CNN_for_image_classification.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "982127ab"
      },
      "source": [
        "# CNN for image classification"
      ],
      "id": "982127ab"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "09b35c83"
      },
      "source": [
        "## Dataset = CIFAR-10"
      ],
      "id": "09b35c83"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ef6eb9b0"
      },
      "source": [
        "In this dataset we have 10 possible classes which are arranges according to their label\n",
        "\n",
        "0.  Airplane \n",
        "1.  Automobile\n",
        "2.  Bird \n",
        "3.  cat\n",
        "4.  Deer\n",
        "5.  Dog\n",
        "6.  Frog\n",
        "7.  Horse\n",
        "8.  Ship \n",
        "9.  Truck"
      ],
      "id": "ef6eb9b0"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "b5e0f445"
      },
      "outputs": [],
      "source": [
        "from keras.datasets import cifar10"
      ],
      "id": "b5e0f445"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "269e95ce",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e690a8b8-058f-483d-b710-77991907b0de"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170500096/170498071 [==============================] - 2s 0us/step\n",
            "170508288/170498071 [==============================] - 2s 0us/step\n"
          ]
        }
      ],
      "source": [
        "dataset = cifar10.load_data()"
      ],
      "id": "269e95ce"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ea822f0f"
      },
      "outputs": [],
      "source": [
        "(X_train, Y_train),(X_test, Y_test) = dataset\n",
        "# Splits the dataset in training and testing"
      ],
      "id": "ea822f0f"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "627147c6"
      },
      "outputs": [],
      "source": [
        "import keras\n",
        "from tensorflow.keras.utils import to_categorical as OneHot \n",
        "Y_train_one_hot = OneHot(Y_train,10)\n",
        "Y_test_one_hot = OneHot(Y_test,10)"
      ],
      "id": "627147c6"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "b89b997c"
      },
      "outputs": [],
      "source": [
        " X_train = X_train.astype('float32')\n",
        " X_test = X_test.astype('float32')\n",
        " X_train = X_train/255\n",
        " X_test = X_test/255"
      ],
      "id": "b89b997c"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "4e306674"
      },
      "outputs": [],
      "source": [
        "import keras\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\n",
        "from keras.layers import BatchNormalization\n",
        "## Importing all the layers to build the CNN"
      ],
      "id": "4e306674"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "8f857b05"
      },
      "outputs": [],
      "source": [
        "model1 = Sequential() "
      ],
      "id": "8f857b05"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "OFvWZNAU5jS0"
      },
      "outputs": [],
      "source": [
        "# # # Model 1\n",
        "\n",
        "model1.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same', input_shape = (32,32,3)))\n",
        "# first layer of the neural network. \n",
        "model1.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same'))\n",
        "# second layer of the neural network in this we dont need to specify input shape as it will take input form the previous layer\n",
        "model1.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model1.add(Dropout(0.25))\n",
        "model1.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model1.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model1.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model1.add(Dropout(0.25))\n",
        "model1.add(Flatten())\n",
        "model1.add(Dense(512, activation = 'relu'))\n",
        "model1.add(Dropout(0.5))\n",
        "model1.add(Dense(10, activation = 'softmax'))\n",
        "model1.compile(loss='categorical_crossentropy', optimizer ='adam', metrics=['accuracy'])\n",
        "hist = model1.fit(X_train, Y_train_one_hot, batch_size = 32, epochs = 25 , validation_split = 0.2, verbose = 1)\n",
        "print(\"Model 1 with validation data for 100 epochs running\")\n",
        "model1.save('model1.h5')\n",
        "print(\"model training complete and saved successfully !\")"
      ],
      "id": "OFvWZNAU5jS0"
    },
    {
      "cell_type": "code",
      "source": [
        "# Model 2\n",
        "import keras\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\n",
        "## Importing all the layers to build the CNN\n",
        "\n",
        "model2 = Sequential() \n",
        "# first layer of the neural network. \n",
        "model2.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same', input_shape = (32,32,3)))\n",
        "# second layer of the neural network in this we dont need to specify input shape as it will  take input form the previous layer\n",
        "model2.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same')) \n",
        "model2.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model2.add(Dropout(0.2))\n",
        "model2.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model2.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model2.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model2.add(Dropout(0.2))\n",
        "model2.add(Conv2D(128,(3,3), activation = 'relu', padding = 'same'))\n",
        "model2.add(Conv2D(128,(3,3), activation = 'relu', padding = 'same'))\n",
        "model2.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model2.add(Dropout(0.2))\n",
        "model2.add(Flatten())\n",
        "model2.add(Dense(128, activation = 'relu'))\n",
        "model2.add(Dropout(0.2))\n",
        "model2.add(Dense(10, activation = 'softmax'))\n",
        "model2.compile(loss='categorical_crossentropy', optimizer = \"sgd\", metrics=['accuracy'])\n",
        "\n",
        "# data Augmentation\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "#ImageDataGenerator from keras is used to modify image\n",
        "datagenerator = ImageDataGenerator(\n",
        "    width_shift_range=0.1,\n",
        "    height_shift_range = 0.1,\n",
        "    horizontal_flip=True,\n",
        "    )\n",
        "\n",
        "# new training set is created with modified images\n",
        "new_train_set = datagenerator.flow(X_train, Y_train_one_hot, batch_size=32)\n",
        "\n",
        "# no.of steps that apply per epoch \n",
        "steps  = int(X_train.shape[0]/32)\n",
        "\n",
        "# new model2 is created with augmented data and fitting it directly on the dataset.\n",
        "augmented_model2 = model2.fit_generator(new_train_set, steps_per_epoch= steps , epochs=50, \n",
        "                                      validation_data=(X_test, Y_test_one_hot), verbose = 1)\n",
        "\n",
        "model2.save('model2.h5')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "l9gK-zU_PD7B",
        "outputId": "1889a6d7-affa-49f8-b3b4-05070de7b44f"
      },
      "id": "l9gK-zU_PD7B",
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:46: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "1562/1562 [==============================] - 64s 33ms/step - loss: 2.1443 - accuracy: 0.2004 - val_loss: 1.9901 - val_accuracy: 0.2788\n",
            "Epoch 2/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.9250 - accuracy: 0.2989 - val_loss: 1.7287 - val_accuracy: 0.3804\n",
            "Epoch 3/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.7367 - accuracy: 0.3600 - val_loss: 1.5861 - val_accuracy: 0.4164\n",
            "Epoch 4/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 1.6072 - accuracy: 0.4099 - val_loss: 1.4087 - val_accuracy: 0.4846\n",
            "Epoch 5/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.5144 - accuracy: 0.4489 - val_loss: 1.3383 - val_accuracy: 0.5137\n",
            "Epoch 6/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.4327 - accuracy: 0.4821 - val_loss: 1.2763 - val_accuracy: 0.5372\n",
            "Epoch 7/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.3585 - accuracy: 0.5106 - val_loss: 1.2074 - val_accuracy: 0.5710\n",
            "Epoch 8/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 1.2911 - accuracy: 0.5395 - val_loss: 1.1796 - val_accuracy: 0.5768\n",
            "Epoch 9/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.2333 - accuracy: 0.5587 - val_loss: 1.0702 - val_accuracy: 0.6168\n",
            "Epoch 10/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.1794 - accuracy: 0.5773 - val_loss: 1.0528 - val_accuracy: 0.6201\n",
            "Epoch 11/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.1312 - accuracy: 0.5955 - val_loss: 0.9958 - val_accuracy: 0.6447\n",
            "Epoch 12/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 1.0900 - accuracy: 0.6124 - val_loss: 0.9387 - val_accuracy: 0.6661\n",
            "Epoch 13/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 1.0500 - accuracy: 0.6284 - val_loss: 0.9606 - val_accuracy: 0.6614\n",
            "Epoch 14/50\n",
            "1562/1562 [==============================] - 51s 32ms/step - loss: 1.0245 - accuracy: 0.6355 - val_loss: 0.9131 - val_accuracy: 0.6818\n",
            "Epoch 15/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.9904 - accuracy: 0.6496 - val_loss: 0.8836 - val_accuracy: 0.6849\n",
            "Epoch 16/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.9607 - accuracy: 0.6586 - val_loss: 0.8979 - val_accuracy: 0.6780\n",
            "Epoch 17/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.9286 - accuracy: 0.6710 - val_loss: 0.7924 - val_accuracy: 0.7205\n",
            "Epoch 18/50\n",
            "1562/1562 [==============================] - 51s 32ms/step - loss: 0.9111 - accuracy: 0.6786 - val_loss: 0.8649 - val_accuracy: 0.6942\n",
            "Epoch 19/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.8847 - accuracy: 0.6897 - val_loss: 0.7598 - val_accuracy: 0.7296\n",
            "Epoch 20/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.8629 - accuracy: 0.6964 - val_loss: 0.7631 - val_accuracy: 0.7297\n",
            "Epoch 21/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.8408 - accuracy: 0.7029 - val_loss: 0.7584 - val_accuracy: 0.7313\n",
            "Epoch 22/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.8222 - accuracy: 0.7117 - val_loss: 0.7320 - val_accuracy: 0.7403\n",
            "Epoch 23/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.8062 - accuracy: 0.7171 - val_loss: 0.7286 - val_accuracy: 0.7481\n",
            "Epoch 24/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.7894 - accuracy: 0.7201 - val_loss: 0.7249 - val_accuracy: 0.7482\n",
            "Epoch 25/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.7697 - accuracy: 0.7318 - val_loss: 0.7239 - val_accuracy: 0.7473\n",
            "Epoch 26/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.7636 - accuracy: 0.7313 - val_loss: 0.6517 - val_accuracy: 0.7727\n",
            "Epoch 27/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.7459 - accuracy: 0.7409 - val_loss: 0.6601 - val_accuracy: 0.7682\n",
            "Epoch 28/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.7285 - accuracy: 0.7450 - val_loss: 0.6501 - val_accuracy: 0.7734\n",
            "Epoch 29/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.7173 - accuracy: 0.7463 - val_loss: 0.6395 - val_accuracy: 0.7753\n",
            "Epoch 30/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.7044 - accuracy: 0.7550 - val_loss: 0.6566 - val_accuracy: 0.7721\n",
            "Epoch 31/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6941 - accuracy: 0.7571 - val_loss: 0.6008 - val_accuracy: 0.7923\n",
            "Epoch 32/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6848 - accuracy: 0.7611 - val_loss: 0.6394 - val_accuracy: 0.7865\n",
            "Epoch 33/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6788 - accuracy: 0.7639 - val_loss: 0.6711 - val_accuracy: 0.7674\n",
            "Epoch 34/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6626 - accuracy: 0.7690 - val_loss: 0.6233 - val_accuracy: 0.7864\n",
            "Epoch 35/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6548 - accuracy: 0.7718 - val_loss: 0.6163 - val_accuracy: 0.7846\n",
            "Epoch 36/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6461 - accuracy: 0.7766 - val_loss: 0.5850 - val_accuracy: 0.7960\n",
            "Epoch 37/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.6382 - accuracy: 0.7770 - val_loss: 0.6297 - val_accuracy: 0.7864\n",
            "Epoch 38/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.6260 - accuracy: 0.7819 - val_loss: 0.5743 - val_accuracy: 0.8051\n",
            "Epoch 39/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.6226 - accuracy: 0.7831 - val_loss: 0.5745 - val_accuracy: 0.8015\n",
            "Epoch 40/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6116 - accuracy: 0.7872 - val_loss: 0.5722 - val_accuracy: 0.8034\n",
            "Epoch 41/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.6084 - accuracy: 0.7880 - val_loss: 0.5675 - val_accuracy: 0.8057\n",
            "Epoch 42/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.5989 - accuracy: 0.7912 - val_loss: 0.5513 - val_accuracy: 0.8087\n",
            "Epoch 43/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5858 - accuracy: 0.7950 - val_loss: 0.5795 - val_accuracy: 0.8048\n",
            "Epoch 44/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5844 - accuracy: 0.7946 - val_loss: 0.5411 - val_accuracy: 0.8140\n",
            "Epoch 45/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5734 - accuracy: 0.8014 - val_loss: 0.5531 - val_accuracy: 0.8129\n",
            "Epoch 46/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5724 - accuracy: 0.8005 - val_loss: 0.5322 - val_accuracy: 0.8202\n",
            "Epoch 47/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5636 - accuracy: 0.8050 - val_loss: 0.5449 - val_accuracy: 0.8171\n",
            "Epoch 48/50\n",
            "1562/1562 [==============================] - 51s 33ms/step - loss: 0.5593 - accuracy: 0.8060 - val_loss: 0.5805 - val_accuracy: 0.8031\n",
            "Epoch 49/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.5506 - accuracy: 0.8073 - val_loss: 0.5207 - val_accuracy: 0.8222\n",
            "Epoch 50/50\n",
            "1562/1562 [==============================] - 52s 33ms/step - loss: 0.5418 - accuracy: 0.8090 - val_loss: 0.5298 - val_accuracy: 0.8225\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ef7746c3"
      },
      "outputs": [],
      "source": [
        "\n",
        "# Model 3\n",
        "\n",
        "model3 = Sequential()\n",
        "# first layer of the neural network. \n",
        "model3.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same', input_shape = (32,32,3)))\n",
        "model3.add(BatchNormalization())\n",
        "# second layer of the neural network in this we dont need to specify input shape as it will take input form the previous layer\n",
        "model3.add(Conv2D(32,(3,3), activation = 'relu', padding = 'same')) \n",
        "model3.add(BatchNormalization())\n",
        "model3.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model3.add(Dropout(0.2))\n",
        "model3.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Conv2D(64,(3,3), activation = 'relu', padding = 'same'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model3.add(Dropout(0.2))\n",
        "model3.add(Conv2D(128,(3,3), activation = 'relu', padding = 'same'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(Conv2D(128,(3,3), activation = 'relu', padding = 'same'))\n",
        "model3.add(BatchNormalization())\n",
        "model3.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model3.add(Dropout(0.2))\n",
        "model3.add(Flatten())\n",
        "model3.add(Dense(128, activation = 'relu'))\n",
        "model3.add(Dropout(0.2))\n",
        "model3.add(Dense(10, activation = 'softmax'))\n",
        "model3.compile(loss='categorical_crossentropy', optimizer = \"sgd\", metrics=['accuracy'])\n",
        "\n",
        "# data Augmentation\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "#ImageDataGenerator is used to modify image\n",
        "datagen = ImageDataGenerator(\n",
        "    width_shift_range=0.1,\n",
        "    height_shift_range = 0.1,\n",
        "    horizontal_flip=True,\n",
        "    )\n",
        "\n",
        "# new training set is created with modified images\n",
        "new_train_set = datagen.flow(X_train, Y_train_one_hot, batch_size=32)\n",
        "\n",
        "# no.of steps that apply per epoch \n",
        "steps  = int(X_train.shape[0]/32)\n",
        "\n",
        "\n",
        "# new model is created with augmented data\n",
        "augmented_model = model3.fit_generator(new_train_set, steps_per_epoch= steps , epochs=50, \n",
        "                                      validation_data=(X_test, Y_test_one_hot), verbose = 1)\n"
      ],
      "id": "ef7746c3"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3d031509"
      },
      "outputs": [],
      "source": [
        "# compile model\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "opt = SGD(lr=0.001, momentum=0.9)\n",
        "model3.compile(loss='categorical_crossentropy', optimizer = \"adam\", metrics=['accuracy'])"
      ],
      "id": "3d031509"
    },
    {
      "cell_type": "markdown",
      "source": [
        ""
      ],
      "metadata": {
        "id": "YcvqKpJR57cA"
      },
      "id": "YcvqKpJR57cA"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Y40ddCjHBqws"
      },
      "outputs": [],
      "source": [
        "hist = model3.fit(X_train, Y_train_one_hot, batch_size = 32, epochs = 25 , validation_split=0.2, verbose = 1)\n",
        "print(\"Model 2 with validation data for 100 epochs running\")\n",
        "# with validation split we do not need to split our dataset into training set and validation set in the begining\n",
        "# In this case 20% of the datset will be used as the validation set\n",
        "\n",
        "hist = model3.fit(X_train, Y_train_one_hot, batch_size = 32, epochs = 25 , \n",
        "                 validation_data = (X_test,Y_test_one_hot), verbose = 1)"
      ],
      "id": "Y40ddCjHBqws"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vCs2jeFZZvzD"
      },
      "outputs": [],
      "source": [
        "X_test.shape"
      ],
      "id": "vCs2jeFZZvzD"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h1bKZJxiB-a-",
        "outputId": "e2e98bd9-ee12-4476-cf75-f48ca8f0678d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 3s 9ms/step - loss: 0.4597 - accuracy: 0.8501\n"
          ]
        }
      ],
      "source": [
        "acc = model3.evaluate(X_test, Y_test_one_hot)"
      ],
      "id": "h1bKZJxiB-a-"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "f772752c",
        "outputId": "1223553e-bebe-4739-bae7-27c07091d5c7"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.plot(augmented_model.history['loss'])\n",
        "plt.plot(augmented_model.history['val_loss'])\n",
        "plt.title('Model loss')\n",
        "plt.ylabel('Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.legend(['Train', 'Val'], loc='upper right')\n",
        "plt.show()"
      ],
      "id": "f772752c"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "690561a4",
        "outputId": "cc70f5bd-9049-4f62-894e-4b598b64a863"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "plt.plot(augmented_model.history['accuracy'])\n",
        "plt.plot(augmented_model.history['val_accuracy'])\n",
        "plt.title('Model accuracy')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.legend(['Train', 'Val'], loc='lower right')\n",
        "plt.show()\n"
      ],
      "id": "690561a4"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "smG2WmlBuYoj"
      },
      "outputs": [],
      "source": [
        "my_img = plt.imread(\"cat-frog.jpg\")"
      ],
      "id": "smG2WmlBuYoj"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "TSY9m859v2L9"
      },
      "outputs": [],
      "source": [
        "from skimage.transform import resize\n",
        "my_img_resized = resize(my_img,(32,32,3))"
      ],
      "id": "TSY9m859v2L9"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ICJ-wj0Vv9M2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 266
        },
        "outputId": "ff26f912-968e-4d2a-e8bb-5702913aedc5"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "img = plt.imshow(my_img_resized)"
      ],
      "id": "ICJ-wj0Vv9M2"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "7UsXzj5cFj8V"
      },
      "outputs": [],
      "source": [
        "import numpy as np \n",
        "probabilities = model.predict(np.array([my_img_resized,]))"
      ],
      "id": "7UsXzj5cFj8V"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "2XmYFWv_FpZy",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "05cb246d-2bb4-4f7d-ceb8-36f273be56a9"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[1.2858708e-05, 1.3957443e-04, 1.2045423e-02, 2.2246581e-02,\n",
              "        4.1112150e-03, 9.5946050e-01, 7.7225978e-04, 1.1991545e-03,\n",
              "        4.1149195e-07, 1.2072728e-05]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 21
        }
      ],
      "source": [
        "probabilities"
      ],
      "id": "2XmYFWv_FpZy"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "gaNrC8XzFqQE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "22309f8f-fb02-4aa3-e7f0-06f305f703eb"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Most likely class: dog -- Possibility: 95.94604969024658 %\n",
            "Second most likely class: cat -- Possibility: 2.2246580570936203 %\n",
            "Third most likely class: bird -- Possibility: 1.2045422568917274 %\n",
            "Fourth most likely class: deer -- Possibility: 0.4111215006560087 %\n",
            "Fifth most likely class: horse -- Possibility: 0.11991545325145125 %\n"
          ]
        }
      ],
      "source": [
        "number_to_class = ['airplane','automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\n",
        "index = np.argsort(probabilities[0,:])\n",
        "print(\"Most likely class:\", number_to_class[index[9]], \"-- Possibility:\", probabilities[0,index[9]]*100, \"%\")\n",
        "print(\"Second most likely class:\", number_to_class[index[8]], \"-- Possibility:\", probabilities[0,index[8]]*100, \"%\")\n",
        "print(\"Third most likely class:\", number_to_class[index[7]], \"-- Possibility:\", probabilities[0,index[7]]*100, \"%\")\n",
        "print(\"Fourth most likely class:\", number_to_class[index[6]], \"-- Possibility:\", probabilities[0,index[6]]*100, \"%\")\n",
        "print(\"Fifth most likely class:\", number_to_class[index[5]], \"-- Possibility:\", probabilities[0,index[5]]*100, \"%\")"
      ],
      "id": "gaNrC8XzFqQE"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Upmh1jUsFtsS"
      },
      "outputs": [],
      "source": [
        ""
      ],
      "id": "Upmh1jUsFtsS"
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "CNN for image classification.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": "3.7.11"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}