COP508 ML CW.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "id": "da8dd0c8",
      "metadata": {
        "id": "da8dd0c8"
      },
      "source": [
        "# COP508 Machine Learning CW\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "f2984791",
      "metadata": {},
      "source": [
        "<b>Name -> Neville Mathew</b>\n",
        "<br>\n",
        "<b>Student No. -> F120315</b>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "5207d0b3",
      "metadata": {
        "id": "5207d0b3"
      },
      "source": [
        "## Table of Contents\n",
        "\n",
        "1. [Data Ingestion](#data-ingestion)\n",
        "\n",
        "   1. [Importing the Libraries](#import_libraries)\n",
        "   2. [Importing the Dataset](#importing_dataset)\n",
        "   3. [Splitting Dataset into Train & Test set](#dataset_split)\n",
        "   4. [Visualising Five Images from the Dataset](#visualize_images)\n",
        "\n",
        "2. [Data Preprocessing](#data_preprocessing)\n",
        "\n",
        "   1. [Grayscale Conversion](#grayscale)\n",
        "   2. [Normalization](#normalization)\n",
        "   3. [Gaussian Blur](#blur)\n",
        "   4. [Image Rotation](#rotate)\n",
        "   5. [One-Hot Encoding](#encoding)\n",
        "\n",
        "3. [Model Learning](#model_learning)\n",
        "\n",
        "   1. [Convolutional Neural Network (CNN) Models](#cnn_models)\n",
        "      1. [Base CNN Model](#base_cnn_model)\n",
        "      2. [Base CNN Model -> Shuffle, 25 Epochs, Kernel Filters, Grayscale & Gaussian Blur](#base_cnn_model_changes)\n",
        "      3. [ResNet50 CNN Model](#resnet50_cnn_model)\n",
        "      4. [Best CNN Model](#best_cnn_model)\n",
        "\n",
        "4. [Model Evaluation](#model_evaluation)\n",
        "\n",
        "   1. [Visualization of Base CNN Model](#visualization_base_cnn_model)\n",
        "   2. [Visualization of Base CNN Model with Changes](#visualization_base_cnn_model_changes)\n",
        "   3. [Visualization of ResNet50 CNN Model](#visualization_resnet50_cnn_model)\n",
        "   4. [Visualization of Best CNN Model](#visualization_best_cnn_model)\n",
        "\n",
        "5. [Model Prediction](#model_prediction)\n",
        "   1. [Prediction of Best CNN Model](#prediction_best_cnn_model)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "33316fd1",
      "metadata": {
        "id": "33316fd1"
      },
      "source": [
        "## <a name=\"data-ingestion\">Data Ingestion</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "7975f6d2",
      "metadata": {
        "id": "7975f6d2"
      },
      "source": [
        "### <a name=\"import_libraries\">Importing the Libraries</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "id": "_f_StMjyUujE",
      "metadata": {
        "id": "_f_StMjyUujE"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "2022-01-13 14:25:28.725141: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n"
          ]
        }
      ],
      "source": [
        "import os\n",
        "import numpy as np\n",
        "import pickle\n",
        "import requests\n",
        "import matplotlib.pyplot as plt\n",
        "import cv2\n",
        "import wget\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.applications.resnet50 import ResNet50\n",
        "from sklearn.metrics import accuracy_score\n",
        "from matplotlib import image\n",
        "from scipy import ndimage\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.layers import Conv2D, MaxPool2D, Flatten, Dense, Dropout, Activation, BatchNormalization\n",
        "from keras.utils import np_utils\n",
        "from tensorflow.keras import optimizers\n",
        "from keras.models import Sequential\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "fe6b960e",
      "metadata": {
        "id": "fe6b960e"
      },
      "source": [
        "### <a name=\"importing_dataset\">Importing the Dataset</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "id": "sVoI7_X6UujG",
      "metadata": {
        "id": "sVoI7_X6UujG"
      },
      "outputs": [],
      "source": [
        "path_to_check = os.path.join('', 'image_recognition_dataset.pth')\n",
        "data_path = path_to_check\n",
        "\n",
        "\n",
        "def load_dataset():\n",
        "    if not os.path.exists(path_to_check):\n",
        "        url = 'https://www.dropbox.com/s/j05hyf667ikzu4s/image_recognition_dataset.pth?dl=1'\n",
        "        filename = wget.download(url)\n",
        "        data_path = os.path.join('', filename)\n",
        "    else:\n",
        "        print(\"Data found on local {:s}\".format(path_to_check))\n",
        "        data_path = path_to_check\n",
        "\n",
        "    data = pickle.load(open(data_path, 'rb'))\n",
        "    train_imgs, train_labels, test_imgs, test_labels = data.values()\n",
        "    return train_imgs, train_labels, test_imgs, test_labels\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "4334cbac",
      "metadata": {
        "id": "4334cbac"
      },
      "source": [
        "### <a name=\"dataset_split\">Splitting the Dataset into the Train & Test set</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "id": "xOE0QVfQUujI",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xOE0QVfQUujI",
        "outputId": "dc335af8-02f1-4eb0-a869-e0ad09daf12a"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Data found on local image_recognition_dataset.pth\n"
          ]
        }
      ],
      "source": [
        "train_imgs, train_labels, test_imgs, test_labels = load_dataset()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "329c6710",
      "metadata": {
        "id": "329c6710"
      },
      "source": [
        "### <a name=\"display_dataset\">Displaying Dataset Information</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "id": "16Vd_aSQUujK",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "16Vd_aSQUujK",
        "outputId": "9d8919c0-d19c-4531-904c-3ea7b4bfaa12"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Total number of classes in dataset -> 8\n",
            "Shape of image training set ->  (40000, 32, 32, 3)\n",
            "Number of training images in dataset ->  40000\n",
            "Number of testing images in dataset ->  8000\n"
          ]
        }
      ],
      "source": [
        "dataset_classes = len(np.unique(train_labels))\n",
        "print(\"Total number of classes in dataset -> {:d}\".format(dataset_classes))\n",
        "print(\"Shape of image training set -> \", train_imgs.shape)\n",
        "print(\"Number of training images in dataset -> \", train_imgs.shape[0])\n",
        "print(\"Number of testing images in dataset -> \", test_imgs.shape[0])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "a383e471",
      "metadata": {
        "id": "a383e471"
      },
      "source": [
        "### <a name=\"visualize_images\">Visualising Five Images from the Dataset</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "id": "HDk6E9rvUujL",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 138
        },
        "id": "HDk6E9rvUujL",
        "outputId": "00bd2b11-52e5-4b7d-f196-5035aa8067ad"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "/home/lunet/conm/.conda/envs/cop508/lib/python3.9/site-packages/matplotlib/text.py:1223: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison\n",
            "  if s != self._text:\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 8 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "def plot_trained_images() -> None:\n",
        "    for i in range(8):\n",
        "        plt.subplot(1, 8, i+1)\n",
        "        imgData = np.array(train_imgs[i])\n",
        "        plt.title(train_labels[i])\n",
        "        plt.imshow(imgData)\n",
        "        plt.axis('off')\n",
        "\n",
        "\n",
        "plot_trained_images()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "id": "5e0bfed7",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 138
        },
        "id": "5e0bfed7",
        "outputId": "932de0f3-868f-4ba5-dfe7-83119ffa0ff4"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 8 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "def plot_test_images() -> None:\n",
        "    for i in range(8):\n",
        "        plt.subplot(1, 8, i+1)\n",
        "        imgData = np.array(test_imgs[i])\n",
        "        plt.title(test_labels[i])\n",
        "        plt.imshow(imgData)\n",
        "        plt.axis('off')\n",
        "\n",
        "\n",
        "plot_test_images()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "e992911c",
      "metadata": {
        "id": "e992911c"
      },
      "source": [
        "## <a name=\"data_preprocessing\">Data Preprocessing</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "f7d56252",
      "metadata": {
        "id": "f7d56252"
      },
      "source": [
        "### <a name=\"grayscale\">Grayscale Conversion</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "id": "e6a22e92",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "e6a22e92",
        "outputId": "598b6b4a-eb1e-414d-95e9-69329e47f58d"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7e346e8fd0>"
            ]
          },
          "execution_count": 7,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Conversion of colour image to grayscale.\n",
        "grayscale_train_img = [cv2.cvtColor(\n",
        "    one_img, cv2.COLOR_BGR2GRAY) for one_img in train_imgs]\n",
        "grayscale_test_img = [cv2.cvtColor(\n",
        "    one_img, cv2.COLOR_BGR2GRAY) for one_img in test_imgs]\n",
        "plt.imshow(grayscale_train_img[0], cmap='gray')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "8810e2d9",
      "metadata": {
        "id": "8810e2d9"
      },
      "source": [
        "### <a name=\"normalization\">Normalization</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "id": "cd6ddd71",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "cd6ddd71",
        "outputId": "8d4e22b6-c05f-4155-b659-ca7440756ce1"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7e34671e20>"
            ]
          },
          "execution_count": 22,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Normalizing all images in the dataset in the range of 0 and 1.\n",
        "normalized_train_img = [one_img/255.0 for one_img in train_imgs]\n",
        "normalized_test_img = [one_img/255.0 for one_img in test_imgs]\n",
        "plt.imshow(normalized_train_img[0])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "f0e1d41c",
      "metadata": {
        "id": "f0e1d41c"
      },
      "source": [
        "### <a name=\"blur\">Gaussian Blur</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "id": "f33f73ba",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "f33f73ba",
        "outputId": "f4729122-0792-4850-c255-0bdd817c24ec"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7e33cd7670>"
            ]
          },
          "execution_count": 9,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Blurring images of sigma 1.\n",
        "blurred_train_img = [ndimage.gaussian_filter(\n",
        "    train_blur_imgs, sigma=0.5) for train_blur_imgs in normalized_train_img]\n",
        "blurred_test_img = [ndimage.gaussian_filter(\n",
        "    test_blur_imgs, sigma=0.5) for test_blur_imgs in normalized_test_img]\n",
        "plt.imshow(blurred_train_img[0])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "272cd6ff",
      "metadata": {
        "id": "272cd6ff"
      },
      "source": [
        "### <a name=\"rotation\">Image Rotation</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "id": "ae327bb7",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 301
        },
        "id": "ae327bb7",
        "outputId": "3d4d0ad9-a33c-4186-fbb2-384bff0f8b58"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7dbb238130>"
            ]
          },
          "execution_count": 23,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Rotation of images to 180 degrees through the use of a list comprehension.\n",
        "rotated_train_img = np.array([ndimage.rotate(\n",
        "    train_rotate_imgs, angle=180) for train_rotate_imgs in normalized_train_img])\n",
        "rotated_test_img = np.array([ndimage.rotate(\n",
        "    test_rotate_imgs, angle=180) for test_rotate_imgs in normalized_test_img])\n",
        "plt.imshow(rotated_train_img[0])\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "cQmiaD1FaE9o",
      "metadata": {
        "id": "cQmiaD1FaE9o"
      },
      "source": [
        "### <a name=\"encoding\">One-Hot Encoding</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "id": "a84df959",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a84df959",
        "outputId": "4b34ab4b-0afe-4f2c-8274-6e1c1b4ebdbc"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Shape of training split before one-hot encoding ->  (40000, 1)\n",
            "Shape of training split after one-hot encoding ->  (40000, 8)\n"
          ]
        }
      ],
      "source": [
        "# Reshaping of rotated_train_img and rotated_test_img which are numpy arrays.\n",
        "rotated_train_img = rotated_train_img.reshape(\n",
        "    rotated_train_img.shape[0], 32, 32, 3)\n",
        "rotated_test_img = rotated_test_img.reshape(\n",
        "    rotated_test_img.shape[0], 32, 32, 3)\n",
        "\n",
        "# one-hot encoding - Used to help in preprocessing categorical features of the images.\n",
        "no_of_classes = 8\n",
        "print(\"Shape of training split before one-hot encoding -> \", train_labels.shape)\n",
        "Y_train = np_utils.to_categorical(train_labels, no_of_classes)\n",
        "Y_test = np_utils.to_categorical(test_labels, no_of_classes)\n",
        "print(\"Shape of training split after one-hot encoding -> \", Y_train.shape)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "d877da98",
      "metadata": {
        "id": "d877da98"
      },
      "source": [
        "## <a name=\"model_learning\">Model Learning</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "b8abfeba",
      "metadata": {
        "id": "b8abfeba"
      },
      "source": [
        "### <a name=\"cnn_models\">Convolutional Neural Network (CNN) Models</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "7b7be011",
      "metadata": {
        "id": "7b7be011"
      },
      "source": [
        "#### <a name=\"base_cnn_model\">Base CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "id": "9b8a60d0",
      "metadata": {
        "id": "9b8a60d0"
      },
      "outputs": [],
      "source": [
        "# Reference: https://www.analyticsvidhya.com/blog/2020/02/learn-image-classification-cnn-convolutional-neural-networks-3-datasets/\n",
        "\n",
        "# To prevent any memory build up from old models or layers.\n",
        "keras.backend.clear_session()\n",
        "np.random.seed(15)\n",
        "tf.random.set_seed(15)\n",
        "\n",
        "\n",
        "def base_cnn_model() -> Sequential:\n",
        "    \"\"\"\n",
        "    Function which returns the base model.\n",
        "    :rtype: Sequential\n",
        "    \"\"\"\n",
        "    # Creating an early stopping function to monitor loss while keeping the best results.\n",
        "    early_callback = tf.keras.callbacks.EarlyStopping(\n",
        "        monitor='loss', restore_best_weights=True, patience=0)\n",
        "\n",
        "    # Initializing Sequential model, creating linear stack of model layers.\n",
        "    base_cnn_model.model1 = Sequential()\n",
        "\n",
        "    # Adding 3 convolutional layers with max pooling and dropout functions.\n",
        "    base_cnn_model.model1.add(Conv2D(32, kernel_size=(3, 3), strides=(1, 1),\n",
        "                                     padding='same', input_shape=(32, 32, 3), activation='relu'))\n",
        "\n",
        "    base_cnn_model.model1.add(Conv2D(64, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding='same', activation='relu'))\n",
        "    base_cnn_model.model1.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    base_cnn_model.model1.add(Dropout(0.25))\n",
        "\n",
        "    base_cnn_model.model1.add(Conv2D(128, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding='same', activation='relu'))\n",
        "    base_cnn_model.model1.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    base_cnn_model.model1.add(Dropout(0.25))\n",
        "\n",
        "    # Flattening convolutional layer output.\n",
        "    base_cnn_model.model1.add(Flatten())\n",
        "\n",
        "    # Adding 2 hidden layers.\n",
        "    base_cnn_model.model1.add(Dense(500, activation='relu'))\n",
        "    base_cnn_model.model1.add(Dropout(0.75))\n",
        "    base_cnn_model.model1.add(Dense(250, activation='relu'))\n",
        "    base_cnn_model.model1.add(Dropout(0.55))\n",
        "\n",
        "    # Final output layer with 8 classes.\n",
        "    base_cnn_model.model1.add(Dense(8, activation='softmax'))\n",
        "\n",
        "    # Compilation of sequential model with suitable parameters.\n",
        "    base_cnn_model.model1.compile(loss='categorical_crossentropy',\n",
        "                                  metrics=['accuracy', 'AUC'], optimizer='adam')\n",
        "\n",
        "    # Training model for 20 epochs\n",
        "    base_cnn_model.base_final_model = base_cnn_model.model1.fit(rotated_train_img, Y_train, batch_size=128, epochs=20, callbacks=[\n",
        "        early_callback], validation_data=(rotated_test_img, Y_test))\n",
        "\n",
        "    return base_cnn_model.model1, base_cnn_model.base_final_model\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "id": "eca4e98c",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 32, 32, 32)        896       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 16, 16, 128)       73856     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 8192)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 500)               4096500   \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 500)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 250)               125250    \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 250)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 8)                 2008      \n",
            "=================================================================\n",
            "Total params: 4,317,006\n",
            "Trainable params: 4,317,006\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "base_cnn_model.model1.summary()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "id": "22769a0f",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 1.9149 - accuracy: 0.2348 - auc: 0.6491 - val_loss: 1.3164 - val_accuracy: 0.5253 - val_auc: 0.8829\n",
            "Epoch 2/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 1.3659 - accuracy: 0.4997 - auc: 0.8620 - val_loss: 1.0954 - val_accuracy: 0.6133 - val_auc: 0.9192\n",
            "Epoch 3/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 1.1715 - accuracy: 0.5716 - auc: 0.9004 - val_loss: 0.9940 - val_accuracy: 0.6404 - val_auc: 0.9294\n",
            "Epoch 4/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 1.0745 - accuracy: 0.6138 - auc: 0.9167 - val_loss: 0.8941 - val_accuracy: 0.6886 - val_auc: 0.9441\n",
            "Epoch 5/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 1.0052 - accuracy: 0.6420 - auc: 0.9271 - val_loss: 0.8490 - val_accuracy: 0.7035 - val_auc: 0.9488\n",
            "Epoch 6/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 0.9345 - accuracy: 0.6684 - auc: 0.9370 - val_loss: 0.8037 - val_accuracy: 0.7138 - val_auc: 0.9534\n",
            "Epoch 7/20\n",
            "313/313 [==============================] - 3s 10ms/step - loss: 0.9026 - accuracy: 0.6784 - auc: 0.9410 - val_loss: 0.7819 - val_accuracy: 0.7304 - val_auc: 0.9574\n",
            "Epoch 8/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 0.8504 - accuracy: 0.7002 - auc: 0.9474 - val_loss: 0.8539 - val_accuracy: 0.6974 - val_auc: 0.9466\n",
            "Epoch 9/20\n",
            "313/313 [==============================] - 4s 11ms/step - loss: 0.8393 - accuracy: 0.7043 - auc: 0.9488 - val_loss: 0.8021 - val_accuracy: 0.7168 - val_auc: 0.9528\n",
            "Epoch 10/20\n",
            "313/313 [==============================] - 4s 11ms/step - loss: 0.7920 - accuracy: 0.7205 - auc: 0.9542 - val_loss: 0.7112 - val_accuracy: 0.7548 - val_auc: 0.9636\n",
            "Epoch 11/20\n",
            "313/313 [==============================] - 3s 10ms/step - loss: 0.7571 - accuracy: 0.7313 - auc: 0.9583 - val_loss: 0.7336 - val_accuracy: 0.7409 - val_auc: 0.9613\n",
            "Epoch 12/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 0.7465 - accuracy: 0.7397 - auc: 0.9590 - val_loss: 0.6991 - val_accuracy: 0.7519 - val_auc: 0.9642\n",
            "Epoch 13/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 0.7249 - accuracy: 0.7464 - auc: 0.9613 - val_loss: 0.6960 - val_accuracy: 0.7596 - val_auc: 0.9644\n",
            "Epoch 14/20\n",
            "313/313 [==============================] - 4s 11ms/step - loss: 0.6865 - accuracy: 0.7582 - auc: 0.9652 - val_loss: 0.6859 - val_accuracy: 0.7611 - val_auc: 0.9651\n",
            "Epoch 15/20\n",
            "313/313 [==============================] - 4s 12ms/step - loss: 0.6794 - accuracy: 0.7607 - auc: 0.9658 - val_loss: 0.6973 - val_accuracy: 0.7552 - val_auc: 0.9638\n",
            "Epoch 16/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 0.6717 - accuracy: 0.7619 - auc: 0.9667 - val_loss: 0.7136 - val_accuracy: 0.7500 - val_auc: 0.9622\n",
            "Epoch 17/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 0.6539 - accuracy: 0.7728 - auc: 0.9681 - val_loss: 0.6717 - val_accuracy: 0.7667 - val_auc: 0.9669\n",
            "Epoch 18/20\n",
            "313/313 [==============================] - 3s 10ms/step - loss: 0.6226 - accuracy: 0.7817 - auc: 0.9711 - val_loss: 0.6951 - val_accuracy: 0.7600 - val_auc: 0.9644\n",
            "Epoch 19/20\n",
            "313/313 [==============================] - 3s 11ms/step - loss: 0.6190 - accuracy: 0.7814 - auc: 0.9713 - val_loss: 0.6643 - val_accuracy: 0.7679 - val_auc: 0.9673\n",
            "Epoch 20/20\n",
            "313/313 [==============================] - 4s 11ms/step - loss: 0.6152 - accuracy: 0.7835 - auc: 0.9716 - val_loss: 0.6599 - val_accuracy: 0.7736 - val_auc: 0.9674\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "(<tensorflow.python.keras.engine.sequential.Sequential at 0x7f7e345f7af0>,\n",
              " <tensorflow.python.keras.callbacks.History at 0x7f7e346414f0>)"
            ]
          },
          "execution_count": 26,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "base_cnn_model()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "98ddb242",
      "metadata": {},
      "source": [
        "#### <a name=\"base_cnn_model_changes\">Base CNN Model -> Shuffle, 25 Epochs, Kernel Filters, Grayscale & Gaussian Blur</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "id": "5ebf61e2",
      "metadata": {},
      "outputs": [],
      "source": [
        "# To prevent any memory build up from old models or layers.\n",
        "keras.backend.clear_session()\n",
        "np.random.seed(15)\n",
        "tf.random.set_seed(15)\n",
        "\n",
        "\n",
        "def base_cnn_model_changes() -> Sequential:\n",
        "    \"\"\"\n",
        "    Function which returns the base model.\n",
        "    :rtype: Sequential\n",
        "    \"\"\"\n",
        "    # Creating an early stopping function to monitor loss while keeping the best results.\n",
        "    early_callback = tf.keras.callbacks.EarlyStopping(\n",
        "        monitor='loss', restore_best_weights=True, patience=0)\n",
        "\n",
        "    # Initializing Sequential model, creating linear stack of model layers.\n",
        "    base_cnn_model_changes.changes_model = Sequential()\n",
        "\n",
        "    # Adding 3 convolutional layers with max pooling and dropout functions.\n",
        "    base_cnn_model_changes.changes_model.add(Conv2D(64, kernel_size=(3, 3), strides=(1, 1),\n",
        "                                                    padding='same', input_shape=(32, 32, 1), activation='relu'))\n",
        "\n",
        "    base_cnn_model_changes.changes_model.add(Conv2D(128, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding='same', activation='relu'))\n",
        "    base_cnn_model_changes.changes_model.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    base_cnn_model_changes.changes_model.add(Dropout(0.25))\n",
        "\n",
        "    base_cnn_model_changes.changes_model.add(Conv2D(256, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding='same', activation='relu'))\n",
        "    base_cnn_model_changes.changes_model.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    base_cnn_model_changes.changes_model.add(Dropout(0.25))\n",
        "\n",
        "    # Flattening convolutional layer output.\n",
        "    base_cnn_model_changes.changes_model.add(Flatten())\n",
        "\n",
        "    # Adding 2 hidden layers.\n",
        "    base_cnn_model_changes.changes_model.add(Dense(500, activation='relu'))\n",
        "    base_cnn_model_changes.changes_model.add(Dropout(0.75))\n",
        "    base_cnn_model_changes.changes_model.add(Dense(250, activation='relu'))\n",
        "    base_cnn_model_changes.changes_model.add(Dropout(0.55))\n",
        "\n",
        "    # Final output layer with 8 classes.\n",
        "    base_cnn_model_changes.changes_model.add(Dense(8, activation='softmax'))\n",
        "\n",
        "    # Compilation of sequential model with suitable parameters.\n",
        "    base_cnn_model_changes.changes_model.compile(loss='categorical_crossentropy',\n",
        "                                                 metrics=['accuracy', 'AUC'], optimizer='adam')\n",
        "\n",
        "    # Training model for 25 epochs with shuffling of data.\n",
        "    base_cnn_model_changes.base_final_model_changes = base_cnn_model_changes.changes_model.fit(rotated_train_img, Y_train, batch_size=128, epochs=25, callbacks=[\n",
        "        early_callback], validation_data=(rotated_test_img, Y_test), shuffle=True)\n",
        "\n",
        "    return base_cnn_model_changes.changes_model, base_cnn_model_changes.base_final_model_changes\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "id": "dbeee39f",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 32, 32, 64)        640       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 128)       73856     \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 16, 16, 128)       0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 16, 16, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 16, 16, 256)       295168    \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 8, 8, 256)         0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 8, 8, 256)         0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 16384)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 500)               8192500   \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 500)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 250)               125250    \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 250)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 8)                 2008      \n",
            "=================================================================\n",
            "Total params: 8,689,422\n",
            "Trainable params: 8,689,422\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "base_cnn_model_changes.changes_model.summary()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "id": "5413e898",
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "2022-01-13 14:26:56.659327: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
            "2022-01-13 14:26:56.659856: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1\n",
            "2022-01-13 14:26:56.691329: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.691624: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
            "pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5\n",
            "coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s\n",
            "2022-01-13 14:26:56.691639: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
            "2022-01-13 14:26:56.692499: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
            "2022-01-13 14:26:56.692524: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
            "2022-01-13 14:26:56.693391: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
            "2022-01-13 14:26:56.693529: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
            "2022-01-13 14:26:56.694308: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
            "2022-01-13 14:26:56.694700: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
            "2022-01-13 14:26:56.696357: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
            "2022-01-13 14:26:56.696429: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.696747: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.697011: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
            "2022-01-13 14:26:56.697465: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2 AVX AVX2 FMA\n",
            "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
            "2022-01-13 14:26:56.697641: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.697915: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
            "pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5\n",
            "coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s\n",
            "2022-01-13 14:26:56.697930: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
            "2022-01-13 14:26:56.697944: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
            "2022-01-13 14:26:56.697956: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
            "2022-01-13 14:26:56.697967: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
            "2022-01-13 14:26:56.697979: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
            "2022-01-13 14:26:56.697990: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
            "2022-01-13 14:26:56.698002: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
            "2022-01-13 14:26:56.698014: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
            "2022-01-13 14:26:56.698050: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.698342: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.698604: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
            "2022-01-13 14:26:56.698624: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
            "2022-01-13 14:26:56.988322: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
            "2022-01-13 14:26:56.988344: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 \n",
            "2022-01-13 14:26:56.988348: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N \n",
            "2022-01-13 14:26:56.988493: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.988797: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.989117: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n",
            "2022-01-13 14:26:56.989373: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10010 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 2080 Ti, pci bus id: 0000:01:00.0, compute capability: 7.5)\n",
            "2022-01-13 14:26:56.989547: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
            "2022-01-13 14:26:57.200902: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
            "2022-01-13 14:26:57.201361: I tensorflow/core/platform/profile_utils/cpu_utils.cc:112] CPU Frequency: 3000000000 Hz\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/25\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "2022-01-13 14:26:57.715069: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
            "2022-01-13 14:26:57.834497: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 7s 19ms/step - loss: 1.9921 - accuracy: 0.2033 - auc: 0.6032 - val_loss: 1.4988 - val_accuracy: 0.4618 - val_auc: 0.8371\n",
            "Epoch 2/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 1.4799 - accuracy: 0.4572 - auc: 0.8334 - val_loss: 1.1825 - val_accuracy: 0.5831 - val_auc: 0.9028\n",
            "Epoch 3/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 1.2515 - accuracy: 0.5477 - auc: 0.8851 - val_loss: 1.0290 - val_accuracy: 0.6309 - val_auc: 0.9235\n",
            "Epoch 4/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 1.1311 - accuracy: 0.5935 - auc: 0.9067 - val_loss: 1.0011 - val_accuracy: 0.6461 - val_auc: 0.9292\n",
            "Epoch 5/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 1.0793 - accuracy: 0.6141 - auc: 0.9156 - val_loss: 0.9373 - val_accuracy: 0.6685 - val_auc: 0.9384\n",
            "Epoch 6/25\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.9964 - accuracy: 0.6481 - auc: 0.9281 - val_loss: 0.8892 - val_accuracy: 0.6820 - val_auc: 0.9431\n",
            "Epoch 7/25\n",
            "313/313 [==============================] - 4s 14ms/step - loss: 0.9618 - accuracy: 0.6558 - auc: 0.9331 - val_loss: 0.8642 - val_accuracy: 0.6990 - val_auc: 0.9469\n",
            "Epoch 8/25\n",
            "313/313 [==============================] - 4s 14ms/step - loss: 0.9205 - accuracy: 0.6773 - auc: 0.9386 - val_loss: 0.8460 - val_accuracy: 0.7032 - val_auc: 0.9486\n",
            "Epoch 9/25\n",
            "313/313 [==============================] - 5s 18ms/step - loss: 0.8955 - accuracy: 0.6859 - auc: 0.9418 - val_loss: 0.8393 - val_accuracy: 0.7013 - val_auc: 0.9499\n",
            "Epoch 10/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.8609 - accuracy: 0.6950 - auc: 0.9462 - val_loss: 0.7917 - val_accuracy: 0.7219 - val_auc: 0.9551\n",
            "Epoch 11/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.8329 - accuracy: 0.7065 - auc: 0.9494 - val_loss: 0.8178 - val_accuracy: 0.7143 - val_auc: 0.9514\n",
            "Epoch 12/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.8173 - accuracy: 0.7118 - auc: 0.9514 - val_loss: 0.7800 - val_accuracy: 0.7231 - val_auc: 0.9554\n",
            "Epoch 13/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7956 - accuracy: 0.7242 - auc: 0.9535 - val_loss: 0.7702 - val_accuracy: 0.7283 - val_auc: 0.9573\n",
            "Epoch 14/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7673 - accuracy: 0.7287 - auc: 0.9569 - val_loss: 0.7748 - val_accuracy: 0.7268 - val_auc: 0.9561\n",
            "Epoch 15/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7561 - accuracy: 0.7353 - auc: 0.9582 - val_loss: 0.7447 - val_accuracy: 0.7394 - val_auc: 0.9593\n",
            "Epoch 16/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7313 - accuracy: 0.7423 - auc: 0.9608 - val_loss: 0.7516 - val_accuracy: 0.7344 - val_auc: 0.9585\n",
            "Epoch 17/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7179 - accuracy: 0.7474 - auc: 0.9622 - val_loss: 0.7516 - val_accuracy: 0.7383 - val_auc: 0.9587\n",
            "Epoch 18/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.7078 - accuracy: 0.7552 - auc: 0.9632 - val_loss: 0.7401 - val_accuracy: 0.7398 - val_auc: 0.9598\n",
            "Epoch 19/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6788 - accuracy: 0.7619 - auc: 0.9659 - val_loss: 0.7432 - val_accuracy: 0.7426 - val_auc: 0.9590\n",
            "Epoch 20/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6735 - accuracy: 0.7650 - auc: 0.9664 - val_loss: 0.7326 - val_accuracy: 0.7458 - val_auc: 0.9601\n",
            "Epoch 21/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6571 - accuracy: 0.7724 - auc: 0.9679 - val_loss: 0.7266 - val_accuracy: 0.7449 - val_auc: 0.9609\n",
            "Epoch 22/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6354 - accuracy: 0.7769 - auc: 0.9699 - val_loss: 0.7268 - val_accuracy: 0.7426 - val_auc: 0.9609\n",
            "Epoch 23/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6420 - accuracy: 0.7734 - auc: 0.9693 - val_loss: 0.7497 - val_accuracy: 0.7409 - val_auc: 0.9586\n",
            "Epoch 24/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6333 - accuracy: 0.7775 - auc: 0.9698 - val_loss: 0.7143 - val_accuracy: 0.7526 - val_auc: 0.9622\n",
            "Epoch 25/25\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 0.6147 - accuracy: 0.7839 - auc: 0.9717 - val_loss: 0.7073 - val_accuracy: 0.7538 - val_auc: 0.9627\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "(<tensorflow.python.keras.engine.sequential.Sequential at 0x7f7e34188100>,\n",
              " <tensorflow.python.keras.callbacks.History at 0x7f7e337ec700>)"
            ]
          },
          "execution_count": 19,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "base_cnn_model_changes()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "926d0beb",
      "metadata": {
        "id": "926d0beb"
      },
      "source": [
        "#### <a name=\"resnet50_cnn_model\">ResNet-50 CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 30,
      "id": "a778b88f",
      "metadata": {},
      "outputs": [],
      "source": [
        "# Reference: https://github.com/y33-j3T/Coursera-Deep-Learning/blob/master/Advanced%20Computer%20Vision%20with%20TensorFlow/Week%201%20-%20Introduction%20to%20Computer%20Vision/Copy%20of%20C3_W1_Lab_2_Transfer_Learning_CIFAR_10.ipynb\n",
        "def preprocess_image_resnet50(data_images: None) -> None:\n",
        "    \"\"\"\n",
        "    Function to preprocess the images for ResNet50 model.\n",
        "    :param data_images: Images to be preprocessed.\n",
        "    :return: Preprocessed images.\n",
        "    :rtype: None\n",
        "    \"\"\"\n",
        "    data_images = data_images.astype('float32')\n",
        "    output_images = tf.keras.applications.resnet50.preprocess_input(\n",
        "        data_images)\n",
        "    return output_images\n",
        "\n",
        "\n",
        "# Using function to preprocess the images.\n",
        "train_X = preprocess_image_resnet50(train_imgs)\n",
        "valid_X = preprocess_image_resnet50(test_imgs)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 35,
      "id": "14b7f218",
      "metadata": {
        "id": "14b7f218",
        "outputId": "d5b29c3c-0ecd-4ea0-f7d2-a8b26202e898"
      },
      "outputs": [],
      "source": [
        "# Reference: https://github.com/y33-j3T/Coursera-Deep-Learning/blob/master/Advanced%20Computer%20Vision%20with%20TensorFlow/Week%201%20-%20Introduction%20to%20Computer%20Vision/Copy%20of%20C3_W1_Lab_2_Transfer_Learning_CIFAR_10.ipynb\n",
        "\n",
        "keras.backend.clear_session()\n",
        "np.random.seed(15)\n",
        "tf.random.set_seed(15)\n",
        "\n",
        "\n",
        "def resnet50_model(image_inputs: None) -> None:\n",
        "    \"\"\"\n",
        "    Pretrained model on imagenet weights with respective input shape.\n",
        "    :param inputs: Inputs to the model.\n",
        "    :return: Outputs of the model.\n",
        "    :rtype: None\n",
        "    \"\"\"\n",
        "    resnet50_model.resnet50_model_features = tf.keras.applications.ResNet50(input_shape=(224, 224, 3),\n",
        "                                                                            include_top=False,\n",
        "                                                                            weights='imagenet')(image_inputs)\n",
        "\n",
        "    return resnet50_model.resnet50_model_features\n",
        "\n",
        "\n",
        "def resnet50_classifier(image_inputs: None) -> None:\n",
        "    \"\"\"\n",
        "    Initialize hidden layer and output layer with softmax activation function.\n",
        "    :param image_inputs: Image inputs to the model.\n",
        "    :return: Outputs of the model.\n",
        "    :rtype: None\n",
        "    \"\"\"\n",
        "    classify = tf.keras.layers.GlobalAveragePooling2D()(image_inputs)\n",
        "    classify = tf.keras.layers.Flatten()(classify)\n",
        "    classify = tf.keras.layers.Dense(1024, activation=\"relu\")(classify)\n",
        "    classify = tf.keras.layers.Dense(512, activation=\"relu\")(classify)\n",
        "    classify = tf.keras.layers.Dense(8, activation=\"softmax\",\n",
        "                                     name=\"classification\")(classify)\n",
        "\n",
        "    return classify\n",
        "\n",
        "\n",
        "def complete_resnet50_model(image_inputs: None) -> None:\n",
        "    \"\"\"\n",
        "    Upsampling in order to transform the images with the input size as (224, 224).\n",
        "    :param image_inputs: Image inputs to the model.\n",
        "    :return: Classification layer of the model.\n",
        "    :rtype: None\n",
        "    \"\"\"\n",
        "\n",
        "    upsample = tf.keras.layers.UpSampling2D(size=(7, 7))(image_inputs)\n",
        "\n",
        "    resnet_extractor = resnet50_model(upsample)\n",
        "    final_classification = resnet50_classifier(resnet_extractor)\n",
        "\n",
        "    return final_classification\n",
        "\n",
        "\n",
        "def compile_resnet50_model() -> None:\n",
        "    \"\"\"\n",
        "    Compile the model with the optimizer, loss function and metrics.\n",
        "    :param model: Model to be compiled.\n",
        "    :return: Compiled model.\n",
        "    :rtype: None\n",
        "    \"\"\"\n",
        "    image_inputs = tf.keras.layers.Input(shape=(32, 32, 3))\n",
        "\n",
        "    final_classification = complete_resnet50_model(image_inputs)\n",
        "    model = tf.keras.Model(inputs=image_inputs, outputs=final_classification)\n",
        "\n",
        "    model.compile(optimizer='SGD',\n",
        "                  loss='sparse_categorical_crossentropy',\n",
        "                  metrics=['accuracy'])\n",
        "\n",
        "    return model\n",
        "\n",
        "\n",
        "model = compile_resnet50_model()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 36,
      "id": "005596d6",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         [(None, 32, 32, 3)]       0         \n",
            "_________________________________________________________________\n",
            "up_sampling2d (UpSampling2D) (None, 224, 224, 3)       0         \n",
            "_________________________________________________________________\n",
            "resnet50 (Functional)        (None, 7, 7, 2048)        23587712  \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d (Gl (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 1024)              2098176   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 512)               524800    \n",
            "_________________________________________________________________\n",
            "classification (Dense)       (None, 8)                 4104      \n",
            "=================================================================\n",
            "Total params: 26,214,792\n",
            "Trainable params: 26,161,672\n",
            "Non-trainable params: 53,120\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "id": "a43fd993",
      "metadata": {
        "id": "a43fd993",
        "outputId": "18ec42dc-408f-436b-b842-d17f06b0e39d"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/4\n",
            "625/625 [==============================] - 160s 252ms/step - loss: 0.7727 - accuracy: 0.7373 - val_loss: 0.2148 - val_accuracy: 0.9274\n",
            "Epoch 2/4\n",
            "625/625 [==============================] - 159s 254ms/step - loss: 0.1176 - accuracy: 0.9633 - val_loss: 0.1616 - val_accuracy: 0.9463\n",
            "Epoch 3/4\n",
            "625/625 [==============================] - 159s 255ms/step - loss: 0.0397 - accuracy: 0.9881 - val_loss: 0.1654 - val_accuracy: 0.9500\n",
            "Epoch 4/4\n",
            "625/625 [==============================] - 159s 255ms/step - loss: 0.0157 - accuracy: 0.9961 - val_loss: 0.1704 - val_accuracy: 0.9521\n"
          ]
        }
      ],
      "source": [
        "# Training ResNet50 model for 4 epochs.\n",
        "resnet_model = model.fit(train_X, train_labels, epochs=4,\n",
        "                         validation_data=(valid_X, test_labels), batch_size=64)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "944356ab",
      "metadata": {
        "id": "944356ab"
      },
      "source": [
        "#### <a name=\"best_cnn_model\">Best CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 57,
      "id": "cd6e24b8",
      "metadata": {
        "id": "cd6e24b8",
        "outputId": "e3183da7-c4ac-4ef2-8fea-c9013b8e112e"
      },
      "outputs": [],
      "source": [
        "# To prevent any memory build up from old models or layers.\n",
        "keras.backend.clear_session()\n",
        "np.random.seed(15)\n",
        "tf.random.set_seed(15)\n",
        "\n",
        "\n",
        "def best_cnn_model() -> Sequential:\n",
        "    \"\"\"\n",
        "    Function which returns the best model.\n",
        "    :rtype: Sequential\n",
        "    \"\"\"\n",
        "    # Creating an early stopping function to monitor loss while keeping the best results.\n",
        "    early_callback = tf.keras.callbacks.EarlyStopping(\n",
        "        monitor='loss', restore_best_weights=True, patience=10)\n",
        "\n",
        "    # Initializing Sequential model, creating linear stack of model layers.\n",
        "    best_cnn_model.model2 = Sequential()\n",
        "\n",
        "    # Adding 6 convolutional layers with max pooling and dropout functions.\n",
        "    best_cnn_model.model2.add(Conv2D(32, input_shape=(32, 32, 3), kernel_size=(\n",
        "        3, 3), strides=(1, 1), padding=\"same\", activation=\"relu\"))\n",
        "\n",
        "    best_cnn_model.model2.add(Conv2D(64, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding=\"same\", activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    best_cnn_model.model2.add(Dropout(0.25))\n",
        "\n",
        "    best_cnn_model.model2.add(Conv2D(128, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding=\"same\", activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    best_cnn_model.model2.add(Dropout(0.25))\n",
        "\n",
        "    best_cnn_model.model2.add(Conv2D(256, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding=\"same\", activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    best_cnn_model.model2.add(Dropout(0.25))\n",
        "\n",
        "    best_cnn_model.model2.add(Conv2D(512, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding=\"same\", activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    best_cnn_model.model2.add(Dropout(0.25))\n",
        "\n",
        "    best_cnn_model.model2.add(Conv2D(512, kernel_size=(3, 3), strides=(\n",
        "        1, 1), padding=\"same\", activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(MaxPool2D(pool_size=(2, 2)))\n",
        "    best_cnn_model.model2.add(Dropout(0.25))\n",
        "\n",
        "    # Flattening convolutional layer output.\n",
        "    best_cnn_model.model2.add(Flatten())\n",
        "\n",
        "    # Adding 2 hidden layers.\n",
        "    best_cnn_model.model2.add(Dense(500, activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(Dropout(0.75))\n",
        "    best_cnn_model.model2.add(Dense(200, activation=\"relu\"))\n",
        "    best_cnn_model.model2.add(Dropout(0.3))\n",
        "\n",
        "    # Final output layer with 8 classes.\n",
        "    best_cnn_model.model2.add(Dense(8, activation=\"softmax\"))\n",
        "\n",
        "    # Compilation of sequential model with suitable parameters.\n",
        "    best_cnn_model.model2.compile(optimizer='adam', loss='categorical_crossentropy',\n",
        "                                  metrics=['accuracy'])\n",
        "\n",
        "    # Training model for 25 epochs\n",
        "    best_cnn_model.best_model = best_cnn_model.model2.fit(rotated_train_img, Y_train, batch_size=128, callbacks=[\n",
        "        early_callback], epochs=20, validation_data=(rotated_test_img, Y_test))\n",
        "\n",
        "    return best_cnn_model.model2, best_cnn_model.best_model\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 50,
      "id": "1977c5df",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 32, 32, 32)        896       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 16, 16, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 16, 16, 128)       73856     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 8, 8, 128)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 8, 8, 256)         295168    \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 4, 4, 256)         0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 4, 4, 256)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 4, 4, 512)         1180160   \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 2, 2, 512)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 2, 2, 512)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 2, 2, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2 (None, 1, 1, 512)         0         \n",
            "_________________________________________________________________\n",
            "dropout_4 (Dropout)          (None, 1, 1, 512)         0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 500)               256500    \n",
            "_________________________________________________________________\n",
            "dropout_5 (Dropout)          (None, 500)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 200)               100200    \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 200)               0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 8)                 1608      \n",
            "=================================================================\n",
            "Total params: 4,286,692\n",
            "Trainable params: 4,286,692\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "best_cnn_model.model2.summary()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 58,
      "id": "38e0431f",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "313/313 [==============================] - 5s 14ms/step - loss: 1.9594 - accuracy: 0.1898 - val_loss: 1.5746 - val_accuracy: 0.3964\n",
            "Epoch 2/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 1.5124 - accuracy: 0.4245 - val_loss: 1.2065 - val_accuracy: 0.5579\n",
            "Epoch 3/20\n",
            "313/313 [==============================] - 5s 14ms/step - loss: 1.2378 - accuracy: 0.5405 - val_loss: 1.1165 - val_accuracy: 0.5974\n",
            "Epoch 4/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 1.0780 - accuracy: 0.6058 - val_loss: 0.9227 - val_accuracy: 0.6614\n",
            "Epoch 5/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.9709 - accuracy: 0.6503 - val_loss: 0.8696 - val_accuracy: 0.6900\n",
            "Epoch 6/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.9040 - accuracy: 0.6811 - val_loss: 0.9158 - val_accuracy: 0.6764\n",
            "Epoch 7/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.8738 - accuracy: 0.6932 - val_loss: 0.7948 - val_accuracy: 0.7272\n",
            "Epoch 8/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.8207 - accuracy: 0.7159 - val_loss: 0.8147 - val_accuracy: 0.7165\n",
            "Epoch 9/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.8007 - accuracy: 0.7235 - val_loss: 0.7042 - val_accuracy: 0.7536\n",
            "Epoch 10/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.7311 - accuracy: 0.7472 - val_loss: 0.6767 - val_accuracy: 0.7679\n",
            "Epoch 11/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.7059 - accuracy: 0.7596 - val_loss: 0.6884 - val_accuracy: 0.7613\n",
            "Epoch 12/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.6960 - accuracy: 0.7613 - val_loss: 0.6557 - val_accuracy: 0.7656\n",
            "Epoch 13/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.6831 - accuracy: 0.7690 - val_loss: 0.6328 - val_accuracy: 0.7799\n",
            "Epoch 14/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.6544 - accuracy: 0.7754 - val_loss: 0.6453 - val_accuracy: 0.7812\n",
            "Epoch 15/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.6375 - accuracy: 0.7816 - val_loss: 0.6492 - val_accuracy: 0.7770\n",
            "Epoch 16/20\n",
            "313/313 [==============================] - 4s 14ms/step - loss: 0.6123 - accuracy: 0.7904 - val_loss: 0.6196 - val_accuracy: 0.7836\n",
            "Epoch 17/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.6001 - accuracy: 0.7968 - val_loss: 0.6111 - val_accuracy: 0.7922\n",
            "Epoch 18/20\n",
            "313/313 [==============================] - 4s 14ms/step - loss: 0.5908 - accuracy: 0.7963 - val_loss: 0.5734 - val_accuracy: 0.8014\n",
            "Epoch 19/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.5820 - accuracy: 0.8019 - val_loss: 0.5891 - val_accuracy: 0.7972\n",
            "Epoch 20/20\n",
            "313/313 [==============================] - 5s 15ms/step - loss: 0.5803 - accuracy: 0.8030 - val_loss: 0.5634 - val_accuracy: 0.8074\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "(<tensorflow.python.keras.engine.sequential.Sequential at 0x7f7e325f8ca0>,\n",
              " <tensorflow.python.keras.callbacks.History at 0x7f7e31f622e0>)"
            ]
          },
          "execution_count": 58,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "best_cnn_model()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "75e92936",
      "metadata": {
        "id": "75e92936"
      },
      "source": [
        "## <a name=\"model_evaluation\">Model Evaluation</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "494300df",
      "metadata": {
        "id": "494300df"
      },
      "source": [
        "### <a name=\"visualization_base_cnn_model\">Visualization of Base CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "id": "384c5987",
      "metadata": {
        "id": "384c5987",
        "outputId": "f014e519-29c6-48fb-90e0-fdbaf372b83c"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "250/250 - 1s - loss: 0.6599 - accuracy: 0.7736 - auc: 0.9674\n",
            "Test accuracy on test data split ->  0.7736250162124634\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Create line graph for displaying test and train accuracy.\n",
        "plt.plot(\n",
        "    base_cnn_model.base_final_model.history['accuracy'], label='Train Accuracy')\n",
        "plt.plot(\n",
        "    base_cnn_model.base_final_model.history['val_accuracy'], label='Test Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='best')\n",
        "\n",
        "# Calculate the test accuracy.\n",
        "test_loss, test_acc, * \\\n",
        "    _ = base_cnn_model.model1.evaluate(rotated_test_img, Y_test, verbose=2)\n",
        "print('Test accuracy on test data split -> ', test_acc)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "43004e8f",
      "metadata": {},
      "source": [
        "### <a name=\"visualization_base_cnn_model_changes\">Visualization of Base CNN Model with Changes</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "id": "d51a4d9a",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "250/250 - 1s - loss: 0.7073 - accuracy: 0.7538 - auc: 0.9627\n",
            "Test accuracy on test data split ->  0.7537500262260437\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Plot line graph for displaying test and train accuracy.\n",
        "plt.plot(\n",
        "    base_cnn_model_changes.base_final_model_changes.history['accuracy'], label='Train Accuracy')\n",
        "plt.plot(\n",
        "    base_cnn_model_changes.base_final_model_changes.history['val_accuracy'], label='Test Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='best')\n",
        "\n",
        "# Calculate the test accuracy.\n",
        "test_loss, test_acc, * \\\n",
        "    _ = base_cnn_model_changes.changes_model.evaluate(\n",
        "        rotated_test_img, Y_test, verbose=2)\n",
        "print('Test accuracy on test data split -> ', test_acc)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "ef1617cc",
      "metadata": {},
      "source": [
        "### <a name=\"visualization_resnet50_cnn_model\">Visualization of ResNet50 CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 42,
      "id": "6d908fed",
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "250/250 [==============================] - 9s 34ms/step - loss: 0.8188 - accuracy: 0.7845\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Create line graph for displaying test and train accuracy.\n",
        "plt.plot(resnet_model.history['accuracy'], label='Train Accuracy')\n",
        "plt.plot(resnet_model.history['val_accuracy'], label='Test Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='best')\n",
        "\n",
        "test_loss, test_acc, *_ = model.evaluate(test_imgs, test_labels)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "f3f54aff",
      "metadata": {
        "id": "f3f54aff"
      },
      "source": [
        "### <a name=\"visualization_best_cnn_model\">Visualization of Best Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 59,
      "id": "32c00bc1",
      "metadata": {
        "id": "32c00bc1",
        "outputId": "d11f284b-54c9-4325-e4ca-009fe65d5c97"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "250/250 [==============================] - 1s 5ms/step - loss: 0.5634 - accuracy: 0.8074\n",
            "Test accuracy on test data split ->  0.8073750138282776\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Create line graph for displaying test and train accuracy.\n",
        "plt.plot(best_cnn_model.best_model.history['accuracy'], label='Train Accuracy')\n",
        "plt.plot(\n",
        "    best_cnn_model.best_model.history['val_accuracy'], label='Test Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='best')\n",
        "\n",
        "# Calculate the test accuracy.\n",
        "test_loss, test_acc, * \\\n",
        "    _ = best_cnn_model.model2.evaluate(rotated_test_img, Y_test)\n",
        "print('Test accuracy on test data split -> ', test_acc)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "8c80f460",
      "metadata": {
        "id": "8c80f460"
      },
      "source": [
        "## <a name=\"model_prediction\">Model Prediction</a>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "7bcf022f",
      "metadata": {
        "id": "7bcf022f"
      },
      "source": [
        "### <a name=\"prediction_best_cnn_model\">Prediction of Best CNN Model</a>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 60,
      "id": "4e464158",
      "metadata": {
        "id": "4e464158"
      },
      "outputs": [],
      "source": [
        "# Reference: https://github.com/y33-j3T/Coursera-Deep-Learning/blob/master/Advanced%20Computer%20Vision%20with%20TensorFlow/Week%201%20-%20Introduction%20to%20Computer%20Vision/Copy%20of%20C3_W1_Lab_2_Transfer_Learning_CIFAR_10.ipynb\n",
        "\n",
        "# Define classes/objects.\n",
        "classes = ['airplane', 'automobile', 'bird',\n",
        "           'cat', 'deer', 'dog', 'frog', 'horse']\n",
        "\n",
        "\n",
        "def display_predictions(predict_test_imgs: None, preds: str, labels: str, title: str) -> str:\n",
        "    \"\"\"\n",
        "    Define function to plot images.\n",
        "    Displays images with actual and predicted labels.\n",
        "    :param None predict_test_imgs: Image to be displayed.\n",
        "    :param str preds: Predicted labels.\n",
        "    :param str labels: Actual labels.\n",
        "    :param str title: Title of figure.\n",
        "    :rtype: str\n",
        "    \"\"\"\n",
        "\n",
        "    n_of_classes = 8\n",
        "\n",
        "    indexes = np.random.choice(len(preds), size=n_of_classes)\n",
        "    n_predict_test_imgs = predict_test_imgs[indexes]\n",
        "    n_preds = preds[indexes]\n",
        "    n_preds = n_preds.reshape((n_of_classes))\n",
        "    n_labels = labels[indexes]\n",
        "\n",
        "    fig = plt.figure(figsize=(20, 3))\n",
        "    plt.title(title)\n",
        "    plt.yticks([])\n",
        "    plt.xticks([])\n",
        "\n",
        "# Display 8 random test images along with their actual and predicted labels.\n",
        "    for z in range(8):\n",
        "        ax = fig.add_subplot(1, 8, z+1)\n",
        "        class_index = n_preds[z]\n",
        "        actual_pred_labels = 'Predicted -> ' + classes[class_index]\n",
        "        plt.xlabel(actual_pred_labels)\n",
        "        plt.xticks([])\n",
        "        plt.yticks([])\n",
        "        plt.imshow(n_predict_test_imgs[z])\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 62,
      "id": "3V6fxzz7UujW",
      "metadata": {
        "id": "3V6fxzz7UujW",
        "outputId": "e5a5c18c-9ac9-4662-a660-86b0bf18cef9"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1440x216 with 9 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Using the trained model to predict the test images.\n",
        "pred_chance = best_cnn_model.model2.predict(rotated_test_img, batch_size=128)\n",
        "pred_chance = np.argmax(pred_chance, axis=1)\n",
        "\n",
        "# Utilizing display_predictions function to display images.\n",
        "display_predictions(\n",
        "    rotated_test_img, pred_chance, Y_test, 'Predictions on test data')\n"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "name": "Copy of image_recognition_dataset.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.7"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}