emotion1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "emotion1.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "UBtcvhCzIKcM",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "ad7a1771-d16c-4827-fab7-124152850ef6"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive',force_remount=True)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/gdrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RJnh3LbyIL5l",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "c2e9b9b4-db8c-4b3c-fc28-923714d51849"
      },
      "source": [
        "import os\n",
        "import zipfile\n",
        "'''\n",
        "local_zip = '/content/gdrive/My Drive/Dockship/dataset.zip'\n",
        "zip_ref = zipfile.ZipFile(local_zip, 'r')\n",
        "\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()\n",
        "'''"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "\"\\nlocal_zip = '/content/gdrive/My Drive/Dockship/dataset.zip'\\nzip_ref = zipfile.ZipFile(local_zip, 'r')\\n\\nzip_ref.extractall('/tmp')\\nzip_ref.close()\\n\""
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 2
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H3pMDg6ITM-I",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# baseline model with dropout and data augmentation on the cifar10 dataset\n",
        "import sys\n",
        "import tensorflow as tf\n",
        "from matplotlib import pyplot\n",
        "from keras.utils import to_categorical\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Conv2D,MaxPooling2D,Dense,Flatten,Dropout,BatchNormalization,AveragePooling2D\n",
        "from keras.optimizers import SGD\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from sklearn.model_selection import train_test_split\n",
        "from keras.preprocessing.image import img_to_array\n",
        "from keras.preprocessing.image import load_img\n",
        "from imutils import paths\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import random\n",
        "import os"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8N63dW1iIoO5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data_train = '/content/gdrive/My Drive/emo'"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HsddBMnmTPSX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "e6463c04-49bb-4682-d27b-5173f3f61d1f"
      },
      "source": [
        "imagepaths = list(paths.list_images(data_train))\n",
        "random.shuffle(imagepaths)\n",
        "random.seed(42)\n",
        "data_images = []\n",
        "labels = []\n",
        "for imgpath in imagepaths:\n",
        "\tlabel = imgpath.split(os.path.sep)[-2]\n",
        "\timg = load_img(imgpath, target_size=(32, 32))\n",
        "\tdata_images.append(img_to_array(img)/255.)\n",
        "\tlabels.append(label) "
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/PIL/Image.py:932: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images\n",
            "  \"Palette images with Transparency expressed in bytes should be \"\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zdd2zWMvDp8d",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.preprocessing import LabelEncoder\n",
        "lb = LabelEncoder()\n",
        "labels_z = lb.fit_transform(np.array(labels))\n",
        "labels_value = to_categorical(labels_z,num_classes=6)\n",
        "data = np.array(data_images,dtype='float32')"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8ZIwxrwYq2ZI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "trainX, testX,trainY, testY = train_test_split(data, labels_value,test_size=0.2, random_state=42,shuffle = True)"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CKhQTfr3oGVT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def define_model():\n",
        "\tmodel = Sequential()\n",
        "\tmodel.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.1))\n",
        "\tmodel.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.2))\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.3))\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.4))\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.5))\n",
        "\tmodel.add(Flatten())\n",
        "\tmodel.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Dropout(0.5))\n",
        "\tmodel.add(Dense(128, activation='relu', kernel_initializer='he_uniform'));return model\n",
        "\n",
        "mobile = define_model()"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wS_ym8KTqOly",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "op_layer = mobile.output\n",
        "final_layer = Dense(128,activation='relu',kernel_initializer='he_uniform')(op_layer)\n",
        "final_layer = Dense(128,activation='relu',kernel_initializer='he_uniform')(final_layer)\n",
        "final_layer = Dense(6,activation= 'softmax')(final_layer)\n",
        "from keras.models  import Model\n",
        "# Define model input and output\n",
        "model = Model(inputs = mobile.input , outputs = final_layer)\n",
        "opt = SGD(lr=0.01, momentum=0.9)\n",
        "import keras\n",
        "optimizer_sgd = keras.optimizers.Adam(learning_rate=0.001)\n",
        "model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IatTCOv4TXlB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "'''\n",
        "# define cnn model\n",
        "def define_model():\n",
        "\tmodel = Sequential()\n",
        "\tmodel.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same', input_shape=(32, 32, 3)))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(32, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.2))\n",
        "\tmodel.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(64, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.3))\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Conv2D(128, (3, 3), activation='relu', kernel_initializer='he_uniform', padding='same'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(MaxPooling2D((2, 2)))\n",
        "\tmodel.add(Dropout(0.4))\n",
        "\tmodel.add(Flatten())\n",
        "\tmodel.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))\n",
        "\tmodel.add(BatchNormalization())\n",
        "\tmodel.add(Dropout(0.5))\n",
        "\tmodel.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))\n",
        "\tmodel.add(Dense(10, activation='softmax'))\n",
        "\t# compile model\n",
        "\topt = SGD(lr=0.001, momentum=0.9)\n",
        "\tmodel.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "\treturn model\n",
        "\n",
        "# define model\n",
        "cnn = define_model()\n",
        "'''"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iGR349OMTejy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# plot diagnostic learning curves\n",
        "import matplotlib.pyplot as plt\n",
        "def summarize_diagnostics(hist):\n",
        "\tplt.plot(hist.history[\"accuracy\"])\n",
        "\tplt.plot(hist.history['val_accuracy'])\n",
        "\tplt.plot(hist.history['loss'])\n",
        "\tplt.plot(hist.history['val_loss'])\n",
        "\tplt.title(\"model accuracy\")\n",
        "\tplt.ylabel(\"Accuracy\")\n",
        "\tplt.xlabel(\"Epoch\")\n",
        "\tplt.legend([\"Accuracy\",\"Validation Accuracy\",\"loss\",\"Validation Loss\"])\n",
        "\tplt.show()"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tE7Cv8uDO7FC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "304764b4-0172-4c9e-de50-452545170f73"
      },
      "source": [
        "# run the test harness for evaluating a model\n",
        "def run_test_harness():\n",
        "\t# create data generator\n",
        "\tdatagen = ImageDataGenerator(width_shift_range=0.1, height_shift_range=0.1, horizontal_flip=True,zoom_range=0.1)\n",
        "\t# prepare iterator\n",
        "\tit_train = datagen.flow(trainX, trainY, batch_size=32)\n",
        "\t# fit model\n",
        "\tsteps = int(trainX.shape[0] / 32)\n",
        "\thist = model.fit_generator(it_train, steps_per_epoch=steps, epochs=100, validation_data=(testX, testY), verbose=1)\n",
        "\t# evaluate model\n",
        "\t_, acc = model.evaluate(testX, testY, verbose=1)\n",
        "\tprint('> %.3f' % (acc * 100.0))\n",
        "\t# learning curves\n",
        "\tsummarize_diagnostics(hist)\n",
        "\n",
        "# entry point, run the test harness\n",
        "run_test_harness()"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.1396 - accuracy: 0.9527 - val_loss: 6.4150 - val_accuracy: 0.4107\n",
            "Epoch 2/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0975 - accuracy: 0.9669 - val_loss: 3.2035 - val_accuracy: 0.5104\n",
            "Epoch 3/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0936 - accuracy: 0.9686 - val_loss: 2.1054 - val_accuracy: 0.6218\n",
            "Epoch 4/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0746 - accuracy: 0.9710 - val_loss: 2.0727 - val_accuracy: 0.6009\n",
            "Epoch 5/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0631 - accuracy: 0.9811 - val_loss: 2.1734 - val_accuracy: 0.6241\n",
            "Epoch 6/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0748 - accuracy: 0.9763 - val_loss: 2.3021 - val_accuracy: 0.6218\n",
            "Epoch 7/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0720 - accuracy: 0.9781 - val_loss: 2.0899 - val_accuracy: 0.6404\n",
            "Epoch 8/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0959 - accuracy: 0.9781 - val_loss: 2.1596 - val_accuracy: 0.6148\n",
            "Epoch 9/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0676 - accuracy: 0.9763 - val_loss: 2.2129 - val_accuracy: 0.6381\n",
            "Epoch 10/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0829 - accuracy: 0.9746 - val_loss: 1.9519 - val_accuracy: 0.6589\n",
            "Epoch 11/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0838 - accuracy: 0.9710 - val_loss: 1.8399 - val_accuracy: 0.6450\n",
            "Epoch 12/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0775 - accuracy: 0.9781 - val_loss: 2.0152 - val_accuracy: 0.6497\n",
            "Epoch 13/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0901 - accuracy: 0.9751 - val_loss: 1.9303 - val_accuracy: 0.6613\n",
            "Epoch 14/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0976 - accuracy: 0.9716 - val_loss: 1.8233 - val_accuracy: 0.6450\n",
            "Epoch 15/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0639 - accuracy: 0.9757 - val_loss: 2.2179 - val_accuracy: 0.6172\n",
            "Epoch 16/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0613 - accuracy: 0.9781 - val_loss: 1.9885 - val_accuracy: 0.6427\n",
            "Epoch 17/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0651 - accuracy: 0.9822 - val_loss: 1.9023 - val_accuracy: 0.6520\n",
            "Epoch 18/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0655 - accuracy: 0.9799 - val_loss: 2.0226 - val_accuracy: 0.6589\n",
            "Epoch 19/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0564 - accuracy: 0.9781 - val_loss: 2.3831 - val_accuracy: 0.6172\n",
            "Epoch 20/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0585 - accuracy: 0.9852 - val_loss: 2.2270 - val_accuracy: 0.6288\n",
            "Epoch 21/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0715 - accuracy: 0.9781 - val_loss: 2.0477 - val_accuracy: 0.6613\n",
            "Epoch 22/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0602 - accuracy: 0.9799 - val_loss: 2.1312 - val_accuracy: 0.6404\n",
            "Epoch 23/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0552 - accuracy: 0.9846 - val_loss: 1.9189 - val_accuracy: 0.6497\n",
            "Epoch 24/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0495 - accuracy: 0.9811 - val_loss: 2.1522 - val_accuracy: 0.6288\n",
            "Epoch 25/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0805 - accuracy: 0.9769 - val_loss: 2.0795 - val_accuracy: 0.6450\n",
            "Epoch 26/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0577 - accuracy: 0.9817 - val_loss: 2.0498 - val_accuracy: 0.6357\n",
            "Epoch 27/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0648 - accuracy: 0.9817 - val_loss: 2.0856 - val_accuracy: 0.6311\n",
            "Epoch 28/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0522 - accuracy: 0.9840 - val_loss: 2.1209 - val_accuracy: 0.6357\n",
            "Epoch 29/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0865 - accuracy: 0.9781 - val_loss: 1.8623 - val_accuracy: 0.6311\n",
            "Epoch 30/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0452 - accuracy: 0.9911 - val_loss: 2.1666 - val_accuracy: 0.6497\n",
            "Epoch 31/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0802 - accuracy: 0.9757 - val_loss: 1.8808 - val_accuracy: 0.6543\n",
            "Epoch 32/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0664 - accuracy: 0.9769 - val_loss: 2.3339 - val_accuracy: 0.6497\n",
            "Epoch 33/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0533 - accuracy: 0.9876 - val_loss: 2.0501 - val_accuracy: 0.6357\n",
            "Epoch 34/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0528 - accuracy: 0.9840 - val_loss: 2.0812 - val_accuracy: 0.6543\n",
            "Epoch 35/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0497 - accuracy: 0.9870 - val_loss: 2.2422 - val_accuracy: 0.6473\n",
            "Epoch 36/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0447 - accuracy: 0.9846 - val_loss: 2.3549 - val_accuracy: 0.6334\n",
            "Epoch 37/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0417 - accuracy: 0.9876 - val_loss: 2.1104 - val_accuracy: 0.6450\n",
            "Epoch 38/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0380 - accuracy: 0.9882 - val_loss: 2.2362 - val_accuracy: 0.6520\n",
            "Epoch 39/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0449 - accuracy: 0.9864 - val_loss: 2.1183 - val_accuracy: 0.6613\n",
            "Epoch 40/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0307 - accuracy: 0.9905 - val_loss: 2.2630 - val_accuracy: 0.6520\n",
            "Epoch 41/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0548 - accuracy: 0.9840 - val_loss: 2.1238 - val_accuracy: 0.6520\n",
            "Epoch 42/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0687 - accuracy: 0.9734 - val_loss: 2.0160 - val_accuracy: 0.6659\n",
            "Epoch 43/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0538 - accuracy: 0.9828 - val_loss: 2.0364 - val_accuracy: 0.6427\n",
            "Epoch 44/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0566 - accuracy: 0.9799 - val_loss: 2.1153 - val_accuracy: 0.6427\n",
            "Epoch 45/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0722 - accuracy: 0.9775 - val_loss: 2.1488 - val_accuracy: 0.6404\n",
            "Epoch 46/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0377 - accuracy: 0.9858 - val_loss: 2.2685 - val_accuracy: 0.6404\n",
            "Epoch 47/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0470 - accuracy: 0.9840 - val_loss: 2.2141 - val_accuracy: 0.6357\n",
            "Epoch 48/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0403 - accuracy: 0.9870 - val_loss: 2.6540 - val_accuracy: 0.6172\n",
            "Epoch 49/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0448 - accuracy: 0.9846 - val_loss: 2.6525 - val_accuracy: 0.6357\n",
            "Epoch 50/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0662 - accuracy: 0.9811 - val_loss: 2.3353 - val_accuracy: 0.6404\n",
            "Epoch 51/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0707 - accuracy: 0.9763 - val_loss: 2.0454 - val_accuracy: 0.6148\n",
            "Epoch 52/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0561 - accuracy: 0.9811 - val_loss: 2.0332 - val_accuracy: 0.6404\n",
            "Epoch 53/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0486 - accuracy: 0.9864 - val_loss: 2.1659 - val_accuracy: 0.6357\n",
            "Epoch 54/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0372 - accuracy: 0.9858 - val_loss: 2.2628 - val_accuracy: 0.6497\n",
            "Epoch 55/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0373 - accuracy: 0.9893 - val_loss: 2.2110 - val_accuracy: 0.6381\n",
            "Epoch 56/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0346 - accuracy: 0.9893 - val_loss: 2.1948 - val_accuracy: 0.6265\n",
            "Epoch 57/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0518 - accuracy: 0.9893 - val_loss: 2.3531 - val_accuracy: 0.6218\n",
            "Epoch 58/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0436 - accuracy: 0.9852 - val_loss: 2.3946 - val_accuracy: 0.6450\n",
            "Epoch 59/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0694 - accuracy: 0.9781 - val_loss: 1.9728 - val_accuracy: 0.6566\n",
            "Epoch 60/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0433 - accuracy: 0.9870 - val_loss: 2.1791 - val_accuracy: 0.6497\n",
            "Epoch 61/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0513 - accuracy: 0.9846 - val_loss: 2.1685 - val_accuracy: 0.6566\n",
            "Epoch 62/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0499 - accuracy: 0.9870 - val_loss: 2.1751 - val_accuracy: 0.6497\n",
            "Epoch 63/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0450 - accuracy: 0.9864 - val_loss: 2.3309 - val_accuracy: 0.6473\n",
            "Epoch 64/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0588 - accuracy: 0.9858 - val_loss: 2.0158 - val_accuracy: 0.6497\n",
            "Epoch 65/100\n",
            "53/53 [==============================] - 1s 18ms/step - loss: 0.0341 - accuracy: 0.9929 - val_loss: 2.1294 - val_accuracy: 0.6357\n",
            "Epoch 66/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0393 - accuracy: 0.9864 - val_loss: 2.4338 - val_accuracy: 0.6102\n",
            "Epoch 67/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0642 - accuracy: 0.9840 - val_loss: 2.0737 - val_accuracy: 0.6497\n",
            "Epoch 68/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0493 - accuracy: 0.9882 - val_loss: 2.3009 - val_accuracy: 0.6172\n",
            "Epoch 69/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0437 - accuracy: 0.9864 - val_loss: 2.2032 - val_accuracy: 0.6450\n",
            "Epoch 70/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0428 - accuracy: 0.9870 - val_loss: 2.1772 - val_accuracy: 0.6729\n",
            "Epoch 71/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0338 - accuracy: 0.9882 - val_loss: 2.2398 - val_accuracy: 0.6636\n",
            "Epoch 72/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0491 - accuracy: 0.9858 - val_loss: 2.1796 - val_accuracy: 0.6497\n",
            "Epoch 73/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0436 - accuracy: 0.9852 - val_loss: 2.1919 - val_accuracy: 0.6357\n",
            "Epoch 74/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0388 - accuracy: 0.9911 - val_loss: 2.2035 - val_accuracy: 0.6404\n",
            "Epoch 75/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0268 - accuracy: 0.9911 - val_loss: 2.3088 - val_accuracy: 0.6288\n",
            "Epoch 76/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0327 - accuracy: 0.9876 - val_loss: 2.4632 - val_accuracy: 0.6172\n",
            "Epoch 77/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0339 - accuracy: 0.9888 - val_loss: 2.4684 - val_accuracy: 0.6311\n",
            "Epoch 78/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0261 - accuracy: 0.9935 - val_loss: 2.5700 - val_accuracy: 0.6288\n",
            "Epoch 79/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0304 - accuracy: 0.9899 - val_loss: 2.4657 - val_accuracy: 0.6450\n",
            "Epoch 80/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0450 - accuracy: 0.9876 - val_loss: 2.3837 - val_accuracy: 0.6450\n",
            "Epoch 81/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0622 - accuracy: 0.9787 - val_loss: 2.0901 - val_accuracy: 0.6543\n",
            "Epoch 82/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0647 - accuracy: 0.9793 - val_loss: 1.8990 - val_accuracy: 0.6404\n",
            "Epoch 83/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0562 - accuracy: 0.9834 - val_loss: 1.9701 - val_accuracy: 0.6473\n",
            "Epoch 84/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0220 - accuracy: 0.9911 - val_loss: 2.2666 - val_accuracy: 0.6427\n",
            "Epoch 85/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0394 - accuracy: 0.9846 - val_loss: 2.2147 - val_accuracy: 0.6543\n",
            "Epoch 86/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0251 - accuracy: 0.9911 - val_loss: 2.5570 - val_accuracy: 0.6357\n",
            "Epoch 87/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0419 - accuracy: 0.9882 - val_loss: 2.2803 - val_accuracy: 0.6288\n",
            "Epoch 88/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0425 - accuracy: 0.9834 - val_loss: 2.2721 - val_accuracy: 0.6589\n",
            "Epoch 89/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0813 - accuracy: 0.9775 - val_loss: 2.1275 - val_accuracy: 0.6334\n",
            "Epoch 90/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0411 - accuracy: 0.9882 - val_loss: 2.2403 - val_accuracy: 0.6427\n",
            "Epoch 91/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0291 - accuracy: 0.9929 - val_loss: 2.3579 - val_accuracy: 0.6357\n",
            "Epoch 92/100\n",
            "53/53 [==============================] - 1s 20ms/step - loss: 0.0882 - accuracy: 0.9757 - val_loss: 2.1148 - val_accuracy: 0.6172\n",
            "Epoch 93/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0451 - accuracy: 0.9870 - val_loss: 2.1535 - val_accuracy: 0.6543\n",
            "Epoch 94/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0415 - accuracy: 0.9846 - val_loss: 2.0376 - val_accuracy: 0.6520\n",
            "Epoch 95/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0267 - accuracy: 0.9923 - val_loss: 2.1481 - val_accuracy: 0.6520\n",
            "Epoch 96/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0303 - accuracy: 0.9893 - val_loss: 2.4766 - val_accuracy: 0.6311\n",
            "Epoch 97/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0287 - accuracy: 0.9893 - val_loss: 2.4564 - val_accuracy: 0.6589\n",
            "Epoch 98/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0480 - accuracy: 0.9876 - val_loss: 2.0711 - val_accuracy: 0.6543\n",
            "Epoch 99/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0272 - accuracy: 0.9882 - val_loss: 2.2237 - val_accuracy: 0.6659\n",
            "Epoch 100/100\n",
            "53/53 [==============================] - 1s 19ms/step - loss: 0.0425 - accuracy: 0.9876 - val_loss: 2.3098 - val_accuracy: 0.6450\n",
            "14/14 [==============================] - 0s 4ms/step - loss: 2.3098 - accuracy: 0.6450\n",
            "> 64.501\n"
          ],
          "name": "stdout"
        },
        {
          "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": {
        "id": "xE96kUIdWnTK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.save('/content/gdrive/My Drive/Dockship/model4.h5')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-ZIcyPJCRvKr",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "ecebc4c3-ef74-4612-a5c2-2da6c728b0ad"
      },
      "source": [
        "from sklearn.metrics import confusion_matrix,classification_report\n",
        "y_pred = model.predict(testX)\n",
        "y_p = np.argmax(y_pred,axis=1)\n",
        "y_true = np.argmax(testY,axis=1)\n",
        "print(confusion_matrix(y_true,y_p))\n",
        "print('Classification report')\n",
        "print(classification_report(y_true,y_p))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[34  1  0  0  0  0]\n",
            " [ 1 46 24 20  4  0]\n",
            " [ 0 11 41  8  7  0]\n",
            " [ 0 23 14 42 12  0]\n",
            " [ 0  9 12  7 25  0]\n",
            " [ 0  0  0  0  0 90]]\n",
            "Classification report\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.97      0.97      0.97        35\n",
            "           1       0.51      0.48      0.50        95\n",
            "           2       0.45      0.61      0.52        67\n",
            "           3       0.55      0.46      0.50        91\n",
            "           4       0.52      0.47      0.50        53\n",
            "           5       1.00      1.00      1.00        90\n",
            "\n",
            "    accuracy                           0.65       431\n",
            "   macro avg       0.67      0.67      0.66       431\n",
            "weighted avg       0.65      0.65      0.64       431\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YqXbyvhF5hdk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.preprocessing import StandardScaler\n",
        "sc = StandardScaler()"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EiLZAB909UhM",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "a49d0749-275f-48b0-b637-7f456bb5c9d7"
      },
      "source": [
        "model_new = Model(inputs = model.input,outputs = op_layer)\n",
        "train_new = sc.fit_transform(model_new.predict(trainX))\n",
        "test_new = sc.fit_transform(model_new.predict(testX))\n",
        "from sklearn.svm import SVC\n",
        "svm = SVC(kernel='rbf')\n",
        "svm.fit(train_new,np.argmax(trainY,axis=1))\n",
        "svm.score(train_new,np.argmax(trainY,axis=1))"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.9953542392566783"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xlkway3w7vkB",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "23c1133f-e0d9-4b71-8825-797c6901b36e"
      },
      "source": [
        "svm.score(test_new,np.argmax(testY,axis=1))"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.642691415313225"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jpYvIb-9ZAiY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "7749c6db-7cc9-4eb3-ef99-fcd69ad52315"
      },
      "source": [
        "from xgboost import XGBClassifier\n",
        "xg = XGBClassifier()\n",
        "xg.fit(train_new,np.argmax(trainY,axis=1))\n",
        "xg.score(train_new,np.argmax(trainY,axis=1))"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1.0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fmow5N56ZiaA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0be0432e-dfa6-4d0d-a30e-60d57ca4ac54"
      },
      "source": [
        "xg.score(test_new,np.argmax(testY,axis=1))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6380510440835266"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CEI8IbAUAKgB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from keras.models import load_model\n",
        "model = load_model('/content/gdrive/My Drive/Dockship/model4.h5',compile = True)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}