Approach_1_on_6000_VGG16.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "VGG 16 dataset3",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "mzQSC1rdTF1h",
        "outputId": "5c12c807-8916-4aa0-abf4-47c9e548c255",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        }
      },
      "source": [
        "!unzip \"/content/drive/My Drive/ADM/dataset3.zip\" -d \"/content/sample_data/dataset\""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Archive:  /content/drive/My Drive/ADM/dataset3.zip\n",
            "replace /content/sample_data/dataset/dataset3/test_images_from_train/00fb450622785388.jpg? [y]es, [n]o, [A]ll, [N]one, [r]ename: "
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jQ0WERIqYH7p"
      },
      "source": [
        "# Make sure we are pointing to the directory that has all the files necessary\n",
        "import numpy as np \n",
        "import pandas as pd \n",
        "import sys, requests, shutil, os\n",
        "import numpy as np\n",
        "from shutil import copyfile\n",
        "import urllib\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dropout, Flatten, Dense, GlobalAveragePooling2D\n",
        "from tensorflow.keras import applications\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras import optimizers\n",
        "#from tensorflow.keras.utils.np_utils import to_categorical\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint\n",
        "from tensorflow.keras.models import Model\n",
        "import csv\n",
        "import os\n",
        "import cv2\n",
        "from tensorflow.keras.models import load_model\n",
        "import matplotlib.pyplot as plt\n",
        "import math\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from sklearn.model_selection import train_test_split\n",
        "from tensorflow.keras.preprocessing.image import img_to_array\n",
        "from tensorflow.keras.utils import to_categorical\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import argparse\n",
        "import random\n",
        "import tensorflow as tf\n",
        "import tensorflow.keras\n",
        "from keras.optimizers import SGD, Adam"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-uR2DbbxZNzK",
        "outputId": "908eb1bb-691c-4525-dcd1-19c3de9eeb33",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "cd \"/content/sample_data/dataset/dataset3\""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/sample_data/dataset/dataset3\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vKoFsKl8ToMq",
        "outputId": "9dec5cdf-4810-4cad-842d-d51acb391c7b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# This is where we fine tune the pretrained model according to our dataset\n",
        "img_width, img_height = 96, 96\n",
        "save_model_weights = \"VGG16_weights.h5\"\n",
        "train_data_dir = 'train_images_model'\n",
        "validation_data_dir = 'validation_images_model'\n",
        "batch_size = 200\n",
        "epochs = 100\n",
        "def train_VGG16():\n",
        "\n",
        "    \n",
        "\n",
        "    base_model = applications.VGG16(weights='imagenet',include_top= False,input_shape=(96,96,3))\n",
        "    top_model = Sequential()\n",
        "    top_model.add(Flatten(input_shape=base_model.output_shape[1:]))\n",
        "    top_model.add(Dense(256, activation='relu'))\n",
        "    top_model.add(Dense(256, activation='relu'))\n",
        "    n_class = 6000\n",
        "    top_model.add(Dense(n_class, activation='softmax'))    \n",
        "\n",
        "    model = Model(base_model.input, top_model(base_model.output))\n",
        "    # set the first 16 layers to non-trainable (weights will not be updated) \n",
        "    # 1 conv layer and three dense layers will be trained\n",
        "    for layer in model.layers[:16]:\n",
        "        layer.trainable = False\n",
        "    #model.load_weights(\"/content/sample_data/dataset/dataset3/VGG16_weights.h5\")\n",
        "    model.compile(loss='categorical_crossentropy',\n",
        "                  optimizer=optimizers.Adam(lr=0.001, beta_1=0.9,beta_2=0.999,epsilon=1e-8, decay=0.0),\n",
        "                  metrics=['accuracy'])\n",
        "    print ('Compilation done.')\n",
        "    \n",
        "    train_datagen = ImageDataGenerator(rescale=1. / 255,\n",
        "                                       rotation_range=90,\n",
        "                                        width_shift_range=0.2,\n",
        "                                        height_shift_range=0.2,\n",
        "                                        zoom_range = 0.5)\n",
        "    \n",
        "    valid_datagen = ImageDataGenerator(rescale=1. / 255)\n",
        "\n",
        "    train_generator = train_datagen.flow_from_directory(\n",
        "        train_data_dir,\n",
        "        target_size=(img_height, img_width),\n",
        "        batch_size=batch_size,\n",
        "        class_mode='categorical')\n",
        "\n",
        "    #np.save('class_indices.npy', train_generator.class_indices)\n",
        "\n",
        "    validation_generator = valid_datagen.flow_from_directory(\n",
        "        validation_data_dir,\n",
        "        target_size=(img_height, img_width),\n",
        "        batch_size=batch_size,\n",
        "        class_mode='categorical')\n",
        "    \n",
        "    print ('Model fit begins...')\n",
        "    history1=model.fit_generator(\n",
        "        train_generator,\n",
        "        steps_per_epoch=150,\n",
        "        epochs=epochs,\n",
        "        validation_data=validation_generator,\n",
        "        validation_steps=75,\n",
        "        callbacks=[ModelCheckpoint(filepath=\"vgg16_weights_tf_dim_ordering_tf_kernels-notop.h5\", \n",
        "                                   save_best_only=True, save_weights_only=True)]\n",
        "        )\n",
        "    \n",
        "    model.save_weights(save_model_weights)\n",
        "    # final weights are saved in bottleneck_fc_model.h5 file\n",
        "    return history1\n",
        "history1=train_VGG16()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Compilation done.\n",
            "Found 246794 images belonging to 6000 classes.\n",
            "Found 54916 images belonging to 6000 classes.\n",
            "Model fit begins...\n"
          ],
          "name": "stdout"
        },
        {
          "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"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 7.4300 - accuracy: 0.0090 - val_loss: 7.1718 - val_accuracy: 0.0139\n",
            "Epoch 2/100\n",
            "150/150 [==============================] - 109s 729ms/step - loss: 6.9714 - accuracy: 0.0192 - val_loss: 6.9208 - val_accuracy: 0.0205\n",
            "Epoch 3/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 6.6725 - accuracy: 0.0275 - val_loss: 6.6716 - val_accuracy: 0.0251\n",
            "Epoch 4/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 6.5231 - accuracy: 0.0335 - val_loss: 6.4537 - val_accuracy: 0.0350\n",
            "Epoch 5/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 6.4007 - accuracy: 0.0410 - val_loss: 6.3077 - val_accuracy: 0.0434\n",
            "Epoch 6/100\n",
            "150/150 [==============================] - 109s 727ms/step - loss: 6.2400 - accuracy: 0.0529 - val_loss: 6.3571 - val_accuracy: 0.0443\n",
            "Epoch 7/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 6.1298 - accuracy: 0.0607 - val_loss: 6.1201 - val_accuracy: 0.0615\n",
            "Epoch 8/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 5.9918 - accuracy: 0.0681 - val_loss: 6.0070 - val_accuracy: 0.0675\n",
            "Epoch 9/100\n",
            "150/150 [==============================] - 109s 727ms/step - loss: 5.8889 - accuracy: 0.0771 - val_loss: 5.9660 - val_accuracy: 0.0698\n",
            "Epoch 10/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 5.7831 - accuracy: 0.0857 - val_loss: 5.6975 - val_accuracy: 0.0823\n",
            "Epoch 11/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.7154 - accuracy: 0.0882 - val_loss: 5.6943 - val_accuracy: 0.0902\n",
            "Epoch 12/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 5.6435 - accuracy: 0.0931 - val_loss: 5.5814 - val_accuracy: 0.0971\n",
            "Epoch 13/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.5917 - accuracy: 0.0958 - val_loss: 5.5662 - val_accuracy: 0.1003\n",
            "Epoch 14/100\n",
            "150/150 [==============================] - 109s 729ms/step - loss: 5.5090 - accuracy: 0.0989 - val_loss: 5.4818 - val_accuracy: 0.1073\n",
            "Epoch 15/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.4479 - accuracy: 0.1081 - val_loss: 5.4121 - val_accuracy: 0.1115\n",
            "Epoch 16/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 5.4146 - accuracy: 0.1098 - val_loss: 5.3636 - val_accuracy: 0.1151\n",
            "Epoch 17/100\n",
            "150/150 [==============================] - 109s 725ms/step - loss: 5.3553 - accuracy: 0.1113 - val_loss: 5.3788 - val_accuracy: 0.1195\n",
            "Epoch 18/100\n",
            "150/150 [==============================] - 110s 730ms/step - loss: 5.3490 - accuracy: 0.1151 - val_loss: 5.4474 - val_accuracy: 0.1133\n",
            "Epoch 19/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.2971 - accuracy: 0.1209 - val_loss: 5.3283 - val_accuracy: 0.1272\n",
            "Epoch 20/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 5.2446 - accuracy: 0.1248 - val_loss: 5.2383 - val_accuracy: 0.1331\n",
            "Epoch 21/100\n",
            "150/150 [==============================] - 110s 730ms/step - loss: 5.2162 - accuracy: 0.1228 - val_loss: 5.1954 - val_accuracy: 0.1388\n",
            "Epoch 22/100\n",
            "150/150 [==============================] - 110s 730ms/step - loss: 5.2105 - accuracy: 0.1253 - val_loss: 5.1831 - val_accuracy: 0.1293\n",
            "Epoch 23/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.1667 - accuracy: 0.1315 - val_loss: 5.1527 - val_accuracy: 0.1325\n",
            "Epoch 24/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 5.1345 - accuracy: 0.1336 - val_loss: 5.2340 - val_accuracy: 0.1307\n",
            "Epoch 25/100\n",
            "150/150 [==============================] - 109s 725ms/step - loss: 5.1241 - accuracy: 0.1340 - val_loss: 5.2166 - val_accuracy: 0.1418\n",
            "Epoch 26/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 5.1025 - accuracy: 0.1361 - val_loss: 5.0651 - val_accuracy: 0.1508\n",
            "Epoch 27/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 5.0498 - accuracy: 0.1392 - val_loss: 5.1113 - val_accuracy: 0.1483\n",
            "Epoch 28/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 5.0443 - accuracy: 0.1408 - val_loss: 5.1296 - val_accuracy: 0.1529\n",
            "Epoch 29/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 5.0499 - accuracy: 0.1398 - val_loss: 5.0342 - val_accuracy: 0.1507\n",
            "Epoch 30/100\n",
            "150/150 [==============================] - 108s 721ms/step - loss: 4.9938 - accuracy: 0.1462 - val_loss: 5.0420 - val_accuracy: 0.1569\n",
            "Epoch 31/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.9983 - accuracy: 0.1471 - val_loss: 5.0137 - val_accuracy: 0.1583\n",
            "Epoch 32/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.9832 - accuracy: 0.1482 - val_loss: 5.0043 - val_accuracy: 0.1606\n",
            "Epoch 33/100\n",
            "150/150 [==============================] - 110s 730ms/step - loss: 4.9514 - accuracy: 0.1527 - val_loss: 4.9065 - val_accuracy: 0.1719\n",
            "Epoch 34/100\n",
            "150/150 [==============================] - 109s 725ms/step - loss: 4.9365 - accuracy: 0.1472 - val_loss: 4.9690 - val_accuracy: 0.1697\n",
            "Epoch 35/100\n",
            "150/150 [==============================] - 110s 732ms/step - loss: 4.9152 - accuracy: 0.1490 - val_loss: 4.8987 - val_accuracy: 0.1738\n",
            "Epoch 36/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 4.8905 - accuracy: 0.1535 - val_loss: 4.8897 - val_accuracy: 0.1688\n",
            "Epoch 37/100\n",
            "150/150 [==============================] - 109s 727ms/step - loss: 4.8836 - accuracy: 0.1577 - val_loss: 5.0219 - val_accuracy: 0.1615\n",
            "Epoch 38/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 4.8446 - accuracy: 0.1567 - val_loss: 4.9855 - val_accuracy: 0.1672\n",
            "Epoch 39/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.8623 - accuracy: 0.1564 - val_loss: 4.9100 - val_accuracy: 0.1745\n",
            "Epoch 40/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 4.8319 - accuracy: 0.1631 - val_loss: 4.9599 - val_accuracy: 0.1672\n",
            "Epoch 41/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 4.8018 - accuracy: 0.1641 - val_loss: 4.8263 - val_accuracy: 0.1813\n",
            "Epoch 42/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.8038 - accuracy: 0.1625 - val_loss: 4.8219 - val_accuracy: 0.1865\n",
            "Epoch 43/100\n",
            "150/150 [==============================] - 109s 727ms/step - loss: 4.8120 - accuracy: 0.1608 - val_loss: 4.8227 - val_accuracy: 0.1813\n",
            "Epoch 44/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.8053 - accuracy: 0.1621 - val_loss: 4.7299 - val_accuracy: 0.1912\n",
            "Epoch 45/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 4.7757 - accuracy: 0.1666 - val_loss: 4.9076 - val_accuracy: 0.1785\n",
            "Epoch 46/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 4.7433 - accuracy: 0.1683 - val_loss: 4.8026 - val_accuracy: 0.1888\n",
            "Epoch 47/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.7614 - accuracy: 0.1668 - val_loss: 4.7838 - val_accuracy: 0.1893\n",
            "Epoch 48/100\n",
            "150/150 [==============================] - 109s 729ms/step - loss: 4.7155 - accuracy: 0.1739 - val_loss: 4.8190 - val_accuracy: 0.1835\n",
            "Epoch 49/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 4.7411 - accuracy: 0.1712 - val_loss: 4.7230 - val_accuracy: 0.1908\n",
            "Epoch 50/100\n",
            "150/150 [==============================] - 109s 729ms/step - loss: 4.7027 - accuracy: 0.1734 - val_loss: 4.7255 - val_accuracy: 0.1964\n",
            "Epoch 51/100\n",
            "150/150 [==============================] - 110s 732ms/step - loss: 4.7041 - accuracy: 0.1739 - val_loss: 4.7440 - val_accuracy: 0.1863\n",
            "Epoch 52/100\n",
            "150/150 [==============================] - 111s 738ms/step - loss: 4.6709 - accuracy: 0.1782 - val_loss: 4.9017 - val_accuracy: 0.1861\n",
            "Epoch 53/100\n",
            "150/150 [==============================] - 110s 734ms/step - loss: 4.6747 - accuracy: 0.1765 - val_loss: 4.6901 - val_accuracy: 0.2022\n",
            "Epoch 54/100\n",
            "150/150 [==============================] - 110s 735ms/step - loss: 4.6791 - accuracy: 0.1747 - val_loss: 4.6284 - val_accuracy: 0.2114\n",
            "Epoch 55/100\n",
            "150/150 [==============================] - 111s 739ms/step - loss: 4.6236 - accuracy: 0.1842 - val_loss: 4.8191 - val_accuracy: 0.1897\n",
            "Epoch 56/100\n",
            "150/150 [==============================] - 110s 735ms/step - loss: 4.6483 - accuracy: 0.1785 - val_loss: 4.8465 - val_accuracy: 0.1816\n",
            "Epoch 57/100\n",
            "150/150 [==============================] - 111s 737ms/step - loss: 4.6349 - accuracy: 0.1788 - val_loss: 4.6319 - val_accuracy: 0.1984\n",
            "Epoch 58/100\n",
            "150/150 [==============================] - 111s 740ms/step - loss: 4.6469 - accuracy: 0.1795 - val_loss: 4.8030 - val_accuracy: 0.1935\n",
            "Epoch 59/100\n",
            "150/150 [==============================] - 111s 738ms/step - loss: 4.6271 - accuracy: 0.1807 - val_loss: 4.6667 - val_accuracy: 0.2045\n",
            "Epoch 60/100\n",
            "150/150 [==============================] - 111s 737ms/step - loss: 4.5969 - accuracy: 0.1824 - val_loss: 4.7158 - val_accuracy: 0.1982\n",
            "Epoch 61/100\n",
            "150/150 [==============================] - 111s 739ms/step - loss: 4.6065 - accuracy: 0.1827 - val_loss: 4.7126 - val_accuracy: 0.2009\n",
            "Epoch 62/100\n",
            "150/150 [==============================] - 111s 739ms/step - loss: 4.5993 - accuracy: 0.1829 - val_loss: 4.8235 - val_accuracy: 0.1881\n",
            "Epoch 63/100\n",
            "150/150 [==============================] - 111s 739ms/step - loss: 4.5754 - accuracy: 0.1884 - val_loss: 4.6701 - val_accuracy: 0.2016\n",
            "Epoch 64/100\n",
            "150/150 [==============================] - 111s 739ms/step - loss: 4.6165 - accuracy: 0.1832 - val_loss: 4.6406 - val_accuracy: 0.2095\n",
            "Epoch 65/100\n",
            "150/150 [==============================] - 110s 734ms/step - loss: 4.5615 - accuracy: 0.1873 - val_loss: 4.6796 - val_accuracy: 0.2053\n",
            "Epoch 66/100\n",
            "150/150 [==============================] - 110s 737ms/step - loss: 4.5861 - accuracy: 0.1844 - val_loss: 4.7908 - val_accuracy: 0.2001\n",
            "Epoch 67/100\n",
            "150/150 [==============================] - 110s 734ms/step - loss: 4.5303 - accuracy: 0.1902 - val_loss: 4.6051 - val_accuracy: 0.2157\n",
            "Epoch 68/100\n",
            "150/150 [==============================] - 110s 734ms/step - loss: 4.5441 - accuracy: 0.1920 - val_loss: 4.6863 - val_accuracy: 0.2141\n",
            "Epoch 69/100\n",
            "150/150 [==============================] - 110s 734ms/step - loss: 4.5555 - accuracy: 0.1838 - val_loss: 4.6632 - val_accuracy: 0.2169\n",
            "Epoch 70/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 4.5475 - accuracy: 0.1881 - val_loss: 4.6371 - val_accuracy: 0.2157\n",
            "Epoch 71/100\n",
            "150/150 [==============================] - 109s 730ms/step - loss: 4.5341 - accuracy: 0.1900 - val_loss: 4.6168 - val_accuracy: 0.2135\n",
            "Epoch 72/100\n",
            "150/150 [==============================] - 110s 735ms/step - loss: 4.5506 - accuracy: 0.1891 - val_loss: 4.6198 - val_accuracy: 0.2141\n",
            "Epoch 73/100\n",
            "150/150 [==============================] - 109s 729ms/step - loss: 4.4899 - accuracy: 0.1952 - val_loss: 4.6055 - val_accuracy: 0.2180\n",
            "Epoch 74/100\n",
            "150/150 [==============================] - 109s 730ms/step - loss: 4.5188 - accuracy: 0.1936 - val_loss: 4.7596 - val_accuracy: 0.2003\n",
            "Epoch 75/100\n",
            "150/150 [==============================] - 110s 730ms/step - loss: 4.5118 - accuracy: 0.1942 - val_loss: 4.7052 - val_accuracy: 0.2155\n",
            "Epoch 76/100\n",
            "150/150 [==============================] - 108s 720ms/step - loss: 4.5407 - accuracy: 0.1928 - val_loss: 4.4593 - val_accuracy: 0.2325\n",
            "Epoch 77/100\n",
            "150/150 [==============================] - 109s 725ms/step - loss: 4.5090 - accuracy: 0.1949 - val_loss: 4.7124 - val_accuracy: 0.2168\n",
            "Epoch 78/100\n",
            "150/150 [==============================] - 108s 718ms/step - loss: 4.4910 - accuracy: 0.1924 - val_loss: 4.6655 - val_accuracy: 0.2123\n",
            "Epoch 79/100\n",
            "150/150 [==============================] - 108s 721ms/step - loss: 4.4860 - accuracy: 0.1936 - val_loss: 4.5779 - val_accuracy: 0.2245\n",
            "Epoch 80/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 4.4877 - accuracy: 0.1953 - val_loss: 4.5497 - val_accuracy: 0.2271\n",
            "Epoch 81/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.4845 - accuracy: 0.1953 - val_loss: 4.7317 - val_accuracy: 0.2072\n",
            "Epoch 82/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 4.4688 - accuracy: 0.1976 - val_loss: 4.5545 - val_accuracy: 0.2273\n",
            "Epoch 83/100\n",
            "150/150 [==============================] - 111s 743ms/step - loss: 4.4392 - accuracy: 0.1977 - val_loss: 4.5983 - val_accuracy: 0.2251\n",
            "Epoch 84/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 4.4504 - accuracy: 0.1988 - val_loss: 4.6340 - val_accuracy: 0.2145\n",
            "Epoch 85/100\n",
            "150/150 [==============================] - 110s 731ms/step - loss: 4.4534 - accuracy: 0.1980 - val_loss: 4.5306 - val_accuracy: 0.2259\n",
            "Epoch 86/100\n",
            "150/150 [==============================] - 110s 732ms/step - loss: 4.4513 - accuracy: 0.1970 - val_loss: 4.5527 - val_accuracy: 0.2299\n",
            "Epoch 87/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.4328 - accuracy: 0.1986 - val_loss: 4.5790 - val_accuracy: 0.2223\n",
            "Epoch 88/100\n",
            "150/150 [==============================] - 110s 732ms/step - loss: 4.4427 - accuracy: 0.1980 - val_loss: 4.5252 - val_accuracy: 0.2321\n",
            "Epoch 89/100\n",
            "150/150 [==============================] - 109s 730ms/step - loss: 4.4634 - accuracy: 0.1975 - val_loss: 4.6184 - val_accuracy: 0.2179\n",
            "Epoch 90/100\n",
            "150/150 [==============================] - 109s 728ms/step - loss: 4.4463 - accuracy: 0.1976 - val_loss: 4.6786 - val_accuracy: 0.2193\n",
            "Epoch 91/100\n",
            "150/150 [==============================] - 108s 722ms/step - loss: 4.4536 - accuracy: 0.1973 - val_loss: 4.6293 - val_accuracy: 0.2164\n",
            "Epoch 92/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.4362 - accuracy: 0.2010 - val_loss: 4.6235 - val_accuracy: 0.2140\n",
            "Epoch 93/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.4205 - accuracy: 0.2005 - val_loss: 4.6258 - val_accuracy: 0.2278\n",
            "Epoch 94/100\n",
            "150/150 [==============================] - 109s 725ms/step - loss: 4.4210 - accuracy: 0.2039 - val_loss: 4.5929 - val_accuracy: 0.2247\n",
            "Epoch 95/100\n",
            "150/150 [==============================] - 108s 721ms/step - loss: 4.4216 - accuracy: 0.2008 - val_loss: 4.5078 - val_accuracy: 0.2268\n",
            "Epoch 96/100\n",
            "150/150 [==============================] - 109s 724ms/step - loss: 4.4129 - accuracy: 0.2039 - val_loss: 4.5469 - val_accuracy: 0.2305\n",
            "Epoch 97/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 4.3943 - accuracy: 0.2042 - val_loss: 4.6411 - val_accuracy: 0.2189\n",
            "Epoch 98/100\n",
            "150/150 [==============================] - 109s 726ms/step - loss: 4.4017 - accuracy: 0.2039 - val_loss: 4.6100 - val_accuracy: 0.2287\n",
            "Epoch 99/100\n",
            "150/150 [==============================] - 108s 723ms/step - loss: 4.4194 - accuracy: 0.2001 - val_loss: 4.5044 - val_accuracy: 0.2341\n",
            "Epoch 100/100\n",
            "150/150 [==============================] - 108s 718ms/step - loss: 4.3767 - accuracy: 0.2115 - val_loss: 4.4924 - val_accuracy: 0.2374\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UNHvzimUZH3v",
        "outputId": "7ebc742d-810d-48a1-c72e-b35fb734ecb6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "history1"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f04acdab198>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8GQvsCEE4zyR",
        "outputId": "3699dcca-f18e-42c3-b48b-b0530d066700",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        }
      },
      "source": [
        "acc = history1.history['accuracy']\n",
        "val_acc = history1.history['val_accuracy']\n",
        "loss = history1.history['loss']\n",
        "val_loss = history1.history['val_loss']\n",
        "epochs = range(60)\n",
        "plt.figure()\n",
        "plt.plot(epochs, acc, 'bo', label='Training accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation accuracy')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "plt.show()\n",
        "epochs = range(60)\n",
        "plt.figure()\n",
        "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
        "plt.title('Training and validation loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "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"
          }
        },
        {
          "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": "QoFWWoes_irQ"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}