Evaluation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Setup"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1 Import required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:211: calling reduce_max_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:211: calling reduce_max_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:213: calling reduce_sum_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:213: calling reduce_sum_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:214: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Deprecated in favor of operator or tf.math.divide.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/mtcnn/layer_factory.py:214: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Deprecated in favor of operator or tf.math.divide.\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import sys\n",
    "import glob\n",
    "import numpy as np\n",
    "import image_preprocessing\n",
    "import cnn\n",
    "import bayesian_network\n",
    "import json\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib import style\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score\n",
    "import itertools\n",
    "\n",
    "%matplotlib inline  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Function to plot the confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    Credit: https://github.com/scikit-learn/scikit-learn/blob/master/examples/model_selection/plot_confusion_matrix.py\n",
    "    \"\"\"\n",
    "    np.set_printoptions(precision=2)\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    print(cm)\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], fmt),\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.3 Function to classify an image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = {\"Positive\": 0, \"Neutral\": 1, \"Negative\": 2, \"None\": 3}\n",
    "def classify_image(image_folder_path, image_name, real_label, cnn_model, bayesian_model, labels_list):\n",
    "    with open('val_labels.json', mode='r', encoding='utf-8') as f:\n",
    "        image_labels_dict = json.load(f)\n",
    "    labels = image_labels_dict[image_name]\n",
    "\n",
    "    print(\"RadhaKrishna\")\n",
    "#     print(labels)\n",
    "\n",
    "    # preprocess the image\n",
    "    image_preprocessing.preprocess(image_folder_path, image_name)\n",
    "\n",
    "    # get mean cnn predictions for the faces from the image\n",
    "    cnn_label, cnn_dict, faces_detected = cnn.predict_image(cnn_model, image_folder_path + \"Aligned/\", image_name)\n",
    "\n",
    "    # get the bayesian and bayesian + cnn predictions for the image\n",
    "    bayesian_label, bayesian_cnn_label, emotion_dict, emotion_cnn_dict = bayesian_network.inference(bayesian_model, labels_list, labels, cnn_label)\n",
    "\n",
    "#     print(\"Faces detected: \" + str(faces_detected))\n",
    "#     print(\"Real Label: \" + str(real_label))\n",
    "#     print(\"CNN Label: \" + str(cnn_label))\n",
    "    print(\"Bayesian Label: \" + str(bayesian_label))\n",
    "#     print(\"Bayesian + CNN Label: \" + str(bayesian_cnn_label))\n",
    "\n",
    "    return classes[real_label], classes[str(cnn_label)], classes[str(bayesian_label)], classes[str(bayesian_cnn_label)], faces_detected"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.4 Load the CNN and bayesian model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RadhaKrishna\n",
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /anaconda3/envs/venv/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded CNN model from disk\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "print(\"RadhaKrishna\")\n",
    "cnn_model = cnn.load_model()\n",
    "bayesian_model, labels_list = bayesian_network.load_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.5 Function to evaluate the test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate(image_folder_path, real_label):\n",
    "    print(\"RadhaKrishna\")\n",
    "    predictions = []\n",
    "    i = 0\n",
    "    for file in sorted(glob.glob(image_folder_path + \"*.jpg\")):\n",
    "        image_name = (file.split('/'))[-1]\n",
    "        print(\"Image: \" + image_name)\n",
    "        prediction = {\"Image\": image_name}\n",
    "        prediction[\"Actual\"], prediction[\"CNN\"], prediction[\"Bayesian\"], prediction[\"Bayesian + CNN\"], prediction[\"Faces Detected\"] = classify_image(image_folder_path, image_name, real_label, cnn_model, bayesian_model, labels_list)\n",
    "        predictions.append(prediction)\n",
    "        i+=1\n",
    "        if (i==100): \n",
    "            break\n",
    "    return predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.1 Perform evaluations on the validation dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RadhaKrishna\n",
      "Image: pos_1.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_10.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_100.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1000.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1001.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1002.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1003.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1004.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1005.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1006.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1007.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1008.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1009.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_101.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1010.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1011.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1012.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1013.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1014.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1015.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1016.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1017.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1018.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1019.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_102.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1020.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1021.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1022.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1023.jpg\n",
      "RadhaKrishna\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/envs/venv/lib/python3.7/site-packages/pgmpy/factors/discrete/DiscreteFactor.py:377: RuntimeWarning: invalid value encountered in true_divide\n",
      "  phi.values = phi.values / phi.values.sum()\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Bayesian Label: Positive\n",
      "Image: pos_1024.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1025.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1026.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1027.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1028.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1029.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_103.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1030.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1031.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1032.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1033.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1034.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1035.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1036.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1037.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1038.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1039.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_104.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1040.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1041.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1042.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1043.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1044.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1045.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Negative\n",
      "Image: pos_1046.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1047.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1048.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1049.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_105.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1050.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1051.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1052.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1053.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1054.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1055.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1056.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1057.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1058.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1059.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_106.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1060.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1061.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1062.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1063.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1064.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1065.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1066.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1067.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1068.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1069.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_107.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1070.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1071.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1072.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1073.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Negative\n",
      "Image: pos_1074.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1075.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1076.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1077.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1078.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1079.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_108.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1080.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1081.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1082.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "Image: pos_1083.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Negative\n",
      "Image: pos_1084.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1085.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1086.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Positive\n",
      "Image: pos_1087.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Negative\n",
      "Image: pos_1088.jpg\n",
      "RadhaKrishna\n",
      "Bayesian Label: Neutral\n",
      "RadhaKrishna\n",
      "Image: neu_1.jpg\n",
      "RadhaKrishna\n"
     ]
    }
   ],
   "source": [
    "class_list = ['Positive', 'Neutral', 'Negative']\n",
    "predictions_list = []\n",
    "for emotion_class in class_list:\n",
    "    predictions = evaluate('input/val/' + emotion_class + '/', emotion_class)\n",
    "    predictions_list += predictions\n",
    "df = pd.DataFrame(predictions_list)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.2 Save the results to a file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_pickle('predictions_300')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Analyse the results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 Read the results from the file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_pickle('predictions_1_2_3_4')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 Extract predictions from the dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = df[[\"Actual\"]]\n",
    "y_pred_cnn = df[[\"CNN\"]]\n",
    "y_pred_bayesian = df[[\"Bayesian\"]]\n",
    "y_pred_bayesian_cnn = df[[\"Bayesian + CNN\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Actual</th>\n",
       "      <th>Bayesian</th>\n",
       "      <th>Bayesian + CNN</th>\n",
       "      <th>CNN</th>\n",
       "      <th>Faces Detected</th>\n",
       "      <th>Image</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "      <td>pos_1.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "      <td>pos_10.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "      <td>pos_100.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>pos_1000.jpg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>True</td>\n",
       "      <td>pos_1001.jpg</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Actual  Bayesian  Bayesian + CNN  CNN  Faces Detected         Image\n",
       "0       0         1               0    0            True     pos_1.jpg\n",
       "1       0         0               0    0            True    pos_10.jpg\n",
       "2       0         1               0    0            True   pos_100.jpg\n",
       "3       0         0               0    1            True  pos_1000.jpg\n",
       "4       0         1               1    0            True  pos_1001.jpg"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.3 Plot the confusion matrix for CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[1473  152   55   67]\n",
      " [ 225  771  261  111]\n",
      " [ 253  351  549   78]\n",
      " [   0    0    0    0]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(y_true, y_pred_cnn)\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cm, classes=classes)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.4 Plot the confusion matrix for Bayesian network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[989 654 104]\n",
      " [209 882 277]\n",
      " [145 304 782]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(y_true, y_pred_bayesian)\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cm, classes={\"Positive\": 0, \"Neutral\": 1, \"Negative\": 2})\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.5 Plot the confusion matrix for Bayesian network + CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[1128  531   88]\n",
      " [ 185  902  281]\n",
      " [ 122  302  807]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(y_true, y_pred_bayesian_cnn)\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cm, classes={\"Positive\": 0, \"Neutral\": 1, \"Negative\": 2})\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.6 CNN Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6426599171652094"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true, y_pred_cnn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.7 Bayesian Network Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6104463874827427"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true, y_pred_bayesian)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.8 Bayesian + CNN Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6527841693511275"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_true, y_pred_bayesian_cnn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}