notebook-code/TN8_Ensemble_v1_randomoversampling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 171532,
     "status": "ok",
     "timestamp": 1610938682229,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "PL21Mz5-OeDg",
    "outputId": "42cee6c4-04dd-4792-f6ad-c91252bc04ce"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n",
      "/content/drive/.shortcut-targets-by-id/1BOvFtzoVhfYNe70lBGxYzhPlDElSrazo/Final_Project_DL\n"
     ]
    }
   ],
   "source": [
    "# Mount \"My Drive\" into /content/drive\n",
    "from google.colab import drive\n",
    "\n",
    "google_drive_dir = \"Final_Project_DL\"  # @param\n",
    "#bml-notebooks/\n",
    "drive.mount('/content/drive')\n",
    "\n",
    "mount_point = \"/content/drive/MyDrive/{}\".format(google_drive_dir)\n",
    "\n",
    "# Change the root directory to your mount_point\n",
    "%cd '$mount_point'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eI1PdIKtLrpv",
    "papermill": {
     "duration": 0.044808,
     "end_time": "2020-12-18T22:02:11.416442",
     "exception": false,
     "start_time": "2020-12-18T22:02:11.371634",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# <span style=\"color: royalblue;\">Load Data</span>\n",
    "Load the image data with labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "executionInfo": {
     "elapsed": 5562,
     "status": "ok",
     "timestamp": 1610938791263,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "HM1rN5gqd5aG"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from src.loadData import loadFer2013\n",
    "from keras.utils import to_categorical\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1816,
     "status": "ok",
     "timestamp": 1610938791265,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "Nt9oKq3mo60M",
    "outputId": "55b9910d-f3f3-4993-ebd2-3f68c80072cf"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Angry', 'Disgust', 'Fear', 'Happy', 'Sad', 'Surprise', 'Neutral']"
      ]
     },
     "execution_count": 3,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emotions = {0: 'Angry', 1:'Disgust', 2:'Fear', 3: 'Happy', 4: 'Sad', 5: 'Surprise', 6: 'Neutral'}\n",
    "class_names = list(emotions.values())\n",
    "class_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "executionInfo": {
     "elapsed": 8934,
     "status": "ok",
     "timestamp": 1610938801583,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "ct5pobN6eBgW"
   },
   "outputs": [],
   "source": [
    "X_train,X_val,X_test,y_train,y_val,y_test = loadFer2013('fer2013/icml_face_data.csv')\n",
    "\n",
    "y_train_onehot = to_categorical(y_train)\n",
    "y_val_onehot = to_categorical(y_val)\n",
    "y_test_onehot = to_categorical(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 871,
     "status": "ok",
     "timestamp": 1610938809269,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "icEw_PExckCo",
    "outputId": "d011031b-0813-40a3-c670-eaa987276bf6"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28623, 48, 48)"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-9KsqSFYycZW"
   },
   "source": [
    "#Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "executionInfo": {
     "elapsed": 948,
     "status": "ok",
     "timestamp": 1610938839265,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "lIAx61wiye5z"
   },
   "outputs": [],
   "source": [
    "def preprocessX(X):\n",
    "    return np.expand_dims(X,axis=-1)\n",
    "\n",
    "X_processed_val = preprocessX(X_val)\n",
    "X_processed_test = preprocessX(X_test)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-DkOqeRecp0l"
   },
   "source": [
    "#Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "executionInfo": {
     "elapsed": 953,
     "status": "ok",
     "timestamp": 1610938841406,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "WfXhrNLKi0sy"
   },
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.preprocessing import image\n",
    "from keras import applications as keras_applications"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 8419,
     "status": "ok",
     "timestamp": 1610938849220,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "zncRqtUejaMI",
    "outputId": "1bb5ca1d-9e55-493c-9b58-f5aa7e08db22"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting keras_applications\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/71/e3/19762fdfc62877ae9102edf6342d71b28fbfd9dea3d2f96a882ce099b03f/Keras_Applications-1.0.8-py3-none-any.whl (50kB)\n",
      "\r",
      "\u001b[K     |██████▌                         | 10kB 28.4MB/s eta 0:00:01\r",
      "\u001b[K     |█████████████                   | 20kB 14.2MB/s eta 0:00:01\r",
      "\u001b[K     |███████████████████▍            | 30kB 12.7MB/s eta 0:00:01\r",
      "\u001b[K     |█████████████████████████▉      | 40kB 12.0MB/s eta 0:00:01\r",
      "\u001b[K     |████████████████████████████████| 51kB 4.6MB/s \n",
      "\u001b[?25hRequirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from keras_applications) (2.10.0)\n",
      "Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python3.6/dist-packages (from keras_applications) (1.19.5)\n",
      "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from h5py->keras_applications) (1.15.0)\n",
      "Installing collected packages: keras-applications\n",
      "Successfully installed keras-applications-1.0.8\n",
      "Collecting git+https://github.com/rcmalli/keras-vggface.git\n",
      "  Cloning https://github.com/rcmalli/keras-vggface.git to /tmp/pip-req-build-5wdl3ihn\n",
      "  Running command git clone -q https://github.com/rcmalli/keras-vggface.git /tmp/pip-req-build-5wdl3ihn\n",
      "Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (1.19.5)\n",
      "Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (1.4.1)\n",
      "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (2.10.0)\n",
      "Requirement already satisfied: pillow in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (7.0.0)\n",
      "Requirement already satisfied: keras in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (2.4.3)\n",
      "Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (1.15.0)\n",
      "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from keras-vggface==0.6) (3.13)\n",
      "Building wheels for collected packages: keras-vggface\n",
      "  Building wheel for keras-vggface (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
      "  Created wheel for keras-vggface: filename=keras_vggface-0.6-cp36-none-any.whl size=8310 sha256=d8183500f9c7472cb0e2ddad5431e57de49c157540ba2aac503dbb32ba15cb88\n",
      "  Stored in directory: /tmp/pip-ephem-wheel-cache-rwz7l5fn/wheels/36/07/46/06c25ce8e9cd396dabe151ea1d8a2bc28dafcb11321c1f3a6d\n",
      "Successfully built keras-vggface\n",
      "Installing collected packages: keras-vggface\n",
      "Successfully installed keras-vggface-0.6\n"
     ]
    }
   ],
   "source": [
    "!pip install keras_applications\n",
    "!pip install git+https://github.com/rcmalli/keras-vggface.git"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "executionInfo": {
     "elapsed": 7640,
     "status": "ok",
     "timestamp": 1610938849222,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "cFzE7T-qjdO1"
   },
   "outputs": [],
   "source": [
    "from keras_vggface.vggface import VGGFace\n",
    "from keras_vggface import utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "executionInfo": {
     "elapsed": 7044,
     "status": "ok",
     "timestamp": 1610938849223,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "c2b8HuspiHVx"
   },
   "outputs": [],
   "source": [
    "from copy import deepcopy as copy\n",
    "\n",
    "def smooth_labels(y, smooth_factor):\n",
    "    '''Convert a matrix of one-hot row-vector labels into smoothed versions.\n",
    "\n",
    "    # Arguments\n",
    "        y: matrix of one-hot row-vector labels to be smoothed\n",
    "        smooth_factor: label smoothing factor (between 0 and 1)\n",
    "\n",
    "    # Returns\n",
    "        A matrix of smoothed labels.\n",
    "    '''\n",
    "    assert len(y.shape) == 2, 'input should be a batch of one-hot-encoded data'\n",
    "    y2 = copy(y)\n",
    "    if 0 <= smooth_factor <= 1:\n",
    "        # label smoothing ref: https://www.robots.ox.ac.uk/~vgg/rg/papers/reinception.pdf\n",
    "        y2 *= 1 - smooth_factor\n",
    "        y2 += smooth_factor / y.shape[1]\n",
    "    else:\n",
    "        raise Exception(\n",
    "            'Invalid label smoothing factor: ' + str(smooth_factor))\n",
    "    return y2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "executionInfo": {
     "elapsed": 7409,
     "status": "ok",
     "timestamp": 1610938850214,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "CHi8FiFwiKuV"
   },
   "outputs": [],
   "source": [
    "from keras.utils import Sequence\n",
    "from keras.utils import to_categorical\n",
    "import cv2\n",
    "from math import floor\n",
    "\n",
    "class data_sequence(Sequence):\n",
    "    '''\n",
    "      yield sequence of data\n",
    "      features -- list of features\n",
    "      labels -- list of labels\n",
    "      target_channels {int} -- 1 (gray) or 3(RGB)\n",
    "    '''\n",
    "    def __init__(self, features, labels, batch_size=128, target_dim=(224,224), \n",
    "                 n_classes=7, shuffle=True, smooth=0.0, ver=2):\n",
    "        'Initialization'\n",
    "        assert len(features)==len(labels), 'number of feature and labels not consistent'\n",
    "        self.features = features\n",
    "        self.labels = labels\n",
    "        self.batch_size = batch_size\n",
    "        self.target_dim = target_dim\n",
    "        self.target_channels = 3\n",
    "        self.n_classes = n_classes\n",
    "        self.shuffle = shuffle\n",
    "        self.smooth = smooth\n",
    "        self.sample_count = len(labels)\n",
    "        self.indexes = np.arange(self.sample_count)\n",
    "        self.on_epoch_end()\n",
    "        self.ver = ver\n",
    "#         self.verbose = verbose\n",
    "\n",
    "    def __len__(self):\n",
    "        'Denotes the number of batches per epoch'\n",
    "        return floor(self.sample_count / self.batch_size)\n",
    "\n",
    "    def __gray2RGB__(self,x):\n",
    "      if len(x.shape)==2:\n",
    "        return np.stack((x,x,x),-1)\n",
    "      else:\n",
    "        assert len(x.shape)==3\n",
    "        if len(x[0,0,:]) == 1:\n",
    "          return np.stack((x[:,:,0],x[:,:,0],x[:,:,0]),-1)\n",
    "        else:\n",
    "          assert len(x[0,0,:])==self.target_channels\n",
    "      return x\n",
    "\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        'Generate one batch of data'\n",
    "        # Generate indexes of the batch\n",
    "        indexes = self.indexes[idx * self.batch_size : (idx + 1) * self.batch_size]\n",
    "        X = np.empty((self.batch_size, *self.target_dim, self.target_channels))\n",
    "        Y = np.empty((self.batch_size, self.n_classes))\n",
    "        for i,ind in enumerate(indexes):\n",
    "          x = self.features[ind]\n",
    "          # resize image to the target size \n",
    "          x = cv2.resize(x,self.target_dim,interpolation=cv2.INTER_CUBIC)\n",
    "          x = self.__gray2RGB__(x)\n",
    "          X[i] = utils.preprocess_input(x, version=self.ver) # or version=2 for VGGFace2 ResNet50  \n",
    "          y = self.labels[ind]\n",
    "          if isinstance(y,int):\n",
    "            Y[i]=to_categorical(y,7)\n",
    "          else:\n",
    "            assert len(y)==self.n_classes\n",
    "            Y[i]=y\n",
    "        X = np.array(X)\n",
    "        Y = np.array(Y)\n",
    "        if self.smooth > 0.0:\n",
    "          smooth_labels(Y, self.smooth)\n",
    "        return X,Y\n",
    "\n",
    "    def on_epoch_end(self):\n",
    "        'Updates indexes after each epoch'\n",
    "        if self.shuffle == True:\n",
    "            np.random.shuffle(self.indexes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SQARMLYepeAF"
   },
   "source": [
    "#Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "executionInfo": {
     "elapsed": 11033,
     "status": "ok",
     "timestamp": 1610938916875,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "WDSeaJn_cjq9"
   },
   "outputs": [],
   "source": [
    "model = keras.models.load_model('Final_model/VGGFACE2_V1_RandomOverSampling.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "executionInfo": {
     "elapsed": 7021,
     "status": "ok",
     "timestamp": 1610938918466,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "kHDz2FECd3Nx"
   },
   "outputs": [],
   "source": [
    "model_vggface1 = keras.models.load_model('Final_model/VGGFACE1.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "nd4dL2rGut6n"
   },
   "outputs": [],
   "source": [
    "val_sequence  = data_sequence(X_processed_val, y_val_onehot, batch_size=37,target_dim=(224,224),n_classes=7,shuffle=False,smooth=0.0,ver=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "executionInfo": {
     "elapsed": 718,
     "status": "ok",
     "timestamp": 1610938923228,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "Ry1_ACAIibmg"
   },
   "outputs": [],
   "source": [
    "test_sequence  = data_sequence(X_processed_test, y_test_onehot, batch_size=37,target_dim=(224,224),n_classes=7,shuffle=False,smooth=0.0,ver=2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SOLRJR7SiWF-"
   },
   "outputs": [],
   "source": [
    "y_pred_v2 = model.predict(val_sequence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ZVM-3AC4jCXx"
   },
   "outputs": [],
   "source": [
    "y_pred_v1 = model_vggface1.predict(val_sequence)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "executionInfo": {
     "elapsed": 1620,
     "status": "ok",
     "timestamp": 1610938975606,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "fL3IhNRNi13e"
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import accuracy_score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "EXkpk1wVjTx3"
   },
   "outputs": [],
   "source": [
    "y_pred = (y_pred_v1+y_pred_v2)/2\n",
    "y_pred = np.argmax(y_pred, axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1075,
     "status": "ok",
     "timestamp": 1610832610094,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "TcDdTaLOj3xk",
    "outputId": "0f1bf121-42e9-476e-afd4-fac97ae33aeb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7283365840066871\n"
     ]
    }
   ],
   "source": [
    "print(accuracy_score(y_pred,y_val))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "oUP8yxbMNfzS"
   },
   "source": [
    "#test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "executionInfo": {
     "elapsed": 18086,
     "status": "ok",
     "timestamp": 1610938946891,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "kyeHv4hXe4_R"
   },
   "outputs": [],
   "source": [
    "y_pred_v2 = model.predict(test_sequence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "executionInfo": {
     "elapsed": 32279,
     "status": "ok",
     "timestamp": 1610938962752,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "R6aaMWv3e6mn"
   },
   "outputs": [],
   "source": [
    "y_pred_v1 = model_vggface1.predict(test_sequence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "executionInfo": {
     "elapsed": 871,
     "status": "ok",
     "timestamp": 1610938965947,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "-QOoC7s0e_MN"
   },
   "outputs": [],
   "source": [
    "y_pred = (y_pred_v1*+y_pred_v2)/2\n",
    "y_pred = np.argmax(y_pred, axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 652,
     "status": "ok",
     "timestamp": 1610938978765,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "8-S1oqFzfD4B",
    "outputId": "6187f713-2747-4b85-c2c0-3d8f0f35e429"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7470047366954583\n"
     ]
    }
   ],
   "source": [
    "print(accuracy_score(y_pred,y_test))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 525,
     "status": "ok",
     "timestamp": 1610940561384,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "Uy6UosBHckwq",
    "outputId": "8bfe9743-9b81-4a58-c467-3a182f76eba1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0 2 4 0 2 0 6 3 2 3]\n",
      "[0 5 6 4 2 0 4 3 0 3]\n"
     ]
    }
   ],
   "source": [
    "print(y_pred[:10])\n",
    "print(y_test[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "executionInfo": {
     "elapsed": 954,
     "status": "ok",
     "timestamp": 1610940609720,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "kAcQ61e4eTUB",
    "outputId": "8b4653cf-d353-4c6b-8fab-5ab9632243ab"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(-0.5, 47.5, 47.5, -0.5)"
      ]
     },
     "execution_count": 37,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.imshow(X_test[6].astype(np.uint8),cmap='gray')\n",
    "plt.axis('off')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "executionInfo": {
     "elapsed": 864,
     "status": "ok",
     "timestamp": 1610940631429,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "eyFqoEBZfhr4",
    "outputId": "c47d2c2b-3057-4503-b27e-e4a52fb0e54b"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'Sad'"
      ]
     },
     "execution_count": 39,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emotions[4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "R7ZwvC--fFSa"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "\n",
    "confusion_matrix = tf.math.confusion_matrix(labels=y_test, predictions=y_pred).numpy()\n",
    "confusion_matrix = np.around(confusion_matrix.astype('float') / confusion_matrix.sum(axis=1)[:, np.newaxis], decimals=2)\n",
    "\n",
    "confusion_matrix = pd.DataFrame(confusion_matrix,\n",
    "                                index = class_names, \n",
    "                                columns = class_names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 599
    },
    "executionInfo": {
     "elapsed": 22827,
     "status": "ok",
     "timestamp": 1610832636176,
     "user": {
      "displayName": "Dũng Nguyễn Minh",
      "photoUrl": "",
      "userId": "05826558414739051995"
     },
     "user_tz": -420
    },
    "id": "ItkQl8AdfWM1",
    "outputId": "10278259-2127-440d-a1ca-8c8f0c726b46"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "figure = plt.figure(figsize=(8, 8))\n",
    "sns.heatmap(confusion_matrix, annot=True,cmap=plt.cm.Blues)\n",
    "plt.tight_layout()\n",
    "plt.ylabel('True label')\n",
    "plt.xlabel('Predicted label')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "q4SRwoWIfdPf"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "TN8_Ensemble_v1_randomoversampling.ipynb",
   "provenance": []
  },
  "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}