Emotion_Recognition(CNN)_FER2013.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Emotion Recognition(CNN) - FER2013",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "dxQDaDGyjcYP",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Facial Emotion Recognition \n",
        "\n",
        "A baby is able to recognize it's parent's faces when it is about a couple of weeks old. The baby grows and matures and by the time it is about two months old, it starts to display social cues. It is able to understand basic emotions like a smile. \n",
        "\n",
        "This project involves training a neural network to do just that, recognize facial emotions. We owe our ability to do that to the wonderful process of evolution. Replicating this is certainly not easy but possible.\n",
        "\n",
        "For this project, I will be using [Google Colab](/https://colab.research.google.com/). For those who are unaware, it is a research project created to disseminate ML education and you can train your neural networks on their Tesla K80 GPUs. It sure speeds up the entire training process.\n"
      ]
    },
    {
      "metadata": {
        "id": "5FQpjWg4Ld6m",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 1.  Connect google Drive.\n",
        "\n",
        "Once connected, we will be able to access the FER2013 dataset.   "
      ]
    },
    {
      "metadata": {
        "id": "SyTEuhg7fR9P",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "7610bf99-8818-4db9-b927-5ab7dafa7abb"
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/gdrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "UQy7KUb6NBs-",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 2. Import Libraries."
      ]
    },
    {
      "metadata": {
        "id": "XRPQRtXGmm2W",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import numpy as np\n",
        "from keras.utils import to_categorical\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from keras.models import Sequential #Initialise our neural network model as a sequential network\n",
        "from keras.layers import Conv2D #Convolution operation\n",
        "from keras.layers.normalization import BatchNormalization\n",
        "from keras.regularizers import l2\n",
        "from keras.layers import Activation#Applies activation function\n",
        "from keras.layers import Dropout#Prevents overfitting by randomly converting few outputs to zero\n",
        "from keras.layers import MaxPooling2D # Maxpooling function\n",
        "from keras.layers import Flatten # Converting 2D arrays into a 1D linear vector\n",
        "from keras.layers import Dense # Regular fully connected neural network\n",
        "from keras import optimizers\n",
        "from keras.callbacks import ReduceLROnPlateau, EarlyStopping, TensorBoard, ModelCheckpoint\n",
        "from sklearn.metrics import accuracy_score"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "xpAjYytuPDaU",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 3. Define Data loading mechanism.\n",
        "\n",
        "the* load_data()* access the FER2013.csv dataset from Google Drive and then we start reading it line by line. Our dataset has 35,887 images, which are classified into seven categories -  Angry, Disgust, Fear, Happy, Sad, Surprsie, and Neutral. Let us have a look at the file.\n",
        "\n",
        "![Dataset overview](https://i.imgur.com/vqlEXw2.png) \n",
        "\n",
        "We are only going to need the Class and the Image Data. Image data is one big string of numbers exactly 2304 numbers. These are pixel intensity values and each number represents the darkness of that pixel in the image. It can take any value from 0 (White) to 255 (Black). We split the string to get hold of individual numbers and then reshape it into a 48 x 48 array and dividing by 255 normalizes the data. \n",
        "\n",
        "Once we go through all the images, we expand the dimension of our data array by one to accommodate for the channel value. We one-hot encode the labels then return the numpy arrays.     \n"
      ]
    },
    {
      "metadata": {
        "id": "H2XHQzzzrRyF",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def load_data(dataset_path):\n",
        "  \n",
        "  #classes = ['Angry', 'Disgust', 'Fear', 'Happy', 'Sad', 'Surprsie', 'Neutral']  #We will be dealing with seven different types of emotions.\n",
        "\n",
        "  data = []\n",
        "  test_data = []\n",
        "  test_labels = []\n",
        "  labels =[]\n",
        "\n",
        "  with open(dataset_path, 'r') as file:\n",
        "      for line_no, line in enumerate(file.readlines()):\n",
        "          if 0 < line_no <= 35887:\n",
        "            curr_class, line, set_type = line.split(',')\n",
        "            image_data = np.asarray([int(x) for x in line.split()]).reshape(48, 48)#Creating a list out of the string then converting it into a 2-Dimensional numpy array.\n",
        "            image_data =image_data.astype(np.uint8)/255.0\n",
        "            \n",
        "            if (set_type.strip() == 'PrivateTest'):\n",
        "              \n",
        "              test_data.append(image_data)\n",
        "              test_labels.append(curr_class)\n",
        "            else:\n",
        "              data.append(image_data)\n",
        "              labels.append(curr_class)\n",
        "      \n",
        "      test_data = np.expand_dims(test_data, -1)\n",
        "      test_labels = to_categorical(test_labels, num_classes = 7)\n",
        "      data = np.expand_dims(data, -1)   \n",
        "      labels = to_categorical(labels, num_classes = 7)\n",
        "    \n",
        "      return np.array(data), np.array(labels), np.array(test_data), np.array(test_labels)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "VFFZrnOgWOII",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 4. Load the data.\n"
      ]
    },
    {
      "metadata": {
        "id": "TpdOghEhxmsd",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        },
        "outputId": "e35f0d60-b20f-408b-bab3-fc1dada3abb2"
      },
      "cell_type": "code",
      "source": [
        "dataset_path = \"/content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Data/fer2013.csv\" \n",
        "train_data, train_labels, test_data, test_labels = load_data(dataset_path)\n",
        "#train_data, test_data, train_labels, test_labels = train_test_split(data, labels, test_size = test_size,random_state = seed)\n",
        "\n",
        "print(\"Number of images in Training set:\", len(train_data))\n",
        "print(\"Number of images in Test set:\", len(test_data))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of images in Training set: 32298\n",
            "Number of images in Test set: 3589\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Rne3AL0_WnF0",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 5. Define and Deploy the neural network.\n",
        "\n",
        "We have built ourselves a Convolutional Neural Network which has convolutional layers armed with filters that extract features out of images. \n",
        "\n",
        "Callback functions are those functions which are called at the end of every epoch and in this model, we use two callback functions - ReduceLROnPlateau and EarlyStopping. \n",
        "\n",
        "\n",
        "\n",
        "> **1. ReduceLROnPlateau:** monitors a certain variable, in this case, validation loss and alters the learning rate when the value stops significantly changing after some certain number of epochs (patience).\n",
        "\n",
        ">  **2. EarlyStopping:** At times our optimiser can land in *local optima* and get stuck there. There is no point in continuing the training as there won't be any further improvements. \n",
        "\n",
        "> **3. ModelCheckpoin**t: Saves the best version of our model along with the weights, so that in case any crash occurs, our model can be recovered.\n",
        "\n",
        "![Network Architecture](https://i.imgur.com/rG5p7JS.png)\n",
        "\n",
        "We have used Keras to deploy our neural network and if you look at the architecture then you will notice that we have used Dropout layers frequently. Dropout layers inhibit overfitting by randomly dropping out units from the neural network. We will use 20 percent of the training data as validation data.\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "F3VAm1A-0fpD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 8235
        },
        "outputId": "4c6adb60-f19c-40de-ff5c-60c25baf65ec"
      },
      "cell_type": "code",
      "source": [
        "#######HYPERPARAMATERS###########\n",
        "epochs = 100\n",
        "batch_size = 64\n",
        "learning_rate = 0.001\n",
        "#################################\n",
        "\n",
        "model = Sequential()\n",
        "    \n",
        "model.add(Conv2D(64, (3, 3), activation='relu', input_shape=(48, 48, 1), kernel_regularizer=l2(0.01)))\n",
        "model.add(Conv2D(64, (3, 3), padding='same',activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(MaxPooling2D(pool_size=(2,2), strides=(2, 2)))\n",
        "model.add(Dropout(0.5))\n",
        "    \n",
        "model.add(Conv2D(128, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(128, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(128, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.5))\n",
        "    \n",
        "model.add(Conv2D(256, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(256, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(256, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.5))\n",
        "    \n",
        "model.add(Conv2D(512, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(512, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Conv2D(512, (3, 3), padding='same', activation='relu'))\n",
        "model.add(BatchNormalization())\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.5))\n",
        "    \n",
        "model.add(Flatten())\n",
        "model.add(Dense(512, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(256, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(64, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(7, activation='softmax'))\n",
        "adam = optimizers.Adam(lr = learning_rate)\n",
        "model.compile(optimizer = adam, loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "    \n",
        "print(model.summary())\n",
        "\n",
        "lr_reducer = ReduceLROnPlateau(monitor='val_loss', factor=0.9, patience=3)\n",
        "early_stopper = EarlyStopping(monitor='val_acc', min_delta=0, patience=6, mode='auto')\n",
        "checkpointer = ModelCheckpoint('/content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5', monitor='val_loss', verbose=1, save_best_only=True)\n",
        "\n",
        "model.fit(\n",
        "          train_data,\n",
        "          train_labels,\n",
        "          epochs = epochs,\n",
        "          batch_size = batch_size,\n",
        "          validation_split = 0.2,\n",
        "          shuffle = True,\n",
        "          callbacks=[lr_reducer, checkpointer, early_stopper]\n",
        "          )"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_1 (Conv2D)            (None, 46, 46, 64)        640       \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 46, 46, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 46, 46, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 23, 23, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 23, 23, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 23, 23, 128)       73856     \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 23, 23, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 23, 23, 128)       147584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 23, 23, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 23, 23, 128)       147584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 23, 23, 128)       512       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 11, 11, 128)       0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 11, 11, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 11, 11, 256)       295168    \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 11, 11, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv2d_7 (Conv2D)            (None, 11, 11, 256)       590080    \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 11, 11, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv2d_8 (Conv2D)            (None, 11, 11, 256)       590080    \n",
            "_________________________________________________________________\n",
            "batch_normalization_7 (Batch (None, 11, 11, 256)       1024      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 5, 5, 256)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 5, 5, 256)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_9 (Conv2D)            (None, 5, 5, 512)         1180160   \n",
            "_________________________________________________________________\n",
            "batch_normalization_8 (Batch (None, 5, 5, 512)         2048      \n",
            "_________________________________________________________________\n",
            "conv2d_10 (Conv2D)           (None, 5, 5, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "batch_normalization_9 (Batch (None, 5, 5, 512)         2048      \n",
            "_________________________________________________________________\n",
            "conv2d_11 (Conv2D)           (None, 5, 5, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "batch_normalization_10 (Batc (None, 5, 5, 512)         2048      \n",
            "_________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2 (None, 2, 2, 512)         0         \n",
            "_________________________________________________________________\n",
            "dropout_4 (Dropout)          (None, 2, 2, 512)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 512)               1049088   \n",
            "_________________________________________________________________\n",
            "dropout_5 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 256)               131328    \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 128)               32896     \n",
            "_________________________________________________________________\n",
            "dropout_7 (Dropout)          (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_4 (Dense)              (None, 64)                8256      \n",
            "_________________________________________________________________\n",
            "dropout_8 (Dropout)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 7)                 455       \n",
            "=================================================================\n",
            "Total params: 9,014,727\n",
            "Trainable params: 9,009,223\n",
            "Non-trainable params: 5,504\n",
            "_________________________________________________________________\n",
            "None\n",
            "Train on 25838 samples, validate on 6460 samples\n",
            "Epoch 1/100\n",
            "25838/25838 [==============================] - 81s 3ms/step - loss: 2.1203 - acc: 0.1996 - val_loss: 1.8632 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 1.86322, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 2/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8604 - acc: 0.2341 - val_loss: 1.8373 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00002: val_loss improved from 1.86322 to 1.83728, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 3/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8411 - acc: 0.2453 - val_loss: 1.8307 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00003: val_loss improved from 1.83728 to 1.83073, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 4/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8316 - acc: 0.2491 - val_loss: 1.8241 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00004: val_loss improved from 1.83073 to 1.82414, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 5/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8253 - acc: 0.2502 - val_loss: 1.8181 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00005: val_loss improved from 1.82414 to 1.81807, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 6/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8205 - acc: 0.2508 - val_loss: 1.8182 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00006: val_loss did not improve from 1.81807\n",
            "Epoch 7/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.8154 - acc: 0.2511 - val_loss: 1.8051 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00007: val_loss improved from 1.81807 to 1.80506, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 8/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8106 - acc: 0.2534 - val_loss: 1.8087 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 1.80506\n",
            "Epoch 9/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.8083 - acc: 0.2547 - val_loss: 1.8026 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00009: val_loss improved from 1.80506 to 1.80262, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 10/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8060 - acc: 0.2526 - val_loss: 1.7916 - val_acc: 0.2497\n",
            "\n",
            "Epoch 00010: val_loss improved from 1.80262 to 1.79161, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 11/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.8031 - acc: 0.2526 - val_loss: 1.7996 - val_acc: 0.2488\n",
            "\n",
            "Epoch 00011: val_loss did not improve from 1.79161\n",
            "Epoch 12/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.7971 - acc: 0.2540 - val_loss: 1.7768 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00012: val_loss improved from 1.79161 to 1.77684, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 13/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.7929 - acc: 0.2551 - val_loss: 1.9112 - val_acc: 0.2489\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 1.77684\n",
            "Epoch 14/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.7664 - acc: 0.2720 - val_loss: 1.7457 - val_acc: 0.2854\n",
            "\n",
            "Epoch 00014: val_loss improved from 1.77684 to 1.74573, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 15/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.7270 - acc: 0.2956 - val_loss: 1.6799 - val_acc: 0.3116\n",
            "\n",
            "Epoch 00015: val_loss improved from 1.74573 to 1.67989, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 16/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.6855 - acc: 0.3106 - val_loss: 1.7334 - val_acc: 0.2690\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 1.67989\n",
            "Epoch 17/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.6343 - acc: 0.3303 - val_loss: 1.5791 - val_acc: 0.3822\n",
            "\n",
            "Epoch 00017: val_loss improved from 1.67989 to 1.57914, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 18/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.5851 - acc: 0.3654 - val_loss: 1.6363 - val_acc: 0.3441\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 1.57914\n",
            "Epoch 19/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.5502 - acc: 0.3827 - val_loss: 1.5124 - val_acc: 0.3831\n",
            "\n",
            "Epoch 00019: val_loss improved from 1.57914 to 1.51241, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 20/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.5147 - acc: 0.3973 - val_loss: 1.4550 - val_acc: 0.4283\n",
            "\n",
            "Epoch 00020: val_loss improved from 1.51241 to 1.45500, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 21/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.4766 - acc: 0.4110 - val_loss: 1.4805 - val_acc: 0.3884\n",
            "\n",
            "Epoch 00021: val_loss did not improve from 1.45500\n",
            "Epoch 22/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.4585 - acc: 0.4205 - val_loss: 1.4034 - val_acc: 0.4317\n",
            "\n",
            "Epoch 00022: val_loss improved from 1.45500 to 1.40341, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 23/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.4316 - acc: 0.4272 - val_loss: 1.3723 - val_acc: 0.4416\n",
            "\n",
            "Epoch 00023: val_loss improved from 1.40341 to 1.37234, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 24/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.4095 - acc: 0.4347 - val_loss: 1.3896 - val_acc: 0.4458\n",
            "\n",
            "Epoch 00024: val_loss did not improve from 1.37234\n",
            "Epoch 25/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.3893 - acc: 0.4411 - val_loss: 1.4201 - val_acc: 0.4022\n",
            "\n",
            "Epoch 00025: val_loss did not improve from 1.37234\n",
            "Epoch 26/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.3725 - acc: 0.4476 - val_loss: 1.3892 - val_acc: 0.4249\n",
            "\n",
            "Epoch 00026: val_loss did not improve from 1.37234\n",
            "Epoch 27/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.3389 - acc: 0.4520 - val_loss: 1.3851 - val_acc: 0.4341\n",
            "\n",
            "Epoch 00027: val_loss did not improve from 1.37234\n",
            "Epoch 28/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.3283 - acc: 0.4602 - val_loss: 1.3637 - val_acc: 0.4519\n",
            "\n",
            "Epoch 00028: val_loss improved from 1.37234 to 1.36367, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 29/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.3166 - acc: 0.4658 - val_loss: 1.3062 - val_acc: 0.4663\n",
            "\n",
            "Epoch 00029: val_loss improved from 1.36367 to 1.30616, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 30/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.2965 - acc: 0.4753 - val_loss: 1.3092 - val_acc: 0.4745\n",
            "\n",
            "Epoch 00030: val_loss did not improve from 1.30616\n",
            "Epoch 31/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.2714 - acc: 0.4883 - val_loss: 1.2831 - val_acc: 0.4983\n",
            "\n",
            "Epoch 00031: val_loss improved from 1.30616 to 1.28308, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 32/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.2644 - acc: 0.4947 - val_loss: 1.2813 - val_acc: 0.5043\n",
            "\n",
            "Epoch 00032: val_loss improved from 1.28308 to 1.28133, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 33/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.2437 - acc: 0.5096 - val_loss: 1.3171 - val_acc: 0.4916\n",
            "\n",
            "Epoch 00033: val_loss did not improve from 1.28133\n",
            "Epoch 34/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.2233 - acc: 0.5218 - val_loss: 1.2500 - val_acc: 0.5138\n",
            "\n",
            "Epoch 00034: val_loss improved from 1.28133 to 1.25005, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 35/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.2040 - acc: 0.5387 - val_loss: 1.2903 - val_acc: 0.4964\n",
            "\n",
            "Epoch 00035: val_loss did not improve from 1.25005\n",
            "Epoch 36/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.1847 - acc: 0.5470 - val_loss: 1.3026 - val_acc: 0.5193\n",
            "\n",
            "Epoch 00036: val_loss did not improve from 1.25005\n",
            "Epoch 37/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.1584 - acc: 0.5594 - val_loss: 1.1968 - val_acc: 0.5372\n",
            "\n",
            "Epoch 00037: val_loss improved from 1.25005 to 1.19675, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 38/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.1414 - acc: 0.5624 - val_loss: 1.2566 - val_acc: 0.5204\n",
            "\n",
            "Epoch 00038: val_loss did not improve from 1.19675\n",
            "Epoch 39/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.1191 - acc: 0.5717 - val_loss: 1.1729 - val_acc: 0.5675\n",
            "\n",
            "Epoch 00039: val_loss improved from 1.19675 to 1.17291, saving model to /content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/Model/weights.hd5\n",
            "Epoch 40/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.0922 - acc: 0.5848 - val_loss: 1.2121 - val_acc: 0.5515\n",
            "\n",
            "Epoch 00040: val_loss did not improve from 1.17291\n",
            "Epoch 41/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.0801 - acc: 0.5878 - val_loss: 1.2705 - val_acc: 0.5341\n",
            "\n",
            "Epoch 00041: val_loss did not improve from 1.17291\n",
            "Epoch 42/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.0705 - acc: 0.5975 - val_loss: 1.1888 - val_acc: 0.5613\n",
            "\n",
            "Epoch 00042: val_loss did not improve from 1.17291\n",
            "Epoch 43/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 1.0274 - acc: 0.6110 - val_loss: 1.1911 - val_acc: 0.5543\n",
            "\n",
            "Epoch 00043: val_loss did not improve from 1.17291\n",
            "Epoch 44/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 1.0098 - acc: 0.6236 - val_loss: 1.1846 - val_acc: 0.5831\n",
            "\n",
            "Epoch 00044: val_loss did not improve from 1.17291\n",
            "Epoch 45/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.9915 - acc: 0.6273 - val_loss: 1.1845 - val_acc: 0.5794\n",
            "\n",
            "Epoch 00045: val_loss did not improve from 1.17291\n",
            "Epoch 46/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.9610 - acc: 0.6414 - val_loss: 1.2226 - val_acc: 0.5760\n",
            "\n",
            "Epoch 00046: val_loss did not improve from 1.17291\n",
            "Epoch 47/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.9421 - acc: 0.6485 - val_loss: 1.1779 - val_acc: 0.5881\n",
            "\n",
            "Epoch 00047: val_loss did not improve from 1.17291\n",
            "Epoch 48/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.9373 - acc: 0.6543 - val_loss: 1.1815 - val_acc: 0.5825\n",
            "\n",
            "Epoch 00048: val_loss did not improve from 1.17291\n",
            "Epoch 49/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.9100 - acc: 0.6643 - val_loss: 1.2371 - val_acc: 0.5865\n",
            "\n",
            "Epoch 00049: val_loss did not improve from 1.17291\n",
            "Epoch 50/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.8927 - acc: 0.6733 - val_loss: 1.2138 - val_acc: 0.5794\n",
            "\n",
            "Epoch 00050: val_loss did not improve from 1.17291\n",
            "Epoch 51/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.8666 - acc: 0.6797 - val_loss: 1.2533 - val_acc: 0.5940\n",
            "\n",
            "Epoch 00051: val_loss did not improve from 1.17291\n",
            "Epoch 52/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.8459 - acc: 0.6930 - val_loss: 1.2457 - val_acc: 0.5918\n",
            "\n",
            "Epoch 00052: val_loss did not improve from 1.17291\n",
            "Epoch 53/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.8250 - acc: 0.7020 - val_loss: 1.2572 - val_acc: 0.5960\n",
            "\n",
            "Epoch 00053: val_loss did not improve from 1.17291\n",
            "Epoch 54/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.8164 - acc: 0.7045 - val_loss: 1.2591 - val_acc: 0.6046\n",
            "\n",
            "Epoch 00054: val_loss did not improve from 1.17291\n",
            "Epoch 55/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.7892 - acc: 0.7100 - val_loss: 1.2804 - val_acc: 0.6022\n",
            "\n",
            "Epoch 00055: val_loss did not improve from 1.17291\n",
            "Epoch 56/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.7681 - acc: 0.7258 - val_loss: 1.3362 - val_acc: 0.5977\n",
            "\n",
            "Epoch 00056: val_loss did not improve from 1.17291\n",
            "Epoch 57/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.7504 - acc: 0.7309 - val_loss: 1.2753 - val_acc: 0.6051\n",
            "\n",
            "Epoch 00057: val_loss did not improve from 1.17291\n",
            "Epoch 58/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.7356 - acc: 0.7350 - val_loss: 1.2741 - val_acc: 0.6128\n",
            "\n",
            "Epoch 00058: val_loss did not improve from 1.17291\n",
            "Epoch 59/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.7153 - acc: 0.7440 - val_loss: 1.3251 - val_acc: 0.6025\n",
            "\n",
            "Epoch 00059: val_loss did not improve from 1.17291\n",
            "Epoch 60/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.6997 - acc: 0.7497 - val_loss: 1.2691 - val_acc: 0.6031\n",
            "\n",
            "Epoch 00060: val_loss did not improve from 1.17291\n",
            "Epoch 61/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.6826 - acc: 0.7591 - val_loss: 1.3048 - val_acc: 0.6042\n",
            "\n",
            "Epoch 00061: val_loss did not improve from 1.17291\n",
            "Epoch 62/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.6571 - acc: 0.7672 - val_loss: 1.3361 - val_acc: 0.6099\n",
            "\n",
            "Epoch 00062: val_loss did not improve from 1.17291\n",
            "Epoch 63/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.6507 - acc: 0.7717 - val_loss: 1.4147 - val_acc: 0.5994\n",
            "\n",
            "Epoch 00063: val_loss did not improve from 1.17291\n",
            "Epoch 64/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.6296 - acc: 0.7799 - val_loss: 1.4144 - val_acc: 0.6173\n",
            "\n",
            "Epoch 00064: val_loss did not improve from 1.17291\n",
            "Epoch 65/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.6034 - acc: 0.7902 - val_loss: 1.4474 - val_acc: 0.6132\n",
            "\n",
            "Epoch 00065: val_loss did not improve from 1.17291\n",
            "Epoch 66/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.5930 - acc: 0.7927 - val_loss: 1.4521 - val_acc: 0.6098\n",
            "\n",
            "Epoch 00066: val_loss did not improve from 1.17291\n",
            "Epoch 67/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.5894 - acc: 0.7915 - val_loss: 1.4289 - val_acc: 0.6193\n",
            "\n",
            "Epoch 00067: val_loss did not improve from 1.17291\n",
            "Epoch 68/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.5699 - acc: 0.8018 - val_loss: 1.4690 - val_acc: 0.6122\n",
            "\n",
            "Epoch 00068: val_loss did not improve from 1.17291\n",
            "Epoch 69/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.5569 - acc: 0.8042 - val_loss: 1.4637 - val_acc: 0.6164\n",
            "\n",
            "Epoch 00069: val_loss did not improve from 1.17291\n",
            "Epoch 70/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.5405 - acc: 0.8119 - val_loss: 1.5473 - val_acc: 0.6195\n",
            "\n",
            "Epoch 00070: val_loss did not improve from 1.17291\n",
            "Epoch 71/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.5275 - acc: 0.8203 - val_loss: 1.4593 - val_acc: 0.6090\n",
            "\n",
            "Epoch 00071: val_loss did not improve from 1.17291\n",
            "Epoch 72/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.5215 - acc: 0.8245 - val_loss: 1.6198 - val_acc: 0.6068\n",
            "\n",
            "Epoch 00072: val_loss did not improve from 1.17291\n",
            "Epoch 73/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.5141 - acc: 0.8234 - val_loss: 1.5702 - val_acc: 0.6119\n",
            "\n",
            "Epoch 00073: val_loss did not improve from 1.17291\n",
            "Epoch 74/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.5007 - acc: 0.8305 - val_loss: 1.5547 - val_acc: 0.6164\n",
            "\n",
            "Epoch 00074: val_loss did not improve from 1.17291\n",
            "Epoch 75/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4870 - acc: 0.8353 - val_loss: 1.5858 - val_acc: 0.6158\n",
            "\n",
            "Epoch 00075: val_loss did not improve from 1.17291\n",
            "Epoch 76/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4682 - acc: 0.8399 - val_loss: 1.6442 - val_acc: 0.6138\n",
            "\n",
            "Epoch 00076: val_loss did not improve from 1.17291\n",
            "Epoch 77/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.4640 - acc: 0.8448 - val_loss: 1.6104 - val_acc: 0.6176\n",
            "\n",
            "Epoch 00077: val_loss did not improve from 1.17291\n",
            "Epoch 78/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4590 - acc: 0.8465 - val_loss: 1.6643 - val_acc: 0.6217\n",
            "\n",
            "Epoch 00078: val_loss did not improve from 1.17291\n",
            "Epoch 79/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4414 - acc: 0.8511 - val_loss: 1.6637 - val_acc: 0.6159\n",
            "\n",
            "Epoch 00079: val_loss did not improve from 1.17291\n",
            "Epoch 80/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.4320 - acc: 0.8564 - val_loss: 1.6958 - val_acc: 0.6211\n",
            "\n",
            "Epoch 00080: val_loss did not improve from 1.17291\n",
            "Epoch 81/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.4267 - acc: 0.8572 - val_loss: 1.6776 - val_acc: 0.6203\n",
            "\n",
            "Epoch 00081: val_loss did not improve from 1.17291\n",
            "Epoch 82/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4169 - acc: 0.8612 - val_loss: 1.7085 - val_acc: 0.6198\n",
            "\n",
            "Epoch 00082: val_loss did not improve from 1.17291\n",
            "Epoch 83/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4133 - acc: 0.8635 - val_loss: 1.7577 - val_acc: 0.6234\n",
            "\n",
            "Epoch 00083: val_loss did not improve from 1.17291\n",
            "Epoch 84/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.4043 - acc: 0.8656 - val_loss: 1.7455 - val_acc: 0.6170\n",
            "\n",
            "Epoch 00084: val_loss did not improve from 1.17291\n",
            "Epoch 85/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.3900 - acc: 0.8736 - val_loss: 1.7401 - val_acc: 0.6200\n",
            "\n",
            "Epoch 00085: val_loss did not improve from 1.17291\n",
            "Epoch 86/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3846 - acc: 0.8735 - val_loss: 1.7579 - val_acc: 0.6198\n",
            "\n",
            "Epoch 00086: val_loss did not improve from 1.17291\n",
            "Epoch 87/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3785 - acc: 0.8743 - val_loss: 1.7705 - val_acc: 0.6263\n",
            "\n",
            "Epoch 00087: val_loss did not improve from 1.17291\n",
            "Epoch 88/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.3690 - acc: 0.8784 - val_loss: 1.8005 - val_acc: 0.6212\n",
            "\n",
            "Epoch 00088: val_loss did not improve from 1.17291\n",
            "Epoch 89/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3641 - acc: 0.8787 - val_loss: 1.8083 - val_acc: 0.6204\n",
            "\n",
            "Epoch 00089: val_loss did not improve from 1.17291\n",
            "Epoch 90/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3559 - acc: 0.8828 - val_loss: 1.8132 - val_acc: 0.6178\n",
            "\n",
            "Epoch 00090: val_loss did not improve from 1.17291\n",
            "Epoch 91/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3491 - acc: 0.8851 - val_loss: 1.8226 - val_acc: 0.6193\n",
            "\n",
            "Epoch 00091: val_loss did not improve from 1.17291\n",
            "Epoch 92/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3431 - acc: 0.8874 - val_loss: 1.8359 - val_acc: 0.6243\n",
            "\n",
            "Epoch 00092: val_loss did not improve from 1.17291\n",
            "Epoch 93/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3399 - acc: 0.8922 - val_loss: 1.8492 - val_acc: 0.6262\n",
            "\n",
            "Epoch 00093: val_loss did not improve from 1.17291\n",
            "Epoch 94/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.3336 - acc: 0.8914 - val_loss: 1.8657 - val_acc: 0.6226\n",
            "\n",
            "Epoch 00094: val_loss did not improve from 1.17291\n",
            "Epoch 95/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3275 - acc: 0.8944 - val_loss: 1.8535 - val_acc: 0.6212\n",
            "\n",
            "Epoch 00095: val_loss did not improve from 1.17291\n",
            "Epoch 96/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3258 - acc: 0.8928 - val_loss: 1.8659 - val_acc: 0.6215\n",
            "\n",
            "Epoch 00096: val_loss did not improve from 1.17291\n",
            "Epoch 97/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3242 - acc: 0.8955 - val_loss: 1.8615 - val_acc: 0.6226\n",
            "\n",
            "Epoch 00097: val_loss did not improve from 1.17291\n",
            "Epoch 98/100\n",
            "25838/25838 [==============================] - 75s 3ms/step - loss: 0.3144 - acc: 0.8955 - val_loss: 1.9016 - val_acc: 0.6203\n",
            "\n",
            "Epoch 00098: val_loss did not improve from 1.17291\n",
            "Epoch 99/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3121 - acc: 0.8989 - val_loss: 1.9755 - val_acc: 0.6226\n",
            "\n",
            "Epoch 00099: val_loss did not improve from 1.17291\n",
            "Epoch 100/100\n",
            "25838/25838 [==============================] - 76s 3ms/step - loss: 0.3057 - acc: 0.9025 - val_loss: 1.9594 - val_acc: 0.6243\n",
            "\n",
            "Epoch 00100: val_loss did not improve from 1.17291\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fd08febc898>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "metadata": {
        "id": "TD8ppILhFUrz",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 6. Test the model.\n",
        "\n",
        "We trained the neural network and we achieved the highest validation accuracy of 62.43%. Now, we will use the test data to check how well our model generalizes. \n",
        "\n",
        "We score an astounding 63.17% on the test set. Though we barely, made it to the first division, we still made it to the 9th rank of the Kaggle competition."
      ]
    },
    {
      "metadata": {
        "id": "OzhQw9NOWjl3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8e086434-a07d-4aad-f173-69dad5d71f32"
      },
      "cell_type": "code",
      "source": [
        "predicted_test_labels = np.argmax(model.predict(test_data), axis=1)\n",
        "test_labels = np.argmax(test_labels, axis=1)\n",
        "print (\"Accuracy score = \", accuracy_score(test_labels, predicted_test_labels))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy score =  0.6316522708275285\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "CGsta6kKHr82",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 7. Save the model."
      ]
    },
    {
      "metadata": {
        "id": "uiMswpPsr4KK",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "cd8fba79-a744-45fc-add0-b46daeca7b44"
      },
      "cell_type": "code",
      "source": [
        "from keras.models import model_from_json\n",
        "model_json = model.to_json()\n",
        "with open(\"/content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/model.json\", \"w\") as json_file:\n",
        "    json_file.write(model_json)\n",
        "# serialize weights to HDF5\n",
        "model.save_weights(\"/content/gdrive/My Drive/Colab Notebooks/Emotion Recognition/model.h5\")\n",
        "print(\"Saved model to disk\")"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "9tQb_TMMH4sb",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 8. Up next?\n",
        "\n",
        "Our neural network is trained and we are ready to use it in an application. In the accompanying IPython Notebook, we will notice an application which analyzes your webcam feed and identifies faces and then recognizes the emotions.\n",
        "\n"
      ]
    }
  ]
}