VGG16_Adaptations_CIFAR10.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [
        "l9zBOHkhH9yK",
        "2q7AvUMNH5OK",
        "yYxGuj8gImXD",
        "HXFRPJTnMTNf"
      ],
      "authorship_tag": "ABX9TyMwDnHHdxdls8Hg+qFtnIli",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "gpuClass": "standard",
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Antonio-Villarreal/Deep-Learning-For-Dummies/blob/main/VGG16_Adaptations_CIFAR10.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# VGG16 Adaptation"
      ],
      "metadata": {
        "id": "i8eGmxl97DBP"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Adapts the VGG16 architecture for a smaller image size - 244 x 244"
      ],
      "metadata": {
        "id": "eTsyqIo5G9l0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.datasets import cifar10\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "# load the CIFAR10 dataset\n",
        "(x_train, y_train), (x_test, y_test) = cifar10.load_data()"
      ],
      "metadata": {
        "id": "-qNYu7v-7Ard",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "96b070fb-8d2b-416f-e806-56ccb56f4b6a"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170498071/170498071 [==============================] - 3s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(x_train.shape)\n",
        "print(y_train.shape)\n",
        "print(x_test.shape)\n",
        "print(x_test.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AvwGbGva9mCS",
        "outputId": "78f317bf-4be0-4510-db63-3049db05330b"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(50000, 32, 32, 3)\n",
            "(50000, 1)\n",
            "(10000, 32, 32, 3)\n",
            "(10000, 32, 32, 3)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "x_train /= 255\n",
        "x_test /= 255"
      ],
      "metadata": {
        "id": "K3ACGStM7O0l"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.utils import to_categorical\n",
        "y_train = to_categorical(y_train, 10)\n",
        "y_test = to_categorical(y_test, 10)"
      ],
      "metadata": {
        "id": "iX8NVPeL9qdA"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Model 1"
      ],
      "metadata": {
        "id": "l9zBOHkhH9yK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, BatchNormalization, Dropout\n",
        "from keras.models import Sequential\n",
        "\n",
        "model = Sequential()\n",
        "\n",
        "# Layer 1: Convolutional layer with 64 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), input_shape=(32, 32, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 2: Convolutional layer with 64 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(64, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "model.add(BatchNormalization())\n",
        "\n",
        "# Layer 3: Max pooling layer with pool size of 2x2\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 4: Convolutional layer with 128 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(128, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 5: Convolutional layer with 128 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(128, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "model.add(BatchNormalization())\n",
        "\n",
        "# Layer 6: Max pooling layer with pool size of 2x2\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 7: Convolutional layer with 256 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(256, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 8: Convolutional layer with 256 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(256, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 9: Convolutional layer with 256 filters, kernel size of 3x3, and 'same' padding\n",
        "model.add(Conv2D(256, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "model.add(BatchNormalization())\n",
        "\n",
        "# Layer 10: Max pooling layer with pool size of 2x2\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 11: Flatten layer to convert the 3D feature maps to 1D feature vectors\n",
        "model.add(Flatten())\n",
        "\n",
        "# Layer 12: Fully connected layer with 512 units and 'relu' activation\n",
        "model.add(Dense(512, activation='relu'))\n",
        "\n",
        "# Layer 13: Fully connected layer with 512 units and 'relu' activation\n",
        "model.add(Dense(512, activation='relu'))\n",
        "\n",
        "# Layer 14: Output layer with 'softmax' activation for multi-class classification\n",
        "model.add(Dense(10, activation='softmax'))"
      ],
      "metadata": {
        "id": "pE1OaQvaDMhp"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pG-DBNEf9-uM",
        "outputId": "75a650eb-61af-4815-90b6-4b390e3452e8"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d (Conv2D)             (None, 32, 32, 64)        1792      \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 32, 32, 64)        36928     \n",
            "                                                                 \n",
            " batch_normalization (BatchN  (None, 32, 32, 64)       256       \n",
            " ormalization)                                                   \n",
            "                                                                 \n",
            " max_pooling2d (MaxPooling2D  (None, 16, 16, 64)       0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " conv2d_2 (Conv2D)           (None, 16, 16, 128)       73856     \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 16, 16, 128)       147584    \n",
            "                                                                 \n",
            " batch_normalization_1 (Batc  (None, 16, 16, 128)      512       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " max_pooling2d_1 (MaxPooling  (None, 8, 8, 128)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_4 (Conv2D)           (None, 8, 8, 256)         295168    \n",
            "                                                                 \n",
            " conv2d_5 (Conv2D)           (None, 8, 8, 256)         590080    \n",
            "                                                                 \n",
            " conv2d_6 (Conv2D)           (None, 8, 8, 256)         590080    \n",
            "                                                                 \n",
            " batch_normalization_2 (Batc  (None, 8, 8, 256)        1024      \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " max_pooling2d_2 (MaxPooling  (None, 4, 4, 256)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 4096)              0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 512)               2097664   \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 512)               262656    \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 10)                5130      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 4,102,730\n",
            "Trainable params: 4,101,834\n",
            "Non-trainable params: 896\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Model 2"
      ],
      "metadata": {
        "id": "2q7AvUMNH5OK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense\n",
        "from keras.models import Sequential\n",
        "\n",
        "model2 = Sequential()\n",
        "\n",
        "# Layer 1: Convolutional layer with 32 filters, kernel size of 3x3, and 'same' padding\n",
        "model2.add(Conv2D(32, kernel_size=(3, 3), input_shape=(32, 32, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 2: Max pooling layer with pool size of 2x2\n",
        "model2.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 3: Convolutional layer with 64 filters, kernel size of 3x3, and 'same' padding\n",
        "model2.add(Conv2D(64, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "\n",
        "# Layer 4: Max pooling layer with pool size of 2x2\n",
        "model2.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 5: Flatten layer to convert the 3D feature maps to 1D feature vectors\n",
        "model2.add(Flatten())\n",
        "\n",
        "# Layer 6: Fully connected layer with 128 units and 'relu' activation\n",
        "model2.add(Dense(128, activation='relu'))\n",
        "\n",
        "# Layer 7: Output layer with 'softmax' activation for multi-class classification\n",
        "model2.add(Dense(10, activation='softmax'))"
      ],
      "metadata": {
        "id": "AQ06Scev7QoG"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model2.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zcmAyCkkHO3s",
        "outputId": "7aa1ca31-19fe-40fc-9f04-77a315405f06"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_7 (Conv2D)           (None, 32, 32, 32)        896       \n",
            "                                                                 \n",
            " max_pooling2d_3 (MaxPooling  (None, 16, 16, 32)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_8 (Conv2D)           (None, 16, 16, 64)        18496     \n",
            "                                                                 \n",
            " max_pooling2d_4 (MaxPooling  (None, 8, 8, 64)         0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " flatten_1 (Flatten)         (None, 4096)              0         \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 128)               524416    \n",
            "                                                                 \n",
            " dense_4 (Dense)             (None, 10)                1290      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 545,098\n",
            "Trainable params: 545,098\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Model 3"
      ],
      "metadata": {
        "id": "yYxGuj8gImXD"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model3 = Sequential()\n",
        "\n",
        "# Layer 1: Convolutional layer with 32 filters, kernel size of 3x3, and 'same' padding\n",
        "model3.add(Conv2D(32, kernel_size=(3, 3), input_shape=(32, 32, 3), padding='same', activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 2: Convolutional layer with 32 filters, kernel size of 3x3, and 'same' padding\n",
        "model3.add(Conv2D(32, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 3: Max pooling layer with pool size of 2x2\n",
        "model3.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 4: Convolutional layer with 64 filters, kernel size of 3x3, and 'same' padding\n",
        "model3.add(Conv2D(64, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 5: Convolutional layer with 64 filters, kernel size of 3x3, and 'same' padding\n",
        "model3.add(Conv2D(64, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 6: Max pooling layer with pool size of 2x2\n",
        "model3.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# Layer 7: Convolutional layer with 128 filters, kernel size of 3x3, and 'same' padding\n",
        "model3.add(Conv2D(128, kernel_size=(3, 3), padding='same', activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 8: Flatten layer to convert the 3D feature maps to 1D feature vectors\n",
        "model3.add(Flatten())\n",
        "\n",
        "# Layer 9: Fully connected layer with 512 units and 'relu' activation\n",
        "model3.add(Dense(512, activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 10: Fully connected layer with 512 units and 'relu' activation\n",
        "model3.add(Dense(512, activation='relu'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Dropout(0.2))\n",
        "\n",
        "# Layer 11: Output layer with 'softmax' activation for multi-class classification\n",
        "model3.add(Dense(10, activation='softmax'))"
      ],
      "metadata": {
        "id": "GA1rXQCqIn3k"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model3.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oGmC2KU8Iqg1",
        "outputId": "d879683b-cd4b-4438-922a-07dabfd2d81f"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_9 (Conv2D)           (None, 32, 32, 32)        896       \n",
            "                                                                 \n",
            " batch_normalization_3 (Batc  (None, 32, 32, 32)       128       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 32, 32, 32)        0         \n",
            "                                                                 \n",
            " conv2d_10 (Conv2D)          (None, 32, 32, 32)        9248      \n",
            "                                                                 \n",
            " batch_normalization_4 (Batc  (None, 32, 32, 32)       128       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_1 (Dropout)         (None, 32, 32, 32)        0         \n",
            "                                                                 \n",
            " max_pooling2d_5 (MaxPooling  (None, 16, 16, 32)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_11 (Conv2D)          (None, 16, 16, 64)        18496     \n",
            "                                                                 \n",
            " batch_normalization_5 (Batc  (None, 16, 16, 64)       256       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_2 (Dropout)         (None, 16, 16, 64)        0         \n",
            "                                                                 \n",
            " conv2d_12 (Conv2D)          (None, 16, 16, 64)        36928     \n",
            "                                                                 \n",
            " batch_normalization_6 (Batc  (None, 16, 16, 64)       256       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_3 (Dropout)         (None, 16, 16, 64)        0         \n",
            "                                                                 \n",
            " max_pooling2d_6 (MaxPooling  (None, 8, 8, 64)         0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_13 (Conv2D)          (None, 8, 8, 128)         73856     \n",
            "                                                                 \n",
            " batch_normalization_7 (Batc  (None, 8, 8, 128)        512       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_4 (Dropout)         (None, 8, 8, 128)         0         \n",
            "                                                                 \n",
            " flatten_2 (Flatten)         (None, 8192)              0         \n",
            "                                                                 \n",
            " dense_5 (Dense)             (None, 512)               4194816   \n",
            "                                                                 \n",
            " batch_normalization_8 (Batc  (None, 512)              2048      \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_5 (Dropout)         (None, 512)               0         \n",
            "                                                                 \n",
            " dense_6 (Dense)             (None, 512)               262656    \n",
            "                                                                 \n",
            " batch_normalization_9 (Batc  (None, 512)              2048      \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_6 (Dropout)         (None, 512)               0         \n",
            "                                                                 \n",
            " dense_7 (Dense)             (None, 10)                5130      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 4,607,402\n",
            "Trainable params: 4,604,714\n",
            "Non-trainable params: 2,688\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Model 4 (Imported)"
      ],
      "metadata": {
        "id": "HXFRPJTnMTNf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.applications.vgg16 import VGG16\n",
        "from keras.applications.vgg16 import preprocess_input\n",
        "from keras.applications.vgg16 import decode_predictions\n",
        "\n",
        "# base_model = VGG16(include_top=False, input_shape=(32,32,3))\n",
        "base_model = VGG16(weights='imagenet', include_top=False, input_shape=(32,32,3))"
      ],
      "metadata": {
        "id": "HUgtfza7MWIi",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "278b0cb6-b47e-41c2-818f-928066cc3510"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58889256/58889256 [==============================] - 0s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Freeze base model layers\n",
        "for layer in base_model.layers:\n",
        "    layer.trainable = False"
      ],
      "metadata": {
        "id": "VKmllHXWMZdg"
      },
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Add new layers\n",
        "x = base_model.output\n",
        "x = Flatten()(x)\n",
        "x = Dense(4096, activation='relu')(x)\n",
        "predictions = Dense(10, activation='softmax')(x)"
      ],
      "metadata": {
        "id": "Eccd83caMdRF"
      },
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.models import Model\n",
        "model4 = Model(inputs=base_model.input, outputs=predictions)"
      ],
      "metadata": {
        "id": "0XZJEg47Meyx"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model4.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZQJhllZCMgqe",
        "outputId": "ac4f75ff-2edc-431b-f05e-1cc0777cac9a"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_1 (InputLayer)        [(None, 32, 32, 3)]       0         \n",
            "                                                                 \n",
            " block1_conv1 (Conv2D)       (None, 32, 32, 64)        1792      \n",
            "                                                                 \n",
            " block1_conv2 (Conv2D)       (None, 32, 32, 64)        36928     \n",
            "                                                                 \n",
            " block1_pool (MaxPooling2D)  (None, 16, 16, 64)        0         \n",
            "                                                                 \n",
            " block2_conv1 (Conv2D)       (None, 16, 16, 128)       73856     \n",
            "                                                                 \n",
            " block2_conv2 (Conv2D)       (None, 16, 16, 128)       147584    \n",
            "                                                                 \n",
            " block2_pool (MaxPooling2D)  (None, 8, 8, 128)         0         \n",
            "                                                                 \n",
            " block3_conv1 (Conv2D)       (None, 8, 8, 256)         295168    \n",
            "                                                                 \n",
            " block3_conv2 (Conv2D)       (None, 8, 8, 256)         590080    \n",
            "                                                                 \n",
            " block3_conv3 (Conv2D)       (None, 8, 8, 256)         590080    \n",
            "                                                                 \n",
            " block3_pool (MaxPooling2D)  (None, 4, 4, 256)         0         \n",
            "                                                                 \n",
            " block4_conv1 (Conv2D)       (None, 4, 4, 512)         1180160   \n",
            "                                                                 \n",
            " block4_conv2 (Conv2D)       (None, 4, 4, 512)         2359808   \n",
            "                                                                 \n",
            " block4_conv3 (Conv2D)       (None, 4, 4, 512)         2359808   \n",
            "                                                                 \n",
            " block4_pool (MaxPooling2D)  (None, 2, 2, 512)         0         \n",
            "                                                                 \n",
            " block5_conv1 (Conv2D)       (None, 2, 2, 512)         2359808   \n",
            "                                                                 \n",
            " block5_conv2 (Conv2D)       (None, 2, 2, 512)         2359808   \n",
            "                                                                 \n",
            " block5_conv3 (Conv2D)       (None, 2, 2, 512)         2359808   \n",
            "                                                                 \n",
            " block5_pool (MaxPooling2D)  (None, 1, 1, 512)         0         \n",
            "                                                                 \n",
            " flatten_3 (Flatten)         (None, 512)               0         \n",
            "                                                                 \n",
            " dense_8 (Dense)             (None, 4096)              2101248   \n",
            "                                                                 \n",
            " dense_9 (Dense)             (None, 10)                40970     \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 16,856,906\n",
            "Trainable params: 2,142,218\n",
            "Non-trainable params: 14,714,688\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Training"
      ],
      "metadata": {
        "id": "lyIaZpXYIhu-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "EWKp1koGBAso"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model2.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "xteiVkBtHRh5"
      },
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model3.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "3dGX_ADqI4sP"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model4.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "i9CP7CLpMo5O"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Epoch = 20"
      ],
      "metadata": {
        "id": "cHDGL092snMt"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(x_train, y_train, batch_size=64, epochs=Epoch, validation_data=(x_test, y_test))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Gv8KPJIz8LXP",
        "outputId": "f811ade9-658e-408a-af86-dfbb26c3a626"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "782/782 [==============================] - 26s 21ms/step - loss: 1.4322 - accuracy: 0.4884 - val_loss: 1.3653 - val_accuracy: 0.5191\n",
            "Epoch 2/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.8914 - accuracy: 0.6859 - val_loss: 1.0936 - val_accuracy: 0.6363\n",
            "Epoch 3/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.6815 - accuracy: 0.7622 - val_loss: 0.7659 - val_accuracy: 0.7404\n",
            "Epoch 4/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.5549 - accuracy: 0.8085 - val_loss: 0.7886 - val_accuracy: 0.7308\n",
            "Epoch 5/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.4570 - accuracy: 0.8412 - val_loss: 0.7531 - val_accuracy: 0.7504\n",
            "Epoch 6/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.3670 - accuracy: 0.8709 - val_loss: 0.7052 - val_accuracy: 0.7737\n",
            "Epoch 7/20\n",
            "782/782 [==============================] - 16s 21ms/step - loss: 0.2906 - accuracy: 0.8991 - val_loss: 0.7181 - val_accuracy: 0.7906\n",
            "Epoch 8/20\n",
            "782/782 [==============================] - 16s 21ms/step - loss: 0.2338 - accuracy: 0.9175 - val_loss: 0.7227 - val_accuracy: 0.7959\n",
            "Epoch 9/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.1828 - accuracy: 0.9356 - val_loss: 0.9712 - val_accuracy: 0.7626\n",
            "Epoch 10/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.1453 - accuracy: 0.9500 - val_loss: 0.7880 - val_accuracy: 0.8128\n",
            "Epoch 11/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.1305 - accuracy: 0.9555 - val_loss: 0.9005 - val_accuracy: 0.8025\n",
            "Epoch 12/20\n",
            "782/782 [==============================] - 16s 21ms/step - loss: 0.1104 - accuracy: 0.9622 - val_loss: 0.7780 - val_accuracy: 0.8220\n",
            "Epoch 13/20\n",
            "782/782 [==============================] - 19s 24ms/step - loss: 0.0942 - accuracy: 0.9685 - val_loss: 1.1020 - val_accuracy: 0.7892\n",
            "Epoch 14/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0865 - accuracy: 0.9699 - val_loss: 0.9083 - val_accuracy: 0.8019\n",
            "Epoch 15/20\n",
            "782/782 [==============================] - 16s 21ms/step - loss: 0.0825 - accuracy: 0.9736 - val_loss: 1.0783 - val_accuracy: 0.7939\n",
            "Epoch 16/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0745 - accuracy: 0.9756 - val_loss: 0.9811 - val_accuracy: 0.8146\n",
            "Epoch 17/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0705 - accuracy: 0.9767 - val_loss: 0.9914 - val_accuracy: 0.8046\n",
            "Epoch 18/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0625 - accuracy: 0.9791 - val_loss: 0.9059 - val_accuracy: 0.8206\n",
            "Epoch 19/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0645 - accuracy: 0.9796 - val_loss: 1.0124 - val_accuracy: 0.8145\n",
            "Epoch 20/20\n",
            "782/782 [==============================] - 16s 20ms/step - loss: 0.0634 - accuracy: 0.9804 - val_loss: 1.5178 - val_accuracy: 0.7910\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history2 = model2.fit(x_train, y_train, batch_size=64, epochs=Epoch, validation_data=(x_test, y_test))"
      ],
      "metadata": {
        "id": "XK_bLj4cHSt2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7c13365c-1c63-448d-f68b-10e9cb0dc5f1"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "782/782 [==============================] - 5s 6ms/step - loss: 1.4698 - accuracy: 0.4730 - val_loss: 1.1678 - val_accuracy: 0.5869\n",
            "Epoch 2/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 1.0782 - accuracy: 0.6201 - val_loss: 1.0641 - val_accuracy: 0.6366\n",
            "Epoch 3/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.9218 - accuracy: 0.6782 - val_loss: 0.9361 - val_accuracy: 0.6715\n",
            "Epoch 4/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.8226 - accuracy: 0.7128 - val_loss: 0.8843 - val_accuracy: 0.6933\n",
            "Epoch 5/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.7490 - accuracy: 0.7391 - val_loss: 0.9103 - val_accuracy: 0.6860\n",
            "Epoch 6/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.6795 - accuracy: 0.7629 - val_loss: 0.8659 - val_accuracy: 0.7041\n",
            "Epoch 7/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.6137 - accuracy: 0.7876 - val_loss: 0.8860 - val_accuracy: 0.7021\n",
            "Epoch 8/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.5554 - accuracy: 0.8054 - val_loss: 0.8830 - val_accuracy: 0.7136\n",
            "Epoch 9/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.4942 - accuracy: 0.8265 - val_loss: 0.9040 - val_accuracy: 0.7126\n",
            "Epoch 10/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.4370 - accuracy: 0.8477 - val_loss: 0.9654 - val_accuracy: 0.7177\n",
            "Epoch 11/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.3900 - accuracy: 0.8640 - val_loss: 0.9772 - val_accuracy: 0.7077\n",
            "Epoch 12/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.3428 - accuracy: 0.8808 - val_loss: 1.0642 - val_accuracy: 0.7077\n",
            "Epoch 13/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.2949 - accuracy: 0.8967 - val_loss: 1.1505 - val_accuracy: 0.7006\n",
            "Epoch 14/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.2549 - accuracy: 0.9119 - val_loss: 1.2046 - val_accuracy: 0.6999\n",
            "Epoch 15/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.2189 - accuracy: 0.9240 - val_loss: 1.3036 - val_accuracy: 0.6957\n",
            "Epoch 16/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.1862 - accuracy: 0.9355 - val_loss: 1.4072 - val_accuracy: 0.6968\n",
            "Epoch 17/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.1610 - accuracy: 0.9454 - val_loss: 1.5509 - val_accuracy: 0.6859\n",
            "Epoch 18/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.1383 - accuracy: 0.9509 - val_loss: 1.6186 - val_accuracy: 0.6898\n",
            "Epoch 19/20\n",
            "782/782 [==============================] - 4s 6ms/step - loss: 0.1198 - accuracy: 0.9599 - val_loss: 1.7211 - val_accuracy: 0.6901\n",
            "Epoch 20/20\n",
            "782/782 [==============================] - 4s 5ms/step - loss: 0.1139 - accuracy: 0.9613 - val_loss: 1.8169 - val_accuracy: 0.6853\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history3 = model3.fit(x_train, y_train, batch_size=64, epochs=Epoch, validation_data=(x_test, y_test))"
      ],
      "metadata": {
        "id": "I51Wqg6mI6oD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "11733d2e-896a-40f8-8684-beb52115bd7a"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "782/782 [==============================] - 13s 14ms/step - loss: 1.4450 - accuracy: 0.5018 - val_loss: 1.5058 - val_accuracy: 0.5114\n",
            "Epoch 2/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.9197 - accuracy: 0.6767 - val_loss: 1.0724 - val_accuracy: 0.6565\n",
            "Epoch 3/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.7329 - accuracy: 0.7415 - val_loss: 0.8786 - val_accuracy: 0.7160\n",
            "Epoch 4/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.6164 - accuracy: 0.7850 - val_loss: 0.9173 - val_accuracy: 0.7152\n",
            "Epoch 5/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.5267 - accuracy: 0.8143 - val_loss: 1.4030 - val_accuracy: 0.6136\n",
            "Epoch 6/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.4526 - accuracy: 0.8416 - val_loss: 0.7045 - val_accuracy: 0.7787\n",
            "Epoch 7/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.3858 - accuracy: 0.8639 - val_loss: 0.7278 - val_accuracy: 0.7880\n",
            "Epoch 8/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.3248 - accuracy: 0.8849 - val_loss: 0.8552 - val_accuracy: 0.7646\n",
            "Epoch 9/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.2861 - accuracy: 0.8989 - val_loss: 0.8082 - val_accuracy: 0.7786\n",
            "Epoch 10/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.2360 - accuracy: 0.9164 - val_loss: 0.7709 - val_accuracy: 0.7883\n",
            "Epoch 11/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.2129 - accuracy: 0.9243 - val_loss: 0.8479 - val_accuracy: 0.7828\n",
            "Epoch 12/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1858 - accuracy: 0.9344 - val_loss: 0.8782 - val_accuracy: 0.7876\n",
            "Epoch 13/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1743 - accuracy: 0.9398 - val_loss: 0.7704 - val_accuracy: 0.8088\n",
            "Epoch 14/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1587 - accuracy: 0.9446 - val_loss: 0.7981 - val_accuracy: 0.8056\n",
            "Epoch 15/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1492 - accuracy: 0.9471 - val_loss: 0.9305 - val_accuracy: 0.7934\n",
            "Epoch 16/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1370 - accuracy: 0.9526 - val_loss: 0.9659 - val_accuracy: 0.7848\n",
            "Epoch 17/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1271 - accuracy: 0.9548 - val_loss: 1.0496 - val_accuracy: 0.7740\n",
            "Epoch 18/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1172 - accuracy: 0.9593 - val_loss: 0.9105 - val_accuracy: 0.7984\n",
            "Epoch 19/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1180 - accuracy: 0.9585 - val_loss: 0.9651 - val_accuracy: 0.7973\n",
            "Epoch 20/20\n",
            "782/782 [==============================] - 10s 13ms/step - loss: 0.1053 - accuracy: 0.9634 - val_loss: 0.9792 - val_accuracy: 0.7953\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "history4 = model4.fit(x_train, y_train, batch_size=64, epochs=Epoch, validation_data=(x_test, y_test))"
      ],
      "metadata": {
        "id": "0o9LuI7rMr8h",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a5a57241-ad40-499d-ea2f-4f00f591c51e"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "782/782 [==============================] - 14s 17ms/step - loss: 1.3145 - accuracy: 0.5372 - val_loss: 1.2177 - val_accuracy: 0.5675\n",
            "Epoch 2/20\n",
            "782/782 [==============================] - 14s 17ms/step - loss: 1.1368 - accuracy: 0.5987 - val_loss: 1.1816 - val_accuracy: 0.5823\n",
            "Epoch 3/20\n",
            "782/782 [==============================] - 13s 17ms/step - loss: 1.0566 - accuracy: 0.6283 - val_loss: 1.1129 - val_accuracy: 0.6121\n",
            "Epoch 4/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.9908 - accuracy: 0.6498 - val_loss: 1.0947 - val_accuracy: 0.6159\n",
            "Epoch 5/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.9305 - accuracy: 0.6723 - val_loss: 1.1531 - val_accuracy: 0.6054\n",
            "Epoch 6/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.8713 - accuracy: 0.6927 - val_loss: 1.0902 - val_accuracy: 0.6223\n",
            "Epoch 7/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.8142 - accuracy: 0.7117 - val_loss: 1.1375 - val_accuracy: 0.6181\n",
            "Epoch 8/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.7577 - accuracy: 0.7305 - val_loss: 1.1453 - val_accuracy: 0.6234\n",
            "Epoch 9/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.7011 - accuracy: 0.7505 - val_loss: 1.1615 - val_accuracy: 0.6201\n",
            "Epoch 10/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.6461 - accuracy: 0.7713 - val_loss: 1.2320 - val_accuracy: 0.6147\n",
            "Epoch 11/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.5945 - accuracy: 0.7909 - val_loss: 1.1871 - val_accuracy: 0.6266\n",
            "Epoch 12/20\n",
            "782/782 [==============================] - 13s 17ms/step - loss: 0.5411 - accuracy: 0.8096 - val_loss: 1.2170 - val_accuracy: 0.6292\n",
            "Epoch 13/20\n",
            "782/782 [==============================] - 13s 17ms/step - loss: 0.4916 - accuracy: 0.8286 - val_loss: 1.2895 - val_accuracy: 0.6262\n",
            "Epoch 14/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.4464 - accuracy: 0.8468 - val_loss: 1.3280 - val_accuracy: 0.6237\n",
            "Epoch 15/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.4053 - accuracy: 0.8611 - val_loss: 1.3583 - val_accuracy: 0.6235\n",
            "Epoch 16/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.3638 - accuracy: 0.8763 - val_loss: 1.4187 - val_accuracy: 0.6190\n",
            "Epoch 17/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.3260 - accuracy: 0.8895 - val_loss: 1.4462 - val_accuracy: 0.6268\n",
            "Epoch 18/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.2913 - accuracy: 0.9024 - val_loss: 1.5403 - val_accuracy: 0.6206\n",
            "Epoch 19/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.2651 - accuracy: 0.9121 - val_loss: 1.6006 - val_accuracy: 0.6170\n",
            "Epoch 20/20\n",
            "782/782 [==============================] - 13s 16ms/step - loss: 0.2342 - accuracy: 0.9232 - val_loss: 1.6657 - val_accuracy: 0.6196\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Analytics"
      ],
      "metadata": {
        "id": "RuZ2h6sN661G"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer',\n",
        "               'dog', 'frog', 'horse', 'ship', 'truck']"
      ],
      "metadata": {
        "id": "R8uQBB6POBMc"
      },
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test_loss4, test_acc4 = model4.evaluate(x_test, y_test, verbose=0)\n",
        "test_loss3, test_acc3 = model3.evaluate(x_test, y_test, verbose=0)\n",
        "test_loss2, test_acc2 = model2.evaluate(x_test, y_test, verbose=0)\n",
        "test_loss, test_acc = model.evaluate(x_test, y_test, verbose=0)\n",
        "print(\"Testing Loss Model 1:\", test_loss)\n",
        "print(\"Testing Accuracy Model 1:\", test_acc)\n",
        "print(\"Testing Loss Model 2:\", test_loss2)\n",
        "print(\"Testing Accuracy Model 2:\", test_acc2)\n",
        "print(\"Testing Loss Model 3:\", test_loss3)\n",
        "print(\"Testing Accuracy Model 3:\", test_acc3)\n",
        "print(\"Testing Loss Model 4:\", test_loss4)\n",
        "print(\"Testing Accuracy Model 4:\", test_acc4)"
      ],
      "metadata": {
        "id": "wmd5jIo1HVR_",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "34d11888-4ea0-4d71-c310-110d2f84c7fa"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Testing Loss Model 1: 1.5177855491638184\n",
            "Testing Accuracy Model 1: 0.7910000085830688\n",
            "Testing Loss Model 2: 1.8168818950653076\n",
            "Testing Accuracy Model 2: 0.6852999925613403\n",
            "Testing Loss Model 3: 0.9791887402534485\n",
            "Testing Accuracy Model 3: 0.7953000068664551\n",
            "Testing Loss Model 4: 1.6656626462936401\n",
            "Testing Accuracy Model 4: 0.6195999979972839\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.plot(history.history['accuracy'], label=\"model 1\")\n",
        "plt.title('Model accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "# plt.show()\n",
        "\n",
        "model2, = plt.plot(history2.history['accuracy'], label=\"model 2\")\n",
        "plt.title('Model accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "# plt.show()\n",
        "\n",
        "plt.plot(history3.history['accuracy'], label=\"model 3\")\n",
        "plt.title('Model accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "# plt.show()\n",
        "\n",
        "plt.plot(history4.history['accuracy'], label=\"model 4\")\n",
        "plt.title('Model accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "leg = plt.legend(loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "s9ptVgLTodgE",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "a48b1d77-a499-427a-8023-9f76c1f7d95d"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}