Mushroom Binary Classification (edible or not).ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Mushroom Binary Classification (edible or not).ipynb",
      "provenance": [],
      "collapsed_sections": [
        "S7eaojUgJfWx",
        "qK4VkMzi4JuX",
        "AIwLkv0u0nQt",
        "cbfJ1Guevbk8",
        "gT-yWrymEK_4",
        "nKL7cjByIZ84",
        "C7o_2EmVIwA9",
        "kDcfZOfCiLah"
      ],
      "mount_file_id": "1Oz0sw_RxHuQQJ-p4xC7jLAM_PVnDGSnF",
      "authorship_tag": "ABX9TyOssW+QQjF/nAsUTi1PulQL"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nX1mP1IKoxQU"
      },
      "source": [
        "Here we will create our model and try to train it\n",
        "\n",
        "---\n",
        "\n",
        "As a pre-trained model Xception was chosen"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S7eaojUgJfWx"
      },
      "source": [
        "## Data Manipulation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qK4VkMzi4JuX"
      },
      "source": [
        "### Data Preparation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xQZFyh0J554Y"
      },
      "source": [
        "As the idea of the project is to detect whether the mushroom is edible or not, poisonous and inedible mushrooms will be labeled as inedible together."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AYJMY6FD5rfz"
      },
      "source": [
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Edible'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Inedible'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZkuO5Qqe4Llv"
      },
      "source": [
        "import os\n",
        "from shutil import copy\n",
        "\n",
        "target_dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data'\n",
        "dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)'\n",
        "\n",
        "ed = '/Edible'\n",
        "ined = '/Inedible'\n",
        "\n",
        "for folder_name in os.listdir(dir + ed):\n",
        "    for image in os.listdir(dir + ed + '/' + folder_name):\n",
        "        copy(dir + ed + '/' + folder_name + '/' + image,\n",
        "             target_dir + ed)\n",
        "        \n",
        "for folder_name in os.listdir(dir + ined):\n",
        "    for image in os.listdir(dir + ined + '/' + folder_name):\n",
        "        copy(dir + ined + '/' + folder_name + '/' + image,\n",
        "             target_dir + ined)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "26bIOU80crvX"
      },
      "source": [
        "target_dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data'\n",
        "dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)'\n",
        "\n",
        "ined = '/Inedible'\n",
        "pois = '/Poisonous'\n",
        "\n",
        "for folder_name in os.listdir(dir + pois):\n",
        "    for image in os.listdir(dir + pois + '/' + folder_name):\n",
        "        copy(dir + pois + '/' + folder_name + '/' + image,\n",
        "             target_dir + ined)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a0NhPP1T7Mht",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f00f1f03-ed50-4a8a-af6e-524cd09fd5b8"
      },
      "source": [
        "import os\n",
        "\n",
        "print('Total number of pictures of edible mushrooms:', len(os.listdir(\n",
        "    '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Edible')))\n",
        "\n",
        "print('Total number of pictures of inedible mushrooms:', len(os.listdir(\n",
        "    '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Inedible')))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Total number of pictures of edible mushrooms: 1479\n",
            "Total number of pictures of inedible mushrooms: 2705\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AIwLkv0u0nQt"
      },
      "source": [
        "### Here we will split our data into train, validation and test"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sB1SOnPQxGc9"
      },
      "source": [
        "**Train:** 750 edible, 1500 inedible\n",
        "\n",
        "\n",
        "---\n",
        "\n",
        "\n",
        "**Validation:** 400 edible, 800 inedible\n",
        "\n",
        "\n",
        "---\n",
        "\n",
        "\n",
        "**Test:** 329 edible, 405 inedible"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sZz9EsMozI09"
      },
      "source": [
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/train'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/train/edible'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/train/inedible'\n",
        "\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/validation'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/validation/edible'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/validation/inedible'\n",
        "\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/test'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/test/edible'\n",
        "! mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/test/inedible'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IHNyxuWx4MPl"
      },
      "source": [
        "import os, shutil\n",
        "\n",
        "original_dataset_dir_ed = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Edible'\n",
        "original_dataset_dir_ined = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Inedible'\n",
        "base_dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data'\n",
        "\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "test_dir = os.path.join(base_dir, 'test')\n",
        "\n",
        "train_ed_dir = os.path.join(train_dir, 'edible')\n",
        "train_ined_dir = os.path.join(train_dir, 'inedible')\n",
        "validation_ed_dir = os.path.join(validation_dir, 'edible')\n",
        "validation_ined_dir = os.path.join(validation_dir, 'inedible')\n",
        "test_ed_dir = os.path.join(test_dir, 'edible')\n",
        "test_ined_dir = os.path.join(test_dir, 'inedible')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vYnJ4p5w0wjz"
      },
      "source": [
        "First we **shuffle** the data and only then distribute it into folders"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "njahnckA4MRe"
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "images_list = os.listdir(original_dataset_dir_ed)\n",
        "images_shuffled = np.random.permutation(images_list)\n",
        "\n",
        "for i in range(750):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ed + '/' + img_name,\n",
        "                train_ed_dir)\n",
        "    \n",
        "for i in range(400):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ed + '/' + img_name,\n",
        "                validation_ed_dir)\n",
        "    \n",
        "for i in range(329):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ed + '/' + img_name,\n",
        "                test_ed_dir)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tzXaZt9n4Lns"
      },
      "source": [
        "images_list = os.listdir(original_dataset_dir_ined)\n",
        "images_shuffled = np.random.permutation(images_list)\n",
        "\n",
        "for i in range(1500):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ined + '/' + img_name,\n",
        "                train_ined_dir)\n",
        "    \n",
        "for i in range(800):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ined + '/' + img_name,\n",
        "                validation_ined_dir)\n",
        "    \n",
        "for i in range(405):\n",
        "    img_name = images_shuffled[i]\n",
        "    shutil.copy(original_dataset_dir_ined + '/' + img_name,\n",
        "                test_ined_dir)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IEqW6QoM4LqY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "be1c88e1-7585-4352-a53b-84b33eb1046d"
      },
      "source": [
        "print('total training edible mushrooms images:', len(os.listdir(train_ed_dir)))\n",
        "print('total training inedible mushrooms images:', len(os.listdir(train_ined_dir)))\n",
        "print('total validation edible mushrooms images:', len(os.listdir(validation_ed_dir)))\n",
        "print('total validation inedible mushrooms images:', len(os.listdir(validation_ined_dir)))\n",
        "print('total test edible mushrooms images:', len(os.listdir(test_ed_dir)))\n",
        "print('total test inedible mushrooms images:', len(os.listdir(test_ined_dir)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "total training edible mushrooms images: 750\n",
            "total training inedible mushrooms images: 1500\n",
            "total validation edible mushrooms images: 400\n",
            "total validation inedible mushrooms images: 800\n",
            "total test edible mushrooms images: 329\n",
            "total test inedible mushrooms images: 405\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cbfJ1Guevbk8"
      },
      "source": [
        "### Data Generation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0F-Gtb1qve8x",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2c14c7b9-5943-4a58-aaf8-537c461153a1"
      },
      "source": [
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "train_datagen = ImageDataGenerator(\n",
        "    rescale=1./255,\n",
        "    rotation_range=40,\n",
        "    width_shift_range=0.2,\n",
        "    height_shift_range=0.2,\n",
        "    shear_range=0.2,\n",
        "    zoom_range=0.2,\n",
        "    horizontal_flip=True,\n",
        "    fill_mode='nearest')\n",
        "\n",
        "test_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "    train_dir, \n",
        "    target_size=(150, 150), \n",
        "    batch_size=20, \n",
        "    class_mode='binary')\n",
        "\n",
        "validation_generator = test_datagen.flow_from_directory(\n",
        "    validation_dir,\n",
        "    target_size=(150, 150),\n",
        "    batch_size=20,\n",
        "    class_mode='binary')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 2250 images belonging to 2 classes.\n",
            "Found 1200 images belonging to 2 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gT-yWrymEK_4"
      },
      "source": [
        "## Complete each time before main model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vZVwIaorJu8t"
      },
      "source": [
        "**TODO:** Add some more Data Augmentation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "omORA921ve-8"
      },
      "source": [
        "import os, shutil\n",
        "\n",
        "original_dataset_dir_ed = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Edible'\n",
        "original_dataset_dir_ined = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data/Inedible'\n",
        "base_dir = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/Prepared Data'\n",
        "\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "test_dir = os.path.join(base_dir, 'test')\n",
        "\n",
        "train_ed_dir = os.path.join(train_dir, 'edible')\n",
        "train_ined_dir = os.path.join(train_dir, 'inedible')\n",
        "validation_ed_dir = os.path.join(validation_dir, 'edible')\n",
        "validation_ined_dir = os.path.join(validation_dir, 'inedible')\n",
        "test_ed_dir = os.path.join(test_dir, 'edible')\n",
        "test_ined_dir = os.path.join(test_dir, 'inedible')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "44rWfQtivfBE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1dc6f1f7-dcf7-4bfb-f900-46481d569001"
      },
      "source": [
        "print('total training edible mushrooms images:', len(os.listdir(train_ed_dir)))\n",
        "print('total training inedible mushrooms images:', len(os.listdir(train_ined_dir)))\n",
        "print('total validation edible mushrooms images:', len(os.listdir(validation_ed_dir)))\n",
        "print('total validation inedible mushrooms images:', len(os.listdir(validation_ined_dir)))\n",
        "print('total test edible mushrooms images:', len(os.listdir(test_ed_dir)))\n",
        "print('total test inedible mushrooms images:', len(os.listdir(test_ined_dir)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "total training edible mushrooms images: 750\n",
            "total training inedible mushrooms images: 1500\n",
            "total validation edible mushrooms images: 400\n",
            "total validation inedible mushrooms images: 800\n",
            "total test edible mushrooms images: 329\n",
            "total test inedible mushrooms images: 405\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UxBpB6b8JRFW",
        "outputId": "60f67a9d-7d9b-403b-8752-7eaea60a20cd"
      },
      "source": [
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "train_datagen = ImageDataGenerator(\n",
        "    rescale=1./255,\n",
        "    rotation_range=40,\n",
        "    width_shift_range=0.2,\n",
        "    height_shift_range=0.2,\n",
        "    shear_range=0.2,\n",
        "    zoom_range=0.2,\n",
        "    horizontal_flip=True,\n",
        "    fill_mode='nearest')\n",
        "\n",
        "test_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "    train_dir, \n",
        "    target_size=(150, 150), \n",
        "    batch_size=20, \n",
        "    class_mode='binary')\n",
        "\n",
        "validation_generator = test_datagen.flow_from_directory(\n",
        "    validation_dir,\n",
        "    target_size=(150, 150),\n",
        "    batch_size=20,\n",
        "    class_mode='binary')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 2250 images belonging to 2 classes.\n",
            "Found 1200 images belonging to 2 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u6LuVcc_4IHl"
      },
      "source": [
        "## Main Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nKL7cjByIZ84"
      },
      "source": [
        "### Xception | Very first simple model (30 epochs, without callbacks)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vBjWggjVTiXa"
      },
      "source": [
        "from keras.applications.xception import Xception"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "90uczQ--owvv",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ecea844e-e78d-4fb8-e4f6-14094c3a233c"
      },
      "source": [
        "conv_base = Xception(weights='imagenet', \n",
        "                     include_top=False, \n",
        "                     input_shape=(150, 150, 3))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/xception/xception_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "83689472/83683744 [==============================] - 1s 0us/step\n",
            "83697664/83683744 [==============================] - 1s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_IqfTS1s3R9q",
        "outputId": "7261b844-bc5c-4fd8-940f-d2cb676add0b"
      },
      "source": [
        "conv_base.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"xception\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 150, 150, 3) 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv1 (Conv2D)           (None, 74, 74, 32)   864         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv1_bn (BatchNormaliza (None, 74, 74, 32)   128         block1_conv1[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv1_act (Activation)   (None, 74, 74, 32)   0           block1_conv1_bn[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv2 (Conv2D)           (None, 72, 72, 64)   18432       block1_conv1_act[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv2_bn (BatchNormaliza (None, 72, 72, 64)   256         block1_conv2[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block1_conv2_act (Activation)   (None, 72, 72, 64)   0           block1_conv2_bn[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block2_sepconv1 (SeparableConv2 (None, 72, 72, 128)  8768        block1_conv2_act[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block2_sepconv1_bn (BatchNormal (None, 72, 72, 128)  512         block2_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block2_sepconv2_act (Activation (None, 72, 72, 128)  0           block2_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block2_sepconv2 (SeparableConv2 (None, 72, 72, 128)  17536       block2_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block2_sepconv2_bn (BatchNormal (None, 72, 72, 128)  512         block2_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 36, 36, 128)  8192        block1_conv2_act[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block2_pool (MaxPooling2D)      (None, 36, 36, 128)  0           block2_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization (BatchNorma (None, 36, 36, 128)  512         conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "add (Add)                       (None, 36, 36, 128)  0           block2_pool[0][0]                \n",
            "                                                                 batch_normalization[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv1_act (Activation (None, 36, 36, 128)  0           add[0][0]                        \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv1 (SeparableConv2 (None, 36, 36, 256)  33920       block3_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv1_bn (BatchNormal (None, 36, 36, 256)  1024        block3_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv2_act (Activation (None, 36, 36, 256)  0           block3_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv2 (SeparableConv2 (None, 36, 36, 256)  67840       block3_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block3_sepconv2_bn (BatchNormal (None, 36, 36, 256)  1024        block3_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 18, 18, 256)  32768       add[0][0]                        \n",
            "__________________________________________________________________________________________________\n",
            "block3_pool (MaxPooling2D)      (None, 18, 18, 256)  0           block3_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_1 (BatchNor (None, 18, 18, 256)  1024        conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "add_1 (Add)                     (None, 18, 18, 256)  0           block3_pool[0][0]                \n",
            "                                                                 batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv1_act (Activation (None, 18, 18, 256)  0           add_1[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv1 (SeparableConv2 (None, 18, 18, 728)  188672      block4_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv1_bn (BatchNormal (None, 18, 18, 728)  2912        block4_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv2_act (Activation (None, 18, 18, 728)  0           block4_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv2 (SeparableConv2 (None, 18, 18, 728)  536536      block4_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block4_sepconv2_bn (BatchNormal (None, 18, 18, 728)  2912        block4_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 9, 9, 728)    186368      add_1[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block4_pool (MaxPooling2D)      (None, 9, 9, 728)    0           block4_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_2 (BatchNor (None, 9, 9, 728)    2912        conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "add_2 (Add)                     (None, 9, 9, 728)    0           block4_pool[0][0]                \n",
            "                                                                 batch_normalization_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv1_act (Activation (None, 9, 9, 728)    0           add_2[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv1 (SeparableConv2 (None, 9, 9, 728)    536536      block5_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv1_bn (BatchNormal (None, 9, 9, 728)    2912        block5_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv2_act (Activation (None, 9, 9, 728)    0           block5_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv2 (SeparableConv2 (None, 9, 9, 728)    536536      block5_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv2_bn (BatchNormal (None, 9, 9, 728)    2912        block5_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv3_act (Activation (None, 9, 9, 728)    0           block5_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv3 (SeparableConv2 (None, 9, 9, 728)    536536      block5_sepconv3_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block5_sepconv3_bn (BatchNormal (None, 9, 9, 728)    2912        block5_sepconv3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "add_3 (Add)                     (None, 9, 9, 728)    0           block5_sepconv3_bn[0][0]         \n",
            "                                                                 add_2[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv1_act (Activation (None, 9, 9, 728)    0           add_3[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv1 (SeparableConv2 (None, 9, 9, 728)    536536      block6_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv1_bn (BatchNormal (None, 9, 9, 728)    2912        block6_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv2_act (Activation (None, 9, 9, 728)    0           block6_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv2 (SeparableConv2 (None, 9, 9, 728)    536536      block6_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv2_bn (BatchNormal (None, 9, 9, 728)    2912        block6_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv3_act (Activation (None, 9, 9, 728)    0           block6_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv3 (SeparableConv2 (None, 9, 9, 728)    536536      block6_sepconv3_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block6_sepconv3_bn (BatchNormal (None, 9, 9, 728)    2912        block6_sepconv3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "add_4 (Add)                     (None, 9, 9, 728)    0           block6_sepconv3_bn[0][0]         \n",
            "                                                                 add_3[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv1_act (Activation (None, 9, 9, 728)    0           add_4[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv1 (SeparableConv2 (None, 9, 9, 728)    536536      block7_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv1_bn (BatchNormal (None, 9, 9, 728)    2912        block7_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv2_act (Activation (None, 9, 9, 728)    0           block7_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv2 (SeparableConv2 (None, 9, 9, 728)    536536      block7_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv2_bn (BatchNormal (None, 9, 9, 728)    2912        block7_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv3_act (Activation (None, 9, 9, 728)    0           block7_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv3 (SeparableConv2 (None, 9, 9, 728)    536536      block7_sepconv3_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block7_sepconv3_bn (BatchNormal (None, 9, 9, 728)    2912        block7_sepconv3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "add_5 (Add)                     (None, 9, 9, 728)    0           block7_sepconv3_bn[0][0]         \n",
            "                                                                 add_4[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv1_act (Activation (None, 9, 9, 728)    0           add_5[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv1 (SeparableConv2 (None, 9, 9, 728)    536536      block8_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv1_bn (BatchNormal (None, 9, 9, 728)    2912        block8_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv2_act (Activation (None, 9, 9, 728)    0           block8_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv2 (SeparableConv2 (None, 9, 9, 728)    536536      block8_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv2_bn (BatchNormal (None, 9, 9, 728)    2912        block8_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv3_act (Activation (None, 9, 9, 728)    0           block8_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv3 (SeparableConv2 (None, 9, 9, 728)    536536      block8_sepconv3_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block8_sepconv3_bn (BatchNormal (None, 9, 9, 728)    2912        block8_sepconv3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "add_6 (Add)                     (None, 9, 9, 728)    0           block8_sepconv3_bn[0][0]         \n",
            "                                                                 add_5[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv1_act (Activation (None, 9, 9, 728)    0           add_6[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv1 (SeparableConv2 (None, 9, 9, 728)    536536      block9_sepconv1_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv1_bn (BatchNormal (None, 9, 9, 728)    2912        block9_sepconv1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv2_act (Activation (None, 9, 9, 728)    0           block9_sepconv1_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv2 (SeparableConv2 (None, 9, 9, 728)    536536      block9_sepconv2_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv2_bn (BatchNormal (None, 9, 9, 728)    2912        block9_sepconv2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv3_act (Activation (None, 9, 9, 728)    0           block9_sepconv2_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv3 (SeparableConv2 (None, 9, 9, 728)    536536      block9_sepconv3_act[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block9_sepconv3_bn (BatchNormal (None, 9, 9, 728)    2912        block9_sepconv3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "add_7 (Add)                     (None, 9, 9, 728)    0           block9_sepconv3_bn[0][0]         \n",
            "                                                                 add_6[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv1_act (Activatio (None, 9, 9, 728)    0           add_7[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv1 (SeparableConv (None, 9, 9, 728)    536536      block10_sepconv1_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv1_bn (BatchNorma (None, 9, 9, 728)    2912        block10_sepconv1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv2_act (Activatio (None, 9, 9, 728)    0           block10_sepconv1_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv2 (SeparableConv (None, 9, 9, 728)    536536      block10_sepconv2_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv2_bn (BatchNorma (None, 9, 9, 728)    2912        block10_sepconv2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv3_act (Activatio (None, 9, 9, 728)    0           block10_sepconv2_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv3 (SeparableConv (None, 9, 9, 728)    536536      block10_sepconv3_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block10_sepconv3_bn (BatchNorma (None, 9, 9, 728)    2912        block10_sepconv3[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "add_8 (Add)                     (None, 9, 9, 728)    0           block10_sepconv3_bn[0][0]        \n",
            "                                                                 add_7[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv1_act (Activatio (None, 9, 9, 728)    0           add_8[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv1 (SeparableConv (None, 9, 9, 728)    536536      block11_sepconv1_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv1_bn (BatchNorma (None, 9, 9, 728)    2912        block11_sepconv1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv2_act (Activatio (None, 9, 9, 728)    0           block11_sepconv1_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv2 (SeparableConv (None, 9, 9, 728)    536536      block11_sepconv2_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv2_bn (BatchNorma (None, 9, 9, 728)    2912        block11_sepconv2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv3_act (Activatio (None, 9, 9, 728)    0           block11_sepconv2_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv3 (SeparableConv (None, 9, 9, 728)    536536      block11_sepconv3_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block11_sepconv3_bn (BatchNorma (None, 9, 9, 728)    2912        block11_sepconv3[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "add_9 (Add)                     (None, 9, 9, 728)    0           block11_sepconv3_bn[0][0]        \n",
            "                                                                 add_8[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv1_act (Activatio (None, 9, 9, 728)    0           add_9[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv1 (SeparableConv (None, 9, 9, 728)    536536      block12_sepconv1_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv1_bn (BatchNorma (None, 9, 9, 728)    2912        block12_sepconv1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv2_act (Activatio (None, 9, 9, 728)    0           block12_sepconv1_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv2 (SeparableConv (None, 9, 9, 728)    536536      block12_sepconv2_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv2_bn (BatchNorma (None, 9, 9, 728)    2912        block12_sepconv2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv3_act (Activatio (None, 9, 9, 728)    0           block12_sepconv2_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv3 (SeparableConv (None, 9, 9, 728)    536536      block12_sepconv3_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block12_sepconv3_bn (BatchNorma (None, 9, 9, 728)    2912        block12_sepconv3[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "add_10 (Add)                    (None, 9, 9, 728)    0           block12_sepconv3_bn[0][0]        \n",
            "                                                                 add_9[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv1_act (Activatio (None, 9, 9, 728)    0           add_10[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv1 (SeparableConv (None, 9, 9, 728)    536536      block13_sepconv1_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv1_bn (BatchNorma (None, 9, 9, 728)    2912        block13_sepconv1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv2_act (Activatio (None, 9, 9, 728)    0           block13_sepconv1_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv2 (SeparableConv (None, 9, 9, 1024)   752024      block13_sepconv2_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block13_sepconv2_bn (BatchNorma (None, 9, 9, 1024)   4096        block13_sepconv2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 5, 5, 1024)   745472      add_10[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "block13_pool (MaxPooling2D)     (None, 5, 5, 1024)   0           block13_sepconv2_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_3 (BatchNor (None, 5, 5, 1024)   4096        conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "add_11 (Add)                    (None, 5, 5, 1024)   0           block13_pool[0][0]               \n",
            "                                                                 batch_normalization_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv1 (SeparableConv (None, 5, 5, 1536)   1582080     add_11[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv1_bn (BatchNorma (None, 5, 5, 1536)   6144        block14_sepconv1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv1_act (Activatio (None, 5, 5, 1536)   0           block14_sepconv1_bn[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv2 (SeparableConv (None, 5, 5, 2048)   3159552     block14_sepconv1_act[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv2_bn (BatchNorma (None, 5, 5, 2048)   8192        block14_sepconv2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block14_sepconv2_act (Activatio (None, 5, 5, 2048)   0           block14_sepconv2_bn[0][0]        \n",
            "==================================================================================================\n",
            "Total params: 20,861,480\n",
            "Trainable params: 20,806,952\n",
            "Non-trainable params: 54,528\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wBfkYpO03SGi",
        "outputId": "d4ef1ea8-5af5-4d28-81ee-4572bb8e6450"
      },
      "source": [
        "from keras.layers import Flatten, Dense\n",
        "from keras.models import Sequential\n",
        "\n",
        "model = Sequential()\n",
        "model.add(conv_base)\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "xception (Functional)        (None, 5, 5, 2048)        20861480  \n",
            "_________________________________________________________________\n",
            "flatten_2 (Flatten)          (None, 51200)             0         \n",
            "_________________________________________________________________\n",
            "dense_4 (Dense)              (None, 128)               6553728   \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 1)                 129       \n",
            "=================================================================\n",
            "Total params: 27,415,337\n",
            "Trainable params: 6,553,857\n",
            "Non-trainable params: 20,861,480\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5eOyEPe73SIz",
        "outputId": "e9fdb68c-4b1e-48e5-8cc5-c7fd9a47f238"
      },
      "source": [
        "print('This is the number of trainable weights', \n",
        "      'before freezing the conv base',\n",
        "      len(model.trainable_weights))\n",
        "\n",
        "conv_base.trainable = False\n",
        "\n",
        "print('This is the number of trainable weights', \n",
        "      'after freezing the conv base',\n",
        "      len(model.trainable_weights))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "This is the number of trainable weights before freezing the conv base 158\n",
            "This is the number of trainable weights after freezing the conv base 4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BDhMG5Tq3SK9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e67fc44f-7987-47fe-87f9-fc7bd57f64ca"
      },
      "source": [
        "model.compile(optimizer='rmsprop', \n",
        "              loss='binary_crossentropy', \n",
        "              metrics=['acc'])\n",
        "\n",
        "history = model.fit(train_generator,\n",
        "                    steps_per_epoch=100,\n",
        "                    epochs=30,\n",
        "                    validation_data=validation_generator,\n",
        "                    validation_steps=50)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "100/100 [==============================] - 37s 340ms/step - loss: 3.8104 - acc: 0.5970 - val_loss: 2.5040 - val_acc: 0.6660\n",
            "Epoch 2/30\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 1.4399 - acc: 0.6525 - val_loss: 1.0398 - val_acc: 0.5960\n",
            "Epoch 3/30\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.9731 - acc: 0.6804 - val_loss: 0.7895 - val_acc: 0.7190\n",
            "Epoch 4/30\n",
            "100/100 [==============================] - 33s 326ms/step - loss: 0.8887 - acc: 0.6739 - val_loss: 0.5259 - val_acc: 0.7310\n",
            "Epoch 5/30\n",
            "100/100 [==============================] - 33s 326ms/step - loss: 0.6395 - acc: 0.6774 - val_loss: 0.5198 - val_acc: 0.7090\n",
            "Epoch 6/30\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.6111 - acc: 0.7040 - val_loss: 0.5081 - val_acc: 0.7150\n",
            "Epoch 7/30\n",
            "100/100 [==============================] - 32s 324ms/step - loss: 0.6296 - acc: 0.7186 - val_loss: 0.5166 - val_acc: 0.7370\n",
            "Epoch 8/30\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.5727 - acc: 0.7176 - val_loss: 0.4928 - val_acc: 0.7690\n",
            "Epoch 9/30\n",
            "100/100 [==============================] - 32s 325ms/step - loss: 0.5477 - acc: 0.7377 - val_loss: 0.5148 - val_acc: 0.7550\n",
            "Epoch 10/30\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.5418 - acc: 0.7327 - val_loss: 0.7459 - val_acc: 0.7370\n",
            "Epoch 11/30\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.5208 - acc: 0.7513 - val_loss: 0.4467 - val_acc: 0.7910\n",
            "Epoch 12/30\n",
            "100/100 [==============================] - 33s 329ms/step - loss: 0.5335 - acc: 0.7518 - val_loss: 0.4666 - val_acc: 0.7850\n",
            "Epoch 13/30\n",
            "100/100 [==============================] - 33s 326ms/step - loss: 0.5092 - acc: 0.7553 - val_loss: 0.4473 - val_acc: 0.7790\n",
            "Epoch 14/30\n",
            "100/100 [==============================] - 33s 329ms/step - loss: 0.4866 - acc: 0.7834 - val_loss: 0.4482 - val_acc: 0.7980\n",
            "Epoch 15/30\n",
            "100/100 [==============================] - 33s 331ms/step - loss: 0.4973 - acc: 0.7568 - val_loss: 0.4584 - val_acc: 0.7780\n",
            "Epoch 16/30\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.4737 - acc: 0.7869 - val_loss: 0.4353 - val_acc: 0.7920\n",
            "Epoch 17/30\n",
            "100/100 [==============================] - 33s 329ms/step - loss: 0.4654 - acc: 0.7824 - val_loss: 0.4311 - val_acc: 0.8130\n",
            "Epoch 18/30\n",
            "100/100 [==============================] - 32s 322ms/step - loss: 0.4674 - acc: 0.7809 - val_loss: 0.4456 - val_acc: 0.8240\n",
            "Epoch 19/30\n",
            "100/100 [==============================] - 32s 324ms/step - loss: 0.4711 - acc: 0.7884 - val_loss: 0.5280 - val_acc: 0.7980\n",
            "Epoch 20/30\n",
            "100/100 [==============================] - 33s 328ms/step - loss: 0.4501 - acc: 0.7980 - val_loss: 0.4122 - val_acc: 0.8050\n",
            "Epoch 21/30\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.4371 - acc: 0.7960 - val_loss: 0.4291 - val_acc: 0.7990\n",
            "Epoch 22/30\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.4520 - acc: 0.7960 - val_loss: 0.3841 - val_acc: 0.8200\n",
            "Epoch 23/30\n",
            "100/100 [==============================] - 33s 329ms/step - loss: 0.4305 - acc: 0.8060 - val_loss: 0.3986 - val_acc: 0.8250\n",
            "Epoch 24/30\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.4129 - acc: 0.8090 - val_loss: 0.4288 - val_acc: 0.8210\n",
            "Epoch 25/30\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.4064 - acc: 0.8240 - val_loss: 0.3764 - val_acc: 0.8270\n",
            "Epoch 26/30\n",
            "100/100 [==============================] - 33s 328ms/step - loss: 0.4175 - acc: 0.8226 - val_loss: 0.3899 - val_acc: 0.8250\n",
            "Epoch 27/30\n",
            "100/100 [==============================] - 32s 320ms/step - loss: 0.4083 - acc: 0.8136 - val_loss: 0.4128 - val_acc: 0.8270\n",
            "Epoch 28/30\n",
            "100/100 [==============================] - 32s 324ms/step - loss: 0.4427 - acc: 0.8131 - val_loss: 0.4149 - val_acc: 0.8340\n",
            "Epoch 29/30\n",
            "100/100 [==============================] - 32s 321ms/step - loss: 0.4167 - acc: 0.8176 - val_loss: 0.4530 - val_acc: 0.8150\n",
            "Epoch 30/30\n",
            "100/100 [==============================] - 33s 327ms/step - loss: 0.4350 - acc: 0.8146 - val_loss: 0.3866 - val_acc: 0.8440\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Weo36URz3SND",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "outputId": "b32c26e6-f9c6-4c8b-e6ff-a3de3b84d625"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\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",
        "\n",
        "plt.figure()\n",
        "\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",
        "\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": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C7o_2EmVIwA9"
      },
      "source": [
        "### Xception | Second model with callbacks"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FfU0dNlgIhrG",
        "outputId": "29a75411-f3d7-4d9d-9ece-1cca2fe58def"
      },
      "source": [
        "from keras.applications.xception import Xception\n",
        "conv_base = Xception(weights='imagenet', \n",
        "                     include_top=False, \n",
        "                     input_shape=(150, 150, 3))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/xception/xception_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "83689472/83683744 [==============================] - 2s 0us/step\n",
            "83697664/83683744 [==============================] - 2s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oB37nvAOIhtb",
        "outputId": "c2eb48d1-79b9-4f33-ac1c-d372562e3dce"
      },
      "source": [
        "from keras.layers import Flatten, Dense\n",
        "from keras.models import Sequential\n",
        "\n",
        "model = Sequential()\n",
        "model.add(conv_base)\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "xception (Functional)        (None, 5, 5, 2048)        20861480  \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 51200)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 128)               6553728   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 1)                 129       \n",
            "=================================================================\n",
            "Total params: 27,415,337\n",
            "Trainable params: 27,360,809\n",
            "Non-trainable params: 54,528\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h16uSOwiIhv4",
        "outputId": "3eacb6d8-5ffb-4bd3-e350-6f3a47b7eda7"
      },
      "source": [
        "print('This is the number of trainable weights', \n",
        "      'before freezing the conv base',\n",
        "      len(model.trainable_weights))\n",
        "\n",
        "conv_base.trainable = False\n",
        "\n",
        "print('This is the number of trainable weights', \n",
        "      'after freezing the conv base',\n",
        "      len(model.trainable_weights))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "This is the number of trainable weights before freezing the conv base 158\n",
            "This is the number of trainable weights after freezing the conv base 4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N__vR4SCJF0G"
      },
      "source": [
        "!mkdir '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/model weights/'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QYyYaU7lJQxa"
      },
      "source": [
        "path = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/model weights/first_weights.hdf5'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RWsKBUx7Isdd"
      },
      "source": [
        "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
        "\n",
        "model_checkpoint_callback = ModelCheckpoint(filepath=path,\n",
        "                                            save_weights_only=True,\n",
        "                                            monitor='val_accuracy',\n",
        "                                            mode='max',\n",
        "                                            save_best_only=True)\n",
        "\n",
        "model_early_stop_callback = EarlyStopping(monitor=\"val_loss\",\n",
        "                                          patience=20,\n",
        "                                          verbose=1,\n",
        "                                          mode=\"min\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SFXUU2CuKi-Y"
      },
      "source": [
        "EPOCHS = 100\n",
        "STEPS_TRAIN = 100\n",
        "STEPS_VAL = 50"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8fc2cyj_Isfv",
        "outputId": "e8366272-a169-4afe-bcd5-887e46836ae3"
      },
      "source": [
        "model.compile(optimizer='rmsprop', \n",
        "              loss='binary_crossentropy', \n",
        "              metrics=['accuracy'])\n",
        "\n",
        "history = model.fit(train_generator,\n",
        "                    steps_per_epoch=STEPS_TRAIN,\n",
        "                    epochs=EPOCHS,\n",
        "                    callbacks=[model_checkpoint_callback,\n",
        "                               model_early_stop_callback],\n",
        "                    validation_data=validation_generator,\n",
        "                    validation_steps=STEPS_VAL)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "100/100 [==============================] - 38s 348ms/step - loss: 0.4562 - accuracy: 0.7920 - val_loss: 0.4141 - val_accuracy: 0.8130\n",
            "Epoch 2/100\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.4228 - accuracy: 0.8075 - val_loss: 0.3829 - val_accuracy: 0.8270\n",
            "Epoch 3/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.4123 - accuracy: 0.8020 - val_loss: 0.3962 - val_accuracy: 0.8030\n",
            "Epoch 4/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.4183 - accuracy: 0.8151 - val_loss: 0.3549 - val_accuracy: 0.8410\n",
            "Epoch 5/100\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.4106 - accuracy: 0.8131 - val_loss: 0.3868 - val_accuracy: 0.8260\n",
            "Epoch 6/100\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.4037 - accuracy: 0.8166 - val_loss: 0.3791 - val_accuracy: 0.8160\n",
            "Epoch 7/100\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.4120 - accuracy: 0.8111 - val_loss: 0.3607 - val_accuracy: 0.8230\n",
            "Epoch 8/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.3988 - accuracy: 0.8221 - val_loss: 0.3523 - val_accuracy: 0.8390\n",
            "Epoch 9/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.4021 - accuracy: 0.8101 - val_loss: 0.3944 - val_accuracy: 0.8250\n",
            "Epoch 10/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.3867 - accuracy: 0.8246 - val_loss: 0.3727 - val_accuracy: 0.8330\n",
            "Epoch 11/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.3929 - accuracy: 0.8201 - val_loss: 0.3271 - val_accuracy: 0.8610\n",
            "Epoch 12/100\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.3797 - accuracy: 0.8335 - val_loss: 0.3542 - val_accuracy: 0.8540\n",
            "Epoch 13/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3731 - accuracy: 0.8291 - val_loss: 0.3900 - val_accuracy: 0.8480\n",
            "Epoch 14/100\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3841 - accuracy: 0.8302 - val_loss: 0.3817 - val_accuracy: 0.8470\n",
            "Epoch 15/100\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.3882 - accuracy: 0.8302 - val_loss: 0.3346 - val_accuracy: 0.8530\n",
            "Epoch 16/100\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.3961 - accuracy: 0.8246 - val_loss: 0.3336 - val_accuracy: 0.8510\n",
            "Epoch 17/100\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.3806 - accuracy: 0.8332 - val_loss: 0.3080 - val_accuracy: 0.8600\n",
            "Epoch 18/100\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3859 - accuracy: 0.8206 - val_loss: 0.3563 - val_accuracy: 0.8390\n",
            "Epoch 19/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.3781 - accuracy: 0.8337 - val_loss: 0.3378 - val_accuracy: 0.8590\n",
            "Epoch 20/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.3812 - accuracy: 0.8340 - val_loss: 0.3286 - val_accuracy: 0.8530\n",
            "Epoch 21/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.3732 - accuracy: 0.8377 - val_loss: 0.3241 - val_accuracy: 0.8450\n",
            "Epoch 22/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3623 - accuracy: 0.8291 - val_loss: 0.3331 - val_accuracy: 0.8480\n",
            "Epoch 23/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.3835 - accuracy: 0.8367 - val_loss: 0.3196 - val_accuracy: 0.8620\n",
            "Epoch 24/100\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.3415 - accuracy: 0.8508 - val_loss: 0.3242 - val_accuracy: 0.8560\n",
            "Epoch 25/100\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.3472 - accuracy: 0.8492 - val_loss: 0.3224 - val_accuracy: 0.8710\n",
            "Epoch 26/100\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3716 - accuracy: 0.8332 - val_loss: 0.3265 - val_accuracy: 0.8580\n",
            "Epoch 27/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.3307 - accuracy: 0.8553 - val_loss: 0.3672 - val_accuracy: 0.8450\n",
            "Epoch 28/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3618 - accuracy: 0.8382 - val_loss: 0.3541 - val_accuracy: 0.8600\n",
            "Epoch 29/100\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.3699 - accuracy: 0.8372 - val_loss: 0.3146 - val_accuracy: 0.8600\n",
            "Epoch 30/100\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.3481 - accuracy: 0.8472 - val_loss: 0.3444 - val_accuracy: 0.8470\n",
            "Epoch 31/100\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3761 - accuracy: 0.8455 - val_loss: 0.3317 - val_accuracy: 0.8660\n",
            "Epoch 32/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.3514 - accuracy: 0.8533 - val_loss: 0.3036 - val_accuracy: 0.8650\n",
            "Epoch 33/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3487 - accuracy: 0.8573 - val_loss: 0.4064 - val_accuracy: 0.8270\n",
            "Epoch 34/100\n",
            "100/100 [==============================] - 33s 328ms/step - loss: 0.3372 - accuracy: 0.8538 - val_loss: 0.4040 - val_accuracy: 0.8290\n",
            "Epoch 35/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.3633 - accuracy: 0.8508 - val_loss: 0.2954 - val_accuracy: 0.8650\n",
            "Epoch 36/100\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.3175 - accuracy: 0.8603 - val_loss: 0.3735 - val_accuracy: 0.8450\n",
            "Epoch 37/100\n",
            "100/100 [==============================] - 33s 327ms/step - loss: 0.3327 - accuracy: 0.8568 - val_loss: 0.2927 - val_accuracy: 0.8830\n",
            "Epoch 38/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3246 - accuracy: 0.8568 - val_loss: 0.4031 - val_accuracy: 0.8310\n",
            "Epoch 39/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3328 - accuracy: 0.8553 - val_loss: 0.3049 - val_accuracy: 0.8660\n",
            "Epoch 40/100\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.3172 - accuracy: 0.8578 - val_loss: 0.3259 - val_accuracy: 0.8640\n",
            "Epoch 41/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3361 - accuracy: 0.8638 - val_loss: 0.3164 - val_accuracy: 0.8610\n",
            "Epoch 42/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3440 - accuracy: 0.8608 - val_loss: 0.3128 - val_accuracy: 0.8690\n",
            "Epoch 43/100\n",
            "100/100 [==============================] - 34s 335ms/step - loss: 0.3083 - accuracy: 0.8630 - val_loss: 0.2875 - val_accuracy: 0.8670\n",
            "Epoch 44/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.3281 - accuracy: 0.8538 - val_loss: 0.2847 - val_accuracy: 0.8860\n",
            "Epoch 45/100\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.3290 - accuracy: 0.8633 - val_loss: 0.2993 - val_accuracy: 0.8750\n",
            "Epoch 46/100\n",
            "100/100 [==============================] - 33s 331ms/step - loss: 0.3338 - accuracy: 0.8495 - val_loss: 0.2671 - val_accuracy: 0.8790\n",
            "Epoch 47/100\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.3219 - accuracy: 0.8615 - val_loss: 0.4041 - val_accuracy: 0.8420\n",
            "Epoch 48/100\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.3170 - accuracy: 0.8683 - val_loss: 0.3077 - val_accuracy: 0.8690\n",
            "Epoch 49/100\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.3352 - accuracy: 0.8553 - val_loss: 0.3083 - val_accuracy: 0.8700\n",
            "Epoch 50/100\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.3095 - accuracy: 0.8710 - val_loss: 0.3017 - val_accuracy: 0.8720\n",
            "Epoch 51/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3203 - accuracy: 0.8693 - val_loss: 0.3632 - val_accuracy: 0.8600\n",
            "Epoch 52/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.3135 - accuracy: 0.8704 - val_loss: 0.3150 - val_accuracy: 0.8780\n",
            "Epoch 53/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.2997 - accuracy: 0.8724 - val_loss: 0.2890 - val_accuracy: 0.8790\n",
            "Epoch 54/100\n",
            "100/100 [==============================] - 36s 358ms/step - loss: 0.3153 - accuracy: 0.8663 - val_loss: 0.2786 - val_accuracy: 0.8890\n",
            "Epoch 55/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.3012 - accuracy: 0.8774 - val_loss: 0.2740 - val_accuracy: 0.8760\n",
            "Epoch 56/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.3027 - accuracy: 0.8779 - val_loss: 0.2692 - val_accuracy: 0.9020\n",
            "Epoch 57/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2793 - accuracy: 0.8754 - val_loss: 0.3038 - val_accuracy: 0.8740\n",
            "Epoch 58/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.3195 - accuracy: 0.8603 - val_loss: 0.2763 - val_accuracy: 0.8840\n",
            "Epoch 59/100\n",
            "100/100 [==============================] - 33s 325ms/step - loss: 0.2949 - accuracy: 0.8804 - val_loss: 0.3098 - val_accuracy: 0.8780\n",
            "Epoch 60/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.3006 - accuracy: 0.8729 - val_loss: 0.2848 - val_accuracy: 0.8900\n",
            "Epoch 61/100\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.3119 - accuracy: 0.8724 - val_loss: 0.2565 - val_accuracy: 0.8880\n",
            "Epoch 62/100\n",
            "100/100 [==============================] - 33s 331ms/step - loss: 0.3075 - accuracy: 0.8714 - val_loss: 0.3241 - val_accuracy: 0.8720\n",
            "Epoch 63/100\n",
            "100/100 [==============================] - 33s 329ms/step - loss: 0.2841 - accuracy: 0.8800 - val_loss: 0.2792 - val_accuracy: 0.8940\n",
            "Epoch 64/100\n",
            "100/100 [==============================] - 33s 331ms/step - loss: 0.2969 - accuracy: 0.8734 - val_loss: 0.2631 - val_accuracy: 0.8930\n",
            "Epoch 65/100\n",
            "100/100 [==============================] - 33s 326ms/step - loss: 0.2852 - accuracy: 0.8789 - val_loss: 0.2991 - val_accuracy: 0.8940\n",
            "Epoch 66/100\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.2783 - accuracy: 0.8879 - val_loss: 0.2580 - val_accuracy: 0.8950\n",
            "Epoch 67/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.2575 - accuracy: 0.8920 - val_loss: 0.2771 - val_accuracy: 0.8890\n",
            "Epoch 68/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.2858 - accuracy: 0.8829 - val_loss: 0.2614 - val_accuracy: 0.8870\n",
            "Epoch 69/100\n",
            "100/100 [==============================] - 33s 328ms/step - loss: 0.2669 - accuracy: 0.8859 - val_loss: 0.2784 - val_accuracy: 0.8840\n",
            "Epoch 70/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.3044 - accuracy: 0.8719 - val_loss: 0.2682 - val_accuracy: 0.8950\n",
            "Epoch 71/100\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.2770 - accuracy: 0.8869 - val_loss: 0.2601 - val_accuracy: 0.8880\n",
            "Epoch 72/100\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.2728 - accuracy: 0.8869 - val_loss: 0.2568 - val_accuracy: 0.8870\n",
            "Epoch 73/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2738 - accuracy: 0.8945 - val_loss: 0.2647 - val_accuracy: 0.8930\n",
            "Epoch 74/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2951 - accuracy: 0.8864 - val_loss: 0.2661 - val_accuracy: 0.8920\n",
            "Epoch 75/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.2671 - accuracy: 0.8775 - val_loss: 0.2808 - val_accuracy: 0.8930\n",
            "Epoch 76/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2609 - accuracy: 0.8854 - val_loss: 0.2711 - val_accuracy: 0.8950\n",
            "Epoch 77/100\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.2803 - accuracy: 0.8849 - val_loss: 0.2525 - val_accuracy: 0.9000\n",
            "Epoch 78/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.3029 - accuracy: 0.8709 - val_loss: 0.2658 - val_accuracy: 0.8980\n",
            "Epoch 79/100\n",
            "100/100 [==============================] - 33s 331ms/step - loss: 0.2518 - accuracy: 0.8930 - val_loss: 0.2618 - val_accuracy: 0.8960\n",
            "Epoch 80/100\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.2814 - accuracy: 0.8889 - val_loss: 0.2554 - val_accuracy: 0.8990\n",
            "Epoch 81/100\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.2672 - accuracy: 0.8799 - val_loss: 0.2232 - val_accuracy: 0.9130\n",
            "Epoch 82/100\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.2852 - accuracy: 0.8859 - val_loss: 0.2475 - val_accuracy: 0.9010\n",
            "Epoch 83/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.2726 - accuracy: 0.8839 - val_loss: 0.3351 - val_accuracy: 0.8540\n",
            "Epoch 84/100\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.2681 - accuracy: 0.8864 - val_loss: 0.2253 - val_accuracy: 0.9110\n",
            "Epoch 85/100\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2628 - accuracy: 0.8955 - val_loss: 0.2760 - val_accuracy: 0.8870\n",
            "Epoch 86/100\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.2598 - accuracy: 0.8945 - val_loss: 0.2815 - val_accuracy: 0.8870\n",
            "Epoch 87/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.2710 - accuracy: 0.8980 - val_loss: 0.2414 - val_accuracy: 0.9080\n",
            "Epoch 88/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2762 - accuracy: 0.8905 - val_loss: 0.2329 - val_accuracy: 0.9110\n",
            "Epoch 89/100\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2472 - accuracy: 0.9035 - val_loss: 0.2518 - val_accuracy: 0.9070\n",
            "Epoch 90/100\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.2785 - accuracy: 0.8844 - val_loss: 0.2935 - val_accuracy: 0.8820\n",
            "Epoch 91/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.2898 - accuracy: 0.8854 - val_loss: 0.2551 - val_accuracy: 0.8950\n",
            "Epoch 92/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2611 - accuracy: 0.8870 - val_loss: 0.2334 - val_accuracy: 0.9080\n",
            "Epoch 93/100\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.2568 - accuracy: 0.8884 - val_loss: 0.2223 - val_accuracy: 0.9090\n",
            "Epoch 94/100\n",
            "100/100 [==============================] - 33s 330ms/step - loss: 0.2727 - accuracy: 0.8985 - val_loss: 0.2448 - val_accuracy: 0.9060\n",
            "Epoch 95/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2565 - accuracy: 0.9030 - val_loss: 0.2271 - val_accuracy: 0.9030\n",
            "Epoch 96/100\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.2571 - accuracy: 0.8945 - val_loss: 0.2190 - val_accuracy: 0.9110\n",
            "Epoch 97/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.2672 - accuracy: 0.8940 - val_loss: 0.2515 - val_accuracy: 0.8940\n",
            "Epoch 98/100\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.2501 - accuracy: 0.9050 - val_loss: 0.2395 - val_accuracy: 0.9040\n",
            "Epoch 99/100\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.2420 - accuracy: 0.9010 - val_loss: 0.2591 - val_accuracy: 0.8970\n",
            "Epoch 100/100\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2571 - accuracy: 0.8975 - val_loss: 0.2613 - val_accuracy: 0.8950\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "Bp7SUuVLIsiF",
        "outputId": "b6fd5250-3f39-4372-a8bc-9bab399c3079"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['accuracy']\n",
        "val_acc = history.history['val_accuracy']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\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",
        "\n",
        "plt.figure()\n",
        "\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",
        "\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": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JmXPwNeQDyFc"
      },
      "source": [
        "## ResNetV2 | Trying smth new"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PeFU6wunFDy2",
        "outputId": "29b78be0-4a6a-4ebd-e086-bec4677e70e4"
      },
      "source": [
        "import keras.applications.resnet_v2 as A\n",
        "dir(A)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['DOC',\n",
              " 'ResNet101V2',\n",
              " 'ResNet152V2',\n",
              " 'ResNet50V2',\n",
              " '__builtins__',\n",
              " '__cached__',\n",
              " '__doc__',\n",
              " '__file__',\n",
              " '__loader__',\n",
              " '__name__',\n",
              " '__package__',\n",
              " '__spec__',\n",
              " 'decode_predictions',\n",
              " 'imagenet_utils',\n",
              " 'keras_export',\n",
              " 'preprocess_input',\n",
              " 'resnet']"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RP5Or_eaD2tk",
        "outputId": "550ccb58-50cf-4f55-b9c7-818d3c8db305"
      },
      "source": [
        "from keras.applications.resnet_v2 import ResNet50V2\n",
        "conv_base = ResNet50V2(weights='imagenet', \n",
        "                       include_top=False, \n",
        "                       input_shape=(150, 150, 3))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50v2_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "94674944/94668760 [==============================] - 1s 0us/step\n",
            "94683136/94668760 [==============================] - 1s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "63RhVOrtD2vo",
        "outputId": "5eabdec5-8cdf-4a1e-99d9-ecc9931a92d0"
      },
      "source": [
        "from keras.layers import Flatten, Dense\n",
        "from keras.models import Sequential\n",
        "\n",
        "model = Sequential()\n",
        "model.add(conv_base)\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "resnet50v2 (Functional)      (None, 5, 5, 2048)        23564800  \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 51200)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 128)               6553728   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 1)                 129       \n",
            "=================================================================\n",
            "Total params: 30,118,657\n",
            "Trainable params: 30,073,217\n",
            "Non-trainable params: 45,440\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "n0erNFpRD2xs",
        "outputId": "8a2f2ef4-6d38-465c-c65d-0f321576f57d"
      },
      "source": [
        "print('This is the number of trainable weights', \n",
        "      'before freezing the conv base',\n",
        "      len(model.trainable_weights))\n",
        "\n",
        "conv_base.trainable = False\n",
        "\n",
        "print('This is the number of trainable weights', \n",
        "      'after freezing the conv base',\n",
        "      len(model.trainable_weights))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "This is the number of trainable weights before freezing the conv base 176\n",
            "This is the number of trainable weights after freezing the conv base 4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XMEAnCl8D20T"
      },
      "source": [
        "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
        "\n",
        "new_weights = 'model weights/second_weights_ResNetV2.hdf5'\n",
        "path = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)/' + \\\n",
        "        new_weights\n",
        "\n",
        "model_checkpoint_callback = ModelCheckpoint(filepath=path,\n",
        "                                            save_weights_only=True,\n",
        "                                            monitor='val_accuracy',\n",
        "                                            mode='max',\n",
        "                                            save_best_only=True)\n",
        "\n",
        "model_early_stop_callback = EarlyStopping(monitor=\"val_loss\",\n",
        "                                          patience=20,\n",
        "                                          verbose=1,\n",
        "                                          mode=\"min\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nznkFnPOGKEv"
      },
      "source": [
        "EPOCHS = 500\n",
        "STEPS_TRAIN = 100\n",
        "STEPS_VAL = 50"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PrV13cC1GKG2",
        "outputId": "2c80cac3-ee21-4025-b5d8-a91e2d6965b5"
      },
      "source": [
        "model.compile(optimizer='rmsprop', \n",
        "              loss='binary_crossentropy', \n",
        "              metrics=['accuracy'])\n",
        "\n",
        "history = model.fit(train_generator,\n",
        "                    steps_per_epoch=STEPS_TRAIN,\n",
        "                    epochs=EPOCHS,\n",
        "                    callbacks=[model_checkpoint_callback,\n",
        "                               model_early_stop_callback],\n",
        "                    validation_data=validation_generator,\n",
        "                    validation_steps=STEPS_VAL)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/500\n",
            "100/100 [==============================] - 679s 7s/step - loss: 2.6786 - accuracy: 0.6447 - val_loss: 0.5673 - val_accuracy: 0.6850\n",
            "Epoch 2/500\n",
            "100/100 [==============================] - 102s 1s/step - loss: 0.6578 - accuracy: 0.6663 - val_loss: 0.5740 - val_accuracy: 0.6940\n",
            "Epoch 3/500\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.6645 - accuracy: 0.6729 - val_loss: 0.5489 - val_accuracy: 0.7190\n",
            "Epoch 4/500\n",
            "100/100 [==============================] - 36s 361ms/step - loss: 0.6413 - accuracy: 0.6910 - val_loss: 0.5970 - val_accuracy: 0.7310\n",
            "Epoch 5/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.6023 - accuracy: 0.7025 - val_loss: 0.5247 - val_accuracy: 0.7250\n",
            "Epoch 6/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.5683 - accuracy: 0.7116 - val_loss: 0.5185 - val_accuracy: 0.7610\n",
            "Epoch 7/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.5675 - accuracy: 0.7291 - val_loss: 0.4993 - val_accuracy: 0.7580\n",
            "Epoch 8/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.5679 - accuracy: 0.7410 - val_loss: 0.5150 - val_accuracy: 0.7730\n",
            "Epoch 9/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.5660 - accuracy: 0.7462 - val_loss: 0.4659 - val_accuracy: 0.7780\n",
            "Epoch 10/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.5070 - accuracy: 0.7633 - val_loss: 0.4284 - val_accuracy: 0.8080\n",
            "Epoch 11/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.5062 - accuracy: 0.7663 - val_loss: 0.4585 - val_accuracy: 0.8070\n",
            "Epoch 12/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.4814 - accuracy: 0.7759 - val_loss: 0.4105 - val_accuracy: 0.8110\n",
            "Epoch 13/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.4856 - accuracy: 0.7905 - val_loss: 0.4433 - val_accuracy: 0.8250\n",
            "Epoch 14/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.4722 - accuracy: 0.7829 - val_loss: 0.6782 - val_accuracy: 0.7120\n",
            "Epoch 15/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.4819 - accuracy: 0.7759 - val_loss: 0.3849 - val_accuracy: 0.8190\n",
            "Epoch 16/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.4483 - accuracy: 0.7970 - val_loss: 0.3723 - val_accuracy: 0.8360\n",
            "Epoch 17/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.4398 - accuracy: 0.8070 - val_loss: 0.3582 - val_accuracy: 0.8350\n",
            "Epoch 18/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.4303 - accuracy: 0.8085 - val_loss: 0.3476 - val_accuracy: 0.8400\n",
            "Epoch 19/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.4307 - accuracy: 0.8141 - val_loss: 0.3413 - val_accuracy: 0.8400\n",
            "Epoch 20/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.4057 - accuracy: 0.8276 - val_loss: 0.3346 - val_accuracy: 0.8450\n",
            "Epoch 21/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.4135 - accuracy: 0.8106 - val_loss: 0.3454 - val_accuracy: 0.8510\n",
            "Epoch 22/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.3941 - accuracy: 0.8231 - val_loss: 0.3484 - val_accuracy: 0.8640\n",
            "Epoch 23/500\n",
            "100/100 [==============================] - 37s 373ms/step - loss: 0.3925 - accuracy: 0.8347 - val_loss: 0.3531 - val_accuracy: 0.8470\n",
            "Epoch 24/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.4001 - accuracy: 0.8261 - val_loss: 0.3581 - val_accuracy: 0.8710\n",
            "Epoch 25/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.3941 - accuracy: 0.8245 - val_loss: 0.3579 - val_accuracy: 0.8360\n",
            "Epoch 26/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3755 - accuracy: 0.8372 - val_loss: 0.2793 - val_accuracy: 0.8850\n",
            "Epoch 27/500\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.3665 - accuracy: 0.8380 - val_loss: 0.3272 - val_accuracy: 0.8670\n",
            "Epoch 28/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3686 - accuracy: 0.8442 - val_loss: 0.3398 - val_accuracy: 0.8580\n",
            "Epoch 29/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.3867 - accuracy: 0.8377 - val_loss: 0.4081 - val_accuracy: 0.8120\n",
            "Epoch 30/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3654 - accuracy: 0.8492 - val_loss: 0.3068 - val_accuracy: 0.8780\n",
            "Epoch 31/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.3641 - accuracy: 0.8442 - val_loss: 0.3418 - val_accuracy: 0.8690\n",
            "Epoch 32/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3505 - accuracy: 0.8492 - val_loss: 0.3134 - val_accuracy: 0.8880\n",
            "Epoch 33/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.3685 - accuracy: 0.8392 - val_loss: 0.3009 - val_accuracy: 0.8720\n",
            "Epoch 34/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.3669 - accuracy: 0.8623 - val_loss: 0.2760 - val_accuracy: 0.8840\n",
            "Epoch 35/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.3871 - accuracy: 0.8462 - val_loss: 0.2671 - val_accuracy: 0.8980\n",
            "Epoch 36/500\n",
            "100/100 [==============================] - 35s 355ms/step - loss: 0.3320 - accuracy: 0.8540 - val_loss: 0.3067 - val_accuracy: 0.8920\n",
            "Epoch 37/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.3600 - accuracy: 0.8588 - val_loss: 0.2647 - val_accuracy: 0.8860\n",
            "Epoch 38/500\n",
            "100/100 [==============================] - 35s 348ms/step - loss: 0.3589 - accuracy: 0.8603 - val_loss: 0.2761 - val_accuracy: 0.8940\n",
            "Epoch 39/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.3327 - accuracy: 0.8598 - val_loss: 0.3019 - val_accuracy: 0.8940\n",
            "Epoch 40/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.2962 - accuracy: 0.8764 - val_loss: 0.3181 - val_accuracy: 0.8970\n",
            "Epoch 41/500\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.3272 - accuracy: 0.8633 - val_loss: 0.4117 - val_accuracy: 0.8400\n",
            "Epoch 42/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.3289 - accuracy: 0.8643 - val_loss: 0.2590 - val_accuracy: 0.9020\n",
            "Epoch 43/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.3486 - accuracy: 0.8553 - val_loss: 0.2273 - val_accuracy: 0.9080\n",
            "Epoch 44/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.3312 - accuracy: 0.8673 - val_loss: 0.2351 - val_accuracy: 0.9060\n",
            "Epoch 45/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.3272 - accuracy: 0.8704 - val_loss: 0.2459 - val_accuracy: 0.9080\n",
            "Epoch 46/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3261 - accuracy: 0.8698 - val_loss: 0.2733 - val_accuracy: 0.9000\n",
            "Epoch 47/500\n",
            "100/100 [==============================] - 35s 343ms/step - loss: 0.3340 - accuracy: 0.8668 - val_loss: 0.2644 - val_accuracy: 0.9050\n",
            "Epoch 48/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.3229 - accuracy: 0.8658 - val_loss: 0.3215 - val_accuracy: 0.8740\n",
            "Epoch 49/500\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.3274 - accuracy: 0.8613 - val_loss: 0.2897 - val_accuracy: 0.8930\n",
            "Epoch 50/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3030 - accuracy: 0.8835 - val_loss: 0.2503 - val_accuracy: 0.9060\n",
            "Epoch 51/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3243 - accuracy: 0.8663 - val_loss: 0.2431 - val_accuracy: 0.9120\n",
            "Epoch 52/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.3060 - accuracy: 0.8789 - val_loss: 0.2760 - val_accuracy: 0.9000\n",
            "Epoch 53/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.2989 - accuracy: 0.8739 - val_loss: 0.2336 - val_accuracy: 0.9140\n",
            "Epoch 54/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2884 - accuracy: 0.8779 - val_loss: 0.2349 - val_accuracy: 0.9200\n",
            "Epoch 55/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2908 - accuracy: 0.8849 - val_loss: 0.3819 - val_accuracy: 0.8660\n",
            "Epoch 56/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.3166 - accuracy: 0.8809 - val_loss: 0.2448 - val_accuracy: 0.9110\n",
            "Epoch 57/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3096 - accuracy: 0.8845 - val_loss: 0.2391 - val_accuracy: 0.9050\n",
            "Epoch 58/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.3065 - accuracy: 0.8744 - val_loss: 0.2400 - val_accuracy: 0.9170\n",
            "Epoch 59/500\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.2975 - accuracy: 0.8794 - val_loss: 0.2343 - val_accuracy: 0.9170\n",
            "Epoch 60/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.2752 - accuracy: 0.8869 - val_loss: 0.2602 - val_accuracy: 0.9040\n",
            "Epoch 61/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.3158 - accuracy: 0.8794 - val_loss: 0.2269 - val_accuracy: 0.9180\n",
            "Epoch 62/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.2819 - accuracy: 0.8920 - val_loss: 0.2340 - val_accuracy: 0.9100\n",
            "Epoch 63/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3022 - accuracy: 0.8789 - val_loss: 0.2594 - val_accuracy: 0.9100\n",
            "Epoch 64/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.2659 - accuracy: 0.8955 - val_loss: 0.2182 - val_accuracy: 0.9200\n",
            "Epoch 65/500\n",
            "100/100 [==============================] - 35s 353ms/step - loss: 0.3079 - accuracy: 0.8724 - val_loss: 0.2230 - val_accuracy: 0.9240\n",
            "Epoch 66/500\n",
            "100/100 [==============================] - 36s 363ms/step - loss: 0.2965 - accuracy: 0.8834 - val_loss: 0.2230 - val_accuracy: 0.9140\n",
            "Epoch 67/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.2742 - accuracy: 0.8990 - val_loss: 0.2571 - val_accuracy: 0.9130\n",
            "Epoch 68/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.3059 - accuracy: 0.8839 - val_loss: 0.2335 - val_accuracy: 0.9190\n",
            "Epoch 69/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.2645 - accuracy: 0.9005 - val_loss: 0.2563 - val_accuracy: 0.9150\n",
            "Epoch 70/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2806 - accuracy: 0.8864 - val_loss: 0.2281 - val_accuracy: 0.9170\n",
            "Epoch 71/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.2912 - accuracy: 0.8915 - val_loss: 0.2367 - val_accuracy: 0.9240\n",
            "Epoch 72/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2771 - accuracy: 0.8965 - val_loss: 0.2038 - val_accuracy: 0.9250\n",
            "Epoch 73/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2724 - accuracy: 0.8965 - val_loss: 0.1936 - val_accuracy: 0.9270\n",
            "Epoch 74/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2610 - accuracy: 0.8970 - val_loss: 0.1951 - val_accuracy: 0.9290\n",
            "Epoch 75/500\n",
            "100/100 [==============================] - 38s 380ms/step - loss: 0.2941 - accuracy: 0.8915 - val_loss: 0.2049 - val_accuracy: 0.9290\n",
            "Epoch 76/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2757 - accuracy: 0.8970 - val_loss: 0.1640 - val_accuracy: 0.9380\n",
            "Epoch 77/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2721 - accuracy: 0.8945 - val_loss: 0.2506 - val_accuracy: 0.9040\n",
            "Epoch 78/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.2553 - accuracy: 0.8980 - val_loss: 0.2255 - val_accuracy: 0.9250\n",
            "Epoch 79/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2919 - accuracy: 0.8844 - val_loss: 0.1916 - val_accuracy: 0.9380\n",
            "Epoch 80/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2827 - accuracy: 0.8995 - val_loss: 0.2365 - val_accuracy: 0.9140\n",
            "Epoch 81/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2461 - accuracy: 0.9075 - val_loss: 0.2285 - val_accuracy: 0.9220\n",
            "Epoch 82/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2969 - accuracy: 0.8950 - val_loss: 0.1831 - val_accuracy: 0.9290\n",
            "Epoch 83/500\n",
            "100/100 [==============================] - 35s 355ms/step - loss: 0.2481 - accuracy: 0.9005 - val_loss: 0.1848 - val_accuracy: 0.9330\n",
            "Epoch 84/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.2673 - accuracy: 0.9005 - val_loss: 0.1727 - val_accuracy: 0.9350\n",
            "Epoch 85/500\n",
            "100/100 [==============================] - 36s 360ms/step - loss: 0.2397 - accuracy: 0.9045 - val_loss: 0.2156 - val_accuracy: 0.9310\n",
            "Epoch 86/500\n",
            "100/100 [==============================] - 36s 363ms/step - loss: 0.2386 - accuracy: 0.9111 - val_loss: 0.2350 - val_accuracy: 0.9290\n",
            "Epoch 87/500\n",
            "100/100 [==============================] - 37s 364ms/step - loss: 0.2472 - accuracy: 0.9151 - val_loss: 0.1774 - val_accuracy: 0.9410\n",
            "Epoch 88/500\n",
            "100/100 [==============================] - 37s 371ms/step - loss: 0.2489 - accuracy: 0.9085 - val_loss: 0.1848 - val_accuracy: 0.9390\n",
            "Epoch 89/500\n",
            "100/100 [==============================] - 37s 366ms/step - loss: 0.2502 - accuracy: 0.9085 - val_loss: 0.1809 - val_accuracy: 0.9430\n",
            "Epoch 90/500\n",
            "100/100 [==============================] - 37s 367ms/step - loss: 0.2926 - accuracy: 0.8990 - val_loss: 0.1512 - val_accuracy: 0.9450\n",
            "Epoch 91/500\n",
            "100/100 [==============================] - 37s 369ms/step - loss: 0.2492 - accuracy: 0.9015 - val_loss: 0.1741 - val_accuracy: 0.9370\n",
            "Epoch 92/500\n",
            "100/100 [==============================] - 37s 367ms/step - loss: 0.2456 - accuracy: 0.9126 - val_loss: 0.2130 - val_accuracy: 0.9180\n",
            "Epoch 93/500\n",
            "100/100 [==============================] - 37s 370ms/step - loss: 0.2391 - accuracy: 0.9141 - val_loss: 0.2216 - val_accuracy: 0.9280\n",
            "Epoch 94/500\n",
            "100/100 [==============================] - 36s 364ms/step - loss: 0.2762 - accuracy: 0.9055 - val_loss: 0.1771 - val_accuracy: 0.9360\n",
            "Epoch 95/500\n",
            "100/100 [==============================] - 37s 366ms/step - loss: 0.2772 - accuracy: 0.9035 - val_loss: 0.1703 - val_accuracy: 0.9370\n",
            "Epoch 96/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.3021 - accuracy: 0.8975 - val_loss: 0.2140 - val_accuracy: 0.9270\n",
            "Epoch 97/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.2696 - accuracy: 0.9020 - val_loss: 0.1858 - val_accuracy: 0.9320\n",
            "Epoch 98/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.2517 - accuracy: 0.9141 - val_loss: 0.1499 - val_accuracy: 0.9520\n",
            "Epoch 99/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2518 - accuracy: 0.9136 - val_loss: 0.2000 - val_accuracy: 0.9330\n",
            "Epoch 100/500\n",
            "100/100 [==============================] - 33s 334ms/step - loss: 0.2592 - accuracy: 0.9080 - val_loss: 0.2993 - val_accuracy: 0.8990\n",
            "Epoch 101/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.2347 - accuracy: 0.9136 - val_loss: 0.1585 - val_accuracy: 0.9390\n",
            "Epoch 102/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.2524 - accuracy: 0.9156 - val_loss: 0.1551 - val_accuracy: 0.9350\n",
            "Epoch 103/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.2394 - accuracy: 0.9156 - val_loss: 0.1770 - val_accuracy: 0.9310\n",
            "Epoch 104/500\n",
            "100/100 [==============================] - 35s 345ms/step - loss: 0.2667 - accuracy: 0.9075 - val_loss: 0.1694 - val_accuracy: 0.9370\n",
            "Epoch 105/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.2480 - accuracy: 0.9136 - val_loss: 0.1520 - val_accuracy: 0.9380\n",
            "Epoch 106/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2367 - accuracy: 0.9151 - val_loss: 0.1514 - val_accuracy: 0.9480\n",
            "Epoch 107/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2362 - accuracy: 0.9211 - val_loss: 0.1906 - val_accuracy: 0.9360\n",
            "Epoch 108/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2361 - accuracy: 0.9130 - val_loss: 0.1739 - val_accuracy: 0.9420\n",
            "Epoch 109/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2075 - accuracy: 0.9271 - val_loss: 0.1683 - val_accuracy: 0.9430\n",
            "Epoch 110/500\n",
            "100/100 [==============================] - 36s 355ms/step - loss: 0.2502 - accuracy: 0.9146 - val_loss: 0.1647 - val_accuracy: 0.9490\n",
            "Epoch 111/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.2466 - accuracy: 0.9080 - val_loss: 0.2247 - val_accuracy: 0.9270\n",
            "Epoch 112/500\n",
            "100/100 [==============================] - 35s 355ms/step - loss: 0.2251 - accuracy: 0.9131 - val_loss: 0.1387 - val_accuracy: 0.9530\n",
            "Epoch 113/500\n",
            "100/100 [==============================] - 36s 362ms/step - loss: 0.2449 - accuracy: 0.9095 - val_loss: 0.1805 - val_accuracy: 0.9330\n",
            "Epoch 114/500\n",
            "100/100 [==============================] - 36s 358ms/step - loss: 0.2496 - accuracy: 0.9085 - val_loss: 0.1467 - val_accuracy: 0.9470\n",
            "Epoch 115/500\n",
            "100/100 [==============================] - 36s 356ms/step - loss: 0.2201 - accuracy: 0.9216 - val_loss: 0.1661 - val_accuracy: 0.9440\n",
            "Epoch 116/500\n",
            "100/100 [==============================] - 36s 358ms/step - loss: 0.2303 - accuracy: 0.9191 - val_loss: 0.1501 - val_accuracy: 0.9420\n",
            "Epoch 117/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.2330 - accuracy: 0.9186 - val_loss: 0.1370 - val_accuracy: 0.9560\n",
            "Epoch 118/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2148 - accuracy: 0.9226 - val_loss: 0.1722 - val_accuracy: 0.9410\n",
            "Epoch 119/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2440 - accuracy: 0.9171 - val_loss: 0.1384 - val_accuracy: 0.9520\n",
            "Epoch 120/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.2377 - accuracy: 0.9131 - val_loss: 0.1521 - val_accuracy: 0.9470\n",
            "Epoch 121/500\n",
            "100/100 [==============================] - 35s 354ms/step - loss: 0.2167 - accuracy: 0.9186 - val_loss: 0.1375 - val_accuracy: 0.9510\n",
            "Epoch 122/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2172 - accuracy: 0.9191 - val_loss: 0.2090 - val_accuracy: 0.9350\n",
            "Epoch 123/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.2310 - accuracy: 0.9196 - val_loss: 0.1446 - val_accuracy: 0.9440\n",
            "Epoch 124/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2310 - accuracy: 0.9191 - val_loss: 0.1567 - val_accuracy: 0.9480\n",
            "Epoch 125/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.2093 - accuracy: 0.9250 - val_loss: 0.1550 - val_accuracy: 0.9420\n",
            "Epoch 126/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2038 - accuracy: 0.9276 - val_loss: 0.1580 - val_accuracy: 0.9380\n",
            "Epoch 127/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2049 - accuracy: 0.9296 - val_loss: 0.1260 - val_accuracy: 0.9530\n",
            "Epoch 128/500\n",
            "100/100 [==============================] - 36s 357ms/step - loss: 0.2126 - accuracy: 0.9211 - val_loss: 0.1563 - val_accuracy: 0.9440\n",
            "Epoch 129/500\n",
            "100/100 [==============================] - 35s 355ms/step - loss: 0.2306 - accuracy: 0.9286 - val_loss: 0.1771 - val_accuracy: 0.9420\n",
            "Epoch 130/500\n",
            "100/100 [==============================] - 36s 354ms/step - loss: 0.2284 - accuracy: 0.9271 - val_loss: 0.1555 - val_accuracy: 0.9420\n",
            "Epoch 131/500\n",
            "100/100 [==============================] - 36s 358ms/step - loss: 0.1962 - accuracy: 0.9251 - val_loss: 0.2246 - val_accuracy: 0.9250\n",
            "Epoch 132/500\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2141 - accuracy: 0.9151 - val_loss: 0.1691 - val_accuracy: 0.9460\n",
            "Epoch 133/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.2299 - accuracy: 0.9221 - val_loss: 0.1660 - val_accuracy: 0.9450\n",
            "Epoch 134/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.2172 - accuracy: 0.9166 - val_loss: 0.1693 - val_accuracy: 0.9420\n",
            "Epoch 135/500\n",
            "100/100 [==============================] - 35s 353ms/step - loss: 0.2118 - accuracy: 0.9181 - val_loss: 0.2139 - val_accuracy: 0.9260\n",
            "Epoch 136/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.1796 - accuracy: 0.9362 - val_loss: 0.1485 - val_accuracy: 0.9380\n",
            "Epoch 137/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.2162 - accuracy: 0.9226 - val_loss: 0.1917 - val_accuracy: 0.9390\n",
            "Epoch 138/500\n",
            "100/100 [==============================] - 35s 355ms/step - loss: 0.2275 - accuracy: 0.9216 - val_loss: 0.1593 - val_accuracy: 0.9550\n",
            "Epoch 139/500\n",
            "100/100 [==============================] - 35s 351ms/step - loss: 0.2153 - accuracy: 0.9281 - val_loss: 0.1609 - val_accuracy: 0.9470\n",
            "Epoch 140/500\n",
            "100/100 [==============================] - 34s 342ms/step - loss: 0.2056 - accuracy: 0.9296 - val_loss: 0.1848 - val_accuracy: 0.9410\n",
            "Epoch 141/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.1808 - accuracy: 0.9377 - val_loss: 0.1683 - val_accuracy: 0.9500\n",
            "Epoch 142/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.2148 - accuracy: 0.9245 - val_loss: 0.1749 - val_accuracy: 0.9460\n",
            "Epoch 143/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.2274 - accuracy: 0.9291 - val_loss: 0.1849 - val_accuracy: 0.9470\n",
            "Epoch 144/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.1897 - accuracy: 0.9382 - val_loss: 0.1251 - val_accuracy: 0.9530\n",
            "Epoch 145/500\n",
            "100/100 [==============================] - 34s 345ms/step - loss: 0.1906 - accuracy: 0.9342 - val_loss: 0.1880 - val_accuracy: 0.9470\n",
            "Epoch 146/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2092 - accuracy: 0.9271 - val_loss: 0.1378 - val_accuracy: 0.9480\n",
            "Epoch 147/500\n",
            "100/100 [==============================] - 34s 343ms/step - loss: 0.2271 - accuracy: 0.9186 - val_loss: 0.1769 - val_accuracy: 0.9470\n",
            "Epoch 148/500\n",
            "100/100 [==============================] - 35s 350ms/step - loss: 0.2063 - accuracy: 0.9276 - val_loss: 0.1317 - val_accuracy: 0.9580\n",
            "Epoch 149/500\n",
            "100/100 [==============================] - 35s 353ms/step - loss: 0.1912 - accuracy: 0.9347 - val_loss: 0.1443 - val_accuracy: 0.9560\n",
            "Epoch 150/500\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.2156 - accuracy: 0.9271 - val_loss: 0.1587 - val_accuracy: 0.9500\n",
            "Epoch 151/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.2307 - accuracy: 0.9317 - val_loss: 0.1871 - val_accuracy: 0.9470\n",
            "Epoch 152/500\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2332 - accuracy: 0.9286 - val_loss: 0.2172 - val_accuracy: 0.9430\n",
            "Epoch 153/500\n",
            "100/100 [==============================] - 33s 335ms/step - loss: 0.2176 - accuracy: 0.9281 - val_loss: 0.1688 - val_accuracy: 0.9470\n",
            "Epoch 154/500\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.1934 - accuracy: 0.9322 - val_loss: 0.1686 - val_accuracy: 0.9430\n",
            "Epoch 155/500\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2218 - accuracy: 0.9281 - val_loss: 0.1205 - val_accuracy: 0.9580\n",
            "Epoch 156/500\n",
            "100/100 [==============================] - 34s 339ms/step - loss: 0.2277 - accuracy: 0.9286 - val_loss: 0.1369 - val_accuracy: 0.9540\n",
            "Epoch 157/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.1982 - accuracy: 0.9342 - val_loss: 0.1611 - val_accuracy: 0.9510\n",
            "Epoch 158/500\n",
            "100/100 [==============================] - 35s 353ms/step - loss: 0.1972 - accuracy: 0.9370 - val_loss: 0.1363 - val_accuracy: 0.9540\n",
            "Epoch 159/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.1808 - accuracy: 0.9382 - val_loss: 0.1177 - val_accuracy: 0.9600\n",
            "Epoch 160/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.2258 - accuracy: 0.9151 - val_loss: 0.1458 - val_accuracy: 0.9510\n",
            "Epoch 161/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2137 - accuracy: 0.9345 - val_loss: 0.1813 - val_accuracy: 0.9460\n",
            "Epoch 162/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.2166 - accuracy: 0.9302 - val_loss: 0.1441 - val_accuracy: 0.9550\n",
            "Epoch 163/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.1960 - accuracy: 0.9392 - val_loss: 0.0997 - val_accuracy: 0.9630\n",
            "Epoch 164/500\n",
            "100/100 [==============================] - 35s 347ms/step - loss: 0.2050 - accuracy: 0.9302 - val_loss: 0.1315 - val_accuracy: 0.9600\n",
            "Epoch 165/500\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.1805 - accuracy: 0.9442 - val_loss: 0.1503 - val_accuracy: 0.9510\n",
            "Epoch 166/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.2097 - accuracy: 0.9362 - val_loss: 0.1809 - val_accuracy: 0.9390\n",
            "Epoch 167/500\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.1969 - accuracy: 0.9302 - val_loss: 0.2171 - val_accuracy: 0.9310\n",
            "Epoch 168/500\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.2044 - accuracy: 0.9276 - val_loss: 0.1550 - val_accuracy: 0.9530\n",
            "Epoch 169/500\n",
            "100/100 [==============================] - 35s 352ms/step - loss: 0.1897 - accuracy: 0.9352 - val_loss: 0.1449 - val_accuracy: 0.9540\n",
            "Epoch 170/500\n",
            "100/100 [==============================] - 34s 341ms/step - loss: 0.1947 - accuracy: 0.9367 - val_loss: 0.2285 - val_accuracy: 0.9450\n",
            "Epoch 171/500\n",
            "100/100 [==============================] - 33s 332ms/step - loss: 0.1581 - accuracy: 0.9442 - val_loss: 0.1660 - val_accuracy: 0.9500\n",
            "Epoch 172/500\n",
            "100/100 [==============================] - 33s 333ms/step - loss: 0.2117 - accuracy: 0.9271 - val_loss: 0.1415 - val_accuracy: 0.9530\n",
            "Epoch 173/500\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.1955 - accuracy: 0.9347 - val_loss: 0.3886 - val_accuracy: 0.9280\n",
            "Epoch 174/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.2124 - accuracy: 0.9372 - val_loss: 0.1651 - val_accuracy: 0.9510\n",
            "Epoch 175/500\n",
            "100/100 [==============================] - 35s 346ms/step - loss: 0.1848 - accuracy: 0.9367 - val_loss: 0.1438 - val_accuracy: 0.9590\n",
            "Epoch 176/500\n",
            "100/100 [==============================] - 34s 337ms/step - loss: 0.2074 - accuracy: 0.9412 - val_loss: 0.2208 - val_accuracy: 0.9350\n",
            "Epoch 177/500\n",
            "100/100 [==============================] - 34s 344ms/step - loss: 0.1752 - accuracy: 0.9402 - val_loss: 0.1743 - val_accuracy: 0.9590\n",
            "Epoch 178/500\n",
            "100/100 [==============================] - 35s 344ms/step - loss: 0.1875 - accuracy: 0.9460 - val_loss: 0.1554 - val_accuracy: 0.9540\n",
            "Epoch 179/500\n",
            "100/100 [==============================] - 34s 338ms/step - loss: 0.2064 - accuracy: 0.9337 - val_loss: 0.1316 - val_accuracy: 0.9610\n",
            "Epoch 180/500\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.1826 - accuracy: 0.9417 - val_loss: 0.1509 - val_accuracy: 0.9620\n",
            "Epoch 181/500\n",
            "100/100 [==============================] - 34s 336ms/step - loss: 0.1933 - accuracy: 0.9377 - val_loss: 0.1808 - val_accuracy: 0.9550\n",
            "Epoch 182/500\n",
            "100/100 [==============================] - 35s 349ms/step - loss: 0.2344 - accuracy: 0.9317 - val_loss: 0.1297 - val_accuracy: 0.9580\n",
            "Epoch 183/500\n",
            "100/100 [==============================] - 34s 340ms/step - loss: 0.1953 - accuracy: 0.9347 - val_loss: 0.1201 - val_accuracy: 0.9670\n",
            "Epoch 00183: early stopping\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "tpuLlwTTGKIp",
        "outputId": "7e5e363d-7085-415d-c243-9cd8f6f4275f"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['accuracy']\n",
        "val_acc = history.history['val_accuracy']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\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",
        "\n",
        "plt.figure()\n",
        "\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",
        "\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": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CRwc2qzgl9Tx"
      },
      "source": [
        "def smooth_curve(points, factor=0.8):\n",
        "    smoothed_points = []\n",
        "    for point in points:\n",
        "        if smoothed_points:\n",
        "            previous = smoothed_points[-1]\n",
        "            smoothed_points.append(previous * factor + point * (1 - factor))\n",
        "        else:\n",
        "            smoothed_points.append(point)\n",
        "    return smoothed_points"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "Vliwhwz9F0vG",
        "outputId": "be45a53e-5d12-4c0c-b773-e14fdf544b41"
      },
      "source": [
        "epochs = range(100, len(acc) + 1)\n",
        "plt.plot(epochs, smooth_curve(acc[99:]), 'bo', label='Training accuracy')\n",
        "plt.plot(epochs, smooth_curve(val_acc[99:]), 'b', label='Validation accuracy')\n",
        "plt.title('Training and Validation accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\n",
        "\n",
        "plt.plot(epochs, smooth_curve(loss[99:]), 'bo', label='Training loss')\n",
        "plt.plot(epochs, smooth_curve(val_loss[99:]), 'b', label='Validation loss')\n",
        "plt.title('Training and Validation loss')\n",
        "plt.legend()\n",
        "\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": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_ZchuhPW7WsH"
      },
      "source": [
        "**TODO: Add TensorBoard for better interpretation**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kDcfZOfCiLah"
      },
      "source": [
        "## Model Evaluation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jyhvAxYtVjt8",
        "outputId": "00910f43-a8bc-4e42-b6d8-d6dfd50204fc"
      },
      "source": [
        "weights_path = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)'\n",
        "os.listdir(weights_path + '/model weights')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['first_weights.hdf5', 'second_weights_ResNetV2.hdf5']"
            ]
          },
          "metadata": {},
          "execution_count": 31
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7gkMSaenkzK0"
      },
      "source": [
        "Here we are loading our weights into model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6dlXc6p6iQwj"
      },
      "source": [
        "weights = new_weights\n",
        "weights_path = '/content/drive/MyDrive/Mushrooms Binary Classification (edible & not)'\n",
        "model.load_weights(weights_path + '/' + weights)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gVhlNtDikwof"
      },
      "source": [
        "`Model.predict_generator` is deprecated and will be removed in a future version. Please use `Model.predict`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "acFfN2xojHSJ",
        "outputId": "6abc1f57-0f29-482c-d439-2fece9088949"
      },
      "source": [
        "test_datagen = ImageDataGenerator(rescale=1./255)\n",
        "test_generator = test_datagen.flow_from_directory(\n",
        "    test_dir,\n",
        "    target_size=(150, 150),\n",
        "    batch_size=20,\n",
        "    shuffle=False,\n",
        "    class_mode='binary')\n",
        "\n",
        "n_files = len(test_generator.filenames)\n",
        "y_pred = model.predict_generator(test_generator, steps=n_files)\n",
        "y_true = test_generator.classes"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 734 images belonging to 2 classes.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py:2035: UserWarning: `Model.predict_generator` is deprecated and will be removed in a future version. Please use `Model.predict`, which supports generators.\n",
            "  warnings.warn('`Model.predict_generator` is deprecated and '\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 734 batches). You may need to use the repeat() function when building your dataset.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QC8ut8TUIsj9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "721e146c-11a5-47f4-d56a-773a8ef22f45"
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "\n",
        "y_pred = np.rint(y_pred.flatten())\n",
        "print(classification_report(y_true, y_pred, target_names=['Edible', 'Inedible']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "      Edible       0.97      0.94      0.96       329\n",
            "    Inedible       0.95      0.98      0.96       405\n",
            "\n",
            "    accuracy                           0.96       734\n",
            "   macro avg       0.96      0.96      0.96       734\n",
            "weighted avg       0.96      0.96      0.96       734\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oEhJJzpgIhx0"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}