/HKaKyJnexP7/kSWVYuQeSC8F3heRbTi/trtz4sHlZeCpzDL3C5QoIg3c8TnuQXsT4Nn/0hGoD6wDXvAoz3AHf4oFmrr7/D+gCnCTx/Qn3IPA5aq6yYtyv++Tqqar6gJVfR7oi7M/scBl7n6uAarjvN8ncZPNWcDmEowvgRObpyKBfR7j093+g3bAqyJyise0zIP1xaq6OFv5Ge76P3BraznFs1FV38Lpc7opp3kKaCTOD43s/XM5yb7fcPK+5yan71V3d31b3fc6Bqut5MqSSvHqBnyiqvVVNUZV6wJbgawOUVXdAvyK0+ST6RVgXOaZMW4H8ElfXlVNA/oBPbPVBPzGberoDpzl7nMMThNg0H4JRaSJiJzuUdQCp0ZwMVDPYz/7kMN+up3XrwDb3SbKkojvb5zO5h7uPKHAHcBJTUSquhqn/+tRb7epqrOB1WTrExSRSiLSIYdYisRtVpvBiTXb3KwCLsxMkm7NuRzenbyQ0/cqFrjK431uifWr5Mqep1K8YnH6TTzNBAZmKxsG/OwxPg7nF9kKETkOJAI/ZZsHgP9r725C6yijMI7/n7YQWsWCUrtRFBQ/SnEhKlVpKFLE6sYuXIguRCUo+NWNK1GsIH7iQhCl0YXUdlFaqaIkoRVJKhSjhYKNSBcFF+1CtBZTC37wujhnmunk1/aBzQAAA7tJREFUJrnaieGG57fKnZn7vnMTcs+dmTvPKaWckLSTeEN7qaX9Ph/ridNv9QvFo8AaSdMuBPeIC4G3s8j/RZwu2gusqC6Op73Aa5L68vFH+ffrA/Yx8xcW5mP/BoA/iQ8nh4kj4SFg+wxjvAockvTyv5h3K7BD0rbaMgHPSnoPOAOcBh6qrd8i6cHa43tr37yay5vEUVhdx/EkPQ18nh9yJoH781RqVxr/V1cAB2vrjkk6ldebAJ6T9Ext/WXdzrMYOabFzMxa49NfZmbWGhcVMzNrjYuKmZm1xkXFzMxa46JiZmatcVGxniXpb00l6e6StOI8xqqn7A5KWjPLthsa6QbdznFOavBcyxvbTM62vsP2Ts+1BeGiYr2siopZC/xBhE6eJek/3YdVSnm0lDIxyyYbiLvQzazBRcUWizHg6jyKGJP0CTAhaamk1yWNK/pkVL09pOhx8oOkfcCl1UDK3hv5812K/i6HJe2XdCVRvKogw/WKaP7dOce4pNvzuZdIGlH0ORmki+wvRc+Ob/M5A411b+Xy/ZJW5bKrJA3lc8aUPXvMForvqLeel0ckm4g7xgFuBNbmnc8DwKlSys15p/tXkkaIPh/XEgnIq4monA8a464CtgH9OdbFpZRfJL0LTJZS3sjtdhCZWAcUib/DwPXAC0QI4VZJ99BdxMjDOcdyYFzS7gzSvAD4ppSyRdLzOfYTRCz8Y6WUo3mH9zvMkD9m9n9wUbFetjyTdCGOVN4nTkt9XWsdcCdwQ3W9BFhJJM72Azszkv64pC86jL8OGK3GyvypTjYSMTTV44sUfXT6iT4clFI+k3Syi9f0lKTN+fPlua8/E4GaVWfP7cCenOM2YFdt7j7MFpCLivWyM5ngfFa+uZ6uLyJ6yww3tru7xf1YQnTprEfHo2mNIGeXQYwbgVtLKb9L+pIOQaKp5Ly/Nn8HZgvJ11RssRsGHs+kYCRdo2hvPEr0E1maQZedenscBPqVbQg0lQT9G9HhsDICPFk9kFS9yY+SvXIkbWJ6HHvTSuBkFpTriCOlyhIi9Zoc80C2STgm6b6cQ5qfzpJmXXNRscVukLheckjSd0TDr2XAx8DRXPch0T72HKWUn4i03z2Z9FudfvoU2FxdqCeaR92UXwSYYOpbaC8SRekIcRrsxzn2dQhYJul74BVqybjE0dct+RruYKrT5gPAI7l/R5i/JGSzrjil2MzMWuMjFTMza42LipmZtcZFxczMWuOiYmZmrXFRMTOz1riomJlZa1xUzMysNf8A/rDdOWS4SJ0AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XqLIJTjW0JE_"
      },
      "source": [
        "##Neural Network Model Evaluation on distance inputs\n",
        "\n",
        "We saw yesterday that building a model on the distances between facial landmarks did better than on raw pixel inputs, is this true for these models as well? "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "k-aKmK6e0a4J"
      },
      "source": [
        "#Load the data (Distances between facial Landmarks)\n",
        "dataX_lm = np.load('./dataX.npy')\n",
        "#dataY_lm = np.load('./dataY.npy')\n",
        "\n",
        "# convert labels to one hot encoding\n",
        "#y_onehot = to_categorical(dataY_lm, len(set(dataY_lm)))\n",
        "\n",
        "# split Data into Train, Test (90-10)\n",
        "X_train_lm, X_test_lm, y_train_lm, y_test_lm = train_test_split(dataX_lm, y_onehot, test_size=0.1, random_state=42)\n",
        "\n",
        "#### Standardize the data ##########\n",
        "lm_scaler = StandardScaler()\n",
        "lm_scaler.fit(X_train_lm)\n",
        "X_train_lm = lm_scaler.transform(X_train_lm)\n",
        "X_test_lm = lm_scaler.transform(X_test_lm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OPub4NX31R3q"
      },
      "source": [
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE\n",
        "############### CALL YOUR MODEL 'lm_model' \n",
        "\n",
        "lm_model = Sequential()\n",
        "\n",
        "lm_model.add(Dense(1024, input_shape = (2278,), activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "lm_model.add(Dropout(rate=0.5))\n",
        "lm_model.add(Dense(512, activation = 'relu', kernel_initializer = 'glorot_normal'))\n",
        "lm_model.add(Dense(5, activation = 'softmax'))\n",
        "\n",
        "lm_model.compile(loss=categorical_crossentropy, optimizer=SGD(lr=0.001), metrics=['accuracy'])\n",
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE\n",
        "#YOUR CODE HERE"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PQZEkMwsvtlw"
      },
      "source": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_QEUcTxh7Ara",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "425cec4a-487e-4491-9fc8-463a260bf3fd"
      },
      "source": [
        "# Compliling the model with SGD optimixer and categorical crossentropy loss\n",
        "lm_model.compile(loss=categorical_crossentropy, optimizer=SGD(lr=0.001), metrics=['accuracy'])\n",
        "\n",
        "#Saves the Best Model Based on Val Loss\n",
        "checkpoint = ModelCheckpoint('best_lm_model.h5', verbose=1, monitor='val_loss',save_best_only=True,  mode='auto')  \n",
        "#training the model\n",
        "lm_history = lm_model.fit(X_train_lm, y_train_lm, batch_size=batch_size, epochs=50, \n",
        "                          verbose=1, callbacks=[checkpoint], validation_data=(X_test_lm, y_test_lm), shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "282/282 [==============================] - 2s 4ms/step - loss: 1.7056 - accuracy: 0.3045 - val_loss: 1.3179 - val_accuracy: 0.4350\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 1.31793, saving model to best_lm_model.h5\n",
            "Epoch 2/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.4741 - accuracy: 0.3995 - val_loss: 1.2737 - val_accuracy: 0.4580\n",
            "\n",
            "Epoch 00002: val_loss improved from 1.31793 to 1.27370, saving model to best_lm_model.h5\n",
            "Epoch 3/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.3915 - accuracy: 0.4344 - val_loss: 1.2566 - val_accuracy: 0.4750\n",
            "\n",
            "Epoch 00003: val_loss improved from 1.27370 to 1.25660, saving model to best_lm_model.h5\n",
            "Epoch 4/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3659 - accuracy: 0.4410 - val_loss: 1.2359 - val_accuracy: 0.4805\n",
            "\n",
            "Epoch 00004: val_loss improved from 1.25660 to 1.23585, saving model to best_lm_model.h5\n",
            "Epoch 5/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3332 - accuracy: 0.4515 - val_loss: 1.2325 - val_accuracy: 0.4805\n",
            "\n",
            "Epoch 00005: val_loss improved from 1.23585 to 1.23252, saving model to best_lm_model.h5\n",
            "Epoch 6/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3260 - accuracy: 0.4600 - val_loss: 1.2246 - val_accuracy: 0.4870\n",
            "\n",
            "Epoch 00006: val_loss improved from 1.23252 to 1.22463, saving model to best_lm_model.h5\n",
            "Epoch 7/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.3058 - accuracy: 0.4656 - val_loss: 1.2162 - val_accuracy: 0.4960\n",
            "\n",
            "Epoch 00007: val_loss improved from 1.22463 to 1.21624, saving model to best_lm_model.h5\n",
            "Epoch 8/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2864 - accuracy: 0.4744 - val_loss: 1.2087 - val_accuracy: 0.4950\n",
            "\n",
            "Epoch 00008: val_loss improved from 1.21624 to 1.20868, saving model to best_lm_model.h5\n",
            "Epoch 9/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2821 - accuracy: 0.4770 - val_loss: 1.2028 - val_accuracy: 0.4960\n",
            "\n",
            "Epoch 00009: val_loss improved from 1.20868 to 1.20284, saving model to best_lm_model.h5\n",
            "Epoch 10/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2533 - accuracy: 0.4943 - val_loss: 1.2004 - val_accuracy: 0.5095\n",
            "\n",
            "Epoch 00010: val_loss improved from 1.20284 to 1.20040, saving model to best_lm_model.h5\n",
            "Epoch 11/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2540 - accuracy: 0.4837 - val_loss: 1.1982 - val_accuracy: 0.5055\n",
            "\n",
            "Epoch 00011: val_loss improved from 1.20040 to 1.19818, saving model to best_lm_model.h5\n",
            "Epoch 12/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2521 - accuracy: 0.4896 - val_loss: 1.1929 - val_accuracy: 0.5075\n",
            "\n",
            "Epoch 00012: val_loss improved from 1.19818 to 1.19293, saving model to best_lm_model.h5\n",
            "Epoch 13/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2351 - accuracy: 0.4914 - val_loss: 1.1860 - val_accuracy: 0.5135\n",
            "\n",
            "Epoch 00013: val_loss improved from 1.19293 to 1.18603, saving model to best_lm_model.h5\n",
            "Epoch 14/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2318 - accuracy: 0.5011 - val_loss: 1.1861 - val_accuracy: 0.5135\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 1.18603\n",
            "Epoch 15/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2326 - accuracy: 0.4998 - val_loss: 1.1816 - val_accuracy: 0.5080\n",
            "\n",
            "Epoch 00015: val_loss improved from 1.18603 to 1.18156, saving model to best_lm_model.h5\n",
            "Epoch 16/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2135 - accuracy: 0.5037 - val_loss: 1.1754 - val_accuracy: 0.5165\n",
            "\n",
            "Epoch 00016: val_loss improved from 1.18156 to 1.17538, saving model to best_lm_model.h5\n",
            "Epoch 17/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2266 - accuracy: 0.4973 - val_loss: 1.1738 - val_accuracy: 0.5170\n",
            "\n",
            "Epoch 00017: val_loss improved from 1.17538 to 1.17376, saving model to best_lm_model.h5\n",
            "Epoch 18/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.2177 - accuracy: 0.5076 - val_loss: 1.1732 - val_accuracy: 0.5130\n",
            "\n",
            "Epoch 00018: val_loss improved from 1.17376 to 1.17316, saving model to best_lm_model.h5\n",
            "Epoch 19/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2068 - accuracy: 0.5124 - val_loss: 1.1684 - val_accuracy: 0.5210\n",
            "\n",
            "Epoch 00019: val_loss improved from 1.17316 to 1.16839, saving model to best_lm_model.h5\n",
            "Epoch 20/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2083 - accuracy: 0.5059 - val_loss: 1.1678 - val_accuracy: 0.5205\n",
            "\n",
            "Epoch 00020: val_loss improved from 1.16839 to 1.16777, saving model to best_lm_model.h5\n",
            "Epoch 21/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.2102 - accuracy: 0.5118 - val_loss: 1.1649 - val_accuracy: 0.5250\n",
            "\n",
            "Epoch 00021: val_loss improved from 1.16777 to 1.16491, saving model to best_lm_model.h5\n",
            "Epoch 22/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1938 - accuracy: 0.5164 - val_loss: 1.1640 - val_accuracy: 0.5210\n",
            "\n",
            "Epoch 00022: val_loss improved from 1.16491 to 1.16397, saving model to best_lm_model.h5\n",
            "Epoch 23/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1874 - accuracy: 0.5234 - val_loss: 1.1599 - val_accuracy: 0.5290\n",
            "\n",
            "Epoch 00023: val_loss improved from 1.16397 to 1.15992, saving model to best_lm_model.h5\n",
            "Epoch 24/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1923 - accuracy: 0.5204 - val_loss: 1.1615 - val_accuracy: 0.5255\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 1.15992\n",
            "Epoch 25/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1850 - accuracy: 0.5192 - val_loss: 1.1610 - val_accuracy: 0.5240\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 1.15992\n",
            "Epoch 26/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1823 - accuracy: 0.5227 - val_loss: 1.1577 - val_accuracy: 0.5300\n",
            "\n",
            "Epoch 00026: val_loss improved from 1.15992 to 1.15769, saving model to best_lm_model.h5\n",
            "Epoch 27/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1870 - accuracy: 0.5227 - val_loss: 1.1537 - val_accuracy: 0.5360\n",
            "\n",
            "Epoch 00027: val_loss improved from 1.15769 to 1.15365, saving model to best_lm_model.h5\n",
            "Epoch 28/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1869 - accuracy: 0.5177 - val_loss: 1.1526 - val_accuracy: 0.5380\n",
            "\n",
            "Epoch 00028: val_loss improved from 1.15365 to 1.15257, saving model to best_lm_model.h5\n",
            "Epoch 29/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1764 - accuracy: 0.5238 - val_loss: 1.1545 - val_accuracy: 0.5300\n",
            "\n",
            "Epoch 00029: val_loss did not improve from 1.15257\n",
            "Epoch 30/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1675 - accuracy: 0.5257 - val_loss: 1.1514 - val_accuracy: 0.5310\n",
            "\n",
            "Epoch 00030: val_loss improved from 1.15257 to 1.15138, saving model to best_lm_model.h5\n",
            "Epoch 31/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1653 - accuracy: 0.5348 - val_loss: 1.1484 - val_accuracy: 0.5395\n",
            "\n",
            "Epoch 00031: val_loss improved from 1.15138 to 1.14839, saving model to best_lm_model.h5\n",
            "Epoch 32/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1681 - accuracy: 0.5298 - val_loss: 1.1485 - val_accuracy: 0.5335\n",
            "\n",
            "Epoch 00032: val_loss did not improve from 1.14839\n",
            "Epoch 33/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1745 - accuracy: 0.5313 - val_loss: 1.1510 - val_accuracy: 0.5265\n",
            "\n",
            "Epoch 00033: val_loss did not improve from 1.14839\n",
            "Epoch 34/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1808 - accuracy: 0.5188 - val_loss: 1.1448 - val_accuracy: 0.5390\n",
            "\n",
            "Epoch 00034: val_loss improved from 1.14839 to 1.14484, saving model to best_lm_model.h5\n",
            "Epoch 35/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1646 - accuracy: 0.5281 - val_loss: 1.1469 - val_accuracy: 0.5350\n",
            "\n",
            "Epoch 00035: val_loss did not improve from 1.14484\n",
            "Epoch 36/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1706 - accuracy: 0.5256 - val_loss: 1.1458 - val_accuracy: 0.5340\n",
            "\n",
            "Epoch 00036: val_loss did not improve from 1.14484\n",
            "Epoch 37/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1742 - accuracy: 0.5274 - val_loss: 1.1466 - val_accuracy: 0.5340\n",
            "\n",
            "Epoch 00037: val_loss did not improve from 1.14484\n",
            "Epoch 38/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1669 - accuracy: 0.5307 - val_loss: 1.1449 - val_accuracy: 0.5340\n",
            "\n",
            "Epoch 00038: val_loss did not improve from 1.14484\n",
            "Epoch 39/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1606 - accuracy: 0.5337 - val_loss: 1.1446 - val_accuracy: 0.5350\n",
            "\n",
            "Epoch 00039: val_loss improved from 1.14484 to 1.14465, saving model to best_lm_model.h5\n",
            "Epoch 40/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1636 - accuracy: 0.5367 - val_loss: 1.1435 - val_accuracy: 0.5415\n",
            "\n",
            "Epoch 00040: val_loss improved from 1.14465 to 1.14349, saving model to best_lm_model.h5\n",
            "Epoch 41/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1547 - accuracy: 0.5282 - val_loss: 1.1425 - val_accuracy: 0.5370\n",
            "\n",
            "Epoch 00041: val_loss improved from 1.14349 to 1.14248, saving model to best_lm_model.h5\n",
            "Epoch 42/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1597 - accuracy: 0.5287 - val_loss: 1.1405 - val_accuracy: 0.5410\n",
            "\n",
            "Epoch 00042: val_loss improved from 1.14248 to 1.14053, saving model to best_lm_model.h5\n",
            "Epoch 43/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1488 - accuracy: 0.5352 - val_loss: 1.1391 - val_accuracy: 0.5365\n",
            "\n",
            "Epoch 00043: val_loss improved from 1.14053 to 1.13907, saving model to best_lm_model.h5\n",
            "Epoch 44/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1496 - accuracy: 0.5363 - val_loss: 1.1384 - val_accuracy: 0.5405\n",
            "\n",
            "Epoch 00044: val_loss improved from 1.13907 to 1.13843, saving model to best_lm_model.h5\n",
            "Epoch 45/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1372 - accuracy: 0.5418 - val_loss: 1.1384 - val_accuracy: 0.5405\n",
            "\n",
            "Epoch 00045: val_loss did not improve from 1.13843\n",
            "Epoch 46/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1520 - accuracy: 0.5398 - val_loss: 1.1393 - val_accuracy: 0.5405\n",
            "\n",
            "Epoch 00046: val_loss did not improve from 1.13843\n",
            "Epoch 47/50\n",
            "282/282 [==============================] - 1s 3ms/step - loss: 1.1462 - accuracy: 0.5453 - val_loss: 1.1361 - val_accuracy: 0.5415\n",
            "\n",
            "Epoch 00047: val_loss improved from 1.13843 to 1.13609, saving model to best_lm_model.h5\n",
            "Epoch 48/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1424 - accuracy: 0.5418 - val_loss: 1.1361 - val_accuracy: 0.5380\n",
            "\n",
            "Epoch 00048: val_loss did not improve from 1.13609\n",
            "Epoch 49/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1452 - accuracy: 0.5354 - val_loss: 1.1380 - val_accuracy: 0.5365\n",
            "\n",
            "Epoch 00049: val_loss did not improve from 1.13609\n",
            "Epoch 50/50\n",
            "282/282 [==============================] - 1s 4ms/step - loss: 1.1293 - accuracy: 0.5464 - val_loss: 1.1358 - val_accuracy: 0.5390\n",
            "\n",
            "Epoch 00050: val_loss improved from 1.13609 to 1.13579, saving model to best_lm_model.h5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Xb2ruZoq1YPO"
      },
      "source": [
        "### Evaluate model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AYdFbtJlxG9S",
        "outputId": "f0067012-3ebd-4da6-b4ae-d44bddf2621b"
      },
      "source": [
        "lm_performance = lm_model.evaluate(X_test_lm, y_test_lm, batch_size=64)\n",
        "lm_performance"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "32/32 [==============================] - 0s 3ms/step - loss: 1.1358 - accuracy: 0.5390\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.1357940435409546, 0.5389999747276306]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yj3XDT2eu31x"
      },
      "source": [
        "###Visualize accuracy and loss over training + display best model's performance"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kG1V_8AIs9Xf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 310
        },
        "outputId": "4277b4cf-1592-4b7a-e0c9-b5e9774e78f4"
      },
      "source": [
        "# YOUR CODE HERE\n",
        "# YOUR CODE HERE\n",
        "#Saves the Best Model Based on Val Loss\n",
        "plot_graphs(lm_history, lm_performance); "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1Xk6GHxUgiY6"
      },
      "source": [
        "## Instructor-Led Discussion\n",
        "\n",
        "How can we explain the difference between these two models?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lX9zvfvrAtrj"
      },
      "source": [
        "#Convolutional Neural Networks for Emotion Detection!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jXuDr1x1CZYP"
      },
      "source": [
        "###Model Hyperparameters"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OvPqbBy8Cc1G"
      },
      "source": [
        "# we'll use the same epochs and batch size as above\n",
        "width, height = 48, 48"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wTH-IlrL0HVE"
      },
      "source": [
        "###Reshape the inputs"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DwTlyCBk6FOd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4d5bdd3a-22e8-498b-9139-5c755af6c0be"
      },
      "source": [
        "# pixels were vectors\n",
        "print(X_train.shape) "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(18000, 2304)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NNSP1-vz6Lnk",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c5b186ef-7bfb-4ffe-c894-97a68e06b940"
      },
      "source": [
        "X_train_cnn = X_train.reshape(len(X_train),height,width)\n",
        "X_test_cnn = X_test.reshape(len(X_test),height,width)\n",
        "\n",
        "# we've converted them to images\n",
        "print(X_train_cnn.shape) \n",
        "# now we add one more dimension for model compatability\n",
        "print(X_test_cnn.shape) \n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(18000, 48, 48)\n",
            "(2000, 48, 48)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RFP5S57i69o7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1b4de6d6-78f6-4e87-db93-c17d1b3fbc46"
      },
      "source": [
        "# now we add one more dimension for model compatability\n",
        "X_train_cnn = np.expand_dims(X_train_cnn,3)\n",
        "X_test_cnn = np.expand_dims(X_test_cnn,3)\n",
        "\n",
        "print(X_train_cnn.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(18000, 48, 48, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fnI75Z1yDylH",
        "cellView": "both"
      },
      "source": [
        "# Change the hyperparameters in the model below \n",
        "\n",
        "# initialize model\n",
        "cnn_model = Sequential()\n",
        "# this conv layer has 64 filters! the input shape needs to be the same dimensions of the image\n",
        "cnn_model.add(Conv2D(64, kernel_size=(3, 3), activation='relu', input_shape=(width, height, 1)))\n",
        "cnn_model.add(Conv2D(64, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
        "cnn_model.add(Dropout(0.5))\n",
        "\n",
        "cnn_model.add(Conv2D(128, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(Conv2D(128, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
        "cnn_model.add(Dropout(0.5))\n",
        "\n",
        "cnn_model.add(Conv2D(256, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(Conv2D(256, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
        "cnn_model.add(Dropout(0.5))\n",
        "\n",
        "cnn_model.add(Conv2D(512, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(Conv2D(512, kernel_size=(3, 3), activation='relu', padding = 'same'))\n",
        "cnn_model.add(BatchNormalization())\n",
        "cnn_model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
        "cnn_model.add(Dropout(0.5))\n",
        "\n",
        "# flatten all the outputs between convolutional and dense layers\n",
        "cnn_model.add(Flatten())\n",
        "# add a \"dense layer\" (i.e. the fully connected layers in MLPs) with dropout\n",
        "cnn_model.add(Dense(512, activation='relu'))\n",
        "cnn_model.add(Dense(256, activation = 'relu'))\n",
        "cnn_model.add(Dense(128, activation = 'relu'))\n",
        "cnn_model.add(Dropout(rate=0.5))\n",
        "# output layer\n",
        "cnn_model.add(Dense(n_labels, activation='softmax'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YqirnWL8wDGR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "18323b56-6f72-4679-f364-2419ef071747"
      },
      "source": [
        "#Saves the Best Model Based on Val Loss\n",
        "checkpoint = ModelCheckpoint('best_cnn_model.h5', verbose=1, monitor='val_loss',save_best_only=True, mode='auto')  \n",
        "\n",
        "# compliling the model with adam optimixer and categorical crossentropy loss\n",
        "cnn_model.compile(loss=categorical_crossentropy, optimizer=Adam(lr=0.001, beta_1=0.9, beta_2=0.999), metrics=['accuracy'])\n",
        "\n",
        "# training the model\n",
        "cnn_history = cnn_model.fit(X_train_cnn, y_train, batch_size=batch_size, epochs=35, verbose=1, \n",
        "                            callbacks=[checkpoint], validation_data=(X_test_cnn, y_test), shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/35\n",
            "282/282 [==============================] - 48s 48ms/step - loss: 1.7723 - accuracy: 0.2411 - val_loss: 1.7167 - val_accuracy: 0.2180\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 1.71672, saving model to best_cnn_model.h5\n",
            "Epoch 2/35\n",
            "282/282 [==============================] - 12s 43ms/step - loss: 1.4621 - accuracy: 0.3364 - val_loss: 1.3432 - val_accuracy: 0.4140\n",
            "\n",
            "Epoch 00002: val_loss improved from 1.71672 to 1.34316, saving model to best_cnn_model.h5\n",
            "Epoch 3/35\n",
            "282/282 [==============================] - 12s 43ms/step - loss: 1.3744 - accuracy: 0.4021 - val_loss: 1.3261 - val_accuracy: 0.4390\n",
            "\n",
            "Epoch 00003: val_loss improved from 1.34316 to 1.32613, saving model to best_cnn_model.h5\n",
            "Epoch 4/35\n",
            "282/282 [==============================] - 12s 44ms/step - loss: 1.2532 - accuracy: 0.4752 - val_loss: 1.1100 - val_accuracy: 0.5445\n",
            "\n",
            "Epoch 00004: val_loss improved from 1.32613 to 1.10997, saving model to best_cnn_model.h5\n",
            "Epoch 5/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 1.1558 - accuracy: 0.5396 - val_loss: 1.0796 - val_accuracy: 0.5680\n",
            "\n",
            "Epoch 00005: val_loss improved from 1.10997 to 1.07963, saving model to best_cnn_model.h5\n",
            "Epoch 6/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 1.0780 - accuracy: 0.5739 - val_loss: 0.9798 - val_accuracy: 0.6040\n",
            "\n",
            "Epoch 00006: val_loss improved from 1.07963 to 0.97980, saving model to best_cnn_model.h5\n",
            "Epoch 7/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 1.0330 - accuracy: 0.5926 - val_loss: 0.9715 - val_accuracy: 0.6015\n",
            "\n",
            "Epoch 00007: val_loss improved from 0.97980 to 0.97152, saving model to best_cnn_model.h5\n",
            "Epoch 8/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.9919 - accuracy: 0.6121 - val_loss: 0.8823 - val_accuracy: 0.6555\n",
            "\n",
            "Epoch 00008: val_loss improved from 0.97152 to 0.88234, saving model to best_cnn_model.h5\n",
            "Epoch 9/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 0.9300 - accuracy: 0.6388 - val_loss: 0.9263 - val_accuracy: 0.6555\n",
            "\n",
            "Epoch 00009: val_loss did not improve from 0.88234\n",
            "Epoch 10/35\n",
            "282/282 [==============================] - 12s 44ms/step - loss: 0.9012 - accuracy: 0.6493 - val_loss: 0.8961 - val_accuracy: 0.6570\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.88234\n",
            "Epoch 11/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 0.8653 - accuracy: 0.6678 - val_loss: 0.9498 - val_accuracy: 0.6315\n",
            "\n",
            "Epoch 00011: val_loss did not improve from 0.88234\n",
            "Epoch 12/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.8312 - accuracy: 0.6894 - val_loss: 0.9036 - val_accuracy: 0.6545\n",
            "\n",
            "Epoch 00012: val_loss did not improve from 0.88234\n",
            "Epoch 13/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.8069 - accuracy: 0.7043 - val_loss: 0.9148 - val_accuracy: 0.6560\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 0.88234\n",
            "Epoch 14/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.7796 - accuracy: 0.7103 - val_loss: 0.8917 - val_accuracy: 0.6655\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.88234\n",
            "Epoch 15/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.7533 - accuracy: 0.7188 - val_loss: 0.8777 - val_accuracy: 0.6850\n",
            "\n",
            "Epoch 00015: val_loss improved from 0.88234 to 0.87770, saving model to best_cnn_model.h5\n",
            "Epoch 16/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.7199 - accuracy: 0.7323 - val_loss: 0.8671 - val_accuracy: 0.6880\n",
            "\n",
            "Epoch 00016: val_loss improved from 0.87770 to 0.86712, saving model to best_cnn_model.h5\n",
            "Epoch 17/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.6826 - accuracy: 0.7484 - val_loss: 0.8497 - val_accuracy: 0.6940\n",
            "\n",
            "Epoch 00017: val_loss improved from 0.86712 to 0.84971, saving model to best_cnn_model.h5\n",
            "Epoch 18/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 0.6532 - accuracy: 0.7651 - val_loss: 0.8961 - val_accuracy: 0.6800\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 0.84971\n",
            "Epoch 19/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 0.6368 - accuracy: 0.7640 - val_loss: 0.8574 - val_accuracy: 0.6990\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.84971\n",
            "Epoch 20/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.6176 - accuracy: 0.7765 - val_loss: 0.8776 - val_accuracy: 0.7055\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.84971\n",
            "Epoch 21/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.5825 - accuracy: 0.7883 - val_loss: 0.9024 - val_accuracy: 0.6905\n",
            "\n",
            "Epoch 00021: val_loss did not improve from 0.84971\n",
            "Epoch 22/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.5480 - accuracy: 0.7989 - val_loss: 0.8703 - val_accuracy: 0.7015\n",
            "\n",
            "Epoch 00022: val_loss did not improve from 0.84971\n",
            "Epoch 23/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.5342 - accuracy: 0.8019 - val_loss: 0.9256 - val_accuracy: 0.7065\n",
            "\n",
            "Epoch 00023: val_loss did not improve from 0.84971\n",
            "Epoch 24/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.4867 - accuracy: 0.8267 - val_loss: 0.9130 - val_accuracy: 0.6980\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 0.84971\n",
            "Epoch 25/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.4723 - accuracy: 0.8257 - val_loss: 0.9328 - val_accuracy: 0.7075\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 0.84971\n",
            "Epoch 26/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.4630 - accuracy: 0.8344 - val_loss: 0.9490 - val_accuracy: 0.6880\n",
            "\n",
            "Epoch 00026: val_loss did not improve from 0.84971\n",
            "Epoch 27/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.4288 - accuracy: 0.8425 - val_loss: 0.9509 - val_accuracy: 0.6980\n",
            "\n",
            "Epoch 00027: val_loss did not improve from 0.84971\n",
            "Epoch 28/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3998 - accuracy: 0.8532 - val_loss: 0.9568 - val_accuracy: 0.6915\n",
            "\n",
            "Epoch 00028: val_loss did not improve from 0.84971\n",
            "Epoch 29/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3909 - accuracy: 0.8596 - val_loss: 1.0166 - val_accuracy: 0.6930\n",
            "\n",
            "Epoch 00029: val_loss did not improve from 0.84971\n",
            "Epoch 30/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3716 - accuracy: 0.8693 - val_loss: 1.1082 - val_accuracy: 0.6865\n",
            "\n",
            "Epoch 00030: val_loss did not improve from 0.84971\n",
            "Epoch 31/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3734 - accuracy: 0.8669 - val_loss: 0.9982 - val_accuracy: 0.7000\n",
            "\n",
            "Epoch 00031: val_loss did not improve from 0.84971\n",
            "Epoch 32/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3440 - accuracy: 0.8764 - val_loss: 1.0371 - val_accuracy: 0.6985\n",
            "\n",
            "Epoch 00032: val_loss did not improve from 0.84971\n",
            "Epoch 33/35\n",
            "282/282 [==============================] - 13s 44ms/step - loss: 0.3293 - accuracy: 0.8842 - val_loss: 0.9988 - val_accuracy: 0.6980\n",
            "\n",
            "Epoch 00033: val_loss did not improve from 0.84971\n",
            "Epoch 34/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.3096 - accuracy: 0.8889 - val_loss: 1.0057 - val_accuracy: 0.7025\n",
            "\n",
            "Epoch 00034: val_loss did not improve from 0.84971\n",
            "Epoch 35/35\n",
            "282/282 [==============================] - 13s 45ms/step - loss: 0.2949 - accuracy: 0.8968 - val_loss: 1.1279 - val_accuracy: 0.6945\n",
            "\n",
            "Epoch 00035: val_loss did not improve from 0.84971\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wXprpaHpxVb-"
      },
      "source": [
        "###Evaluate the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JvCbnwZ_CBYv",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "98a32a7f-516d-42b3-d2f2-70ee7f305a2e"
      },
      "source": [
        "cnn_performance = cnn_model.evaluate(X_test_cnn, y_test, batch_size=64)\n",
        "cnn_performance"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "32/32 [==============================] - 0s 14ms/step - loss: 1.1279 - accuracy: 0.6945\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.1278973817825317, 0.6945000290870667]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cDmHi_05BKnr"
      },
      "source": [
        "###Plot Accuracy and Loss"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YdRZz5w2ixwO",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 310
        },
        "outputId": "d34dd7b2-53f7-49e2-999c-c3ce4152dcb0"
      },
      "source": [
        "plot_graphs(cnn_history, cnn_performance); "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hzLSZVH0jdG0"
      },
      "source": [
        "## Instructor-Led Discussion\n",
        "\n",
        "How does this model compare with the others? Did this perform as well as we wanted it to? What might be some of the reasons why?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_eqKTSlRmEGF"
      },
      "source": [
        "#Transfer Learning"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OPpE58jJmIdY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d840b632-eb3f-4ef6-81be-c0612eb91a8a"
      },
      "source": [
        "#@title Run this to build your transfer learning model { display-mode: \"form\" }\n",
        "\n",
        "import keras\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Activation, MaxPooling2D, Dropout, Flatten, Reshape, Dense, Conv2D, GlobalAveragePooling2D\n",
        "from keras.wrappers.scikit_learn import KerasClassifier\n",
        "import keras.optimizers as optimizers\n",
        "from keras.applications.vgg16 import VGG16\n",
        "\n",
        "# load the vgg network that is an 'expert' at 'imagenet' but do not include the FC layers\n",
        "vgg_expert = VGG16(weights = 'imagenet', include_top = False, input_shape = (48, 48, 3))\n",
        "\n",
        "# we add the first 12 layers of vgg to our own model vgg_model\n",
        "vgg_model = Sequential()\n",
        "vgg_model.add(vgg_expert)\n",
        "\n",
        "# and then add our own layers on top of it\n",
        "vgg_model.add(GlobalAveragePooling2D())\n",
        "vgg_model.add(Dense(1024, activation = 'relu'))\n",
        "vgg_model.add(Dropout(0.3))\n",
        "vgg_model.add(Dense(512, activation = 'relu'))\n",
        "vgg_model.add(Dropout(0.3))\n",
        "vgg_model.add(Dense(5, activation = 'sigmoid'))\n",
        "\n",
        "# finally, we build the vgg model and turn it on so we can use it!\n",
        "vgg_model.compile(loss = 'categorical_crossentropy', \n",
        "          optimizer = optimizers.SGD(lr=1e-4, momentum=0.95), \n",
        "          metrics=['accuracy'])\n",
        "\n",
        "X_TRAIN = np.array([np.transpose(np.array([X_train_cnn[ix].squeeze() for i in range(3)]), (1,2,0)) for ix in range(len(X_train))])\n",
        "X_TEST = np.array([np.transpose(np.array([X_test_cnn[ix].squeeze() for i in range(3)]), (1,2,0)) for ix in range(len(X_test))])\n",
        "\n",
        "#training the model\n",
        "vgg_history = vgg_model.fit(X_TRAIN, y_train,\n",
        "          batch_size=batch_size,\n",
        "          epochs=epochs,\n",
        "          verbose=1,\n",
        "          callbacks=[checkpoint],\n",
        "          validation_data=(X_TEST, y_test),\n",
        "          shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 0s 0us/step\n",
            "58900480/58889256 [==============================] - 0s 0us/step\n",
            "Epoch 1/20\n",
            "282/282 [==============================] - 22s 67ms/step - loss: 1.6100 - accuracy: 0.2565 - val_loss: 1.3201 - val_accuracy: 0.4385\n",
            "\n",
            "Epoch 00001: val_loss did not improve from 0.84971\n",
            "Epoch 2/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 1.3062 - accuracy: 0.4333 - val_loss: 1.1445 - val_accuracy: 0.5240\n",
            "\n",
            "Epoch 00002: val_loss did not improve from 0.84971\n",
            "Epoch 3/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 1.1536 - accuracy: 0.5264 - val_loss: 1.0388 - val_accuracy: 0.5830\n",
            "\n",
            "Epoch 00003: val_loss did not improve from 0.84971\n",
            "Epoch 4/20\n",
            "282/282 [==============================] - 18s 62ms/step - loss: 1.0497 - accuracy: 0.5809 - val_loss: 1.0141 - val_accuracy: 0.5970\n",
            "\n",
            "Epoch 00004: val_loss did not improve from 0.84971\n",
            "Epoch 5/20\n",
            "282/282 [==============================] - 18s 62ms/step - loss: 1.0010 - accuracy: 0.6017 - val_loss: 0.9748 - val_accuracy: 0.6125\n",
            "\n",
            "Epoch 00005: val_loss did not improve from 0.84971\n",
            "Epoch 6/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.9540 - accuracy: 0.6212 - val_loss: 0.9866 - val_accuracy: 0.6080\n",
            "\n",
            "Epoch 00006: val_loss did not improve from 0.84971\n",
            "Epoch 7/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.9055 - accuracy: 0.6409 - val_loss: 0.9217 - val_accuracy: 0.6360\n",
            "\n",
            "Epoch 00007: val_loss did not improve from 0.84971\n",
            "Epoch 8/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.8571 - accuracy: 0.6609 - val_loss: 0.8944 - val_accuracy: 0.6390\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 0.84971\n",
            "Epoch 9/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.8281 - accuracy: 0.6771 - val_loss: 0.9760 - val_accuracy: 0.6275\n",
            "\n",
            "Epoch 00009: val_loss did not improve from 0.84971\n",
            "Epoch 10/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.8130 - accuracy: 0.6840 - val_loss: 0.8607 - val_accuracy: 0.6665\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.84971\n",
            "Epoch 11/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.7640 - accuracy: 0.7073 - val_loss: 0.9073 - val_accuracy: 0.6435\n",
            "\n",
            "Epoch 00011: val_loss did not improve from 0.84971\n",
            "Epoch 12/20\n",
            "282/282 [==============================] - 18s 62ms/step - loss: 0.7444 - accuracy: 0.7174 - val_loss: 0.8703 - val_accuracy: 0.6655\n",
            "\n",
            "Epoch 00012: val_loss did not improve from 0.84971\n",
            "Epoch 13/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.7100 - accuracy: 0.7325 - val_loss: 0.8655 - val_accuracy: 0.6730\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 0.84971\n",
            "Epoch 14/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.6775 - accuracy: 0.7478 - val_loss: 0.8604 - val_accuracy: 0.6695\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.84971\n",
            "Epoch 15/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.6393 - accuracy: 0.7558 - val_loss: 0.8700 - val_accuracy: 0.6660\n",
            "\n",
            "Epoch 00015: val_loss did not improve from 0.84971\n",
            "Epoch 16/20\n",
            "282/282 [==============================] - 18s 62ms/step - loss: 0.6009 - accuracy: 0.7715 - val_loss: 0.8906 - val_accuracy: 0.6750\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 0.84971\n",
            "Epoch 17/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.5762 - accuracy: 0.7841 - val_loss: 0.8887 - val_accuracy: 0.6820\n",
            "\n",
            "Epoch 00017: val_loss did not improve from 0.84971\n",
            "Epoch 18/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.5349 - accuracy: 0.8008 - val_loss: 0.9233 - val_accuracy: 0.6805\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 0.84971\n",
            "Epoch 19/20\n",
            "282/282 [==============================] - 17s 62ms/step - loss: 0.4975 - accuracy: 0.8187 - val_loss: 0.9740 - val_accuracy: 0.6670\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.84971\n",
            "Epoch 20/20\n",
            "282/282 [==============================] - 18s 63ms/step - loss: 0.4746 - accuracy: 0.8268 - val_loss: 0.9177 - val_accuracy: 0.6840\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.84971\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rwABV9sHAclF"
      },
      "source": [
        "###Evaluate model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EjquMpMmAZEo"
      },
      "source": [
        "###Visualize accuracy and loss over training + display best model's performance"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OflGd8RFsrTX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7765dacc-49eb-4b0d-f617-67f2dfb635ec"
      },
      "source": [
        "vgg_performance = vgg_model.evaluate(X_TEST, y_test, batch_size=64)\n",
        "plot_graphs(vgg_history, vgg_performance); \n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "32/32 [==============================] - 1s 18ms/step - loss: 0.9177 - accuracy: 0.6840\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "60FJChkPc7iY",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "3b2115a3-38fc-45fe-d37a-936e1b0aa53f"
      },
      "source": [
        "compare_learning(mlp_history, lm_history, cnn_history, vgg_history)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-kmm2fWDBtfW"
      },
      "source": [
        "###How can we explain this pattern of performance across all these models?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SRW9UiFbBoAK"
      },
      "source": [
        "# Finished :D"
      ]
    }
  ]
}