Emotion_Detection_1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Emotion_Detection_1.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JxB0El3YTeeY"
      },
      "source": [
        "**Import Required Libraries**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rIiu3R4bzamb"
      },
      "source": [
        "import sys,os\n",
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import keras\n",
        "from keras.models import Sequential \n",
        "from keras.layers import Dense, Dropout, Activation, Flatten  \n",
        "from keras.layers import Conv2D, MaxPooling2D, BatchNormalization,AveragePooling2D\n",
        "from keras.losses import categorical_crossentropy  \n",
        "from keras.optimizers import Adam  \n",
        "from keras.regularizers import l2 \n",
        "from keras.utils import np_utils"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hmVKriscTp9v"
      },
      "source": [
        "**Loading, Reading, and Pre-Processing Dataset**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nX7XB-Eiz8dC",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1dc5222e-dfb4-49da-ff99-4b75b276a7e1"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RHBkBf-10i_z"
      },
      "source": [
        "os.chdir(\"/content/drive/My Drive/Emotion Detection\")"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C88o_Lmf0v_T",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "039c6907-ef0b-4ebd-9227-3ded489d4862"
      },
      "source": [
        "os.listdir()"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['archive.zip',\n",
              " 'm.json',\n",
              " 'Detection_Emotion.h5',\n",
              " 'Emotion_detection.mp4',\n",
              " 'Emotion_Detection_1.ipynb',\n",
              " 'Real_Time_Detection.py',\n",
              " 'AyushKesarwani_Emotion_Detection_Synopsis.pdf',\n",
              " 'Explanation_Emotion_Detection_AyushKesarwani_1018652.pdf']"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WNsyp0Jd0x41",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4f1b3c0f-342b-4852-e337-cad3d1fba52e"
      },
      "source": [
        "df = pd.read_csv(\"archive.zip\")\n",
        "print(df)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "       emotion                                             pixels        Usage\n",
            "0            0  70 80 82 72 58 58 60 63 54 58 60 48 89 115 121...     Training\n",
            "1            0  151 150 147 155 148 133 111 140 170 174 182 15...     Training\n",
            "2            2  231 212 156 164 174 138 161 173 182 200 106 38...     Training\n",
            "3            4  24 32 36 30 32 23 19 20 30 41 21 22 32 34 21 1...     Training\n",
            "4            6  4 0 0 0 0 0 0 0 0 0 0 0 3 15 23 28 48 50 58 84...     Training\n",
            "...        ...                                                ...          ...\n",
            "35882        6  50 36 17 22 23 29 33 39 34 37 37 37 39 43 48 5...  PrivateTest\n",
            "35883        3  178 174 172 173 181 188 191 194 196 199 200 20...  PrivateTest\n",
            "35884        0  17 17 16 23 28 22 19 17 25 26 20 24 31 19 27 9...  PrivateTest\n",
            "35885        3  30 28 28 29 31 30 42 68 79 81 77 67 67 71 63 6...  PrivateTest\n",
            "35886        2  19 13 14 12 13 16 21 33 50 57 71 84 97 108 122...  PrivateTest\n",
            "\n",
            "[35887 rows x 3 columns]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CFEZP00j01t3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "328556cf-42c4-4e4a-b6b5-92d0fab16303"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 35887 entries, 0 to 35886\n",
            "Data columns (total 3 columns):\n",
            " #   Column   Non-Null Count  Dtype \n",
            "---  ------   --------------  ----- \n",
            " 0   emotion  35887 non-null  int64 \n",
            " 1   pixels   35887 non-null  object\n",
            " 2   Usage    35887 non-null  object\n",
            "dtypes: int64(1), object(2)\n",
            "memory usage: 841.2+ KB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_V8cgNmp05eg"
      },
      "source": [
        "X_train,train_y,X_test,test_y=[],[],[],[]  \n",
        "\n",
        "for index, row in df.iterrows():  \n",
        "    val=row['pixels'].split(\" \")  \n",
        "    if 'Training' in row['Usage']:\n",
        "      X_train.append(np.array(val,'float32'))  \n",
        "      train_y.append(row['emotion'])  \n",
        "    elif 'PublicTest' in row['Usage']:  \n",
        "      X_test.append(np.array(val,'float32'))  \n",
        "      test_y.append(row['emotion']) "
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "As7tSc3G09d8"
      },
      "source": [
        "num_features = 64  \n",
        "num_labels = 7  \n",
        "batch_size = 64  \n",
        "epochs = 175\n",
        "width, height = 48, 48  \n",
        "\n",
        "\n",
        "X_train = np.array(X_train,'float32')  \n",
        "train_y = np.array(train_y,'float32')  \n",
        "X_test = np.array(X_test,'float32')  \n",
        "test_y = np.array(test_y,'float32')  \n",
        "\n",
        "train_y=np_utils.to_categorical(train_y, num_classes=num_labels)  \n",
        "test_y=np_utils.to_categorical(test_y, num_classes=num_labels)  "
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TpQHVvMs1Ck-"
      },
      "source": [
        "X_train -= np.mean(X_train, axis=0)  \n",
        "X_train /= np.std(X_train, axis=0)  \n",
        "\n",
        "X_test -= np.mean(X_test, axis=0)  \n",
        "X_test /= np.std(X_test, axis=0)  \n",
        "\n",
        "X_train = X_train.reshape(X_train.shape[0], 48, 48, 1)  \n",
        "\n",
        "X_test = X_test.reshape(X_test.shape[0], 48, 48, 1) "
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CogiK5dPUvoU"
      },
      "source": [
        "**Building Model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vKIZx0_I1GJ_",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2f896fbd-b5e4-49c3-d879-466ad374e731"
      },
      "source": [
        "model = Sequential()  \n",
        "\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), activation='relu', input_shape=(X_train.shape[1:]), padding='same'))  \n",
        "model.add(Conv2D(64,kernel_size= (3, 3), activation='relu', padding='same'))  \n",
        "model.add(BatchNormalization())  \n",
        "model.add(MaxPooling2D(pool_size=(2,2)))  \n",
        "model.add(Dropout(0.3))  \n",
        "\n",
        "#2nd convolution layer  \n",
        "model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))  \n",
        "model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))  \n",
        "model.add(BatchNormalization())  \n",
        "model.add(MaxPooling2D(pool_size=(2,2)))  \n",
        "model.add(Dropout(0.3))  \n",
        "\n",
        "#3rd convolution layer  \n",
        "model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))  \n",
        "model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))  \n",
        "model.add(BatchNormalization())  \n",
        "model.add(MaxPooling2D(pool_size=(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(num_labels, activation='softmax'))  \n",
        "\n",
        "model.summary() "
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 48, 48, 64)        640       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 48, 48, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization (BatchNo (None, 48, 48, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 24, 24, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 24, 24, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 24, 24, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 12, 12, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 12, 12, 128)       73856     \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 12, 12, 128)       147584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 12, 12, 128)       512       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 6, 6, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 4608)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 7)                 32263     \n",
            "=================================================================\n",
            "Total params: 366,151\n",
            "Trainable params: 365,639\n",
            "Non-trainable params: 512\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0VTKhzC8VRKN"
      },
      "source": [
        "**Compiling and Training Model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hZ3r1hfi1QgZ"
      },
      "source": [
        "model.compile(loss=categorical_crossentropy,  \n",
        "              optimizer=Adam(),  \n",
        "              metrics=['accuracy'])  "
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ASF0SP0KZop1"
      },
      "source": [
        "class myCallback(tf.keras.callbacks.Callback):\n",
        "        def on_epoch_end(self, epoch, logs={}):\n",
        "            if(logs.get('accuracy')>=0.99):\n",
        "                print(\"Reached 99% accuracy so cancelling training!\")\n",
        "                self.model.stop_training = True\n",
        "\n",
        "callback = myCallback()"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o8lb5Oj11X6y",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "27012330-258a-4670-e505-eda7afbaa026"
      },
      "source": [
        "history = model.fit(X_train, train_y, \n",
        "          steps_per_epoch=len(X_train) / batch_size, \n",
        "          batch_size=batch_size,  \n",
        "          epochs=epochs,  \n",
        "          verbose=1,  \n",
        "           callbacks=[callback],\n",
        "          validation_data=(X_test, test_y)\n",
        "          )  "
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/175\n",
            "448/448 [==============================] - 14s 23ms/step - loss: 2.1030 - accuracy: 0.2962 - val_loss: 1.6188 - val_accuracy: 0.4074\n",
            "Epoch 2/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 1.5318 - accuracy: 0.4428 - val_loss: 1.5544 - val_accuracy: 0.4466\n",
            "Epoch 3/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 1.3614 - accuracy: 0.4981 - val_loss: 1.2433 - val_accuracy: 0.5258\n",
            "Epoch 4/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 1.2176 - accuracy: 0.5456 - val_loss: 1.2731 - val_accuracy: 0.5230\n",
            "Epoch 5/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 1.1311 - accuracy: 0.5816 - val_loss: 1.2343 - val_accuracy: 0.5475\n",
            "Epoch 6/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 1.0442 - accuracy: 0.6165 - val_loss: 1.1790 - val_accuracy: 0.5701\n",
            "Epoch 7/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.9750 - accuracy: 0.6365 - val_loss: 1.2135 - val_accuracy: 0.5745\n",
            "Epoch 8/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.8942 - accuracy: 0.6679 - val_loss: 1.1762 - val_accuracy: 0.5815\n",
            "Epoch 9/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.8146 - accuracy: 0.7016 - val_loss: 1.1986 - val_accuracy: 0.5690\n",
            "Epoch 10/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.7744 - accuracy: 0.7205 - val_loss: 1.2441 - val_accuracy: 0.5704\n",
            "Epoch 11/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.6951 - accuracy: 0.7460 - val_loss: 1.2834 - val_accuracy: 0.5731\n",
            "Epoch 12/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.6471 - accuracy: 0.7650 - val_loss: 1.2998 - val_accuracy: 0.5740\n",
            "Epoch 13/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.5815 - accuracy: 0.7881 - val_loss: 1.3326 - val_accuracy: 0.5798\n",
            "Epoch 14/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.5208 - accuracy: 0.8085 - val_loss: 1.3084 - val_accuracy: 0.5801\n",
            "Epoch 15/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.4643 - accuracy: 0.8313 - val_loss: 1.3942 - val_accuracy: 0.5731\n",
            "Epoch 16/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.4409 - accuracy: 0.8410 - val_loss: 1.4291 - val_accuracy: 0.5795\n",
            "Epoch 17/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.4020 - accuracy: 0.8544 - val_loss: 1.5519 - val_accuracy: 0.5729\n",
            "Epoch 18/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.3710 - accuracy: 0.8663 - val_loss: 1.5940 - val_accuracy: 0.5745\n",
            "Epoch 19/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.3509 - accuracy: 0.8715 - val_loss: 1.5932 - val_accuracy: 0.5818\n",
            "Epoch 20/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.3249 - accuracy: 0.8826 - val_loss: 1.8073 - val_accuracy: 0.5603\n",
            "Epoch 21/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.3188 - accuracy: 0.8841 - val_loss: 1.7788 - val_accuracy: 0.5676\n",
            "Epoch 22/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2852 - accuracy: 0.8981 - val_loss: 1.7317 - val_accuracy: 0.5874\n",
            "Epoch 23/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2617 - accuracy: 0.9082 - val_loss: 1.7531 - val_accuracy: 0.5885\n",
            "Epoch 24/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2581 - accuracy: 0.9082 - val_loss: 1.7539 - val_accuracy: 0.5809\n",
            "Epoch 25/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2522 - accuracy: 0.9109 - val_loss: 1.8336 - val_accuracy: 0.5745\n",
            "Epoch 26/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2356 - accuracy: 0.9171 - val_loss: 1.8056 - val_accuracy: 0.5812\n",
            "Epoch 27/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2215 - accuracy: 0.9207 - val_loss: 1.9242 - val_accuracy: 0.5851\n",
            "Epoch 28/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2178 - accuracy: 0.9223 - val_loss: 1.9023 - val_accuracy: 0.5868\n",
            "Epoch 29/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2019 - accuracy: 0.9289 - val_loss: 1.9178 - val_accuracy: 0.5860\n",
            "Epoch 30/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2095 - accuracy: 0.9277 - val_loss: 1.9674 - val_accuracy: 0.5804\n",
            "Epoch 31/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.2036 - accuracy: 0.9262 - val_loss: 1.9545 - val_accuracy: 0.5834\n",
            "Epoch 32/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1849 - accuracy: 0.9351 - val_loss: 2.0142 - val_accuracy: 0.5798\n",
            "Epoch 33/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1789 - accuracy: 0.9362 - val_loss: 1.9951 - val_accuracy: 0.5929\n",
            "Epoch 34/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1847 - accuracy: 0.9362 - val_loss: 1.9546 - val_accuracy: 0.5776\n",
            "Epoch 35/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1774 - accuracy: 0.9400 - val_loss: 2.1043 - val_accuracy: 0.5821\n",
            "Epoch 36/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1768 - accuracy: 0.9387 - val_loss: 1.9951 - val_accuracy: 0.5882\n",
            "Epoch 37/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1713 - accuracy: 0.9379 - val_loss: 2.0249 - val_accuracy: 0.5823\n",
            "Epoch 38/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1679 - accuracy: 0.9415 - val_loss: 2.0939 - val_accuracy: 0.5862\n",
            "Epoch 39/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1525 - accuracy: 0.9487 - val_loss: 2.0674 - val_accuracy: 0.5854\n",
            "Epoch 40/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1590 - accuracy: 0.9452 - val_loss: 2.0984 - val_accuracy: 0.5901\n",
            "Epoch 41/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1549 - accuracy: 0.9446 - val_loss: 2.1710 - val_accuracy: 0.5840\n",
            "Epoch 42/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1537 - accuracy: 0.9469 - val_loss: 2.2029 - val_accuracy: 0.5890\n",
            "Epoch 43/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1441 - accuracy: 0.9508 - val_loss: 2.1612 - val_accuracy: 0.5882\n",
            "Epoch 44/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1437 - accuracy: 0.9514 - val_loss: 2.0816 - val_accuracy: 0.5907\n",
            "Epoch 45/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1370 - accuracy: 0.9520 - val_loss: 2.1847 - val_accuracy: 0.5929\n",
            "Epoch 46/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1314 - accuracy: 0.9539 - val_loss: 2.1702 - val_accuracy: 0.5837\n",
            "Epoch 47/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1463 - accuracy: 0.9490 - val_loss: 2.2037 - val_accuracy: 0.5938\n",
            "Epoch 48/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1431 - accuracy: 0.9505 - val_loss: 2.2224 - val_accuracy: 0.5815\n",
            "Epoch 49/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1263 - accuracy: 0.9577 - val_loss: 2.2449 - val_accuracy: 0.5790\n",
            "Epoch 50/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1378 - accuracy: 0.9526 - val_loss: 2.2753 - val_accuracy: 0.5904\n",
            "Epoch 51/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1365 - accuracy: 0.9540 - val_loss: 2.2374 - val_accuracy: 0.5834\n",
            "Epoch 52/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1253 - accuracy: 0.9572 - val_loss: 2.2559 - val_accuracy: 0.5815\n",
            "Epoch 53/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1247 - accuracy: 0.9561 - val_loss: 2.2243 - val_accuracy: 0.5957\n",
            "Epoch 54/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1211 - accuracy: 0.9592 - val_loss: 2.3347 - val_accuracy: 0.5885\n",
            "Epoch 55/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1153 - accuracy: 0.9606 - val_loss: 2.3351 - val_accuracy: 0.5787\n",
            "Epoch 56/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1187 - accuracy: 0.9596 - val_loss: 2.3312 - val_accuracy: 0.5952\n",
            "Epoch 57/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1152 - accuracy: 0.9594 - val_loss: 2.2875 - val_accuracy: 0.5935\n",
            "Epoch 58/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1107 - accuracy: 0.9610 - val_loss: 2.3014 - val_accuracy: 0.5943\n",
            "Epoch 59/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1125 - accuracy: 0.9649 - val_loss: 2.4076 - val_accuracy: 0.5913\n",
            "Epoch 60/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1144 - accuracy: 0.9624 - val_loss: 2.3952 - val_accuracy: 0.5854\n",
            "Epoch 61/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1150 - accuracy: 0.9614 - val_loss: 2.3421 - val_accuracy: 0.5918\n",
            "Epoch 62/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1045 - accuracy: 0.9664 - val_loss: 2.4105 - val_accuracy: 0.5862\n",
            "Epoch 63/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1135 - accuracy: 0.9624 - val_loss: 2.3686 - val_accuracy: 0.5821\n",
            "Epoch 64/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1102 - accuracy: 0.9627 - val_loss: 2.2913 - val_accuracy: 0.5832\n",
            "Epoch 65/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1046 - accuracy: 0.9645 - val_loss: 2.3618 - val_accuracy: 0.5940\n",
            "Epoch 66/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0961 - accuracy: 0.9684 - val_loss: 2.3797 - val_accuracy: 0.5921\n",
            "Epoch 67/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0989 - accuracy: 0.9687 - val_loss: 2.3950 - val_accuracy: 0.5779\n",
            "Epoch 68/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0971 - accuracy: 0.9655 - val_loss: 2.4394 - val_accuracy: 0.5862\n",
            "Epoch 69/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.1078 - accuracy: 0.9622 - val_loss: 2.3600 - val_accuracy: 0.5837\n",
            "Epoch 70/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0985 - accuracy: 0.9670 - val_loss: 2.4183 - val_accuracy: 0.5949\n",
            "Epoch 71/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0998 - accuracy: 0.9681 - val_loss: 2.3575 - val_accuracy: 0.5834\n",
            "Epoch 72/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0903 - accuracy: 0.9697 - val_loss: 2.4655 - val_accuracy: 0.5832\n",
            "Epoch 73/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0920 - accuracy: 0.9702 - val_loss: 2.4930 - val_accuracy: 0.5860\n",
            "Epoch 74/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0986 - accuracy: 0.9679 - val_loss: 2.4088 - val_accuracy: 0.5848\n",
            "Epoch 75/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0948 - accuracy: 0.9698 - val_loss: 2.4980 - val_accuracy: 0.5801\n",
            "Epoch 76/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0842 - accuracy: 0.9728 - val_loss: 2.5413 - val_accuracy: 0.5896\n",
            "Epoch 77/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0965 - accuracy: 0.9680 - val_loss: 2.4521 - val_accuracy: 0.5921\n",
            "Epoch 78/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0914 - accuracy: 0.9714 - val_loss: 2.4600 - val_accuracy: 0.6002\n",
            "Epoch 79/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0873 - accuracy: 0.9704 - val_loss: 2.4500 - val_accuracy: 0.5857\n",
            "Epoch 80/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0922 - accuracy: 0.9703 - val_loss: 2.3614 - val_accuracy: 0.5879\n",
            "Epoch 81/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0904 - accuracy: 0.9702 - val_loss: 2.4569 - val_accuracy: 0.5887\n",
            "Epoch 82/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0829 - accuracy: 0.9742 - val_loss: 2.4872 - val_accuracy: 0.5834\n",
            "Epoch 83/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0820 - accuracy: 0.9733 - val_loss: 2.5014 - val_accuracy: 0.5692\n",
            "Epoch 84/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0867 - accuracy: 0.9722 - val_loss: 2.4806 - val_accuracy: 0.5865\n",
            "Epoch 85/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0775 - accuracy: 0.9742 - val_loss: 2.4820 - val_accuracy: 0.5926\n",
            "Epoch 86/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0802 - accuracy: 0.9732 - val_loss: 2.5363 - val_accuracy: 0.5840\n",
            "Epoch 87/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0823 - accuracy: 0.9703 - val_loss: 2.4834 - val_accuracy: 0.5890\n",
            "Epoch 88/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0786 - accuracy: 0.9751 - val_loss: 2.5238 - val_accuracy: 0.5832\n",
            "Epoch 89/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0877 - accuracy: 0.9729 - val_loss: 2.5548 - val_accuracy: 0.5974\n",
            "Epoch 90/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0764 - accuracy: 0.9741 - val_loss: 2.5885 - val_accuracy: 0.5684\n",
            "Epoch 91/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0754 - accuracy: 0.9757 - val_loss: 2.4675 - val_accuracy: 0.5946\n",
            "Epoch 92/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0801 - accuracy: 0.9745 - val_loss: 2.4621 - val_accuracy: 0.5885\n",
            "Epoch 93/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0778 - accuracy: 0.9749 - val_loss: 2.4947 - val_accuracy: 0.5862\n",
            "Epoch 94/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0753 - accuracy: 0.9764 - val_loss: 2.5615 - val_accuracy: 0.5991\n",
            "Epoch 95/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0755 - accuracy: 0.9754 - val_loss: 2.4937 - val_accuracy: 0.5960\n",
            "Epoch 96/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0764 - accuracy: 0.9765 - val_loss: 2.4410 - val_accuracy: 0.5924\n",
            "Epoch 97/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0752 - accuracy: 0.9745 - val_loss: 2.5866 - val_accuracy: 0.5804\n",
            "Epoch 98/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0763 - accuracy: 0.9749 - val_loss: 2.5120 - val_accuracy: 0.5860\n",
            "Epoch 99/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0708 - accuracy: 0.9765 - val_loss: 2.5351 - val_accuracy: 0.5926\n",
            "Epoch 100/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0760 - accuracy: 0.9761 - val_loss: 2.5937 - val_accuracy: 0.5801\n",
            "Epoch 101/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0833 - accuracy: 0.9751 - val_loss: 2.5526 - val_accuracy: 0.5949\n",
            "Epoch 102/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0726 - accuracy: 0.9772 - val_loss: 2.5597 - val_accuracy: 0.5918\n",
            "Epoch 103/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0684 - accuracy: 0.9783 - val_loss: 2.5496 - val_accuracy: 0.5954\n",
            "Epoch 104/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0678 - accuracy: 0.9794 - val_loss: 2.5854 - val_accuracy: 0.5904\n",
            "Epoch 105/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0683 - accuracy: 0.9794 - val_loss: 2.5114 - val_accuracy: 0.5809\n",
            "Epoch 106/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0721 - accuracy: 0.9770 - val_loss: 2.5543 - val_accuracy: 0.5907\n",
            "Epoch 107/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0720 - accuracy: 0.9756 - val_loss: 2.5956 - val_accuracy: 0.5985\n",
            "Epoch 108/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0713 - accuracy: 0.9763 - val_loss: 2.4407 - val_accuracy: 0.6032\n",
            "Epoch 109/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0749 - accuracy: 0.9771 - val_loss: 2.5437 - val_accuracy: 0.6032\n",
            "Epoch 110/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0701 - accuracy: 0.9789 - val_loss: 2.6110 - val_accuracy: 0.5882\n",
            "Epoch 111/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0683 - accuracy: 0.9782 - val_loss: 2.5349 - val_accuracy: 0.5885\n",
            "Epoch 112/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0707 - accuracy: 0.9776 - val_loss: 2.5652 - val_accuracy: 0.5893\n",
            "Epoch 113/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0667 - accuracy: 0.9793 - val_loss: 2.5547 - val_accuracy: 0.5832\n",
            "Epoch 114/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0718 - accuracy: 0.9782 - val_loss: 2.6136 - val_accuracy: 0.5879\n",
            "Epoch 115/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0677 - accuracy: 0.9796 - val_loss: 2.5939 - val_accuracy: 0.5874\n",
            "Epoch 116/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0608 - accuracy: 0.9811 - val_loss: 2.5155 - val_accuracy: 0.5952\n",
            "Epoch 117/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0607 - accuracy: 0.9813 - val_loss: 2.6159 - val_accuracy: 0.5868\n",
            "Epoch 118/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0594 - accuracy: 0.9823 - val_loss: 2.5419 - val_accuracy: 0.5929\n",
            "Epoch 119/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0607 - accuracy: 0.9813 - val_loss: 2.5272 - val_accuracy: 0.5862\n",
            "Epoch 120/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0647 - accuracy: 0.9805 - val_loss: 2.5995 - val_accuracy: 0.5929\n",
            "Epoch 121/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0646 - accuracy: 0.9801 - val_loss: 2.5635 - val_accuracy: 0.5887\n",
            "Epoch 122/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0632 - accuracy: 0.9798 - val_loss: 2.5753 - val_accuracy: 0.5926\n",
            "Epoch 123/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0739 - accuracy: 0.9790 - val_loss: 2.5697 - val_accuracy: 0.5929\n",
            "Epoch 124/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0604 - accuracy: 0.9805 - val_loss: 2.5625 - val_accuracy: 0.5940\n",
            "Epoch 125/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0654 - accuracy: 0.9810 - val_loss: 2.5608 - val_accuracy: 0.5921\n",
            "Epoch 126/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0560 - accuracy: 0.9811 - val_loss: 2.5938 - val_accuracy: 0.5893\n",
            "Epoch 127/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0630 - accuracy: 0.9816 - val_loss: 2.5341 - val_accuracy: 0.5893\n",
            "Epoch 128/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0597 - accuracy: 0.9822 - val_loss: 2.6072 - val_accuracy: 0.5921\n",
            "Epoch 129/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0614 - accuracy: 0.9814 - val_loss: 2.6022 - val_accuracy: 0.5949\n",
            "Epoch 130/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0587 - accuracy: 0.9824 - val_loss: 2.5470 - val_accuracy: 0.5949\n",
            "Epoch 131/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0621 - accuracy: 0.9802 - val_loss: 2.5863 - val_accuracy: 0.5949\n",
            "Epoch 132/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0695 - accuracy: 0.9778 - val_loss: 2.5675 - val_accuracy: 0.5865\n",
            "Epoch 133/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0563 - accuracy: 0.9823 - val_loss: 2.5328 - val_accuracy: 0.5979\n",
            "Epoch 134/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0631 - accuracy: 0.9806 - val_loss: 2.5914 - val_accuracy: 0.5940\n",
            "Epoch 135/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0605 - accuracy: 0.9818 - val_loss: 2.5848 - val_accuracy: 0.5985\n",
            "Epoch 136/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0535 - accuracy: 0.9833 - val_loss: 2.6080 - val_accuracy: 0.5946\n",
            "Epoch 137/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0573 - accuracy: 0.9834 - val_loss: 2.5446 - val_accuracy: 0.5977\n",
            "Epoch 138/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0552 - accuracy: 0.9831 - val_loss: 2.6069 - val_accuracy: 0.5924\n",
            "Epoch 139/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0555 - accuracy: 0.9811 - val_loss: 2.5421 - val_accuracy: 0.5968\n",
            "Epoch 140/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0569 - accuracy: 0.9814 - val_loss: 2.6587 - val_accuracy: 0.5932\n",
            "Epoch 141/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0571 - accuracy: 0.9825 - val_loss: 2.6256 - val_accuracy: 0.5943\n",
            "Epoch 142/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0603 - accuracy: 0.9828 - val_loss: 2.5892 - val_accuracy: 0.5991\n",
            "Epoch 143/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0586 - accuracy: 0.9825 - val_loss: 2.6054 - val_accuracy: 0.5885\n",
            "Epoch 144/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0599 - accuracy: 0.9823 - val_loss: 2.6200 - val_accuracy: 0.5899\n",
            "Epoch 145/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0593 - accuracy: 0.9818 - val_loss: 2.5854 - val_accuracy: 0.5868\n",
            "Epoch 146/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0555 - accuracy: 0.9843 - val_loss: 2.5812 - val_accuracy: 0.5896\n",
            "Epoch 147/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0527 - accuracy: 0.9821 - val_loss: 2.6322 - val_accuracy: 0.5965\n",
            "Epoch 148/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0547 - accuracy: 0.9842 - val_loss: 2.6936 - val_accuracy: 0.5938\n",
            "Epoch 149/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0519 - accuracy: 0.9837 - val_loss: 2.6553 - val_accuracy: 0.5862\n",
            "Epoch 150/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0555 - accuracy: 0.9836 - val_loss: 2.6852 - val_accuracy: 0.5896\n",
            "Epoch 151/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0598 - accuracy: 0.9810 - val_loss: 2.6507 - val_accuracy: 0.5921\n",
            "Epoch 152/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0555 - accuracy: 0.9832 - val_loss: 2.7183 - val_accuracy: 0.5938\n",
            "Epoch 153/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0490 - accuracy: 0.9847 - val_loss: 2.6140 - val_accuracy: 0.5885\n",
            "Epoch 154/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0528 - accuracy: 0.9826 - val_loss: 2.5729 - val_accuracy: 0.5896\n",
            "Epoch 155/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0556 - accuracy: 0.9819 - val_loss: 2.6662 - val_accuracy: 0.5907\n",
            "Epoch 156/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0576 - accuracy: 0.9838 - val_loss: 2.6423 - val_accuracy: 0.5893\n",
            "Epoch 157/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0522 - accuracy: 0.9842 - val_loss: 2.5882 - val_accuracy: 0.5915\n",
            "Epoch 158/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0508 - accuracy: 0.9857 - val_loss: 2.6159 - val_accuracy: 0.5940\n",
            "Epoch 159/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0545 - accuracy: 0.9846 - val_loss: 2.8219 - val_accuracy: 0.5829\n",
            "Epoch 160/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0456 - accuracy: 0.9853 - val_loss: 2.6936 - val_accuracy: 0.5837\n",
            "Epoch 161/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0504 - accuracy: 0.9869 - val_loss: 2.5804 - val_accuracy: 0.5940\n",
            "Epoch 162/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0509 - accuracy: 0.9852 - val_loss: 2.6976 - val_accuracy: 0.5851\n",
            "Epoch 163/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0481 - accuracy: 0.9846 - val_loss: 2.6349 - val_accuracy: 0.5899\n",
            "Epoch 164/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0525 - accuracy: 0.9852 - val_loss: 2.6719 - val_accuracy: 0.5826\n",
            "Epoch 165/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0515 - accuracy: 0.9845 - val_loss: 2.6654 - val_accuracy: 0.5879\n",
            "Epoch 166/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0492 - accuracy: 0.9857 - val_loss: 2.6665 - val_accuracy: 0.5879\n",
            "Epoch 167/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0501 - accuracy: 0.9864 - val_loss: 2.6680 - val_accuracy: 0.5882\n",
            "Epoch 168/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0522 - accuracy: 0.9840 - val_loss: 2.5555 - val_accuracy: 0.5940\n",
            "Epoch 169/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0554 - accuracy: 0.9849 - val_loss: 2.5600 - val_accuracy: 0.5974\n",
            "Epoch 170/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0553 - accuracy: 0.9834 - val_loss: 2.6138 - val_accuracy: 0.5896\n",
            "Epoch 171/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0539 - accuracy: 0.9848 - val_loss: 2.5796 - val_accuracy: 0.5921\n",
            "Epoch 172/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0520 - accuracy: 0.9853 - val_loss: 2.6579 - val_accuracy: 0.5899\n",
            "Epoch 173/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0473 - accuracy: 0.9869 - val_loss: 2.6866 - val_accuracy: 0.5874\n",
            "Epoch 174/175\n",
            "448/448 [==============================] - 10s 22ms/step - loss: 0.0505 - accuracy: 0.9854 - val_loss: 2.5912 - val_accuracy: 0.5879\n",
            "Epoch 175/175\n",
            "448/448 [==============================] - 10s 23ms/step - loss: 0.0534 - accuracy: 0.9837 - val_loss: 2.6600 - val_accuracy: 0.5848\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "WyzY058uOeEP",
        "outputId": "d4f50066-c6b0-4246-bf80-e332b8cfc5ab"
      },
      "source": [
        "plt.plot(history.history['accuracy'][0:100])\n",
        "plt.plot(history.history['val_accuracy'][0:100])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training_accuracy', 'validation_accuracy'], loc='lower right')\n",
        "plt.show()"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "-XlKR69VR--4",
        "outputId": "24fa7654-5a4a-4430-d79e-dc0806f6512a"
      },
      "source": [
        "plt.plot(history.history['loss'][0:100])\n",
        "plt.plot(history.history['val_loss'][0:100])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training_loss', 'validation_loss'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AJ85iiv0VYG4"
      },
      "source": [
        "**Save Model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cm95vbYu1h6K"
      },
      "source": [
        "m_json = model.to_json()  \n",
        "with open(\"m.json\", \"w\") as json_file:  \n",
        "    json_file.write(m_json)  \n",
        "model.save_weights(\"Detection_Emotion.h5\") "
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C1aQsmjfWEEU"
      },
      "source": [
        "## After Saving the model, I’ll use Opencv to predict Emotions in Real-Time. But first of all, download the saved weights and model. We’ll work on our local machines now. why?? Because VideoCapture doesn’t work in google colab, in other words, using a webcam in google colab isn’t easy.\n",
        "\n",
        "\n",
        "Go to Real_Time_Detection.py file."
      ]
    }
  ]
}