419Project.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "419Project.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/liyujiel/419project/blob/master/419Project.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "Yrd5GjAt97RZ",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# CMPT419 Project"
      ]
    },
    {
      "metadata": {
        "id": "e2b9tYJSoJmM",
        "colab_type": "code",
        "outputId": "6d43c484-73e3-4218-cde6-8bb50cd617d2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "rtpM9euFpf2M",
        "outputId": "bd447b5d-3c83-4168-9ed2-3c04125be65a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "66NZWBbLErsq",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### import libs"
      ]
    },
    {
      "metadata": {
        "id": "PW7WMcAjxpZ6",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import keras\n",
        "from keras.models import Sequential, model_from_json, Input, Model\n",
        "from keras.layers import Average, MaxPool2D\n",
        "from keras.layers.core import Dense, Dropout, Activation, Flatten\n",
        "from keras.callbacks import EarlyStopping,ModelCheckpoint, TensorBoard\n",
        "from keras.layers.convolutional import Conv2D, MaxPooling2D, ZeroPadding2D, AveragePooling2D\n",
        "from keras.layers.normalization import BatchNormalization\n",
        "from keras.regularizers import l2\n",
        "\n",
        "from keras.preprocessing import image\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "gOFq3LWyEvht",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))\n",
        "sess.run(tf.global_variables_initializer())\n",
        "keras.backend.set_session(sess)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "KPSuG_nc-71h",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Baseline"
      ]
    },
    {
      "metadata": {
        "id": "kN9fDoAHu4HN",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Gender model"
      ]
    },
    {
      "metadata": {
        "id": "FRm7_5ZYHbQM",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "gender_model.load_weights(\"/content/gdrive/My Drive/gender_mini_XCEPTION.21-0.95.hdf5\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "3F7uIDzJLWMp",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Pre-trained model"
      ]
    },
    {
      "metadata": {
        "id": "hXCJxgc3MmD4",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def baseline_model_saved(model_name):\n",
        "    #load json and create model\n",
        "    json_file = open('{}.json'.format(model_name), 'r')\n",
        "    loaded_model_json = json_file.read()\n",
        "    json_file.close()\n",
        "    model = model_from_json(loaded_model_json)\n",
        "    #load weights from h5 file\n",
        "    model.load_weights(\"{}.h5\".format(model_name))\n",
        "    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[\"accuracy\"])\n",
        "    return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Ai5kdBlOLVJQ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "pre_train_f = baseline_model_saved(\"/content/gdrive/My Drive/model_2layer_2_2_pool\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5riAP5aHLfbh",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "pre_train_s = baseline_model_saved(\"/content/gdrive/My Drive/model_4layer_2_2_pool\") # load pre_train model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_7u9ywnQWKz3",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from keras.utils import plot_model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "woMmdCZDWP7M",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# https://pypi.python.org/pypi/pydot\n",
        "!apt-get -qq install -y graphviz && pip install -q pydot\n",
        "import pydot"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "73mN36sDXCG1",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "!pip install -q pydot\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "74eqavNdbc3X",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1020
        },
        "outputId": "c2b69e59-64aa-4947-a738-165cd79dfed6"
      },
      "cell_type": "code",
      "source": [
        "model = Sequential()\n",
        "\n",
        "model.add(Conv2D(64, 3, data_format=\"channels_last\", kernel_initializer=\"he_normal\", \n",
        "                 input_shape=(48, 48, 1)))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Conv2D(64, 3))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(MaxPool2D(pool_size=(2, 2), strides=2))\n",
        "model.add(Dropout(0.6))\n",
        "\n",
        "model.add(Conv2D(32, 3))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Conv2D(32, 3))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Conv2D(32, 3))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(MaxPool2D(pool_size=(2, 2), strides=2))\n",
        "model.add(Dropout(0.6))\n",
        "\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128))\n",
        "model.add(BatchNormalization())\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(Dropout(0.6))\n",
        "\n",
        "model.add(Dense(7))\n",
        "model.add(Activation('softmax'))\n",
        "\n",
        "#model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "model.compile(loss='squared_hinge', optimizer='adam', metrics=['accuracy'])\n",
        "model.summary()"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_3 (Conv2D)            (None, 46, 46, 64)        640       \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 46, 46, 64)        256       \n",
            "_________________________________________________________________\n",
            "activation_3 (Activation)    (None, 46, 46, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 44, 44, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 44, 44, 64)        256       \n",
            "_________________________________________________________________\n",
            "activation_4 (Activation)    (None, 44, 44, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 22, 22, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 22, 22, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 20, 20, 32)        18464     \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 20, 20, 32)        128       \n",
            "_________________________________________________________________\n",
            "activation_5 (Activation)    (None, 20, 20, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 18, 18, 32)        9248      \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 18, 18, 32)        128       \n",
            "_________________________________________________________________\n",
            "activation_6 (Activation)    (None, 18, 18, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_7 (Conv2D)            (None, 16, 16, 32)        9248      \n",
            "_________________________________________________________________\n",
            "batch_normalization_7 (Batch (None, 16, 16, 32)        128       \n",
            "_________________________________________________________________\n",
            "activation_7 (Activation)    (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 8, 8, 32)          0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 8, 8, 32)          0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 128)               262272    \n",
            "_________________________________________________________________\n",
            "batch_normalization_8 (Batch (None, 128)               512       \n",
            "_________________________________________________________________\n",
            "activation_8 (Activation)    (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 7)                 903       \n",
            "_________________________________________________________________\n",
            "activation_9 (Activation)    (None, 7)                 0         \n",
            "=================================================================\n",
            "Total params: 339,111\n",
            "Trainable params: 338,407\n",
            "Non-trainable params: 704\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "CgY06oH5WGjb",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "plot_model(model, to_file='model.png', show_shapes=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "UljnTFHOVlUU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1156
        },
        "outputId": "7626a4d0-24a0-4d93-acaf-3ce534adc46c"
      },
      "cell_type": "code",
      "source": [
        "pre_train_s.summary()"
      ],
      "execution_count": 74,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_1 (Conv2D)            (None, 48, 48, 64)        640       \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 48, 48, 64)        256       \n",
            "_________________________________________________________________\n",
            "activation_1 (Activation)    (None, 48, 48, 64)        0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 24, 24, 128)       204928    \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 24, 24, 128)       512       \n",
            "_________________________________________________________________\n",
            "activation_2 (Activation)    (None, 24, 24, 128)       0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 12, 12, 128)       0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 12, 12, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 12, 12, 512)       590336    \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 12, 12, 512)       2048      \n",
            "_________________________________________________________________\n",
            "activation_3 (Activation)    (None, 12, 12, 512)       0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 6, 6, 512)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 6, 6, 512)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 6, 6, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 6, 6, 512)         2048      \n",
            "_________________________________________________________________\n",
            "activation_4 (Activation)    (None, 6, 6, 512)         0         \n",
            "_________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2 (None, 3, 3, 512)         0         \n",
            "_________________________________________________________________\n",
            "dropout_4 (Dropout)          (None, 3, 3, 512)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 4608)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 256)               1179904   \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 256)               1024      \n",
            "_________________________________________________________________\n",
            "activation_5 (Activation)    (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dropout_5 (Dropout)          (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 512)               131584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 512)               2048      \n",
            "_________________________________________________________________\n",
            "activation_6 (Activation)    (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 7)                 3591      \n",
            "=================================================================\n",
            "Total params: 4,478,727\n",
            "Trainable params: 4,474,759\n",
            "Non-trainable params: 3,968\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "W0L3MVIDE0K5",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Emotion model"
      ]
    },
    {
      "metadata": {
        "id": "kzHrdC0C-K4y",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "num_classes = 7 #angry, disgust, fear, happy, sad, surprise, neutral\n",
        "labels = [\"angry\", \"disgust\", \"fear\", \"happy\", \"sad\", \"surprise\", \"neutral\"]\n",
        "batch_size = 256\n",
        "epochs = 15"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "AbQHooVq-K1p",
        "colab_type": "code",
        "outputId": "8785b016-1f34-4dd3-d3fb-15b1870ced9e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "cell_type": "code",
      "source": [
        "with open(\"/content/gdrive/My Drive/fer2013.csv\") as f:\n",
        "    content = f.readlines()\n",
        "\n",
        "lines = np.array(content) \n",
        "\n",
        "num_of_instances = lines.size\n",
        "print(\"number of instances: \",num_of_instances)\n",
        "print(\"instance length: \",len(lines[1].split(\",\")[1].split(\" \")))"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "number of instances:  35888\n",
            "instance length:  2304\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "3nYNuVJd_9kj",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### initialize trainset and test set"
      ]
    },
    {
      "metadata": {
        "id": "GB42iuQi_69a",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "x_train, y_train, x_test, y_test, x_valid, y_valid = [], [], [], [], [], []"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "bs3qsIF8_67D",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for i in range(1,num_of_instances):\n",
        "    try:\n",
        "        emotion, img, usage = lines[i].split(\",\")\n",
        "          \n",
        "        val = img.split(\" \")\n",
        "        pixels = np.array(val, 'float32')\n",
        "\n",
        "        emotion = keras.utils.to_categorical(emotion, num_classes)\n",
        "\n",
        "        if 'Training' in usage:\n",
        "            y_train.append(emotion)\n",
        "            x_train.append(pixels)\n",
        "        elif 'PublicTest' in usage:\n",
        "            y_test.append(emotion)\n",
        "            x_test.append(pixels)\n",
        "        elif 'PrivateTest' in usage:\n",
        "            y_valid.append(emotion)\n",
        "            x_valid.append(pixels)\n",
        "\n",
        "    except:\n",
        "    \tprint(\"\",end=\"\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "25Bd5QPK_66N",
        "colab_type": "code",
        "outputId": "6e1c1e2a-1b7d-4e8b-b1f7-e07e41d207a3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "cell_type": "code",
      "source": [
        "x_train = np.array(x_train, 'float32')\n",
        "y_train = np.array(y_train, 'float32')\n",
        "x_test = np.array(x_test, 'float32')\n",
        "y_test = np.array(y_test, 'float32')\n",
        "x_valid = np.array(x_valid, 'float32')\n",
        "y_valid = np.array(y_valid, 'float32')\n",
        "\n",
        "x_train /= 255 #normalize inputs between [0, 1]\n",
        "x_test /= 255\n",
        "x_valid /= 255\n",
        "\n",
        "x_train = x_train.reshape(x_train.shape[0], 48, 48, 1)\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.reshape(x_test.shape[0], 48, 48, 1)\n",
        "x_test = x_test.astype('float32')\n",
        "x_valid = x_valid.reshape(x_valid.shape[0], 48, 48, 1)\n",
        "x_valid = x_valid.astype('float32')\n",
        "\n",
        "print(x_train.shape[0], 'train samples')\n",
        "print(x_test.shape[0], 'test samples')\n",
        "print(x_valid.shape[0], 'valid samples')"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "28709 train samples\n",
            "3589 test samples\n",
            "3589 valid samples\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Ts7cuAc4P2L3",
        "colab_type": "code",
        "outputId": "ee6301d9-5a42-439c-b757-012d0263e5b6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "print(x_train.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(28709, 48, 48, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "rAYQ_INUP7TV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "input_shape = x_train[0,:,:,:].shape"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "8Zb388UlP8gQ",
        "colab_type": "code",
        "outputId": "ad85c32c-cbf3-48d2-b179-66c446531be4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "input_shape"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(48, 48, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 66
        }
      ]
    },
    {
      "metadata": {
        "id": "xS9TgnqANxaa",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model_input = Input(shape=input_shape)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "vmLnozk2E_UX",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# m_x_train, m_y_train, m_x_test, m_y_test, m_x_valid, m_y_valid = [], [], [], [], [], []"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mspS05TZE_JA",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# f_x_train, f_y_train, f_x_test, f_y_test, f_x_valid, f_y_valid = [], [], [], [], [], []"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "VQW_-cg9Pios",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# import cv2"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mkqbK_vSE_FJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for i in range(0,len(x_train)):\n",
        "    per_x_train = x_train[i]\n",
        "    per_y_train = y_train[i]\n",
        "    per_x_train = cv2.resize(per_x_train,(64,64),1)\n",
        "    per_x_train = per_x_train.reshape(1, 64, 64, -1)\n",
        "    print(gender_model.predict(per_x_train))\n",
        "    if gender_model.predict(per_x_train)[0][0] >= 0.5:\n",
        "        m_x_train.append(per_x_train)\n",
        "    else:\n",
        "        f_x_train.append(per_x_train)   "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "skiawFPjFzG8",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for per_y_train in y_train:\n",
        "    if gender_model.predict(per_y_train)[0] == 1:\n",
        "        m_y_train.append(per_y_train)\n",
        "    else:\n",
        "        f_y_train.append(per_y_train)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "KmYoRDVkFyj0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for per_x_test in x_test:\n",
        "    if gender_model.predict(per_x_test)[0] == 1:\n",
        "        m_x_test.append(per_x_test)\n",
        "    else:\n",
        "        f_x_test.append(per_x_test)  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "NPu2ZlQLFyfO",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for per_y_test in y_test:\n",
        "    if gender_model.predict(per_y_test)[0] == 1:\n",
        "        m_y_test.append(per_y_test)\n",
        "    else:\n",
        "        f_y_test.append(per_y_test)  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5nSP1LDMGhYB",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for per_x_valid in x_valid:\n",
        "    if gender_model.predict(per_x_valid)[0] == 1:\n",
        "        m_x_valid.append(per_x_valid)\n",
        "    else:\n",
        "        f_x_valid.append(per_x_valid)  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "f6zCMibbGhUZ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for per_y_valid in y_valid:\n",
        "    if gender_model.predict(per_y_valid)[0] == 1:\n",
        "        m_y_valid.append(per_y_valid)\n",
        "    else:\n",
        "        f_y_valid.append(per_y_valid)  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "s4dRC1i8G1k2",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "m_x_train = np.array(m_x_train, 'float32')\n",
        "m_y_train = np.array(m_y_train, 'float32')\n",
        "m_x_test = np.array(m_x_test, 'float32')\n",
        "m_y_test = np.array(m_y_test, 'float32')\n",
        "m_x_valid = np.array(m_x_valid, 'float32')\n",
        "m_y_valid = np.array(m_y_valid, 'float32')\n",
        "\n",
        "f_x_train = np.array(f_x_train, 'float32')\n",
        "f_y_train = np.array(f_y_train, 'float32')\n",
        "f_x_test = np.array(f_x_test, 'float32')\n",
        "f_y_test = np.array(f_y_test, 'float32')\n",
        "f_x_valid = np.array(f_x_valid, 'float32')\n",
        "f_y_valid = np.array(f_y_valid, 'float32')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "kG6h1Ws4nQqU",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "valid = (x_valid, y_valid)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Q6jDxoV52VS7",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Baseline Net"
      ]
    },
    {
      "metadata": {
        "id": "4vORZLnX_62d",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def base_cnn_model(img_shape=(48, 48, 1), n_classes=7):\n",
        "    #construct CNN structure\n",
        "    model = Sequential()\n",
        "\n",
        "    #1st convolution layer\n",
        "    model.add(Conv2D(64, (5, 5), activation='relu', input_shape=img_shape))\n",
        "    model.add(MaxPooling2D(pool_size=(5,5), strides=(2, 2)))\n",
        "\n",
        "    #2nd convolution layer\n",
        "    model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "    model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "    model.add(AveragePooling2D(pool_size=(3,3), strides=(2, 2)))\n",
        "\n",
        "    #3rd convolution layer\n",
        "    model.add(Conv2D(128, (3, 3), activation='relu'))\n",
        "    model.add(Conv2D(128, (3, 3), activation='relu'))\n",
        "    model.add(AveragePooling2D(pool_size=(3,3), strides=(2, 2)))\n",
        "\n",
        "    model.add(Flatten())\n",
        "\n",
        "    #fully connected neural networks\n",
        "    model.add(Dense(1024, activation='relu'))\n",
        "    model.add(Dropout(0.2))\n",
        "    model.add(Dense(1024, activation='relu'))\n",
        "    model.add(Dropout(0.2))\n",
        "\n",
        "    model.add(Dense(n_classes, activation='softmax'))\n",
        "    \n",
        "    return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "W3OgjufxcFDI",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# model = base_cnn_model()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "TKorh7aC2cRb",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### AlexNet"
      ]
    },
    {
      "metadata": {
        "id": "Ht5rwmPf20ty",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def alexnet_model(img_shape=(48, 48, 1), n_classes=7, l2_reg=0.1,\n",
        "\tweights=None):\n",
        "\n",
        "\t# Initialize model\n",
        "\talexnet = Sequential()\n",
        "\n",
        "\t# Layer 1\n",
        "\talexnet.add(Conv2D(96, (11, 11), input_shape=img_shape,\n",
        "\t\tpadding='same', kernel_regularizer=l2(l2_reg)))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 2\n",
        "\talexnet.add(Conv2D(256, (5, 5), padding='same'))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 3\n",
        "\talexnet.add(ZeroPadding2D((1, 1)))\n",
        "\talexnet.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 4\n",
        "\talexnet.add(ZeroPadding2D((1, 1)))\n",
        "\talexnet.add(Conv2D(1024, (3, 3), padding='same'))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\n",
        "\t# Layer 5\n",
        "\talexnet.add(ZeroPadding2D((1, 1)))\n",
        "\talexnet.add(Conv2D(1024, (3, 3), padding='same'))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 6\n",
        "\talexnet.add(Flatten())\n",
        "\talexnet.add(Dense(3072))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(Dropout(0.5))\n",
        "\n",
        "\t# Layer 7\n",
        "\talexnet.add(Dense(4096))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('relu'))\n",
        "\talexnet.add(Dropout(0.5))\n",
        "\n",
        "\t# Layer 8\n",
        "\talexnet.add(Dense(n_classes))\n",
        "\talexnet.add(BatchNormalization())\n",
        "\talexnet.add(Activation('softmax'))\n",
        "\n",
        "\tif weights is not None:\n",
        "\t\talexnet.load_weights(weights)\n",
        "\n",
        "\treturn alexnet"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "FmUZi7_GHNXL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model = alexnet_model()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "bOCx_iBCnLGR",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### ResNet50"
      ]
    },
    {
      "metadata": {
        "id": "zooD9aSKqMgZ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# from keras import applications"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "U75fQkkdqgM6",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# model = keras.applications.resnet50.ResNet50(include_top=True, weights=None, input_tensor=None, input_shape=(48, 48, 1), pooling=None, classes=7)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "BGQ6kzVmHJK4",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### VGG19"
      ]
    },
    {
      "metadata": {
        "id": "HOQL4Qxhq7Et",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def vgg19_model(img_shape=(48, 48, 1), n_classes=7, l2_reg=0.1,\n",
        "\tweights=None):\n",
        "\n",
        "\t# Initialize model\n",
        "\tvgg19 = Sequential()\n",
        "\n",
        "\t# Layer 1 & 2\n",
        "\tvgg19.add(Conv2D(64, (3, 3), padding='same',\n",
        "\t\tinput_shape=img_shape, kernel_regularizer=l2(l2_reg)))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(64, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 3 & 4\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(128, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(128, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layer 5, 6, 7, & 8\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(256, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(256, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(256, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(256, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layers 9, 10, 11, & 12\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layers 13, 14, 15, & 16\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(ZeroPadding2D((1, 1)))\n",
        "\tvgg19.add(Conv2D(512, (3, 3), padding='same'))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\t# Layers 17, 18, & 19\n",
        "\tvgg19.add(Flatten())\n",
        "\tvgg19.add(Dense(4096))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(Dropout(0.5))\n",
        "\tvgg19.add(Dense(4096))\n",
        "\tvgg19.add(Activation('relu'))\n",
        "\tvgg19.add(Dropout(0.5))\n",
        "\tvgg19.add(Dense(n_classes))\n",
        "\tvgg19.add(Activation('softmax'))\n",
        "\n",
        "\tif weights is not None:\n",
        "\t\tvgg19.load_weights(weights)\n",
        "\n",
        "\treturn vgg19"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "NxmtoOZL8DL0",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# model = vgg19_model()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "UhAxipS_YCy9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### VGG16"
      ]
    },
    {
      "metadata": {
        "id": "lxc2HAn1YCTZ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# model = keras.applications.vgg16.VGG16(include_top=True, weights=None, input_tensor=None, input_shape=(48,48,1), pooling=None, classes=7)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "-T_oq0rJ_Ekz",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Model init"
      ]
    },
    {
      "metadata": {
        "id": "bAfmN3uC_6zq",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#batch process\n",
        "gen = ImageDataGenerator()\n",
        "train_generator = gen.flow(x_train, y_train, batch_size=batch_size)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "N6odbVHuAdRL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "filepath=\"/content/gdrive/My Drive/weights_base_new_opt.best.hdf5\"\n",
        "checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6t9TMnEqTwap",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ZrEcAZIdALPA",
        "colab_type": "code",
        "outputId": "e596e740-d8f0-4fc9-b139-77efde91ef61",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 215
        }
      },
      "cell_type": "code",
      "source": [
        "model.compile(loss='categorical_crossentropy'\n",
        "    , optimizer=keras.optimizers.Adam()\n",
        "    , metrics=['accuracy']\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-16-8c6fee90c7c5>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m model.compile(loss='categorical_crossentropy'\n\u001b[0m\u001b[1;32m      2\u001b[0m     \u001b[0;34m,\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptimizers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mAdam\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m     \u001b[0;34m,\u001b[0m \u001b[0mmetrics\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'accuracy'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m )\n",
            "\u001b[0;31mNameError\u001b[0m: name 'model' is not defined"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "uSOfnGivK_NA",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "tbCallBack = TensorBoard(log_dir='./log', histogram_freq=1,\n",
        "                         write_graph=True,\n",
        "                         write_grads=True,\n",
        "                         batch_size=batch_size,\n",
        "                         write_images=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "qU1WZYYvwDwd",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model = keras.models.load_model(\"/content/gdrive/My Drive/weights_base.best.hdf5\") # 99% acc"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "x82vInNkxzyk",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### TPU model"
      ]
    },
    {
      "metadata": {
        "id": "I2P7C1pC05RK",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# tf.keras.backend.clear_session()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "90zw0lm8xq25",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# tpu_model = tf.contrib.tpu.keras_support.tpu_model(\n",
        "#     model,\n",
        "#     strategy=tf.contrib.tpu.TPUDistributionStrategy(\n",
        "#         tf.contrib.cluster_resolver.TPUClusterResolver(TPU_WORKER)))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "OwJ7cejZx3T0",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Train model"
      ]
    },
    {
      "metadata": {
        "id": "Wse_wzO8ALMV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "fit = False\n",
        "\n",
        "if fit == True:\n",
        "\t#model.fit_generator(x_train, y_train, epochs=epochs) #train for all trainset\n",
        "\tmodel.fit_generator(train_generator, validation_data=valid, steps_per_epoch=batch_size, epochs=epochs, callbacks=[checkpoint, tensor_board]) #train for randomly selected one\n",
        "else:\n",
        "\tmodel.load_weights('/content/gdrive/My Drive/alex_net_epoch25.h5') #load weights"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "vz3W-EjbRE3A",
        "colab_type": "code",
        "outputId": "d4b7da1f-61c8-4f40-cd5b-8ce50de25dbf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "model_input"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tf.Tensor 'input_9:0' shape=(?, 48, 48, 1) dtype=float32>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 70
        }
      ]
    },
    {
      "metadata": {
        "id": "k_HewBrXNS3u",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "models = [model, pre_train_f, pre_train_s]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "8xcBhbl9TG_L",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from keras import layers"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ehb_rCvhNSzh",
        "colab_type": "code",
        "outputId": "5a2a1e9d-4ad3-494c-c1fc-ad0f7ec6c8e8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 793
        }
      },
      "cell_type": "code",
      "source": [
        "def ensemble(models, model_input):\n",
        "    \n",
        "    outputs = [model(model_input) for model in models]\n",
        "\n",
        "    y = layers.average(outputs) \n",
        "    \n",
        "    model = Model(model_input, y, name='ensemble')\n",
        "    \n",
        "    return model\n",
        "\n",
        "ensemble_model = ensemble(models, model_input)"
      ],
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "error",
          "ename": "ValueError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-70-bdc317777f43>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m \u001b[0mensemble_model\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mensemble\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodels\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel_input\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m<ipython-input-70-bdc317777f43>\u001b[0m in \u001b[0;36mensemble\u001b[0;34m(models, model_input)\u001b[0m\n\u001b[1;32m      5\u001b[0m     \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayers\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0maverage\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m     \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mModel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_input\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'semble'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/legacy/interfaces.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     89\u001b[0m                 warnings.warn('Update your `' + object_name + '` call to the ' +\n\u001b[1;32m     90\u001b[0m                               'Keras 2 API: ' + signature, stacklevel=2)\n\u001b[0;32m---> 91\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     92\u001b[0m         \u001b[0mwrapper\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_original_function\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m     91\u001b[0m                 'inputs' in kwargs and 'outputs' in kwargs):\n\u001b[1;32m     92\u001b[0m             \u001b[0;31m# Graph network\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 93\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_init_graph_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     94\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     95\u001b[0m             \u001b[0;31m# Subclassed network\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m_init_graph_network\u001b[0;34m(self, inputs, outputs, name)\u001b[0m\n\u001b[1;32m    229\u001b[0m         \u001b[0;31m# Keep track of the network's nodes and layers.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    230\u001b[0m         nodes, nodes_by_depth, layers, layers_by_depth = _map_graph_network(\n\u001b[0;32m--> 231\u001b[0;31m             self.inputs, self.outputs)\n\u001b[0m\u001b[1;32m    232\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_network_nodes\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnodes\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    233\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_nodes_by_depth\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnodes_by_depth\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m_map_graph_network\u001b[0;34m(inputs, outputs)\u001b[0m\n\u001b[1;32m   1453\u001b[0m             raise ValueError('The name \"' + name + '\" is used ' +\n\u001b[1;32m   1454\u001b[0m                              \u001b[0mstr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mall_names\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcount\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m+\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1455\u001b[0;31m                              \u001b[0;34m' times in the model. '\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1456\u001b[0m                              'All layer names should be unique.')\n\u001b[1;32m   1457\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mnetwork_nodes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnodes_by_depth\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_by_depth\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mValueError\u001b[0m: The name \"sequential_1\" is used 2 times in the model. All layer names should be unique."
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "5L36ZP7LThW6",
        "colab_type": "code",
        "outputId": "973b2645-700c-4f70-8d00-fd9613534962",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "print(ensemble_model)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<keras.engine.training.Model object at 0x7fc70244c898>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "O9bwdkXOT-wa",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "ensemble_model.compile(loss='categorical_crossentropy'\n",
        "    , optimizer=keras.optimizers.Adam()\n",
        "    , metrics=['accuracy']\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ZDxo0pkQUEal",
        "colab_type": "code",
        "outputId": "73e98a4d-ff06-4e35-a254-520d8f861e56",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1074
        }
      },
      "cell_type": "code",
      "source": [
        "best_model.fit_generator(train_generator, validation_data=valid, steps_per_epoch=batch_size, epochs=epochs, callbacks=[checkpoint]) #train for randomly selected one"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/15\n",
            "256/256 [==============================] - 222s 866ms/step - loss: 0.0078 - acc: 0.9970 - val_loss: 2.5201 - val_acc: 0.5272\n",
            "\n",
            "Epoch 00001: val_acc did not improve from 0.61215\n",
            "Epoch 2/15\n",
            "256/256 [==============================] - 222s 867ms/step - loss: 0.0847 - acc: 0.9720 - val_loss: 1.8328 - val_acc: 0.5704\n",
            "\n",
            "Epoch 00002: val_acc did not improve from 0.61215\n",
            "Epoch 3/15\n",
            "256/256 [==============================] - 221s 865ms/step - loss: 0.0158 - acc: 0.9948 - val_loss: 2.0212 - val_acc: 0.6124\n",
            "\n",
            "Epoch 00003: val_acc improved from 0.61215 to 0.61243, saving model to /content/gdrive/My Drive/weights_base.best.hdf5\n",
            "Epoch 4/15\n",
            "256/256 [==============================] - 221s 865ms/step - loss: 0.0067 - acc: 0.9974 - val_loss: 2.0024 - val_acc: 0.6127\n",
            "\n",
            "Epoch 00004: val_acc improved from 0.61243 to 0.61271, saving model to /content/gdrive/My Drive/weights_base.best.hdf5\n",
            "Epoch 5/15\n",
            "256/256 [==============================] - 222s 865ms/step - loss: 0.0051 - acc: 0.9978 - val_loss: 2.0497 - val_acc: 0.6105\n",
            "\n",
            "Epoch 00005: val_acc did not improve from 0.61271\n",
            "Epoch 6/15\n",
            "256/256 [==============================] - 222s 866ms/step - loss: 0.0050 - acc: 0.9973 - val_loss: 2.0552 - val_acc: 0.6138\n",
            "\n",
            "Epoch 00006: val_acc improved from 0.61271 to 0.61382, saving model to /content/gdrive/My Drive/weights_base.best.hdf5\n",
            "Epoch 7/15\n",
            "256/256 [==============================] - 221s 864ms/step - loss: 0.0047 - acc: 0.9975 - val_loss: 2.0628 - val_acc: 0.6102\n",
            "\n",
            "Epoch 00007: val_acc did not improve from 0.61382\n",
            "Epoch 8/15\n",
            "256/256 [==============================] - 222s 866ms/step - loss: 0.0047 - acc: 0.9976 - val_loss: 2.0932 - val_acc: 0.6177\n",
            "\n",
            "Epoch 00008: val_acc improved from 0.61382 to 0.61772, saving model to /content/gdrive/My Drive/weights_base.best.hdf5\n",
            "Epoch 9/15\n",
            "256/256 [==============================] - 222s 866ms/step - loss: 0.0048 - acc: 0.9973 - val_loss: 2.0839 - val_acc: 0.6116\n",
            "\n",
            "Epoch 00009: val_acc did not improve from 0.61772\n",
            "Epoch 10/15\n",
            "256/256 [==============================] - 222s 866ms/step - loss: 0.0143 - acc: 0.9952 - val_loss: 2.6912 - val_acc: 0.4943\n",
            "\n",
            "Epoch 00010: val_acc did not improve from 0.61772\n",
            "Epoch 11/15\n",
            "256/256 [==============================] - 221s 863ms/step - loss: 0.0440 - acc: 0.9850 - val_loss: 2.2318 - val_acc: 0.5782\n",
            "\n",
            "Epoch 00011: val_acc did not improve from 0.61772\n",
            "Epoch 12/15\n",
            "256/256 [==============================] - 221s 865ms/step - loss: 0.0190 - acc: 0.9941 - val_loss: 2.1244 - val_acc: 0.5982\n",
            "\n",
            "Epoch 00012: val_acc did not improve from 0.61772\n",
            "Epoch 13/15\n",
            "256/256 [==============================] - 221s 864ms/step - loss: 0.0067 - acc: 0.9974 - val_loss: 2.1193 - val_acc: 0.6085\n",
            "\n",
            "Epoch 00013: val_acc did not improve from 0.61772\n",
            "Epoch 14/15\n",
            "256/256 [==============================] - 221s 864ms/step - loss: 0.0174 - acc: 0.9939 - val_loss: 2.1646 - val_acc: 0.6002\n",
            "\n",
            "Epoch 00014: val_acc did not improve from 0.61772\n",
            "Epoch 15/15\n",
            "256/256 [==============================] - 221s 863ms/step - loss: 0.0090 - acc: 0.9967 - val_loss: 2.5467 - val_acc: 0.5561\n",
            "\n",
            "Epoch 00015: val_acc did not improve from 0.61772\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fe47648b400>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "metadata": {
        "id": "BdhKIyXYiCNQ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "best_model = keras.models.load_model(\"/content/gdrive/My Drive/weights_base.best.hdf5\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "nW3Beb7q8qK1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 136
        },
        "outputId": "af20b422-6628-4bb1-f633-d3995ff4b3ad"
      },
      "cell_type": "code",
      "source": [
        "y_prob = best_model.predict(x_test)\n",
        "y_pred = best_model.predict_classes(x_test)\n",
        "y_test_true = [ np.argmax(item) for item in y_test]\n",
        "from sklearn.metrics import confusion_matrix\n",
        "confusion_matrix(y_test_true, y_pred)"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[256,   3,  53,  29,  78,  10,  38],\n",
              "       [ 17,  27,   3,   3,   2,   0,   4],\n",
              "       [ 53,   2, 212,  24, 105,  29,  71],\n",
              "       [ 34,   1,  25, 702,  48,  19,  66],\n",
              "       [ 91,   1,  76,  50, 340,  10,  85],\n",
              "       [ 17,   0,  42,  18,  10, 309,  19],\n",
              "       [ 73,   1,  53,  56, 116,   4, 304]])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 71
        }
      ]
    },
    {
      "metadata": {
        "id": "VU-BOyZbLyjV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "labels = ['Angry', 'Disgust','Fear', 'Happy', 'Sad', 'Surprise', 'Neutral']\n",
        "#0=Angry, 1=Disgust, 2=Fear, 3=Happy, 4=Sad, 5=Surprise, 6=Neutral\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib\n",
        "def plot_confusion_matrix(y_true, y_pred, cmap=plt.cm.Blues):\n",
        "    cm = confusion_matrix(y_true, y_pred)\n",
        "    fig = plt.figure(figsize=(7,7))\n",
        "    matplotlib.rcParams.update({'font.size': 16})\n",
        "    ax  = fig.add_subplot(111)\n",
        "    matrix = ax.imshow(cm, interpolation='nearest', cmap=cmap)\n",
        "    fig.colorbar(matrix) \n",
        "    for i in range(0,7):\n",
        "        for j in range(0,7):  \n",
        "            ax.text(j,i,cm[i,j],va='center', ha='center')\n",
        "    # ax.set_title('Confusion Matrix')\n",
        "    ticks = np.arange(len(labels))\n",
        "    ax.set_xticks(ticks)\n",
        "    ax.set_xticklabels(labels, rotation=45)\n",
        "    ax.set_yticks(ticks)\n",
        "    ax.set_yticklabels(labels)\n",
        "    ax.grid(False)\n",
        "    plt.tight_layout()\n",
        "    plt.ylabel('True label')\n",
        "    plt.xlabel('Predicted label')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "WGNOKeYFL1G8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 500
        },
        "outputId": "c858bcfb-ea64-4daa-8532-3aa87dd60e09"
      },
      "cell_type": "code",
      "source": [
        "plot_confusion_matrix(y_test_true, y_pred, cmap=plt.cm.OrRd)"
      ],
      "execution_count": 73,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeAAAAHjCAYAAAD7QMw7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ntOD/ARLMcB8L7FEpAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f06d7819d30>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "9Xgj84H3iRso",
        "colab_type": "code",
        "outputId": "2bb2beec-3a53-483b-bcc5-330839bc2a82",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "cell_type": "code",
      "source": [
        "score = best_model.evaluate(x_test, y_test)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', 100*score[1])"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3589/3589 [==============================] - 10s 3ms/step\n",
            "Test loss: 2.1044667149759264\n",
            "Test accuracy: 59.905266091248286\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "-1laTJX9ALJy",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "y_pred = best_model.predict(x_test)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "1D9abwgwze9l",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "y_true = [np.argmax(true) for true in y_test]\n",
        "y_pred_true = [np.argmax(true) for true in y_pred]\n",
        "accs = list()\n",
        "precs = list()\n",
        "recalls = list()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "8R_trqyEZtIr",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from sklearn.metrics import confusion_matrix"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "3nrp7jQkXKD_",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def class_precision(y_true, y_pred, emotion):\n",
        "    cm = confusion_matrix(y_true, y_pred)\n",
        "    i = [i for i, label in enumerate(labels) if label == emotion][0]\n",
        "    col = [cm[j,i] for j in range(0,len(labels))]\n",
        "    return float(col[i])/sum(col)\n",
        "\n",
        "def class_recall(y_true, y_pred, emotion):\n",
        "    cm = confusion_matrix(y_true, y_pred)\n",
        "    i = [i for i, label in enumerate(labels) if label == emotion][0]\n",
        "    row = [cm[i,j] for j in range(0,len(labels))]\n",
        "    return float(row[i])/sum(row)\n",
        "\n",
        "def class_accuracy(y_true, y_pred, emotion):\n",
        "    cm = confusion_matrix(y_true, y_pred)\n",
        "    i = [i for i, label in enumerate(labels) if label == emotion][0]\n",
        "    tp = cm[i,i]\n",
        "    fn = sum([cm[i,j] for j in range(0,6) if j != i])\n",
        "    fp = sum([cm[j,i] for j in range(0,6) if j != i])\n",
        "    tn = sum([cm[i,j] for j in range(0,6) for i in range(0,6)]) -(tp+fp+fn)\n",
        "    return float(tp + tn)/sum([tp, fn, fp, tn])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "BXGzgF_cB3Zl",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from sklearn.metrics import precision_score, recall_score"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "MlIXMNkuD3gx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "4b6752b0-87e0-4de4-a27b-ee2e173b35b2"
      },
      "cell_type": "code",
      "source": [
        "recall_score(y_true, y_pred_true, average='weighted')"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.599052660908331"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 58
        }
      ]
    },
    {
      "metadata": {
        "id": "s4LK8whsA34N",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def prec_precision(y_true, y_pred_true):\n",
        "    return precision_score(y_true, y_pred_true, average='macro')\n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "qj2YhRA2B8CP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "58b245ed-5ac1-4b03-f01e-0be869f3fd9c"
      },
      "cell_type": "code",
      "source": [
        "prec_precision(y_true, y_pred_true)"
      ],
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6161088020362141"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 53
        }
      ]
    },
    {
      "metadata": {
        "id": "xH_isQU8AE4E",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def prec_acc(y_true, y_pred_true):\n",
        "    count = 0\n",
        "    for i in range(0, len(y_true)):\n",
        "        if y_true[i] == y_pred_true[i]:\n",
        "            count += 1\n",
        "    return count/len(y_true)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "-P0KoOjYAsOP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0882ff34-e5b5-4232-f949-0d127202896e"
      },
      "cell_type": "code",
      "source": [
        "prec_acc(y_true, y_pred_true)"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.599052660908331"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "metadata": {
        "id": "hqL9tECd5haw",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "plot_vals = list()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "0M0TTdYwXLqf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 612
        },
        "outputId": "b879cd92-f1c7-4ee5-ddef-814220e4dcb5"
      },
      "cell_type": "code",
      "source": [
        "for emotion in labels:\n",
        "    print(emotion.upper())\n",
        "    acc = class_accuracy(y_true, y_pred_true, emotion)\n",
        "    acc_list = ['acc', emotion, acc]\n",
        "    prec = class_precision(y_true, y_pred_true, emotion)\n",
        "    prec_list = ['prec', emotion, prec]\n",
        "    recall = class_recall(y_true, y_pred_true, emotion)\n",
        "    recall_list = ['recall', emotion, recall]\n",
        "    plot_vals.append(acc_list)\n",
        "    plot_vals.append(prec_list)\n",
        "    plot_vals.append(recall_list)\n",
        "    print('   acc = {}'.format(acc))\n",
        "    print('  prec = {}'.format(prec))\n",
        "    print('recall = {}\\n'.format(recall))"
      ],
      "execution_count": 134,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ANGRY\n",
            "   acc = 0.8573545757688033\n",
            "  prec = 0.4731977818853974\n",
            "recall = 0.5481798715203426\n",
            "\n",
            "DISGUST\n",
            "   acc = 0.9881437569470174\n",
            "  prec = 0.7714285714285715\n",
            "recall = 0.48214285714285715\n",
            "\n",
            "FEAR\n",
            "   acc = 0.8473508706928492\n",
            "  prec = 0.45689655172413796\n",
            "recall = 0.4274193548387097\n",
            "\n",
            "HAPPY\n",
            "   acc = 0.9070025935531678\n",
            "  prec = 0.7959183673469388\n",
            "recall = 0.7843575418994413\n",
            "\n",
            "SAD\n",
            "   acc = 0.8254909225639125\n",
            "  prec = 0.4864091559370529\n",
            "recall = 0.5206738131699847\n",
            "\n",
            "SURPRISE\n",
            "   acc = 0.9425713227121156\n",
            "  prec = 0.8110236220472441\n",
            "recall = 0.744578313253012\n",
            "\n",
            "NEUTRAL\n",
            "   acc = 0.7828825490922564\n",
            "  prec = 0.5178875638841567\n",
            "recall = 0.500823723228995\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "oYyzWhDKy5qj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 136
        },
        "outputId": "76be2e89-059c-41c2-d1ca-0b2a6f25cac4"
      },
      "cell_type": "code",
      "source": [
        "precs"
      ],
      "execution_count": 86,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.4731977818853974,\n",
              " 0.7714285714285715,\n",
              " 0.45689655172413796,\n",
              " 0.7959183673469388,\n",
              " 0.4864091559370529,\n",
              " 0.8110236220472441,\n",
              " 0.5178875638841567]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 86
        }
      ]
    },
    {
      "metadata": {
        "id": "B-jWO4Tu3AhS",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "%matplotlib inline"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "JqJIv3YT3WfI",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        },
        "outputId": "0219ef1f-a90a-40ad-d557-634fd500cf65"
      },
      "cell_type": "code",
      "source": [
        "plot_vals"
      ],
      "execution_count": 135,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[['acc', 'angry', 0.8573545757688033],\n",
              " ['prec', 'angry', 0.4731977818853974],\n",
              " ['recall', 'angry', 0.5481798715203426],\n",
              " ['acc', 'disgust', 0.9881437569470174],\n",
              " ['prec', 'disgust', 0.7714285714285715],\n",
              " ['recall', 'disgust', 0.48214285714285715],\n",
              " ['acc', 'fear', 0.8473508706928492],\n",
              " ['prec', 'fear', 0.45689655172413796],\n",
              " ['recall', 'fear', 0.4274193548387097],\n",
              " ['acc', 'happy', 0.9070025935531678],\n",
              " ['prec', 'happy', 0.7959183673469388],\n",
              " ['recall', 'happy', 0.7843575418994413],\n",
              " ['acc', 'sad', 0.8254909225639125],\n",
              " ['prec', 'sad', 0.4864091559370529],\n",
              " ['recall', 'sad', 0.5206738131699847],\n",
              " ['acc', 'surprise', 0.9425713227121156],\n",
              " ['prec', 'surprise', 0.8110236220472441],\n",
              " ['recall', 'surprise', 0.744578313253012],\n",
              " ['acc', 'neutral', 0.7828825490922564],\n",
              " ['prec', 'neutral', 0.5178875638841567],\n",
              " ['recall', 'neutral', 0.500823723228995]]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 135
        }
      ]
    },
    {
      "metadata": {
        "id": "YwtVS9bP5HxF",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import pandas as pd"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "GhX1JnCJ3OU1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 305
        },
        "outputId": "332af26c-cdcc-4e8c-ef47-27b0e342e1ea"
      },
      "cell_type": "code",
      "source": [
        "df = pd.DataFrame(plot_vals, columns=['','emtions','percentage'])\n",
        "\n",
        "pivot = df.pivot(\"emtions\", \"\", \"percentage\")\n",
        "\n",
        "pivot.plot(kind='bar', color=['lightskyblue', 'orange', 'forestgreen'])\n",
        "\n",
        "plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)\n",
        "plt.show()"
      ],
      "execution_count": 153,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAb4AAAEgCAYAAADL4RbaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAHBZJREFUeJzt3XucXWV97/FPJFyMJhp1lIsoBfF3\nQFsroBKQO1gRahVDq2gFDp6jEhU9RUUQtVpBrTQIeLcHsRrByxGxgqJyUUEogqAV/GG5CBgsgwZI\nS7mEzPljrSmbSTKzd2bvWXvN83m/XnnN3mvt2fPbk5n57udZz2XO2NgYkiSV4lFNFyBJ0kwy+CRJ\nRTH4JElFMfgkSUUx+CRJRTH4JElFmTvTX3B0dOVA508sXDiPFSvuHeSXGCjrb1ab629z7WD9UxkZ\nmT9nYE9emFnX4ps7d4OmS5gW629Wm+tvc+1g/Zo5sy74JEmajMEnSSpKV9f4IuLZwDeBpZl52oRz\n+wInAA8B52bmB/pepSRJfTJl8EXEY4BTgR+s4yGnAH8G/Ba4OCK+npnX9q9ESdIgfOhnD/Z1sOEx\nz92wFQNwumnx3Q+8BHjnxBMRsTXwh8y8tb5/LrAPYPBJkh4hIhYAy4DHAPOANwOP4+FewzMz8+SI\n2G/isX7WMeU1vsxclZn/tY7TmwKjHffvADbrR2GSpFlnU+BzmbkX8C6qBtUnqBpXuwL7RsSj13Gs\nb/o9j2/KZu7ChfMGPux3ZGT+QJ9/0Ky/WW2uv821g/UX4N+B4yPiaGBjqpbffZk53oA6MCKePPFY\nv4uYbvAtp0rwcVvUx9Zp0BNUR0bmMzq6cqBfY5Csv1ltrr/NtYP1d/P8s8Bbgd9m5l9HxE7A6azZ\n8/jQWo711bSCLzNvjogFEbEVcBtVMr+6H4XNFv942yaTnj/iqffNUCWS1LgnAT+vb78cWAk8ISLG\nG03fAl4DbDDxWGbe1a8iuhnVuSNwErAV8GBELAbOAW7KzG8AbwS+XD/8rMy8vl/FSZJmlS8AX4iI\ng4HTgFcBHwS+Vp//SmbeFRFHTjzWzyLmjI0NdOnMNQx6rc5h6y7ptcU3bPX3yvqb0+bawfq7eP5W\nTBVoA1dukSQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVpd9rdUrSrFDEqkvL5vR3XvUh\nY62Ya2jwSZJmREQcBrwYWAA8FVgKHAucS7W7z+nAPwIbUa3Z+brMvCUi/hp4C7Aa+IfMPGs6ddjV\nKUmaSc8CXgrsDfwd1S4N52XmB4EPACdl5j7AyVQ7OcwH3gPsTrXp+SHTLcAWnyRpJl2cmauAOyNi\nBbA18C/1uV2AiIh3AxtQ7fe6HfCrel/Y/wL+YroFGHySpJnU2dM4BxgDHqjvPwAcnJm3jz+g3iih\nr72TBp8kaSYtiogNgIXAfOD3HecuB14GfDIi9qba7/WbVK3AxwKrqLYpelFmrvfAHK/xSZJm0s3A\nV4ELgOOoBqyMex/wsoj4IfBe4CeZ+Z9U1/i+D1wEfG46oQe2+DTLFTEkXVpfzUw/uCEzj+64/0/j\nNzJzOdUAlkfIzGXAsn4VYItPklQUW3ySpBmRmZ9vugawxSdJKozBJ0kqisEnSSqKwSdJKorBJ0lq\nnYi4s/54UUQ8u5fPdVSnJBVqm49u09dtiW44+ga3JZIkaVy9LdH+wObAd4CXUK3ccnZmnhQRjwe+\nRLVt0d3AK4HH8/Ak9w2BQzPzhunUYVenJGkmPQ14LbAv8EKq7YZeERFPA44GvpuZuwE/qB+zGfD+\nzNwL+L/AkdMtYOhbfC45JUmzyhXA84FtgQvrY/OBrYAdgOMBMnMpQERsCZwSEX9LtbD1ldMtYOiD\nT5I0qzxQ//t2Zr6+80REvJ01eyLfT9UK/FRELAYOnG4BBp8mZYtb0gBcCXw4IuZRbS57MnAMVWtw\nb+CKiHg9cB/wJOCGiJhDtQntBtP94l7jkyTNqMy8hSrsfghcBvyu3mH9Y8AuEXERVcvu/wGfBk4F\nzgPOBPaIiBdN5+vb4pOkQs309IPORaoz8xPAJyacv5uqVdfpn+t/47aoPz6p/pw9e63DFp8kqSi2\n+CQNhNeHNaxs8UmSitJViy8ilgI7A2PAUZl5Rce5JcBrgIeAn2bmWwdRqCRJ/TBl8EXEHsC2mbko\nIrajmjm/qD63AHg78IzMXBUR50fEzpl52UCrlqSGjXxvwZrHOm6P7nfPzBWjnnTT1bkPcDZAZl4H\nLKwDDx6eiPjYiJgLzAP+MIhCJUnqh26Cb1NgtOP+aH2MzLwP+FvgRuA3wOWZeX2/i5QkqV/WZ1Tn\nf8/7qFt+xwLPBO4BLoiI52TmNev65IUL5zF3bg8T7297cNLTIyPzuzrWGOtvVtvr79FQ1d727/0U\n9U9lqF6LHqGb4FtO3cKrbQ7cXt/eDrgxM8c3BPwRsCOwzuBbseLeHkucfEj06OjKR9wfGZm/xrFm\nWX+zJq+fZZPP323TdZq2fe9b/7MzhX6/FoO0f7rp6jwfWAwQETsAyzNz/H/0ZmC7iHh0fX8n4Nf9\nLlKSpH6ZssWXmZdGxJURcSnVhoFL6s0E787Mb0TE3wMXRsQq4NLM/NFgS5Zmzs5n/Omk5y879OqB\nfW0ngEuD0dU1vsw8ZsKhazrOfZpqEVFJkoaeK7dIkoriWp0NcxKsJM0sW3ySpKIYfJKkohh8kqSi\nGHySpKI4uEWSBqDJOaCanC0+SVJRDD5JUlEMPklSUQw+SVJRDD5JUlEMPklSUQw+SVJRDD5JUlEM\nPklSUQw+SVJRDD5JUlEMPklSUQw+SVJRDD5JUlEMPklSUQw+SVJRDD5JUlEMPklSUQw+SVJRDD5J\nUlEMPklSUQw+SVJRDD5JUlEMPklSUQw+SVJRDD5JUlHmdvOgiFgK7AyMAUdl5hUd57YEvgxsBFyV\nmW8YRKGSJPXDlC2+iNgD2DYzFwFHAKdMeMhJwEmZ+XzgoYh4Wv/LlCSpP7rp6twHOBsgM68DFkbE\nAoCIeBSwG3BOfX5JZt4yoFolSZq2boJvU2C04/5ofQxgBFgJLI2IH0fEiX2uT5KkvurqGt8Ecybc\n3gL4GHAz8O2IOCAzv72uT164cB5z527Q/Ve77cFJT4+MzO/qWGOmqH8qjb+Wwr//Uxnoa53l3/u2\n1z9dQ/VaC9NN8C3n4RYewObA7fXtO4HfZOYNABHxA+BZwDqDb8WKe3sscZNJz46OrnzE/ZGR+Wsc\na9bk9U+l+ddS9vd/KoN9rbP7e9/2+qer19dqUPZPN12d5wOLASJiB2B5Zq4EyMxVwI0RsW392B2B\nHEShkiT1w5Qtvsy8NCKujIhLgdXAkog4DLg7M78BvBX4fD3Q5RfAtwZZsCRJ09HVNb7MPGbCoWs6\nzv0b8MJ+FiVJ0qC4coskqSgGnySpKAafJKkoBp8kqSgGnySpKAafJKko67NkWavsfMafTnr+skOv\nnqFKJEnDwBafJKkoBp8kqSgGnySpKAafJKkos35wi6R2cmCaBsXgk1pq5HsL1jzWcXt0v3tmrhip\nRezqlCQVxeCTJBXFrk5plvIambR2tvgkSUUx+CRJRbGrU1IjphqVClvPVCkqjME35LxOI0n9ZVen\nJKkorW/x2V0iSeqFLT5JUlEMPklSUQw+SVJRWn+NT81yoWRJbWOLT5JUFINPklQUg0+SVBSDT5JU\nFINPklQUR3VqoFxrVNKwscUnSSpKVy2+iFgK7AyMAUdl5hVrecyJwKLM3LOvFUqS1EdTtvgiYg9g\n28xcBBwBnLKWx2wP7N7/8iRJ6q9uujr3Ac4GyMzrgIURMXG5jpOA4/pcmyRJfddN8G0KjHbcH62P\nARARhwEXAzf3szBJkgZhfUZ1zhm/ERFPAA4H9gW26OaTFy6cx9y5G3T/1W57sMfyejMyMn+gz2/9\nk7P+SbS5drD+KQy8fq1TN8G3nI4WHrA5cHt9e2+qNYl/BGwMbBMRSzPzbet6shUr7u2xxE16fHxv\nRkdXDvT5rX9y1j+ZNtcO1j+5Xus3KPunm67O84HFABGxA7A8M1cCZObXMnP7zNwZeDlw1WShJ0lS\n06YMvsy8FLgyIi6lGtG5JCIOi4iXD7w6SZL6rKtrfJl5zIRD16zlMTcDe06/JEmSBseVWyRJRTH4\nJElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJ\nRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx\n+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUUx+CRJRTH4JElFMfgkSUWZ282DImIpsDMwBhyV\nmVd0nNsLOBF4CEjgdZm5egC1SpI0bVO2+CJiD2DbzFwEHAGcMuEhnwEWZ+auwHzgxX2vUpKkPumm\nq3Mf4GyAzLwOWBgRCzrO75iZt9W3R4En9rdESZL6p5vg25Qq0MaN1scAyMx7ACJiM+BFwLn9LFCS\npH7q6hrfBHMmHoiIJwPfAo7MzN9P9skLF85j7twNuv9qtz3Ya309GRmZP9Dnt/7JWf8k2lw7WP8U\nBl6/1qmb4FtORwsP2By4ffxO3e15HnBcZp4/1ZOtWHFvjyVu0uPjezM6unKgz2/9k7P+ybS5drD+\nyfVav0HZP910dZ4PLAaIiB2A5ZnZ+T92ErA0M78zgPokSeqrKVt8mXlpRFwZEZcCq4ElEXEYcDfw\nXeC1wLYR8br6U5Zl5mcGVbAkSdPR1TW+zDxmwqFrOm5v3L9yJEkaLFdukSQVxeCTJBXF4JMkFcXg\nkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMk\nFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF\n4JMkFcXgkyQVxeCTJBXF4JMkFcXgkyQVxeCTJBXF4JMkFWVuNw+KiKXAzsAYcFRmXtFxbl/gBOAh\n4NzM/MAgCpUkqR+mbPFFxB7Atpm5CDgCOGXCQ04BXgHsCrwoIrbve5WSJPVJN12d+wBnA2TmdcDC\niFgAEBFbA3/IzFszczVwbv14SZKGUjddnZsCV3bcH62P3VN/HO04dwewzWRPNjIyf04vBR4zMsUD\nnjs26ekbevliA2D9zWpz/W2uHay/6fq1buszuGWy4Oop1CRJmmndBN9yqpbduM2B29dxbov6mCRJ\nQ6mb4DsfWAwQETsAyzNzJUBm3gwsiIitImIucGD9eEmShtKcsbHJ+6kBIuJDwO7AamAJ8Fzg7sz8\nRkTsDny4fujXM/OjgypWkqTp6ir4JEmaLVy5RZJUFINPklQUg0+SVJTWB19EfLQebdpKEXHgWo69\nqola1kdEbNZ0DdMREa3/HZDUm64WqR5yVwHvjIitgH8GvpSZNzZb0tQi4nnA84G3RMTTOk5tCLwd\n+HIjhfXuTGCPpouYhusj4lyqn5vLmy6mWxFxBdWi8RPNAcYy8/kzXFJPIuK1k53PzC/MVC3rKyKe\nCrwHWJiZB0fEK4GfZOZvGi5NU2h98GXmMmBZRGwI7A18OSJWA58CvpCZwzps9XfAfwAbAZ2LI60G\nDmuioPV0e0RcAlwBPDB+MDPf0VxJPdmean3ZwyPi74GLgGWZ+atGq5ra4knOLZixKtbfH9cftwae\nAVxC1QO1K/ALYOiDD/gc8DHgmPr+HcDngb2aKkjdaX3wAUTEzsArgT2BHwJnAfvVH/+yucrWLTNv\nBc6IiG8Dj8rMOyIigO2AHzdbXU/OW8ux1vxcZeYDwHkRcT6wL/B+4FURcRPwtsz8ZaMFrsN4qyIi\nHg+8GnhifWoj4FBgy4ZK60pmvh2g/vnfMTNX1fc3BL7SZG092CAzz4uIdwBk5gUR8d6mi9LUWn99\nIyIS+D/AD4CdMvMtmXlJZr4PmNdocd05Ddil7qr9KvAs4IxGK+pBZp4B/BS4qf63HHhro0X1ICL2\niohPA9cAfwa8MTO3Bd5EO/4fvgo8mSr8/hNYRFV7W2wJPK7j/qOBP2qoll49GBF7AxtExFMi4g3A\nfzVdlKbWmnfmkzgjM09Y24nMXGPgyBB6SmaeHRHHAKdm5mcj4ntNF9WtiPgUVSv1fwD/AuwIfKTR\nonrzeqputSMz86Hxg5l5fUR8trmyuvaozHxvROyRmSdFxGlUPR3fbLqwLn0EuCoi7qG6ZrkAeF+j\nFXXvCOADwJOA7wCXA4c3WpG6MhuCbyQi9mPNa0z3NldST+ZFxK7Aa4A9666rhQ3X1ItnZeZuEXFR\nZv55RGwJHN90UT14B1UL9ciIGAOuBU7OzNsz89PNltaVjSLiOcC99e/BjVTXzFohM78IfDEinkg1\nMOf3wKQDX4bIHcBxmfnvHZcpVjRck7owG4LvQOBlE46NUV00b4N3U/3x/VBm3hkR72bNXe6H2dyO\njYlHMvPW+g9xW5wJLAO+RPWHdxHwdWCXJovqwRKqrs53Ug20eGL9sRUiYieq2juvUW5KO7qZvwSc\nGRFXU3U5nwW8CvirRqvSlFoffPX1mDZL4M0A9bSGNoxm63Qq1QCiU4FfRMSDQGu6aoH7MvO0jvs/\njYiXNFZN7w7IzBPr23s3Wsn6ORU4lmqh+zcCLwcua7Si7rX6MkXJWh98EbG2OXsPUW2AfGxmXjXD\nJfXq61Qt1DlUc/i2Bn5GS+bG1dNJAIiIc4D5mfmHBkvq1U/rUXnfpxrstRvwq4jYHiAzr22yuC48\nueVd/fdm5oURcX9mXglcGRHfoZqTO+zafpmiWK0PPuCzwF3AOVQB8hKqeXEXUnUZvrC50qaWmc/r\nvB8Rm1JdMG+FiHg28A9UgbcoIl4bET9swRuOcePf//0nHP841c/TsLeiDqDdXf33RsRLgZsi4gSq\nN6xPm+JzhsXxtPsyRbFmQ/Dtn5m7d9z/XERckJknVteb2yUzf9eya2SnAkcCn6jvnw98hiF/wzEu\nM/eKiCcA21AtHvDrzLyn4bJ68erMvKLzQD3Evi2WUAXdm6gGGf0J8L8arWgKEbFxZt5PNd/2x/Wx\neVRvANUCsyH47ouIpVQrP6ymege/Ud398x+NVtaFCUtPzQGeQtXt1harMvO68TcZmXltvXJOK0TE\nu6j+0P4rVVfndhHxyWHfUDkingEEcEJ9jWlOfWouVatjq4ZK69U/AUdRraCzJ1Ur6j1UcyqH1enA\nIcAveeSycXNoV2u7WLMh+BZTDX/ei+oH79+AvwAeQztGV3UuPTUG3JOZdzVVzHq4KyL+J/CYiHgB\n1eCEOxquqReLge3qd/BExCZU7+KHOvioJnrvRDWis3N1otW0Zx4cVG+crq6Xizs5My+JiKH+u5SZ\nh9Q335SZ3260GK2Xof4B69JDwC1AZ/fUAW1Y5La2xhJHETE+OOdTwxqCEXF6Zh4OrAQ2A+6kWrPw\ncqols9riFtZcwej6JgrpRWb+gmoU7deGdVm1Ls2NiOOAlwLH14u3P7bhmrq1JCIuGdbfUa3bbAi+\n71MtlfXbjmPDujD12twJPJ2HB+fsD4yPilxGNVhnGG0XEVdRXRvrDIqnAgdR7TzRBhsDN0fE5cAG\nwHOB6yLiKwCZOZRrvXa4qJ54D9Wo4PnATS2a5vMaqlb3QZl5X0RsDbyh4Zq6tQC4NSJuoBpR24qd\nMTQ7gu+Bjq6HNtoxM/fpuL8sIs7LzP0jYuJIw2HyQmBzqgv6f9NwLdPx4bUc25Rq94yhl5mdO3sQ\nEX9CFSatUC/WvrTj/lkNltOrVzddgNbPbAi+b0XEAcCPgFXjB1s0j2lhPZz7Uh4enPPUeprAoxut\nbBL1avq3MPn2OG1wCdVAis6VQ96Vmds0V9L6y8yfR0RbVp1pu8dTXap4JlVvzbVUu3toyM2G4Hs9\nVRfVRG0ZWXUo1S/PiTw8OOd1VINz/neDdZXiK1TXKfek6m7eixYNDomIr/LIrv3NqXZp0OCdTjUC\n9SdUv7u7AF+k6i7XEJsNwXcIa1/rry2uBZZ0LHS7PXBNZt7XcF2lWJiZB9WLbL+5Xn3jU1TD7Nug\nc7m1MapBXtc0VEtpfp+ZnSvMnBMRQz0HUZXW78dHNWfp41R77x1NtWLLUY1W1JsvAYs69uPbnnYs\n0DtbbBwRTwdWRcQzgfup5se1xSXAFsDzMvOHVN39s+ENbRv8KiI+EREHRcTiiPgYsDwiXtKy9V6L\nMxuC797MvJBqkMuVmflu2rUR51My82yqHeRPzcwPAk9ouKaSHE81H+4DwLlU1y3bspcdVEv2PQc4\nuL6/J+1b6LytHkt1Hf7PqZaOW0DV43Qw7b/2PavNhneGbV7rD1zotmm7UO+OwcOrnxxJe67zbZmZ\nh0fEhQCZeVpEHDzVJ6kvfr2uTbA13GZD8B1CdU1vfK2/59CejSzBhW6bthj4o8xs64CQjeo3S2MA\nEbEd1dxEDV7bN8EuVuuDLzNXUo3KgxYNJXah26HxczqmwbTQscAPgG0j4rr62BEN1lOStu+MUazW\nB1+LrWuhW+r7rZxH1hYd0wDmA1mvQtM5D3TYV2wZ90yqa/V3Ui0dt4JqSL1/fAcsM5/ZdA1aPwZf\ncxZ1bKI7Z8K5Ni251lanTf2QVjiaqtXx26keqP6KiJtY83f1oRYtF1csg685z6YKvGOBq4GLqN65\n70X1Ll4DlJkXN11Dn1yfmUO/qPYs9eyO2xsCu9GuqTDFmjM2ZuOiSRFxcWbuMeHY9zJzv6ZqUntE\nxGep/gD/hEd21b6jsaIKVm+C3aaNgItki69590fESTxyrc61LcEmrc1/D47SzKr3EOxsOWxGdc1Y\nQ87ga94rqOfwUXV9JtVmrtKUMtNVfpqzkmqln99RBeCLcceGVjD4GlZPx/hk03VI6tk+VMsjbgKc\nALwR+CjVbh8aYrNhyTJJasKqzLyaqtfm5My8BBsTrWDwSdL6mRsRxwEvBc6PiOdRrd+pIWfwSdL6\neQ1wL3BQvY3Y1sAbmi1J3XA6gySpKLb4JElFMfgkSUUx+FS0iNglIraub58cETs2XZOkwXLorUp3\nOHAWcGNmvrXpYiQNnoNbNCtExJuBv6R6M/cr4CPA14Dzgd2BUartel4LbAUcTDUK73TgN8DbgPcA\nf5eZ3683BD4QeBD4V+AtwBbAOcB3gRdQLU91AHAH8DmqBYrHgJ9l5pJBv2ZJ68euTrVeRDyfapm3\n3TNzEXAXsC9VEH0yM3esb2+dmS8ClgGHZ+Y3qHbG+JvMvKDj+RZRTUreLTN3A0ao9k4E2B74fGbu\nXn/uXwF/DLwgMxdl5i7A1RHxuIG/cEnrxa5OzQZ7As8ALowIgMdQtc7u7Niy57dUC4ED3AY8fZLn\newFwcWY+WN+/iGrx8Ivr5/xlffw3wBOA64A7I+Jc4FvAVzLz7um/LEmDYPBpNrgfOCcz3zR+ICK2\nYs1dC1Z13J64+W+nif3/czqOrZp4rp68vFtE7EDVPXpFROyambd3Wb+kGWRXp2aDS4D9I+KxABFx\nJNUWMd1YTbWJaKfLgL0iYvz4PvWxtYqInSLi0My8KjPfD1yJmwlLQ8vgU+tl5k+BjwMXRcSPqbo+\nu+1q/B7w6Yg4qOP5LgfOBH4UEZcAtwJfnuQ5bgAWR8SlEXEB1TXGS3p+IZJmhKM6JUlFscUnSSqK\nwSdJKorBJ0kqisEnSSqKwSdJKorBJ0kqisEnSSqKwSdJKsr/B/wQMn2nOh5uAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fe470fbde48>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "JhY61eWdZVtk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 364
        },
        "outputId": "88730f51-d2a4-48ef-b7c1-d83674050ac1"
      },
      "cell_type": "code",
      "source": [
        "x = np.arange(num_classes)\n",
        "plt.hist((accs, precs, recalls), label = (\"acc\", \"prec\", \"recall\"))\n",
        "plt.legend()\n",
        "\n",
        "# labels\n",
        "# plt.xticks(x, tuple(labels))"
      ],
      "execution_count": 73,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fe4727b1a20>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 73
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeEAAAFKCAYAAAAqkecjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fe4727b1be0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "avZLwEJEi8pb",
        "colab_type": "code",
        "outputId": "904dace4-1e52-4d39-f75a-a77ad3ceb77e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "cell_type": "code",
      "source": [
        "!npm install -g localtunnel"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[K\u001b[?25h/tools/node/bin/lt -> /tools/node/lib/node_modules/localtunnel/bin/client\n",
            "\u001b[K\u001b[?25h+ localtunnel@1.9.1\n",
            "added 54 packages from 32 contributors in 2.808s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "dR4F-SFEi8hy",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "LOG_DIR = './log'\n",
        "get_ipython().system_raw(\n",
        "    'tensorboard --logdir {} --host 0.0.0.0 --port 6006 &'\n",
        "    .format(LOG_DIR)\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "daDDXbT8i8ap",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "get_ipython().system_raw('./ngrok http 6006 &')\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "0Z4f7OoIjBg8",
        "colab_type": "code",
        "outputId": "a11fab33-d027-40b2-9070-f30a52c92ad7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        }
      },
      "cell_type": "code",
      "source": [
        "! wget https://bin.equinox.io/c/4VmDzA7iaHb/ngrok-stable-linux-amd64.zip\n",
        "! unzip ngrok-stable-linux-amd64.zip"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2018-12-09 02:37:22--  https://bin.equinox.io/c/4VmDzA7iaHb/ngrok-stable-linux-amd64.zip\n",
            "Resolving bin.equinox.io (bin.equinox.io)... 52.202.60.111, 52.55.191.55, 52.201.75.180, ...\n",
            "Connecting to bin.equinox.io (bin.equinox.io)|52.202.60.111|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 5363700 (5.1M) [application/octet-stream]\n",
            "Saving to: ‘ngrok-stable-linux-amd64.zip’\n",
            "\n",
            "ngrok-stable-linux- 100%[===================>]   5.11M  3.53MB/s    in 1.5s    \n",
            "\n",
            "2018-12-09 02:37:25 (3.53 MB/s) - ‘ngrok-stable-linux-amd64.zip’ saved [5363700/5363700]\n",
            "\n",
            "Archive:  ngrok-stable-linux-amd64.zip\n",
            "  inflating: ngrok                   \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "dE7rN34w074a",
        "colab_type": "code",
        "outputId": "ff28eb1f-c69d-4ca9-b1bf-97d691598eff",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "! curl -s http://localhost:4040/api/tunnels | python3 -c \\\n",
        "    \"import sys, json; print(json.load(sys.stdin)['tunnels'][0]['public_url'])\""
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "http://da497423.ngrok.io\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "b-c2L0Q1jDlj",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "rm -r '/content/gdrive/My Drive/419Project.tbgraph'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "0Z-Y0tcnBXN9",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "y_classes"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5jeRekm1_kTN",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "score = model.predict(x_test)\n",
        "y_classes = score.argmax(axis=-1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Kayh7PT5CsUo",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model.save('/content/gdrive/My Drive/alexnet_epoch15_batch256.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "rgVlAXS3ALHQ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def emotion_analysis(emotions):\n",
        "    objects = ('angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral')\n",
        "    y_pos = np.arange(len(objects))\n",
        "    \n",
        "    plt.bar(y_pos, emotions, align='center', alpha=0.5)\n",
        "    plt.xticks(y_pos, objects)\n",
        "    plt.ylabel('percentage')\n",
        "    plt.title('emotion')\n",
        "    \n",
        "    plt.show()\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "hYhMPW3vAYBM",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "monitor_testset_results = False\n",
        "\n",
        "if monitor_testset_results == True:\n",
        "\t#make predictions for test set\n",
        "\tpredictions = model.predict(x_test)\n",
        "\n",
        "\tindex = 0\n",
        "\tfor i in predictions:\n",
        "\t\tif index < 30 and index >= 20:\n",
        "\t\t\t#print(i) #predicted scores\n",
        "\t\t\t#print(y_test[index]) #actual scores\n",
        "\t\t\t\n",
        "\t\t\ttesting_img = np.array(x_test[index], 'float32')\n",
        "\t\t\ttesting_img = testing_img.reshape([48, 48]);\n",
        "\t\t\t\n",
        "\t\t\tplt.gray()\n",
        "\t\t\tplt.imshow(testing_img)\n",
        "\t\t\tplt.show()\n",
        "\t\t\t\n",
        "\t\t\tprint(i)\n",
        "\t\t\t\n",
        "\t\t\temotion_analysis(i)\n",
        "\t\t\tprint(\"----------------------------------------------\")\n",
        "\t\tindex = index + 1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "vXUROIc0AcQj",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### Test Picture"
      ]
    },
    {
      "metadata": {
        "id": "tIF4WBot0RXf",
        "colab_type": "code",
        "outputId": "85b5cde2-2d0f-463e-9d72-bc06491e610f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        }
      },
      "cell_type": "code",
      "source": [
        "!wget https://nricafe.com/wp-content/uploads/2015/09/Happy-Indian.jpg"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2018-12-04 06:20:36--  https://nricafe.com/wp-content/uploads/2015/09/Happy-Indian.jpg\n",
            "Resolving nricafe.com (nricafe.com)... 104.27.184.42, 104.27.185.42, 2606:4700:30::681b:b92a, ...\n",
            "Connecting to nricafe.com (nricafe.com)|104.27.184.42|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 69751 (68K) [image/jpeg]\n",
            "Saving to: ‘Happy-Indian.jpg’\n",
            "\n",
            "\rHappy-Indian.jpg      0%[                    ]       0  --.-KB/s               \rHappy-Indian.jpg    100%[===================>]  68.12K  --.-KB/s    in 0.007s  \n",
            "\n",
            "2018-12-04 06:20:37 (8.93 MB/s) - ‘Happy-Indian.jpg’ saved [69751/69751]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "c-lafymXAX-W",
        "colab_type": "code",
        "outputId": "7ccae9da-4034-4265-c394-f05469608efa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 747
        }
      },
      "cell_type": "code",
      "source": [
        "img = image.load_img(\"Happy-Indian.jpg\", grayscale=True, target_size=(48, 48))\n",
        "\n",
        "x = image.img_to_array(img)\n",
        "x = np.expand_dims(x, axis = 0)\n",
        "\n",
        "x /= 255\n",
        "\n",
        "custom = model.predict(x)\n",
        "emotion_analysis(custom[0])\n",
        "\n",
        "x = np.array(x, 'float32')\n",
        "x = x.reshape([48, 48]);\n",
        "\n",
        "plt.gray()\n",
        "plt.imshow(x)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras_preprocessing/image.py:492: UserWarning: grayscale is deprecated. Please use color_mode = \"grayscale\"\n",
            "  warnings.warn('grayscale is deprecated. Please use '\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFZCAYAAACxGqelAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f7b4e7f9fd0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUsAAAFLCAYAAABft66eAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZSki4kGVpYiIh/8Fu0lsi4KPKWAAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f7b4e766b38>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "NRIRAI_zAX7n",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "tutqH9EiAX4u",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}