Cat_And_Dog_classifier.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/sainisatish/DeepLearningFromScratch/blob/main/Cat_And_Dog_classifier.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "v6WCy5o10UD0"
      },
      "outputs": [],
      "source": [
        "from google.colab import files"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wY4uzgFD1_N6"
      },
      "outputs": [],
      "source": [
        "files.upload()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Y1OY37Vw2CmZ"
      },
      "outputs": [],
      "source": [
        "!mkdir ~/.kaggle\n",
        "!cp kaggle.json ~/.kaggle/\n",
        "!chmod 600 ~/.kaggle/kaggle.json"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pfSvnDVW2QjP",
        "outputId": "98eb258d-9650-4584-d350-ff547f1244d1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading dogs-vs-cats.zip to /content\n",
            " 99% 802M/812M [00:04<00:00, 209MB/s]\n",
            "100% 812M/812M [00:04<00:00, 205MB/s]\n"
          ]
        }
      ],
      "source": [
        "!kaggle competitions download -c dogs-vs-cats"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "s-GP942i2QVs"
      },
      "outputs": [],
      "source": [
        "!unzip -qq dogs-vs-cats.zip"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ejhoTSpz2dzn"
      },
      "outputs": [],
      "source": [
        "!unzip -qq train.zip"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Preparing Dataset\n"
      ],
      "metadata": {
        "id": "AlEt3DznQb3u"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "PTNNfy1p2qfB"
      },
      "outputs": [],
      "source": [
        "import os,shutil,pathlib"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wmgO2jhi384k"
      },
      "outputs": [],
      "source": [
        "original_dir = pathlib.Path(\"train\")\n",
        "new_base_dir = pathlib.Path(\"cats_vs_dogs_small\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "BYBrOlP-_xde"
      },
      "outputs": [],
      "source": [
        "def make_subset(subset_name, start_index, end_index): \n",
        "    for category in (\"cat\", \"dog\"):\n",
        "        dir = new_base_dir / subset_name / category\n",
        "        os.makedirs(dir)\n",
        "        fnames = [f\"{category}.{i}.jpg\" \n",
        "                  for i in range(start_index, end_index)]\n",
        "        for fname in fnames:\n",
        "            shutil.copyfile(src=original_dir / fname,\n",
        "                            dst=dir / fname)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "WWxlw-MpAKLA"
      },
      "outputs": [],
      "source": [
        "make_subset(\"train\", start_index=0, end_index=1000)\n",
        "make_subset(\"validation\", start_index=1000, end_index=1500) \n",
        "make_subset(\"test\", start_index=1500, end_index=2500)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_ma2ABCrHMbS"
      },
      "outputs": [],
      "source": [
        "from tensorflow.keras.utils import image_dataset_from_directory "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "q2J8ZMjzHbdJ",
        "outputId": "2e8127b1-824d-4d07-f3a4-48fe0d646cd9"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 2000 files belonging to 2 classes.\n",
            "Found 1000 files belonging to 2 classes.\n",
            "Found 2000 files belonging to 2 classes.\n"
          ]
        }
      ],
      "source": [
        "train_dataset=image_dataset_from_directory(new_base_dir/\"train\",\n",
        "                                           image_size=(180,180),\n",
        "                                           batch_size=32)\n",
        "\n",
        "validation_dataset = image_dataset_from_directory(new_base_dir /\"validation\",\n",
        "                                                  image_size=(180,180),\n",
        "                                                  batch_size=32)\n",
        "\n",
        "test_size = image_dataset_from_directory(new_base_dir /\"test\",\n",
        "                                         image_size=(180,180),\n",
        "                                         batch_size=32)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KH2wI9kpIhUO",
        "outputId": "d81b3d3f-8381-4535-9c3b-16b3f4b4bf2e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "data batch shape (32, 180, 180, 3)\n",
            "label batch shape (32,)\n"
          ]
        }
      ],
      "source": [
        "for data_batch,label_batch in train_dataset:\n",
        "    print(f\"data batch shape {data_batch.shape}\")\n",
        "    print(f\"label batch shape { label_batch.shape}\")\n",
        "    break"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Creating Base model from Scratch"
      ],
      "metadata": {
        "id": "bJgUv-_OQgtA"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "CiFz33z2AV4P"
      },
      "outputs": [],
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "XGYZ4kr9Be9q"
      },
      "outputs": [],
      "source": [
        "inputs = keras.Input(shape=(180,180,3))\n",
        "x=layers.Rescaling(1./255)(inputs)\n",
        "x=layers.Conv2D(filters=32,kernel_size=3,activation='relu')(x)\n",
        "x=layers.MaxPooling2D(pool_size=2)(x)\n",
        "x=layers.Conv2D(filters=64,kernel_size=3,activation='relu')(x)\n",
        "x=layers.MaxPooling2D(pool_size=2)(x)\n",
        "x=layers.Conv2D(filters=128,kernel_size=3,activation='relu')(x)\n",
        "x=layers.MaxPooling2D(pool_size=2)(x)\n",
        "x=layers.Conv2D(filters=256,kernel_size=3,activation='relu')(x)\n",
        "x=layers.MaxPooling2D(pool_size=2)(x)\n",
        "x=layers.Conv2D(filters=256,kernel_size=3,activation='relu')(x)\n",
        "x=layers.Flatten()(x)\n",
        "outputs=layers.Dense(1,activation='sigmoid')(x)\n",
        "model = keras.Model(inputs=inputs,outputs=outputs)\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MO--_VAnDjd8",
        "outputId": "1e0f70e6-2c9c-4c2d-ccd0-7f1e387885dc"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_1 (InputLayer)        [(None, 180, 180, 3)]     0         \n",
            "                                                                 \n",
            " rescaling (Rescaling)       (None, 180, 180, 3)       0         \n",
            "                                                                 \n",
            " conv2d (Conv2D)             (None, 178, 178, 32)      896       \n",
            "                                                                 \n",
            " max_pooling2d (MaxPooling2D  (None, 89, 89, 32)       0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 87, 87, 64)        18496     \n",
            "                                                                 \n",
            " max_pooling2d_1 (MaxPooling  (None, 43, 43, 64)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_2 (Conv2D)           (None, 41, 41, 128)       73856     \n",
            "                                                                 \n",
            " max_pooling2d_2 (MaxPooling  (None, 20, 20, 128)      0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 18, 18, 256)       295168    \n",
            "                                                                 \n",
            " max_pooling2d_3 (MaxPooling  (None, 9, 9, 256)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_4 (Conv2D)           (None, 7, 7, 256)         590080    \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 12544)             0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 1)                 12545     \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 991,041\n",
            "Trainable params: 991,041\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Jdh1vBHCD1ru"
      },
      "outputs": [],
      "source": [
        "model.compile(optimizer='rmsprop',\n",
        "              loss='binary_crossentropy',\n",
        "              metrics='accuracy')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "dvnIRiVj-yKl"
      },
      "outputs": [],
      "source": [
        "callbacks = [keras.callbacks.ModelCheckpoint(filepath=\"convnet_from_scratch.keras\",save_best_only=True,monitor=\"val_loss\")]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IJoP8run-3SD",
        "outputId": "c4b1b7cc-c280-473c-9295-6139613a3719"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "63/63 [==============================] - 15s 73ms/step - loss: 0.6978 - accuracy: 0.5120 - val_loss: 0.6917 - val_accuracy: 0.5000\n",
            "Epoch 2/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.6937 - accuracy: 0.5370 - val_loss: 0.6956 - val_accuracy: 0.5070\n",
            "Epoch 3/30\n",
            "63/63 [==============================] - 5s 75ms/step - loss: 0.6744 - accuracy: 0.6025 - val_loss: 0.6554 - val_accuracy: 0.6310\n",
            "Epoch 4/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.6519 - accuracy: 0.6435 - val_loss: 0.6247 - val_accuracy: 0.6510\n",
            "Epoch 5/30\n",
            "63/63 [==============================] - 6s 85ms/step - loss: 0.6288 - accuracy: 0.6535 - val_loss: 0.6228 - val_accuracy: 0.6460\n",
            "Epoch 6/30\n",
            "63/63 [==============================] - 4s 63ms/step - loss: 0.5947 - accuracy: 0.6805 - val_loss: 0.5969 - val_accuracy: 0.6820\n",
            "Epoch 7/30\n",
            "63/63 [==============================] - 5s 79ms/step - loss: 0.5757 - accuracy: 0.6945 - val_loss: 0.6570 - val_accuracy: 0.6540\n",
            "Epoch 8/30\n",
            "63/63 [==============================] - 5s 72ms/step - loss: 0.5241 - accuracy: 0.7325 - val_loss: 0.5953 - val_accuracy: 0.6870\n",
            "Epoch 9/30\n",
            "63/63 [==============================] - 5s 64ms/step - loss: 0.5529 - accuracy: 0.7575 - val_loss: 0.5359 - val_accuracy: 0.7420\n",
            "Epoch 10/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.4332 - accuracy: 0.8070 - val_loss: 0.6178 - val_accuracy: 0.7200\n",
            "Epoch 11/30\n",
            "63/63 [==============================] - 5s 75ms/step - loss: 0.4113 - accuracy: 0.8195 - val_loss: 0.6234 - val_accuracy: 0.6950\n",
            "Epoch 12/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.3568 - accuracy: 0.8405 - val_loss: 0.7227 - val_accuracy: 0.6990\n",
            "Epoch 13/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.2931 - accuracy: 0.8785 - val_loss: 0.8533 - val_accuracy: 0.6940\n",
            "Epoch 14/30\n",
            "63/63 [==============================] - 5s 79ms/step - loss: 0.2455 - accuracy: 0.8975 - val_loss: 0.9554 - val_accuracy: 0.7100\n",
            "Epoch 15/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.1833 - accuracy: 0.9240 - val_loss: 0.8342 - val_accuracy: 0.7360\n",
            "Epoch 16/30\n",
            "63/63 [==============================] - 5s 70ms/step - loss: 0.1444 - accuracy: 0.9395 - val_loss: 0.8707 - val_accuracy: 0.7370\n",
            "Epoch 17/30\n",
            "63/63 [==============================] - 5s 63ms/step - loss: 0.1211 - accuracy: 0.9590 - val_loss: 0.9328 - val_accuracy: 0.7410\n",
            "Epoch 18/30\n",
            "63/63 [==============================] - 4s 60ms/step - loss: 0.0874 - accuracy: 0.9665 - val_loss: 1.1053 - val_accuracy: 0.7460\n",
            "Epoch 19/30\n",
            "63/63 [==============================] - 5s 76ms/step - loss: 0.0985 - accuracy: 0.9645 - val_loss: 1.1478 - val_accuracy: 0.7260\n",
            "Epoch 20/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.0743 - accuracy: 0.9760 - val_loss: 1.2079 - val_accuracy: 0.7300\n",
            "Epoch 21/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.0666 - accuracy: 0.9790 - val_loss: 1.2162 - val_accuracy: 0.7410\n",
            "Epoch 22/30\n",
            "63/63 [==============================] - 5s 78ms/step - loss: 0.0635 - accuracy: 0.9770 - val_loss: 1.3987 - val_accuracy: 0.7360\n",
            "Epoch 23/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.0353 - accuracy: 0.9900 - val_loss: 1.6204 - val_accuracy: 0.7290\n",
            "Epoch 24/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.0547 - accuracy: 0.9835 - val_loss: 1.6473 - val_accuracy: 0.7310\n",
            "Epoch 25/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.0333 - accuracy: 0.9900 - val_loss: 1.7759 - val_accuracy: 0.7340\n",
            "Epoch 26/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.0677 - accuracy: 0.9790 - val_loss: 1.6256 - val_accuracy: 0.7310\n",
            "Epoch 27/30\n",
            "63/63 [==============================] - 5s 77ms/step - loss: 0.0297 - accuracy: 0.9910 - val_loss: 2.4296 - val_accuracy: 0.7060\n",
            "Epoch 28/30\n",
            "63/63 [==============================] - 4s 62ms/step - loss: 0.0568 - accuracy: 0.9835 - val_loss: 1.7396 - val_accuracy: 0.7060\n",
            "Epoch 29/30\n",
            "63/63 [==============================] - 4s 61ms/step - loss: 0.0626 - accuracy: 0.9800 - val_loss: 2.3901 - val_accuracy: 0.6830\n",
            "Epoch 30/30\n",
            "63/63 [==============================] - 5s 77ms/step - loss: 0.0293 - accuracy: 0.9895 - val_loss: 2.7800 - val_accuracy: 0.7050\n"
          ]
        }
      ],
      "source": [
        "history = model.fit(train_dataset,epochs=30,validation_data=validation_dataset,callbacks=callbacks)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tlgYKvuVF2b6",
        "outputId": "29ad2205-b004-4f62-91ee-4ebdb49311a3"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
          ]
        }
      ],
      "source": [
        "print(history.history.keys())"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aTblE9GCC8Au"
      },
      "outputs": [],
      "source": [
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Zie1CIucDLhZ"
      },
      "outputs": [],
      "source": [
        "accuracy,val_accuracy,loss,val_loss = history.history[\"accuracy\"],history.history[\"val_accuracy\"],history.history[\"loss\"],history.history[\"val_loss\"]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "vSerenHcDscn",
        "outputId": "57c3d667-8efd-447f-da40-8f752ea62697"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "epochs = range(1,len(accuracy)+1)\n",
        "plt.plot(epochs, accuracy, \"bo\", label=\"Training accuracy\")\n",
        "plt.plot(epochs, val_accuracy, \"b\", label=\"Validation accuracy\")\n",
        "plt.title(\"Training and validation accuracy\")\n",
        "plt.legend()\n",
        "plt.figure()\n",
        "plt.plot(epochs, loss, \"bo\", label=\"Training loss\")\n",
        "plt.plot(epochs, val_loss, \"b\", label=\"Validation loss\")\n",
        "plt.title(\"Training and validation loss\")\n",
        "plt.legend()\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3uVwh7iKD_Fr",
        "outputId": "fc0c9b2d-5563-4cb4-98ee-7aa0dfe6608a"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "63/63 [==============================] - 4s 43ms/step - loss: 2.9484 - accuracy: 0.6895\n"
          ]
        }
      ],
      "source": [
        "test_loss,test_acc = model.evaluate(test_size)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Performing data augmentation  to increase dataset"
      ],
      "metadata": {
        "id": "E39skB2dQp1g"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "F2oWmkDBEyHP"
      },
      "outputs": [],
      "source": [
        "data_augmentation = keras.Sequential(\n",
        "    [\n",
        "        layers.RandomFlip(\"horizontal\"),\n",
        "        layers.RandomRotation(0.1),\n",
        "        layers.RandomZoom(0.2),\n",
        "    ]\n",
        "    )"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1LVCtGcuIvaG",
        "outputId": "3c62ef0f-cc9a-4ec7-e9f3-45f00ff67eef"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n"
          ]
        }
      ],
      "source": [
        "inputs = keras.Input(shape=(180,180,3))\n",
        "x= data_augmentation(inputs)\n",
        "x= layers.Rescaling(1./255)(x)\n",
        "x= layers.Conv2D(filters=32,kernel_size=3,activation='relu')(x)\n",
        "x= layers.MaxPooling2D(pool_size=2)(x)\n",
        "x= layers.Conv2D(filters=64,kernel_size=3,activation='relu')(x)\n",
        "x= layers.MaxPooling2D(pool_size=2)(x)\n",
        "x= layers.Conv2D(filters=128,kernel_size=3,activation='relu')(x)\n",
        "x= layers.MaxPooling2D(pool_size=2)(x)\n",
        "x= layers.Conv2D(filters=256,kernel_size=3,activation='relu')(x)\n",
        "x= layers.MaxPooling2D(pool_size=2)(x)\n",
        "x= layers.Conv2D(filters=256,kernel_size=3,activation='relu')(x)\n",
        "x= layers.Flatten()(x)\n",
        "outputs= layers.Dense(1,activation='sigmoid')(x)\n",
        "model_aug = keras.Model(inputs=inputs,outputs=outputs)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "6ay8sMZ5IvGG"
      },
      "outputs": [],
      "source": [
        "model_aug.compile(loss=\"binary_crossentropy\",\n",
        "              optimizer=\"rmsprop\",\n",
        "              metrics=[\"accuracy\"])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "srrYECBCIu7Q"
      },
      "outputs": [],
      "source": [
        " callbacks = [(keras.callbacks.ModelCheckpoint(filepath=\"augment_model.keras\",save_best_only=True,monitor=\"val_loss\"))]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-dQCND-zIuwW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "854e0277-7ef4-4f99-8853-e985464f1479"
      },
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n"
          ]
        },
        {
          "metadata": {
            "tags": null
          },
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "63/63 [==============================] - 27s 234ms/step - loss: 0.6970 - accuracy: 0.4970 - val_loss: 0.6923 - val_accuracy: 0.5020\n",
            "Epoch 2/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.6927 - accuracy: 0.5195 - val_loss: 0.6867 - val_accuracy: 0.5880\n",
            "Epoch 3/100\n",
            "63/63 [==============================] - 14s 215ms/step - loss: 0.6850 - accuracy: 0.5715 - val_loss: 0.6612 - val_accuracy: 0.5990\n",
            "Epoch 4/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.6635 - accuracy: 0.6215 - val_loss: 0.6706 - val_accuracy: 0.6270\n",
            "Epoch 5/100\n",
            "63/63 [==============================] - 15s 229ms/step - loss: 0.6219 - accuracy: 0.6595 - val_loss: 0.6228 - val_accuracy: 0.6570\n",
            "Epoch 6/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.6108 - accuracy: 0.6830 - val_loss: 0.6341 - val_accuracy: 0.6430\n",
            "Epoch 7/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.5915 - accuracy: 0.6800 - val_loss: 0.6330 - val_accuracy: 0.6750\n",
            "Epoch 8/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.5912 - accuracy: 0.7010 - val_loss: 0.5789 - val_accuracy: 0.6880\n",
            "Epoch 9/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.5699 - accuracy: 0.6980 - val_loss: 0.5636 - val_accuracy: 0.7140\n",
            "Epoch 10/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.5737 - accuracy: 0.7085 - val_loss: 0.5716 - val_accuracy: 0.7080\n",
            "Epoch 11/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.5410 - accuracy: 0.7205 - val_loss: 0.5379 - val_accuracy: 0.7340\n",
            "Epoch 12/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.5224 - accuracy: 0.7465 - val_loss: 0.5508 - val_accuracy: 0.7410\n",
            "Epoch 13/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.5175 - accuracy: 0.7475 - val_loss: 0.5275 - val_accuracy: 0.7330\n",
            "Epoch 14/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.5075 - accuracy: 0.7480 - val_loss: 0.9244 - val_accuracy: 0.6610\n",
            "Epoch 15/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.5002 - accuracy: 0.7600 - val_loss: 0.5309 - val_accuracy: 0.7450\n",
            "Epoch 16/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.4842 - accuracy: 0.7680 - val_loss: 0.6949 - val_accuracy: 0.6850\n",
            "Epoch 17/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.4865 - accuracy: 0.7685 - val_loss: 0.5183 - val_accuracy: 0.7450\n",
            "Epoch 18/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.4480 - accuracy: 0.7800 - val_loss: 0.5630 - val_accuracy: 0.7580\n",
            "Epoch 19/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.4484 - accuracy: 0.7885 - val_loss: 0.4779 - val_accuracy: 0.7750\n",
            "Epoch 20/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.4420 - accuracy: 0.7885 - val_loss: 0.5879 - val_accuracy: 0.7230\n",
            "Epoch 21/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.4151 - accuracy: 0.8100 - val_loss: 0.4890 - val_accuracy: 0.7820\n",
            "Epoch 22/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.4090 - accuracy: 0.8100 - val_loss: 0.5286 - val_accuracy: 0.7580\n",
            "Epoch 23/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.3991 - accuracy: 0.8210 - val_loss: 0.5062 - val_accuracy: 0.7830\n",
            "Epoch 24/100\n",
            "63/63 [==============================] - 13s 210ms/step - loss: 0.3888 - accuracy: 0.8295 - val_loss: 0.6364 - val_accuracy: 0.7440\n",
            "Epoch 25/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.3732 - accuracy: 0.8365 - val_loss: 0.5859 - val_accuracy: 0.7540\n",
            "Epoch 26/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.3475 - accuracy: 0.8475 - val_loss: 0.6952 - val_accuracy: 0.7400\n",
            "Epoch 27/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.3572 - accuracy: 0.8435 - val_loss: 0.6586 - val_accuracy: 0.7560\n",
            "Epoch 28/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.3320 - accuracy: 0.8565 - val_loss: 0.6201 - val_accuracy: 0.7250\n",
            "Epoch 29/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.3482 - accuracy: 0.8480 - val_loss: 0.6160 - val_accuracy: 0.7670\n",
            "Epoch 30/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.3210 - accuracy: 0.8635 - val_loss: 0.4737 - val_accuracy: 0.8060\n",
            "Epoch 31/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.2882 - accuracy: 0.8750 - val_loss: 0.6870 - val_accuracy: 0.7760\n",
            "Epoch 32/100\n",
            "63/63 [==============================] - 15s 232ms/step - loss: 0.3017 - accuracy: 0.8735 - val_loss: 0.5451 - val_accuracy: 0.8030\n",
            "Epoch 33/100\n",
            "63/63 [==============================] - 13s 209ms/step - loss: 0.2833 - accuracy: 0.8890 - val_loss: 0.5480 - val_accuracy: 0.8030\n",
            "Epoch 34/100\n",
            "63/63 [==============================] - 14s 214ms/step - loss: 0.2835 - accuracy: 0.8880 - val_loss: 0.5921 - val_accuracy: 0.7920\n",
            "Epoch 35/100\n",
            "63/63 [==============================] - 13s 210ms/step - loss: 0.2681 - accuracy: 0.8970 - val_loss: 0.5782 - val_accuracy: 0.8060\n",
            "Epoch 36/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.2657 - accuracy: 0.8945 - val_loss: 0.6912 - val_accuracy: 0.7630\n",
            "Epoch 37/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.2557 - accuracy: 0.8960 - val_loss: 0.6320 - val_accuracy: 0.7910\n",
            "Epoch 38/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.2466 - accuracy: 0.8990 - val_loss: 0.6703 - val_accuracy: 0.7830\n",
            "Epoch 39/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.2405 - accuracy: 0.9040 - val_loss: 0.5746 - val_accuracy: 0.7740\n",
            "Epoch 40/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.2154 - accuracy: 0.9080 - val_loss: 0.9902 - val_accuracy: 0.7770\n",
            "Epoch 41/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.2141 - accuracy: 0.9225 - val_loss: 0.6428 - val_accuracy: 0.7850\n",
            "Epoch 42/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.1985 - accuracy: 0.9210 - val_loss: 0.6527 - val_accuracy: 0.8140\n",
            "Epoch 43/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.2007 - accuracy: 0.9240 - val_loss: 0.7746 - val_accuracy: 0.7810\n",
            "Epoch 44/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.2085 - accuracy: 0.9240 - val_loss: 0.7128 - val_accuracy: 0.7850\n",
            "Epoch 45/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.2007 - accuracy: 0.9140 - val_loss: 0.6215 - val_accuracy: 0.8120\n",
            "Epoch 46/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1753 - accuracy: 0.9335 - val_loss: 0.6717 - val_accuracy: 0.7990\n",
            "Epoch 47/100\n",
            "63/63 [==============================] - 14s 211ms/step - loss: 0.1608 - accuracy: 0.9425 - val_loss: 0.8112 - val_accuracy: 0.7640\n",
            "Epoch 48/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.1566 - accuracy: 0.9350 - val_loss: 0.6642 - val_accuracy: 0.8020\n",
            "Epoch 49/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.1783 - accuracy: 0.9375 - val_loss: 0.8663 - val_accuracy: 0.7910\n",
            "Epoch 50/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.1697 - accuracy: 0.9390 - val_loss: 0.6945 - val_accuracy: 0.7980\n",
            "Epoch 51/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1634 - accuracy: 0.9405 - val_loss: 0.9105 - val_accuracy: 0.7860\n",
            "Epoch 52/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1485 - accuracy: 0.9465 - val_loss: 0.6337 - val_accuracy: 0.7960\n",
            "Epoch 53/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.1440 - accuracy: 0.9440 - val_loss: 0.6270 - val_accuracy: 0.8100\n",
            "Epoch 54/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.1326 - accuracy: 0.9500 - val_loss: 0.8960 - val_accuracy: 0.7770\n",
            "Epoch 55/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.1640 - accuracy: 0.9425 - val_loss: 0.6704 - val_accuracy: 0.8160\n",
            "Epoch 56/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.1318 - accuracy: 0.9465 - val_loss: 0.8876 - val_accuracy: 0.8050\n",
            "Epoch 57/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1528 - accuracy: 0.9500 - val_loss: 0.7811 - val_accuracy: 0.8120\n",
            "Epoch 58/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1227 - accuracy: 0.9535 - val_loss: 0.7534 - val_accuracy: 0.7870\n",
            "Epoch 59/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1302 - accuracy: 0.9550 - val_loss: 0.9930 - val_accuracy: 0.7800\n",
            "Epoch 60/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1210 - accuracy: 0.9555 - val_loss: 0.7985 - val_accuracy: 0.8090\n",
            "Epoch 61/100\n",
            "63/63 [==============================] - 13s 205ms/step - loss: 0.1076 - accuracy: 0.9635 - val_loss: 0.8968 - val_accuracy: 0.8100\n",
            "Epoch 62/100\n",
            "63/63 [==============================] - 15s 227ms/step - loss: 0.1038 - accuracy: 0.9625 - val_loss: 1.0589 - val_accuracy: 0.7800\n",
            "Epoch 63/100\n",
            "63/63 [==============================] - 13s 208ms/step - loss: 0.1250 - accuracy: 0.9550 - val_loss: 1.0676 - val_accuracy: 0.7340\n",
            "Epoch 64/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.1150 - accuracy: 0.9560 - val_loss: 0.9520 - val_accuracy: 0.7820\n",
            "Epoch 65/100\n",
            "63/63 [==============================] - 13s 206ms/step - loss: 0.1338 - accuracy: 0.9535 - val_loss: 0.9258 - val_accuracy: 0.8080\n",
            "Epoch 66/100\n",
            "63/63 [==============================] - 14s 213ms/step - loss: 0.0886 - accuracy: 0.9705 - val_loss: 1.3108 - val_accuracy: 0.7880\n",
            "Epoch 67/100\n",
            "63/63 [==============================] - 13s 207ms/step - loss: 0.1285 - accuracy: 0.9590 - val_loss: 0.9757 - val_accuracy: 0.7970\n",
            "Epoch 68/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1168 - accuracy: 0.9535 - val_loss: 1.0262 - val_accuracy: 0.7800\n",
            "Epoch 69/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1021 - accuracy: 0.9625 - val_loss: 0.9514 - val_accuracy: 0.7990\n",
            "Epoch 70/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.1183 - accuracy: 0.9625 - val_loss: 1.2326 - val_accuracy: 0.7810\n",
            "Epoch 71/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1031 - accuracy: 0.9665 - val_loss: 0.9443 - val_accuracy: 0.8160\n",
            "Epoch 72/100\n",
            "63/63 [==============================] - 13s 202ms/step - loss: 0.1011 - accuracy: 0.9685 - val_loss: 0.9262 - val_accuracy: 0.8180\n",
            "Epoch 73/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.0879 - accuracy: 0.9725 - val_loss: 1.5521 - val_accuracy: 0.7710\n",
            "Epoch 74/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.1224 - accuracy: 0.9615 - val_loss: 1.0413 - val_accuracy: 0.7890\n",
            "Epoch 75/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.0970 - accuracy: 0.9630 - val_loss: 0.9867 - val_accuracy: 0.8030\n",
            "Epoch 76/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.0927 - accuracy: 0.9670 - val_loss: 1.1988 - val_accuracy: 0.7840\n",
            "Epoch 77/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.1075 - accuracy: 0.9620 - val_loss: 0.7191 - val_accuracy: 0.8130\n",
            "Epoch 78/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.0873 - accuracy: 0.9680 - val_loss: 1.1592 - val_accuracy: 0.7940\n",
            "Epoch 79/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.1091 - accuracy: 0.9645 - val_loss: 1.0450 - val_accuracy: 0.7990\n",
            "Epoch 80/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.0991 - accuracy: 0.9675 - val_loss: 0.9264 - val_accuracy: 0.8310\n",
            "Epoch 81/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.0732 - accuracy: 0.9760 - val_loss: 4.2227 - val_accuracy: 0.6850\n",
            "Epoch 82/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.1158 - accuracy: 0.9690 - val_loss: 1.3219 - val_accuracy: 0.7630\n",
            "Epoch 83/100\n",
            "63/63 [==============================] - 13s 204ms/step - loss: 0.0993 - accuracy: 0.9645 - val_loss: 1.0687 - val_accuracy: 0.8090\n",
            "Epoch 84/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.1135 - accuracy: 0.9675 - val_loss: 1.0056 - val_accuracy: 0.8040\n",
            "Epoch 85/100\n",
            "63/63 [==============================] - 13s 199ms/step - loss: 0.0997 - accuracy: 0.9715 - val_loss: 1.3252 - val_accuracy: 0.7760\n",
            "Epoch 86/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.0947 - accuracy: 0.9675 - val_loss: 1.0793 - val_accuracy: 0.7980\n",
            "Epoch 87/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.0835 - accuracy: 0.9720 - val_loss: 1.1795 - val_accuracy: 0.7890\n",
            "Epoch 88/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.1038 - accuracy: 0.9720 - val_loss: 1.0748 - val_accuracy: 0.8030\n",
            "Epoch 89/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.0782 - accuracy: 0.9725 - val_loss: 1.1267 - val_accuracy: 0.8220\n",
            "Epoch 90/100\n",
            "63/63 [==============================] - 13s 199ms/step - loss: 0.0832 - accuracy: 0.9680 - val_loss: 1.3205 - val_accuracy: 0.8340\n",
            "Epoch 91/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.1106 - accuracy: 0.9695 - val_loss: 1.1187 - val_accuracy: 0.8140\n",
            "Epoch 92/100\n",
            "63/63 [==============================] - 13s 199ms/step - loss: 0.0923 - accuracy: 0.9740 - val_loss: 1.3150 - val_accuracy: 0.8030\n",
            "Epoch 93/100\n",
            "63/63 [==============================] - 14s 207ms/step - loss: 0.0841 - accuracy: 0.9730 - val_loss: 1.3577 - val_accuracy: 0.8040\n",
            "Epoch 94/100\n",
            "63/63 [==============================] - 13s 201ms/step - loss: 0.1114 - accuracy: 0.9670 - val_loss: 0.9918 - val_accuracy: 0.8180\n",
            "Epoch 95/100\n",
            "63/63 [==============================] - 14s 218ms/step - loss: 0.1014 - accuracy: 0.9710 - val_loss: 1.2790 - val_accuracy: 0.7970\n",
            "Epoch 96/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.0916 - accuracy: 0.9690 - val_loss: 1.3343 - val_accuracy: 0.8020\n",
            "Epoch 97/100\n",
            "63/63 [==============================] - 13s 200ms/step - loss: 0.0952 - accuracy: 0.9740 - val_loss: 1.2039 - val_accuracy: 0.8180\n",
            "Epoch 98/100\n",
            "63/63 [==============================] - 13s 202ms/step - loss: 0.0949 - accuracy: 0.9750 - val_loss: 1.3264 - val_accuracy: 0.8130\n",
            "Epoch 99/100\n",
            "63/63 [==============================] - 13s 203ms/step - loss: 0.0764 - accuracy: 0.9830 - val_loss: 1.0668 - val_accuracy: 0.8150\n",
            "Epoch 100/100\n",
            "63/63 [==============================] - 13s 202ms/step - loss: 0.1050 - accuracy: 0.9715 - val_loss: 1.6346 - val_accuracy: 0.7760\n"
          ]
        }
      ],
      "source": [
        "history = model_aug.fit(train_dataset,epochs=100,validation_data=validation_dataset,callbacks=callbacks)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "j7mlJ_LxI922",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "284c166d-7e51-470c-a12c-1952adfacfe2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
          ]
        }
      ],
      "source": [
        "print(history.history.keys())"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Vt2zVXOaI9uu"
      },
      "outputs": [],
      "source": [
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "a-puAGhVI9ke",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "outputId": "69475f75-adb6-4353-efa4-944f601f4dab"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEICAYAAABPgw/pAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO2deXgUVdbG35OwBpAlgGCAAMoWhBCIoKAsroAOCKMoIIKoCIrbjAqO6+Awo5/OCI46DuqIIoq4jIOKCyi44UJklU22BMNO2IKsSc73x+mbru6u6i3d6XTn/J6nn666devWrarut06de+69xMxQFEVR4p+kWFdAURRFiQwq6IqiKAmCCrqiKEqCoIKuKIqSIKigK4qiJAgq6IqiKAmCCnoCQ0QfE9HoSOeNJUSUS0QXR6FcJqKzXMsvENFDweQN4zgjieizcOupKP4gjUOvWBDREctqCoATAIpd67cw8+zyr1XFgYhyAdzEzAsjXC4DaMPMmyKVl4haAtgKoCozF0WinorijyqxroDiCTPXNsv+xIuIqqhIKBUF/T1WDNTlEicQUV8iyieiSUS0C8ArRFSfiD4kor1EdMC13Myyz2Iiusm1PIaIviGip1x5txLRgDDztiKir4iokIgWEtFzRPS6Q72DqeNjRPStq7zPiKihZfsoIsojogIiesDP9elBRLuIKNmSNoSIVrmWuxPRd0R0kIh2EtGzRFTNoayZRPQXy/q9rn12ENFYr7yXE9FyIjpMRL8S0aOWzV+5vg8S0REiOs9cW8v+PYloKREdcn33DPbahHidGxDRK65zOEBE71u2DSaiFa5z2ExE/V3pHu4tInrU3GciaulyPd1IRNsAfOFKf9t1Hw65fiMdLfvXJKK/u+7nIddvrCYRfUREt3udzyoiGmJ3roozKujxRRMADQCkAxgHuX+vuNZbADgG4Fk/+/cAsAFAQwD/B+BlIqIw8r4B4EcAqQAeBTDKzzGDqeMIADcAaAygGoB7AICIMgD8y1X+Ga7jNYMNzPwDgN8AXOhV7huu5WIAd7vO5zwAFwG41U+94apDf1d9LgHQBoC3//43ANcDqAfgcgATiOhK17beru96zFybmb/zKrsBgI8APOM6t38A+IiIUr3Owefa2BDoOs+CuPA6usp62lWH7gBeA3Cv6xx6A8h1uh429AHQAcBlrvWPIdepMYBlAKwuwqcAdAPQE/I7vg9ACYBXAVxnMhFRJoA0yLVRQoGZ9VNBP5A/1sWu5b4ATgKo4Sd/FwAHLOuLIS4bABgDYJNlWwoABtAklLwQsSgCkGLZ/jqA14M8J7s6PmhZvxXAJ67lhwHMsWyr5boGFzuU/RcA/3Et14GIbbpD3rsA/NeyzgDOci3PBPAX1/J/ADxuydfWmtem3GkAnnYtt3TlrWLZPgbAN67lUQB+9Nr/OwBjAl2bUK4zgKYQ4axvk+/fpr7+fn+u9UfNfbacW2s/dajnylMX8sA5BiDTJl8NAAcg7RKACP/z5f1/S4SPWujxxV5mPm5WiCiFiP7teoU9DHnFr2d1O3ixyyww81HXYu0Q854BYL8lDQB+dapwkHXcZVk+aqnTGdaymfk3AAVOx4JY40OJqDqAoQCWMXOeqx5tXW6IXa56/BVirQfCow4A8rzOrwcRLXK5Og4BGB9kuabsPK+0PIh1anC6Nh4EuM7NIffsgM2uzQFsDrK+dpReGyJKJqLHXW6bw3Bb+g1dnxp2x3L9pt8CcB0RJQEYDnmjUEJEBT2+8A5J+iOAdgB6MPNpcL/iO7lRIsFOAA2IKMWS1txP/rLUcae1bNcxU50yM/NaiCAOgKe7BRDXzXqIFXgagD+FUwfIG4qVNwDMA9CcmesCeMFSbqAQsh0QF4mVFgC2B1Evb/xd518h96yezX6/AjjToczfIG9nhiY2eaznOALAYIhbqi7Eijd12AfguJ9jvQpgJMQVdpS93FNKcKigxzd1IK+xB13+2EeifUCXxZsD4FEiqkZE5wH4XZTq+A6AK4jofFcD5hQE/s2+AeBOiKC97VWPwwCOEFF7ABOCrMNcAGOIKMP1QPGufx2I9Xvc5Y8eYdm2F+LqaO1Q9nwAbYloBBFVIaJrAGQA+DDIunnXw/Y6M/NOiG/7eVfjaVUiMoL/MoAbiOgiIkoiojTX9QGAFQCudeXPBnBVEHU4AXmLSoG8BZk6lEDcV/8gojNc1vx5rrcpuAS8BMDfodZ52KigxzfTANSEWD/fA/iknI47EtKwWADxW78F+SPbEXYdmXkNgNsgIr0T4mfND7Dbm5CGui+YeZ8l/R6I2BYCeNFV52Dq8LHrHL4AsMn1beVWAFOIqBDi859r2fcogKkAviWJrjnXq+wCAFdArOsCSCPhFV71DpZA13kUgFOQt5Q9kDYEMPOPkEbXpwEcAvAl3G8ND0Es6gMA/gzPNx47XoO8IW0HsNZVDyv3AFgNYCmA/QCegKcGvQagE6RNRgkD7ViklBkiegvAemaO+huCkrgQ0fUAxjHz+bGuS7yiFroSMkR0DhGd6XpF7w/xm74faD9FccLlzroVwIxY1yWeUUFXwqEJJKTuCCSGegIzL49pjZS4hYgug7Q37EZgt47iB3W5KIqiJAhqoSuKoiQIAQfnIqL/QFri9zDz2TbbCcB0AAMhHR/GMPOyQOU2bNiQW7ZsGXKFFUVRKjM//fTTPmZuZLctmNEWZ0LGhHjNYfsAyNgNbSDjf/zL9e2Xli1bIicnJ4jDK4qiKAYi8u5dXEpAlwszfwWJGXViMIDXWPge0t24aejVVBRFUcpCJHzoafAc6yIfnmNRlEJE44goh4hy9u7dG4FDK4qiKIZybRRl5hnMnM3M2Y0a2bqAFEVRlDCJhKBvh+fgRc0Q3uBCiqIoShmIhKDPA3A9CecCOOQaDEhRFEUpRwIKOhG9CRl0vx3JFGg3EtF4IhrvyjIfwBbIwEUvIohZYBRFUSoLs2cDLVsCSUnyPTuK07wHDFtk5uEBtjNkRDxFURTFwuzZwLhxwFHXdDB5ebIOACNHRv542lNUURQlSjzwgFvMDUePSno0UEFXFCWmlKdLorww55Tn0AVo27boHFcFXVGUmGFcEnl5ALPbJRELUbc+WBo2lE+wDxnvfceOdRZzAGjhPZFhhFBBV5QEIVqWbjQt6HBdEk51chLlQALt/WApKJCP90PG7rh2+5486Vx3IskblbcRZo7Jp1u3bqwoSmR4/XXmlBRmkRT5pKRIerjlpadLOUThl2vKIZJv7/28y7Z+7PL7O9cJE3zTnT7muKmp8glmn9RU3/L91d/fcctyjwDksIOuxmw89OzsbNbBuRQlMjj5a9PTgdzcwPvPni1W8bZtQIMGQGGhfyvTu1zr/i1aAFOnSro1wgMAUlKAGa45iR54wL9bwi7/tm1iHRcXBz6nikZysn29g71HBiL6iZmzbbepoCtKxcFOGIMJb0tKEpvPGyKgpCTwMb2FNxBEwKxZblEm8jx+SgpQs6a4H5z2D0V6Qs1f0UhJcb6+wdwjz/zOgq4+dEWpINg1EI4aJX/4QP5Wp0Y273Q7H/N114Um5oDUb9Qot4XtLbZHjzqLuV3+YI4XT1StCqSmyr1LT5e3jPR0+7yRbCANZjx0RVHKAbsGQiNkgTqkTJ3qa2VbG9/sXCD+BDcY4k1kQyE1Vb7DuUbp6c5vVnYuKHNvIoKTcz3aH20UVRTB2gAZ6GMaCu0aGgM1ZAbb+BdMY140PnaNjv4+ycmBy0tNlbpbGz/9nYt3I6W/+xJOA2egRuJggJ9GURV0RSknrH/mYAUmGCEx61axL6u4Vq3qFkNTbjRF3YhhsPW35g81usfuPjgJrL/yIyHO4aCCrigxxk4YoiWMZS3DSZwCCa0/wfcX8heskDrlj7awxkq4nVBBV5QIE8yf3JonkHsgVIH09wnnWMFatoFENpIWbUUT0oqCCrqihImdqAQjWmURZKtAhus+CcZSD+RuCPZ6hJNHCR9/gq5x6IrigF18tr946NRU4Nix0EMAvTGdaUyUhL9BnuwwURb+Ou6E2plFqThoHLpSqQl3LBJ/YYR2FBSEL+ZE8m1ilq0hb1Onisjb5TffBhMGN3KkCPbrr/vuG/FQOaXi4GS6R/ujLhelPAg2CsIpAiXSn+Tk4KIr7M7DX6iiukAqD1CXi1JZ8TfGidUtUR5dy71dKYoSDupyUSodgSYYMN3qnbquRwp/rhRFiTTa9V9JOIIdbCpUEQ91ACp/XcAVJRqoha4kHHaNmWUlPV1GF0xP9xxwafp0+0bH11+XRkkVc6U8UQtdSTgiPV+jNXLESaDDGfJWUSKNWuhKXBHMvI9Ow5EmJ/sv2/i7U1N9hz71J9AmRLCkRK1yJbaooCsVEjvhJnI3ZDI7z/toF7edkiLbneK5jUuFGdi3Tz4q0Eq8oS4XpcLh3ahpbXT015BpJhc2PSDt3CC9eql7RElcNA5dqXCE2tXdSqjTeSlKvKFx6EpcUZZGzaSk0Lv4K0qioIKuVDjKMsdicbGvT11RKgsq6EqFw65R04pdNIpdBIvxqStKZUEFXakwmMiWUaOkB6YRa+8wQrtoFCe/eaRj0hWlIqNRLkpMmT3bfoCsggKx0mfNCi4KpUUL+4bUsrhvFCXeUAtdiRkmPNFpgKxQXCZOsec67rdSmQhK0ImoPxFtIKJNRDTZZns6EX1ORKuIaDERNYt8VZVEwbhWrrsu8JgrwbpMRo6UHp3eY61ojLlSmQgo6ESUDOA5AAMAZAAYTkQZXtmeAvAaM3cGMAXA3yJdUSU+CDQ7kLdVHohQXCbaBV+p7ARjoXcHsImZtzDzSQBzAAz2ypMB4AvX8iKb7UolwCrWJnRw1CixmE33/WCscoO6TBQlNIIR9DQAv1rW811pVlYCGOpaHgKgDhGlehdEROOIKIeIcvbu3RtOfZUKjL85OM24K4HQCSEUJXwi1Sh6D4A+RLQcQB8A2wEUe2di5hnMnM3M2Y0aNYrQoZVYYOdaKWuIoDUkUV0mihI6wYQtbgfQ3LLezJVWCjPvgMtCJ6LaAH7PzAcjVUmlYuE9eJZxrYQ7LJDOtakokSEYC30pgDZE1IqIqgG4FsA8awYiakhEpqz7AfwnstVUKhL+XCuhoq4VRYkcAQWdmYsATATwKYB1AOYy8xoimkJEg1zZ+gLYQES/ADgdgDZlJTDBulaMP9wOnaZNUSKPDp+rhEwww9sSiT/cjD3eoIGk79+v45ArSlnwN3yudv1XQmbqVE8fuh0tWvifg1NRlMijXf8Vv9hFs1h7ZQK+rhWNH1eU2KCCrjjir6PQAw+IaDOLa0W73CtK7FGXi+KIv2gWM4EEoK4VRakoqIWuOBIomkUnkFCUioUKuuJIMANj6QQSilJxUEFXPLA2gh45AlSr5j+/TiChKBUHFXSlFO9G0IIC+U51DbOm0SyKUrFRQVdKsWsEPXUKqF1bo1kUJR5QQa+khDJaoknXCSQUpWKjYYuVELvREseNk+75dmOWq59cUeIDtdArIXauFbOuEy0rSvyigp7ghOJaKSgAataURlD1kytK/KEulwQmVNcKIOkpKdIAqkKuKPGFWugJjJNrpaDA/1jl2gNUUeITFfQExl8vTmb/oq49QBUl/lBBT2ACRacwA8nJ4e2rKErFQwU9ATENoXl5/q1wACgu1sgWRUkUVNATDGv3fSCwa8VEsmgPUEWJfzTKJcFwGsM8NRU4dsxzm7HEdTxzRUkM1EJPEKxuFjv271dLXFEAGbpi+HDg8cdjXZPIQ2ymoClnsrOzOScnJybHTjS8483tSE+X8VcUpaLw9dfAsmXAnXeW73Fffhm46SagXj1g506gRo3yPX5ZIaKfmDnbbpta6AmAnZvFijZyKhWRxx8HHn64fI+5ezdwzz1As2bAwYPAvHnle/xoo4IexwRyswDqWlEqJsXFwLffAoWF7nlqy4O77xbj59NPgebNgVdeKb9jlwfaKBqnqJtFiWfWrAEOHZLl336TMfejzSefAG++CTzyCJCRAYweDfz1r8D27UBaWvSPXx6ohR6nqJtFiWe+/tq9XFgY/eMVFwO33w60awfcf7+kjRkjDaSvvRb945cXKuhxir+u+epmUcoCM7BpE/Duu8DJk9E5xjffuJfLQ9AXLJBzevRRoHp1STvzTKB3b3G7BHL77NkjLiJ/nDoFfPRR9K5ZMKigxylOXfONm0XFvHzJywOeekoswXilsBC45RagVSugTRvgqqtE1CMNs1jop53mPm60efFFoGFDYMgQz/QxY4CNG4HvvvO//6hRQN++wN69znn+/nfgiiuAc86R6B1DSQmwY0f5tBWooMcB1jHNGzaUj123/kR2s6xeLT7PWFo//pgxA7j3XmDatFjXJHw++kjO4+yzgX/+U8b5+fnn8Mr65Rfnt8i8PPFbX3KJrB8+HN4xgmXnTolmGTPGbZ0brr4aqFUL+Pe/nfdfuBD47DOgqAiYO9c+z7FjwNNPA5mZIvrduwMTJ0q8e9Om4qP/+OOInZIzzByTT7du3VgJzOuvM6ekMMvz3fdDJN/p6ZI3Ubn9djnPb75xzpOfz9ymDfOXX0avHqdOMR854pvev7/Ur1o15p9/dqe/8QZzp07MeXnBlV9UxLxjB3NxcWTqGwp33cVcowbzyZOy3q4d85Ah4ZXVpg2z01981iy5Vi+/LN//+194xwiWv/5VjrNhg/32P/xBtn/+ue+24mLmrl2ZW7Rg7tiRuUcP+zKee07KWLSIef9+5jFjZL1pU+brrmOuUoV58uTInA+AHHbQVRX0Ck56urOYm096eqxrGX0yMuRcn3jCOY/5U3XsKMIbDcaOZT7rLOaSEndaSQlz48bMAwcyN2rEnJXFfOIE87PPuh+4Tz7pXOapU8wjRogIVq0q+R9/PPg6bd/OvHVr2KdUynnnMffq5V6/8krm9u1DL2frVvdvMyfHd/u4ccx16zKvWyd5/BkieXkijqtWhV4PZhHkVq2Y+/Z1zvPbb/Lwat6c+eBBz21vvil1fO015v/7P1n+5RfPPKdOMbdsyXzuuZ6/iwMH3OtdujBfdll45+CNCnocYwTB34cossc8epT5hRfEWgyHd991tobCYccO97kOHuycb9AgsTAB5meeidzxDcuXu+/H5s3u9O3bJW36dOb//leWe/SQ70GD5AFzwQXO5S5ZInkvvph50iTmM89k7t07+Hr16yeCUhar/sQJ5urVmf/4R3fa/feLZXniRGhlGcs7KYn5llt8t2dkMA8Y4L6vzz/vXNZTT0meqlWZ//xn99tDsHz2mez/xhv+8/3wA3NyMvPo0e60EyeYW7dm7txZ/gu//ir3/5FHPPc1bxz+3jRuuEEe9lbBD5cyCzqA/gA2ANgEYLLN9hYAFgFYDmAVgIGBylRBD45YWOivvirlfvSR77Zvv2X+8EP5LFjgawmvWCE/+uuvj1x9Zs+W+nTq5PynOHGCuVYt5vHjmS+6iLlePea9eyNXB2Zxq1Sr5rbYDB98IGlffy3ro0fL+ujRcn0efFDEbd8++3L/8hfJb+p7zz1ynKNHA9dp/34RIoB54cLwz+3HH6WMuXPdaa+9Jmlr1jjvt2ED8+HDnmkjRjCffrr8BmrXZi4sdG/bt0/K/OtfJT3QW9fYscypqczDh0verCxfK9ofV10l+x87Fjjvww/LMR55RB4eAwfK+vz57jwXXigPXPMbLC6WB3bHjv4fqP/8p5SVnx983Z0ok6ADSAawGUBrANUArASQ4ZVnBoAJruUMALmBylVBD45APvSUlMj7zu+8U8q+917P9OXLfY9/ww2eeYwvuUuXyNVn7FgR6BdeYEdf6BdfuK2kn38WkZswwTdfSYlYzTNmhFaHzz93i89pp8mDwzBlijzEjLAdPcr8ySfuP7gRy1mz7Mvu18/zen34ITv6dL156y23NTxyZGjnZMUIjtXXv3SppL3zjv0+eXnyRmT9DZSUMDdpIgL8zTey/0svubf/73+S9tVXkpdIHnhOnHuuXB9mtyVsfZj6Y+NGecP4wx+Cy3/yJHN2tvutt1kz5okTPQ2IV16R7d99Jw/rhx4Krk7mWsybF1xd/FFWQT8PwKeW9fsB3O+V598AJlnyLwlUrgp68Lz+uljhRGJtpKbKcrQaQnv3ll9GdrZn+tSpbqH58UfmO+5wiyizW/QaN5bX90j4sUtK5DyHDBFLEZA/lTeTJsmf14jq7beLyHn7XvPzubTxcvVqz207d4rf05viYmnga95cLL3LLpO3BcOVVzK3bet8DsXFInJXX+277ehRX1fHwYNSd39CZxg9mrlBA+abb2auWTM463XjRnkLsLovrrtO6mgVL2NBP/aYfTmm4a9mTfd1M/fopZekrIwM5u7dZVtxsdS3WjW3xXzaaWJA2FFSIttvu829f6NGUldv8vJ8LeRrrhGDZ8cOv5fDg8OHmdevd7boDx2Sh9igQfKbAJiHDQvsCioslP/so48GXxcnyiroVwF4ybI+CsCzXnmaAlgNIB/AAQDdHMoaByAHQE6LFi3KfmZKxCkuZq5TR8QxKclTIC64QF55DSdOMGdmioDv3i0PgObN3Zb02rVlr8+mTVLWs89K3erVY77pJt98mZmeDV/79tn7O42FWLWqu/GSWSJjTjtNyvH+c86ZI/vMnCnrf/6zlG1ELD2d+dpr/Z/HzTfLdfX2Ry9YwLburXPOYT7/fP9lFhfLtR8xQnzAgOebx4svitXuLXTGJTRnjjvtrLPkweRNixZSvjerVsk1uPRS9/1hlrYLgHnLFlmfNk3W335bfj+A5/1LS/N9yzOYh+9zz7nTrr3W98GzapW8kd10kzvdvF0E81AMlWuukbIbNfJ0UQWifXv/bUDBUh6C/gcAf2S3hb4WQJK/ctVC96W8LXE7Nm6UX8V118n3Bx9I+sGDIvL33++Zf+VKsbjatHGL3k8/yfJbb5W9PjNmeD4cBg4Uq8+KaVz72988088+W1xAVh5+WB5Uxi//4IPyGlyjhoSYeft0t28X0czMdDcSL1wo+T7+2O0T/r//838e8+ZJvs8+80w3DY9WPzOzuLuqVpUIDCeMK+f1193W8HnnuY9nGnCtD4v9+90Nx6ah1pyD9/VjlrcR60PccPnlEqlSUCBhfZ07Sx0GD5aoEkNBgbyBAJJ/5kxPMW7fXvzcdpgGzUWL3GmmwdX65jV5Mpe6AP/xDyn/wgvl/3PokPP1C5d16+SYe/aEtt/w4WLwlJXycLmsAdDcsr4FQGN/5aqgexILX7kdxif77beeroD33pP0xYt993niCS5ttCwqEjeCnctg3brQo1+uuUaE1oiAcfsUFLjzzJwpacuXe+57003ijrAKyOWXux8IY8ZIPZOT5e1i716xUmvWFAvz1ClxP6WkeDYMHj4s+z30kNvCXrDA/3n89puUe/vtnundu3uGChrmzw9c7iOPiGibxlQTVjd3rjRGdusmFvBFF7n3Mb7yESPke+VK97G++ML3GHfdJfW2WvmLF0t+E1pp3siWLBHR9n6DmjJFLGu7BsHu3Z3D+Yx1v3u3O23bNkn7+99lvaREInwuvZT597+X6/HHP0qeadOcr10sePJJqVdZG+vLKuhVXALdytIo2tErz8cAxriWOwDYAdfkGU4fFXRPKkq8+aRJYhkePy4ujK5dJf2WW+xdBswi4g89xLxsmTutXTvfV/i2bUVofvghuLrY+UwXLZJr8eGH7jS713BmcTkAnnHDTZowjxolywcPypvFJZe4fe+//irnedllYj07NXhlZYkVaETUKYLFyu9+J+Jj6nnggDwYHn7YN+/hw/Kg+dOfnMs75xy3Rc4sbQAm4uWMM+Tt4vHHZX3FCjlup04i9AUFYqmPG+d+a/GOVmFm/ve/ZX8T515SIiGZaWnuKJxDhyTCqGtXyfvmm4GvheGii5h79rTfNm6cWNne97VdO/eb1/ffu98Mjxxx16FlS/kNVyTMm533W1qoRCJscSCAX1zRLg+40qYAGORazgDwrUvsVwC4NFCZKuiexCLe3I5LL3W/Yj/6qNtXnJ5u72N14qqrJLzLYDqbJCcz168fXEeRlStln//8x53222/iojBCV1QkVrhdmOSqVZ6CbOLFrZbbqVO+gmH8wADzjTfa123iRBGxq64SP3MwmAeMiV4x/ny7tx5mEU4nsdu1i20bLM0bhunQs3+/vGGMHu0Wv3//W7bdeKNsO/dcz0ZeK199xR5umy+/lPUXXvDMd9NN7mu2a5ffy+BT37PPtt92/vn28fsTJ8o5Hj/OfPfd4vIzbT35+fJWZRdyG2sKCtjHpRcO2rGoAmP85oHE3MlC/+23yFkiJSXMDRtKmCCz+89rrNB//Sv4skzDoekmb3zhH30k1t3pp/v2uPPGhE96d5s/5xzmPn3k1XXiRMkze7bv/kVF8kZgoiSMH9vEiztRVCRC0rWrcyz4G29IWdWrB/+g279fOqrUqiWulDvucAuTHSZyx9u/zux2M1nfipjFWs7N9UybOFHeui6/XI5tLPFly9y/LbuGZma5xoB08GGWiI769X19+8af7yTOTlx/vf3vuqREjmPXMck8CBcskDeRUAyNWJOeLm7EsqCCXkEJ5De3fqxWqZXu3cUXHAmMf9JELBw/Lq/l9etLuolcCAbjczfulauvlj9fSYk0cDZsKK/FTg1Lxt94882+2+68U6yyOnXEVTB2rLMo9uvnDr80Pmc7gfTm1Cn/YZd5ee57M2VK4PIMO3aINVytmlyDSy91zvvJJ1L+p5/6bhs2zLNtwR8bN7rfAL2vZ8+ekv7ii877N2ok1vz27fI7tIZYGkpKmIcOZX766cD1sXLbbfKG5c3OnVKv6dN9tx06JPXo1UvyWKN1KjpDhoibryyooFdQ/Fnm1iiX5s0lzfvPv3mzpFujCsqCsXyWLHGnXXihpLVrF1pZJlrmpZfcbhFrt+offpCHRa9evmJs/LbDhtkPP2CEzsSm+2PyZPnzHz3KfMUVzB06hHYe/mjWTOphIoGCZf9+t5D6e/0uLLSPLGIWS3/YsOCPOXiwHG/pUs/0d9+VY/h7W+rdW+prHoibNgV/3EBMnixvDwHizeIAACAASURBVN4PJtOnwalR+PzzZXtKiv1gaRWVxx6Tepcl+safoOvwuTHEaXhRImDfPvmUlADz50v6F1/IxLaGDz6Q761bgQMHyl6fZcvk2J07u9P69pXv/v1DK6t1axnOd/VqYPlyYP9+93CpgAwvOnOmTBowbpw8xrZtA/70J2D8eGDgQGDWLBnC1ZvLLpPr8N57MpWYP3r0kGFPly8HfvoJ6NYttPPwR8+e8t21a2j71a8vw7FOmybjjztRuzbQoQOwapVn+vHjcs87dAj+mP/4B/D8877nP3SoDPfapo3zvh06AGvXyhCzAwbIxBCRok4dmRjixAnP9LVr5btjR/v9zG9p0CAZ/jZeyMqS75Uro1O+Cno5Yx3bPMnh6p9+uuf61q3yXVQkY1Yb5s0DqrhmhV2xoux1W7YMaN/e8w8yYICI/JVXhlZWUpL8GVevltliAODiiz3zXHMN8Oc/yxRg2dlyXR5/XCZWeOcdoFo15/Lr1g2uHj16yPf778u42JEU9PHjgVtvlfGuQ6VWLeDOOwOfR0aGW9wMGzfKAzAUQW/dGpgwwXcMfQCoV8//vh06yAN01y7gttuCP2YwOE1ysXat1KtJE/v9fvc7OZfrr49sfaKNefgvXx6d8lXQywEj4kQy80lenvwhnWa3Of98z3Uj6KedJsIEyB/sq69k0gfAc4aUcFm+3NfazM6WP7Kx1EOhUycR9M8+E6vf+0EFAA89JBMP7N0LPPignOvcuUDNmuGcgS9Nm8rs7v/5j6xHUtD79QOee85eJCNFRobMQPXbb+609evlu3376B3XinlwtG4d+ptaIOrUkW87Qc/IcL62WVnygB4wILL1iTZNmwJLlgBjx0anfBX0KDN7trgU8vJkndk3j3ErGIs9JcVz+9atkjZ8uMx6cuyYzGBeVCQ/jGbNyi7oe/YA+fnuV0IrjRuHV2anTiLUX38NXHqpfR4imdNx2zZgyhSZQi/SnHsuUFAgx7I7v4pMRob8ZjZscKetWyfn0rZt+dShc2egalV5o3B6qwyXQILuDzsDIR447zxxp0UDFfQo88ADwNGj/vOUlIg/uaREXjN/+cVze26uWPhDhoiltnCh+M8bNhSXQteuZX+FM/uH6g/2R6dO8l1c7Ok/L2+M26Vdu+j9kaKFETWr22X9evk9ROotJhBNmgCbNwO33x75so2gW6eh27tXPoEEXfFFBT1KGDeLscz90by5NFqdfbbMcbhxo+f2rVtl4t5+/cTn+vbb0lB6+eVi3XftKn9y62t5qJhGmi5dwi/DGyPo1asDF1wQuXJDxQh6JN0t5cVZZ0k7ibegl5e7xdC8eXRcS3Y+9HXr5FsFPXRU0KOAt5vFHykpwIgREsnwhz/Ia3RBgTtqhdkt6NWqiYjPni0+9EGDJE9WluQrS8v5qlXyp61fP/wyvGncWKy7888vP2vSjm7dxHd52WWxq0O4VKsmEShG0EtKYiPo0cLO5WIEPZRGX0VQQY8CgdwsxtJJT5dZ1levFn/giBHu8DFjpR84IK+jrVrJ+pVXyp+6WjW3XzoSLeerVnmGK0aKuXOBZ5+NfLmhULMmsGOHNEjHIx07ugX911+lDSXRBN3qcsnPF199Wlps6hTPqKBHAaf4ckBEfNYssahzc8Ut8dFHEv5Wvbpb0I0fPTdXvo2gDxgg+fr1c/uD09KARo2CaxgtLpaYcCsnT4pVFA1Bv+CCxBGfWJGRIT7s48fdES6JYr3aWeg7d4qBY9cHQfFPlVhXIBFp0cLe3ZKe7hZoQET9rruABg2AiRMl7cwzxYI3FroJWWzZUr5r15YONUbgAXf0hlXQd+wAjhzxjYSYNEkaYHfscMd5r18vETPREHSl7GRkyFvZhg3lH7IYbZwEPZzYfkUt9LCwdg5q2FA+1uW8PN8GpJQUYOpUz7T33gMWLQIee0xEHRDrOz3dV9CtAj5woK+F1rUrsGaN9LgrKJDQqHPP9exBunev9BYsKAByctzppieiCnrFxBrpsm6d/FYaNoxtnSJFcrL8N7wF3alDkeIfFfQQsTZ4Mos4FhR4LgOy7u0rHznSXc6xY8Af/ygul3HjPI/Rpo2noNerF7g3X1aWdKFevVo6G+3cKQ2nTzzhzjN9ury2A9IpybBqlVjr5RXXrIRG27ZiMKxdKxZ6hw7R7cxU3tSp4+lDVws9fFTQQySYuHIDs9vNYhVzAHjqKXkoPPOMu/u+wQi68bNbrXMnTMPo+PHik3/6aTnm9OnSyHT4sDRODh0qguAt6B07+tZDqRhUry7hi0bQE8XdYqhTx22hFxdLJzcV9PBQQQ8Rfw2eduTl+Xbx37kT+NvfZMwSuy71bdsChw6Ji8SELAaidWuJ6f3pJ4llv/VWceWUlACPPAK88IKUOXky0KcP8M034jcHJNxR3S0Vm4wM4LvvgN27E0/QTzvNLeh798pvVgU9PFTQQ6RFi9D32bLFc/2bb8TlMmmSfX5rpIvpJRqIpCQZ/a9NG+Cll+SVvGVLGUxp5kxxvVx8sYzN0ru3/IFWrhRraNcuFfSKTkaGGAJA4kS4GKwWujlHFfTwUEEPkalTg+8kY1wYpmHTYCJdnHzWRtC//VaEPxgLHZCY75wcd+87QIajrV1bQhXvv1/SeveW7y+/FJ87AGRmBncMJTZYe00mmoVu9aGroJcN9ZqGyCWXSA9IE5aYlCTd8Q8edEeq7N8vlnxamoysZg1VBETgGzTwFF4rLVtK6/9nn8l6sIJuQsCsNGwo/vQlSyR2HZB6nXmm+NFN45pa6BUbI+jVqwf3xhZPWF0uRtA1yiU81EIPgbw86ca+Z4+MevjVV+4ww/37PSelyM11W/Legh7IjVK1qvjEv/lG1oMVdCfGjnW7YQx9+sgoiCtWyJ+nUaOyHUOJLu3auUdYTLQON1aXy65d8q2CHh4q6CFw880i5gsWyLjQF1wgseRr1khDpDe7d8t3qIIOiNvl5ElZjoZF1ru3PITmzVPrPB5ISZFIpHgb/jcYvH3o9esDNWrEtk7xigp6kBw+LJ2Axo8HevVyp/fvL+NEf/GFO2rEYATd6kM3oYjBCDog7h3v8dEjQZ8+8n3woAp6vPDZZxLmmmjUqSNtRUVFGoNeVlTQg2TBAvnBDRzou611a3GzGAEHJO++fbJstdD37AmuodMIelndLU6kp8voioAKerzQtGnwU+/FE9YhdFXQy4YKepD885/iw+zTR6zr2bPd28yocNu3u9P27RNrvGlT8QseOybpRtyDtdCjJejmXAAVdCW2WMdz0W7/ZUMFPQhef11C/Mz0cXl50l3fiLqdoBtr3UyuYKJighV0E9IYLUEHpCdpr16JF9esxBfWIXR37VILvSyooAfBvff6ph09KsMAAMEJuhHyYAW9RQuZ8mvYsDAqHCT9+0skjRl1UVFigRH0/HwZa0gFPXw0Dj0ITCiVN2YYgMaNpRNRsILesGHguS2TkhKzAUxRvDE+dDMHgAp6+KiFboP32CtOFqwZBiApSX6EdoKelSVx5UbQt25NvI4hilIWjIW+YYN8q6CHjwq6Db16yVjizzwjkSAmHtyK9/jmaWkyaYRh926Jpa1bVyJKTOhisGOzKEplwQi6WuhlR10uXmzcCPzwgyybbwORe0jcqVM9h8RNS5MORobdu2UaLTNIVm6u7JuXB/zud9E+C0WJH4ygmzkANMolfNRC9+Ljj+W7cWPfbf7GN09L83W5nH66LBtB371bGn2iGbmiKPGGEfRt22S4DKcxjpTABCXoRNSfiDYQ0SYimmyz/WkiWuH6/EJEByNf1ehhnVJu0iTgjDNkXGY7nMZDT0uTOFrThdkq6K1aSYciY8Gry0VR3FSvLu1Upt9GIs3GVN4EFHQiSgbwHIABADIADCeiDGseZr6bmbswcxcA/wTwXjQqGw28p5Q7flx84Sbm3Bun8dDPOEO+jZXubaEDwOLFnuuKogjGSlf/edkIxkLvDmATM29h5pMA5gAY7Cf/cABvRqJy5UEoU8rZTfRssMail5SIhe8k6Onp4dZWURITFfTIEIygpwH41bKe70rzgYjSAbQC8IXD9nFElENEOXudfBrlTLBTytlN9GzFKugFBRL66C3oP/wgvvlatcpUZUVJOIzfXAW9bEQ6yuVaAO8wc7HdRmaeAWAGAGRnZzs4NcqXFi3c3fKdIPIdAtcbI+g7drhj0I2gN2kifsITJ9Tdoih2GAtdI1zKRjAW+nYAzS3rzVxpdlyLOHK3AOJCCTQ8bTDziNaqJTHn27f7CnpSktvNooKuKL6oyyUyBCPoSwG0IaJWRFQNItrzvDMRUXsA9QF8F9kqRpeRI8WVUq+e/XZ/fnNvTOiit6ADbiFXQVcUX1TQI0NAQWfmIgATAXwKYB2Aucy8hoimENEgS9ZrAcxhdooPiT3W8ETrELgjRwKdOgFdusjIiunp4mYJ5Df3RgVdUcJDfeiRISgfOjPPBzDfK+1hr/VHI1etyGPCE01ES14eMGoUcN114rfbtQt47DER72AF3Ju0NGDtWhH0atU8rX7TmUg7FSmKL2qhR4ZK01PULjzRvEvs2iUWeVl/TGlpUtaOHRLNYu0gkZkpIzLq2OOK4kvLljIKacOGsa5JfFNpBD1QeCKzWOhlIS1NwhVXr/Z0twAy9nh+vsagK4odt94qoy0mVRpFig6V5vIFE6kSbEy6EyZ08eeffQWdyDdNURShalWgQYNY1yL+qTSCHqnwRH+Y7v9FRSreiqKUP5VG0E14opPLI5TwRCfSLP1nVdAVRSlvKo2gAyLqZlzySZPc6aGGJzrRuDGQnCzLKuiKopQ3lXaCi+rVpQHmwIHIjb+cnCyRMvn5KuiKopQ/lcpCt/L11xJKGOnB9I3bRQVdUZTyJuEF3a536KlTwPffA+efH/njqaArihIrEtrlYtc7dNw4YPNm4Ngx4IILIn9MFXRFUWJFQgu6Xe/Qo0eBadNkORoWer9+wPLlGlOrKEr5k9AuF6eOQgcOAGeeGZ1xI4YMEf+89nhTFKW8SWjZceoolJQUHetcURQlliS0oE+d6o4LN9SoIXN+RsN/riiKEksSWtBHjpTR22rWlPUqVYBrrpFltdAVRUk0ElrQ8/NlbPKpU4H582WMlQ8+ABo1Atq2jXXtFEVRIktCC/rixfLdr58MX9u3L7B/v1jn1rHKFUVREoGEFvRFi4D69YHOnUXAn3hCvi+8MNY1UxRFiTwJHYe+eDHQp487hLB7d2DNGglZVBRFSTQS1kLftg3YskXcLFY6dJD5PhVFURKNhBV0q/9cURSlMpCwgr5oEZCaCpx9dqxroiiKUj4ktKBb/eeKoiiJTkLK3bRpMrLie++5h8xVFEVJdBJO0GfP9pxezgyZq6KuKEqik3CC/sADwMmTnmlHj0q6oihKIpNwgp6XZ5/uNJSuoihKopBwgt6kiX2601C6iqIoiULCCXrv3r5pKSkyQJeiKEoik3CCvm8f0Lw5kJ4u47akpwMzZshQuoqiKIlMQo3lcvQo8M03wMSJwN//HuvaKIqilC8JZaF/9ZVEuFxySaxroiiKUv4EJehE1J+INhDRJiKa7JBnGBGtJaI1RPRGZKsZHAsWyMBbdn50RVGURCegy4WIkgE8B+ASAPkAlhLRPGZea8nTBsD9AHox8wEiahytCvtjwQKgVy9pBFUURalsBGOhdwewiZm3MPNJAHMADPbKczOA55j5AAAw857IVjMwu3YBq1eru0VRlMpLMIKeBuBXy3q+K81KWwBtiehbIvqeiPrbFURE44goh4hy9u7dG16NHXjmGfkeMCCixSqKosQNkWoUrQKgDYC+AIYDeJGI6nlnYuYZzJzNzNmNGjWK0KGB776T6eXGjgW6dIlYsYqiKHFFMIK+HUBzy3ozV5qVfADzmPkUM28F8AtE4KPOb78B118vsedPP10eR1QURamYBCPoSwG0IaJWRFQNwLUA5nnleR9inYOIGkJcMFsiWE9HJk0CNm8GXn0VOO208jiioihKxSSgoDNzEYCJAD4FsA7AXGZeQ0RTiGiQK9unAAqIaC2ARQDuZeaCaFXa8N13wHPPAXffLZNZKIqiVGaC6inKzPMBzPdKe9iyzAD+4PqUG99/L9/331+eR1UURamYxHVP0T17gKpVZe5QRVGUyk5cC/ru3UDjxjIIl6IoSmUnrgV9zx4RdEVRFCXOBX33buD002W+0JYtgaQknRRaUZTKS1wPn7tnD1CjhkwCffSopJlJoQEdA11RlMpF3FrozGKhr1rlFnODTgqtKEplJG4t9MJC4MQJ+dihk0IrilLZiFsLffdu+XYKWdRJoRVFqWzErYW+xzVA7403As8+6+l20UmhlXjj1KlTyM/Px/Hjx2NdFaWCUKNGDTRr1gxVq1YNep+4FXRjoQ8fDnTuLD7zbdvEMp86VRtElfgiPz8fderUQcuWLUHasaLSw8woKChAfn4+WrVqFfR+cSvoxkJv3FiGzFUBV+KZ48ePq5grpRARUlNTEeq8EXHvQ4/gsOqKElNUzBUr4fwe4lbQ9+wBGjSQsVwURVGUOBZ000tUUSojke4dXVBQgC5duqBLly5o0qQJ0tLSStdPnjzpd9+cnBzccccdAY/Rs2fPslVSCUhc+9B1HBelMjJ7duR7R6empmLFihUAgEcffRS1a9fGPffcU7q9qKgIVarYy0V2djays7MDHmPJkiXhVS6GFBcXIzk5OdbVCBq10BUlznjggfLpHT1mzBiMHz8ePXr0wH333Ycff/wR5513HrKystCzZ09s2LABALB48WJcccUVAORhMHbsWPTt2xetW7fGM2b2dgC1a9cuzd+3b19cddVVaN++PUaOHAmZUgGYP38+2rdvj27duuGOO+4oLddKbm4uLrjgAnTt2hVdu3b1eFA88cQT6NSpEzIzMzF58mQAwKZNm3DxxRcjMzMTXbt2xebNmz3qDAATJ07EzJkzAQAtW7bEpEmT0LVrV7z99tt48cUXcc455yAzMxO///3vcdR18Xfv3o0hQ4YgMzMTmZmZWLJkCR5++GFMmzattNwHHngA06dPL/O9CBa10BUlznDqBR2N3tH5+flYsmQJkpOTcfjwYXz99deoUqUKFi5ciD/96U949913ffZZv349Fi1ahMLCQrRr1w4TJkzwiaVevnw51qxZgzPOOAO9evXCt99+i+zsbNxyyy346quv0KpVKwwfPty2To0bN8aCBQtQo0YNbNy4EcOHD0dOTg4+/vhj/O9//8MPP/yAlJQU7N+/HwAwcuRITJ48GUOGDMHx48dRUlKCX3/91e95p6amYtmyZQDEHXXzzTcDAB588EG8/PLLuP3223HHHXegT58++O9//4vi4mIcOXIEZ5xxBoYOHYq77roLJSUlmDNnDn788ceQr3u4xKWgnzgBHDyoFrpSOWnRQtwsdumR5uqrry51ORw6dAijR4/Gxo0bQUQ4deqU7T6XX345qlevjurVq6Nx48bYvXs3mjVr5pGne/fupWldunRBbm4uateujdatW5fGXQ8fPhwzZszwKf/UqVOYOHEiVqxYgeTkZPzyyy8AgIULF+KGG25ASkoKAKBBgwYoLCzE9u3bMWTIEADSWScYrrnmmtLln3/+GQ8++CAOHjyII0eO4LLLLgMAfPHFF3jttdcAAMnJyahbty7q1q2L1NRULF++HLt370ZWVhZSy3EGnrgUdBOaqRa6UhmZOtXThw5Er3d0rVq1Spcfeugh9OvXD//973+Rm5uLvn372u5TvXr10uXk5GQUFRWFlceJp59+GqeffjpWrlyJkpKSoEXaSpUqVVBSUlK67t1D13reY8aMwfvvv4/MzEzMnDkTixcv9lv2TTfdhJkzZ2LXrl0YO3ZsyHUrC3HpQzcx6GqhK5WRkSOBGTOA9HSZrSs9Xdaj3bnu0KFDSEtLA4BSf3MkadeuHbZs2YLc3FwAwFtvveVYj6ZNmyIpKQmzZs1CcXExAOCSSy7BK6+8Uurj3r9/P+rUqYNmzZrh/fffBwCcOHECR48eRXp6OtauXYsTJ07g4MGD+Pzzzx3rVVhYiKZNm+LUqVOYbQknuuiii/Cvf/0LgDSeHjp0CAAwZMgQfPLJJ1i6dGmpNV9exKWgW3uJKkplZORIIDcXKCmR7/LoKX3ffffh/vvvR1ZWVkgWdbDUrFkTzz//PPr3749u3bqhTp06qFu3rk++W2+9Fa+++ioyMzOxfv36Umu6f//+GDRoELKzs9GlSxc89dRTAIBZs2bhmWeeQefOndGzZ0/s2rULzZs3x7Bhw3D22Wdj2LBhyMrKcqzXY489hh49eqBXr15o3759afr06dOxaNEidOrUCd26dcPatWsBANWqVUO/fv0wbNiwco+QIdO6XN5kZ2dzTk5OWPvOnAnccAOweTPQunVk66UosWDdunXo0KFDrKsRc44cOYLatWuDmXHbbbehTZs2uPvuu2NdrZAoKSkpjZBp06ZNmcqy+10Q0U/MbBsnqha6oigVhhdffBFdunRBx44dcejQIdxyyy2xrlJIrF27FmeddRYuuuiiMot5OMRlo+ju3dII5AprVRQlQbj77rvjziK3kpGRgS1btsTs+HFrodeqpRNDK4qiWIlLC33FCqCgwB2+qBNDK4qixKmFvmGDtO5b0YmhFUWp7MSloDt0UNOJoRVFqdTEnaB7W+ZWdGJoRQmPfv364dNPP/VImzZtGiZMmOC4T9++fWFCjwcOHIiDBw/65Hn00UdL48GdeP/990tjuAHg4YcfxsKFC0OpvuIi7gTdNd6Oz8QWOjG0ooTP8OHDMWfOHI+0OXPmOA6Q5c38+fNRr169sI7tLehTpkzBxRdfHFZZscL0Vo01cSfoJgZ93Ljy7/qsKOXBXXcBfftG9nPXXf6PedVVV+Gjjz4qncwiNzcXO3bswAUXXIAJEyYgOzsbHTt2xCOPPGK7f8uWLbFv3z4AwNSpU9G2bVucf/75pUPsArAdhnbJkiWYN28e7r33XnTp0gWbN2/GmDFj8M477wAAPv/8c2RlZaFTp04YO3YsTpw4UXq8Rx55BF27dkWnTp2wfv16nzpVxmF2407QzTguQ4eWf9dnRUlUGjRogO7du+Pjjz8GINb5sGHDQESYOnUqcnJysGrVKnz55ZdYtWqVYzk//fQT5syZgxUrVmD+/PlYunRp6bahQ4di6dKlWLlyJTp06ICXX34ZPXv2xKBBg/Dkk09ixYoVOPPMM0vzHz9+HGPGjMFbb72F1atXo6ioqHTsFABo2LAhli1bhgkTJti6dcwwu8uWLcNbb71VOquSdZjdlStX4r777gMgw+zedtttWLlyJZYsWYKmTZsGvG5mmN1rr73W9vwAlA6zu3LlSixbtgwdO3bE2LFjS0dqNMPsXnfddQGPF4igwhaJqD+A6QCSAbzEzI97bR8D4EkA211JzzLzS2WunQ3GQteBuZRExWK4lSvG7TJ48GDMmTOnVJDmzp2LGTNmoKioCDt37sTatWvRuXNn2zK+/vprDBkypHQI20GDBpVucxqG1okNGzagVatWaNu2LQBg9OjReO6553CX63Vj6NChAIBu3brhvffe89m/Mg6zG1DQiSgZwHMALgGQD2ApEc1j5rVeWd9i5ollrpEfZs8Gbr9dlvv3Bx5/XC1zRYkUgwcPxt13341ly5bh6NGj6NatG7Zu3YqnnnoKS5cuRf369TFmzBifoWaDJdRhaANhhuB1Gn63Mg6zG4zLpTuATcy8hZlPApgDYHBEjh4CZh7FAwdkPT9f1rWHqKJEhtq1a6Nfv34YO3ZsaWPo4cOHUatWLdStWxe7d+8udck40bt3b7z//vs4duwYCgsL8cEHH5RucxqGtk6dOigsLPQpq127dsjNzcWmTZsAyKiJffr0Cfp8KuMwu8EIehoA63xN+a40b35PRKuI6B0iam5XEBGNI6IcIsrZa7p5Bkl5zaOoKJWZ4cOHY+XKlaWCnpmZiaysLLRv3x4jRoxAr169/O7ftWtXXHPNNcjMzMSAAQNwzjnnlG5zGob22muvxZNPPomsrCxs3ry5NL1GjRp45ZVXcPXVV6NTp05ISkrC+PHjgz6XyjjMbsDhc4noKgD9mfkm1/ooAD2s7hUiSgVwhJlPENEtAK5h5gv9lRvq8LlJSYBdVYn8x6YrSjygw+dWPoIZZjcaw+duB2C1uJvB3fgJAGDmAmY+4Vp9CUC3IMoNCadOQ9qZSFGUeCNaw+wGE+WyFEAbImoFEfJrAYywZiCipsy807U6CMC6iNXQRXnOo6goihJNojXMbkALnZmLAEwE8ClEqOcy8xoimkJEJibpDiJaQ0QrAdwBYEykKxqreRQVpbyI1exhSsUknN9DXE5BpyiJxtatW1GnTh2kpqaCiGJdHSXGMDMKCgpQWFiIVq1aeWzz50OPy/HQFSXRaNasGfLz8xFq9JeSuNSoUQPNmjULaR8VdEWpAFStWtXHElOUUIm7sVwURVEUe1TQFUVREgQVdEVRlAQhZlEuRLQXQF4IuzQEsC9K1anIVMbzroznDFTO866M5wyU7bzTmbmR3YaYCXqoEFGOU6hOIlMZz7synjNQOc+7Mp4zEL3zVpeLoihKgqCCriiKkiDEk6DPiHUFYkRlPO/KeM5A5TzvynjOQJTOO2586IqiKIp/4slCVxRFUfyggq4oipIgxIWgE1F/ItpARJuIaHKs6xMNiKg5ES0iorWuoYjvdKU3IKIFRLTR9V0/1nWNNESUTETLiehD13orIvrBdb/fIqJqsa5jpCGieq7pGtcT0ToiOq+S3Ou7Xb/vn4noTSKqkWj3m4j+Q0R7iOhnS5rtvSXhGde5ryKirmU5doUXdCJKBvAcgAEAMgAMJ6KM2NYqKhQB+CMzZwA4F8BtrvOcDOBzZm4D4HPXeqJxJzwnRXkCwNPMEpT6GgAAAuZJREFUfBaAAwBujEmtost0AJ8wc3sAmZDzT+h7TURpkPkSspn5bADJkAlzEu1+zwTQ3yvN6d4OANDG9RkH4F9lOXCFF3QA3QFsYuYtzHwSwBwAg2Ncp4jDzDuZeZlruRDyB0+DnOurrmyvArgyNjWMDkTUDMDlkKkLQTIY+IUA3nFlScRzrgugN4CXAYCZTzLzQST4vXZRBUBNIqoCIAXATiTY/WbmrwDs90p2ureDAbzGwvcA6hFR03CPHQ+CngbgV8t6vistYSGilgCyAPwA4HTL9H67AJweo2pFi2kA7gNgpvpOBXDQNVMWkJj3uxWAvQBecbmaXiKiWkjwe83M2wE8BWAbRMgPAfgJiX+/Aed7G1F9iwdBr1QQUW0A7wK4i5kPW7exxJgmTJwpEV0BYA8z/xTrupQzVQB0BfAvZs4C8Bu83CuJdq8BwOU3Hgx5oJ0BoBZ8XRMJTzTvbTwI+nYAzS3rzVxpCQcRVYWI+Wxmfs+VvNu8grm+98SqflGgF4BBRJQLcaVdCPEt13O9kgOJeb/zAeQz8w+u9XcgAp/I9xoALgawlZn3MvMpAO9BfgOJfr8B53sbUX2LB0FfCqCNqyW8GqQRZV6M6xRxXL7jlwGsY+Z/WDbNAzDatTwawP/Ku27RgpnvZ+ZmzNwScl+/YOaRABYBuMqVLaHOGQCYeReAX4monSvpIgBrkcD32sU2AOcSUYrr927OO6HvtwunezsPwPWuaJdzARyyuGZCh5kr/AfAQAC/ANgM4IFY1ydK53g+5DVsFYAVrs9AiE/5cwAbASwE0CDWdY3S+fcF8KFruTWAHwFsAvA2gOqxrl8UzrcLgBzX/X4fQP3KcK8B/BnAegA/A5gFoHqi3W8Ab0LaCE5B3sZudLq3AAgSxbcZwGpIBFDYx9au/4qiKAlCPLhcFEVRlCBQQVcURUkQVNAVRVESBBV0RVGUBEEFXVEUJUFQQVcURUkQVNAVRVEShP8Hxb7tyI3KAX0AAAAASUVORK5CYII=\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "accuracy,val_accuracy,loss,val_loss = history.history[\"accuracy\"],history.history[\"val_accuracy\"],history.history[\"loss\"],history.history[\"val_loss\"]\n",
        "epochs = range(1,len(accuracy)+1)\n",
        "plt.plot(epochs, accuracy, \"bo\", label=\"Training accuracy\")\n",
        "plt.plot(epochs, val_accuracy, \"b\", label=\"Validation accuracy\")\n",
        "plt.title(\"Training and validation accuracy\")\n",
        "plt.legend()\n",
        "plt.figure()\n",
        "plt.plot(epochs, loss, \"bo\", label=\"Training loss\")\n",
        "plt.plot(epochs, val_loss, \"b\", label=\"Validation loss\")\n",
        "plt.title(\"Training and validation loss\")\n",
        "plt.legend()\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "test_model = keras.models.load_model(\n",
        "    \"augment_model.keras\")\n",
        "test_loss, test_acc = test_model.evaluate(test_size)\n",
        "print(f\"Test accuracy: {test_acc:.3f}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OAi857wpAKa9",
        "outputId": "4233d8d8-913a-4582-8f21-3795f35a370b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "63/63 [==============================] - 4s 50ms/step - loss: 0.5484 - accuracy: 0.7660\n",
            "Test accuracy: 0.766\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Pretrained model to increase accuracy - VGG16\n"
      ],
      "metadata": {
        "id": "UEUxI5AdQS72"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "conv_base = keras.applications.vgg16.VGG16(weights=\"imagenet\",include_top=False,input_shape=(180,180,3))"
      ],
      "metadata": {
        "id": "Lk7e2NzQA1bX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "cf124842-f6ce-4819-b078-9db6ee4b5da0"
      },
      "execution_count": null,
      "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 [==============================] - 2s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "conv_base.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Fn9hflt2Q0O-",
        "outputId": "8dc9620f-3ffc-4f7e-d577-de04b6b4073d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"vgg16\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_1 (InputLayer)        [(None, 180, 180, 3)]     0         \n",
            "                                                                 \n",
            " block1_conv1 (Conv2D)       (None, 180, 180, 64)      1792      \n",
            "                                                                 \n",
            " block1_conv2 (Conv2D)       (None, 180, 180, 64)      36928     \n",
            "                                                                 \n",
            " block1_pool (MaxPooling2D)  (None, 90, 90, 64)        0         \n",
            "                                                                 \n",
            " block2_conv1 (Conv2D)       (None, 90, 90, 128)       73856     \n",
            "                                                                 \n",
            " block2_conv2 (Conv2D)       (None, 90, 90, 128)       147584    \n",
            "                                                                 \n",
            " block2_pool (MaxPooling2D)  (None, 45, 45, 128)       0         \n",
            "                                                                 \n",
            " block3_conv1 (Conv2D)       (None, 45, 45, 256)       295168    \n",
            "                                                                 \n",
            " block3_conv2 (Conv2D)       (None, 45, 45, 256)       590080    \n",
            "                                                                 \n",
            " block3_conv3 (Conv2D)       (None, 45, 45, 256)       590080    \n",
            "                                                                 \n",
            " block3_pool (MaxPooling2D)  (None, 22, 22, 256)       0         \n",
            "                                                                 \n",
            " block4_conv1 (Conv2D)       (None, 22, 22, 512)       1180160   \n",
            "                                                                 \n",
            " block4_conv2 (Conv2D)       (None, 22, 22, 512)       2359808   \n",
            "                                                                 \n",
            " block4_conv3 (Conv2D)       (None, 22, 22, 512)       2359808   \n",
            "                                                                 \n",
            " block4_pool (MaxPooling2D)  (None, 11, 11, 512)       0         \n",
            "                                                                 \n",
            " block5_conv1 (Conv2D)       (None, 11, 11, 512)       2359808   \n",
            "                                                                 \n",
            " block5_conv2 (Conv2D)       (None, 11, 11, 512)       2359808   \n",
            "                                                                 \n",
            " block5_conv3 (Conv2D)       (None, 11, 11, 512)       2359808   \n",
            "                                                                 \n",
            " block5_pool (MaxPooling2D)  (None, 5, 5, 512)         0         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 14,714,688\n",
            "Trainable params: 14,714,688\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np"
      ],
      "metadata": {
        "id": "YWCfiuZ-RL0r"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# first Method to use pretrained model"
      ],
      "metadata": {
        "id": "gvmiMSdKZx7T"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def get_features_and_labels(dataset):\n",
        "    all_features = []\n",
        "    all_labels = []\n",
        "    for images, labels in dataset:\n",
        "        preprocessed_images = keras.applications.vgg16.preprocess_input(images)\n",
        "        features = conv_base.predict(preprocessed_images)\n",
        "        all_features.append(features)\n",
        "        all_labels.append(labels)\n",
        "    return np.concatenate(all_features), np.concatenate(all_labels)"
      ],
      "metadata": {
        "id": "wNXuSzK-3vpf"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_features, train_labels =  get_features_and_labels(train_dataset)\n",
        "val_features, val_labels =  get_features_and_labels(validation_dataset)\n",
        "test_features, test_labels =  get_features_and_labels(test_size) "
      ],
      "metadata": {
        "id": "tUDBNYJt4gzH"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "inputs = keras.Input(shape=(5, 5, 512))\n",
        "x = layers.Flatten()(inputs)        \n",
        "x = layers.Dense(256)(x)\n",
        "x = layers.Dropout(0.5)(x)\n",
        "outputs = layers.Dense(1, activation=\"sigmoid\")(x)\n",
        "model = keras.Model(inputs, outputs)\n",
        "model.compile(loss=\"binary_crossentropy\",\n",
        "              optimizer=\"rmsprop\",\n",
        "              metrics=[\"accuracy\"])"
      ],
      "metadata": {
        "id": "KOjrihJ_4rCu"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "callbacks = [\n",
        "    keras.callbacks.ModelCheckpoint(\n",
        "        filepath=\"feature_extraction.keras\",\n",
        "        save_best_only=True,\n",
        "        monitor=\"val_loss\")]"
      ],
      "metadata": {
        "id": "Rp1jdnwM7DYK"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(\n",
        "    train_features, train_labels,\n",
        "    epochs=20,\n",
        "    validation_data=(val_features, val_labels),\n",
        "    callbacks=callbacks)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mvTn5Xff8D6t",
        "outputId": "7d7f0913-0f97-4d67-be9e-84cf2a0a4075"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "63/63 [==============================] - 3s 11ms/step - loss: 20.3978 - accuracy: 0.9190 - val_loss: 6.2690 - val_accuracy: 0.9480\n",
            "Epoch 2/20\n",
            "63/63 [==============================] - 0s 8ms/step - loss: 2.4782 - accuracy: 0.9765 - val_loss: 4.3505 - val_accuracy: 0.9780\n",
            "Epoch 3/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 1.6212 - accuracy: 0.9880 - val_loss: 5.1702 - val_accuracy: 0.9700\n",
            "Epoch 4/20\n",
            "63/63 [==============================] - 1s 9ms/step - loss: 1.0563 - accuracy: 0.9890 - val_loss: 4.9310 - val_accuracy: 0.9760\n",
            "Epoch 5/20\n",
            "63/63 [==============================] - 1s 8ms/step - loss: 0.8361 - accuracy: 0.9920 - val_loss: 6.0496 - val_accuracy: 0.9740\n",
            "Epoch 6/20\n",
            "63/63 [==============================] - 1s 8ms/step - loss: 0.5249 - accuracy: 0.9935 - val_loss: 4.4637 - val_accuracy: 0.9660\n",
            "Epoch 7/20\n",
            "63/63 [==============================] - 1s 9ms/step - loss: 0.4439 - accuracy: 0.9950 - val_loss: 4.1633 - val_accuracy: 0.9780\n",
            "Epoch 8/20\n",
            "63/63 [==============================] - 0s 8ms/step - loss: 0.5956 - accuracy: 0.9945 - val_loss: 5.4188 - val_accuracy: 0.9740\n",
            "Epoch 9/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 0.2525 - accuracy: 0.9970 - val_loss: 4.5272 - val_accuracy: 0.9770\n",
            "Epoch 10/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 0.2434 - accuracy: 0.9960 - val_loss: 4.3112 - val_accuracy: 0.9810\n",
            "Epoch 11/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.1399 - accuracy: 0.9975 - val_loss: 5.2055 - val_accuracy: 0.9820\n",
            "Epoch 12/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.3155 - accuracy: 0.9970 - val_loss: 4.8338 - val_accuracy: 0.9750\n",
            "Epoch 13/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.2924 - accuracy: 0.9965 - val_loss: 6.5383 - val_accuracy: 0.9770\n",
            "Epoch 14/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.0910 - accuracy: 0.9975 - val_loss: 4.2382 - val_accuracy: 0.9780\n",
            "Epoch 15/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 0.1801 - accuracy: 0.9980 - val_loss: 8.0991 - val_accuracy: 0.9750\n",
            "Epoch 16/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.0456 - accuracy: 0.9995 - val_loss: 5.2427 - val_accuracy: 0.9790\n",
            "Epoch 17/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 0.1068 - accuracy: 0.9990 - val_loss: 5.0748 - val_accuracy: 0.9770\n",
            "Epoch 18/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 0.0168 - accuracy: 0.9995 - val_loss: 6.5086 - val_accuracy: 0.9770\n",
            "Epoch 19/20\n",
            "63/63 [==============================] - 0s 6ms/step - loss: 0.2997 - accuracy: 0.9975 - val_loss: 5.5456 - val_accuracy: 0.9790\n",
            "Epoch 20/20\n",
            "63/63 [==============================] - 0s 7ms/step - loss: 1.5199e-18 - accuracy: 1.0000 - val_loss: 5.5456 - val_accuracy: 0.9790\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "acc = history.history[\"accuracy\"]\n",
        "val_acc = history.history[\"val_accuracy\"]\n",
        "loss = history.history[\"loss\"]\n",
        "val_loss = history.history[\"val_loss\"]\n",
        "epochs = range(1, len(acc) + 1)\n",
        "plt.plot(epochs, acc, \"bo\", label=\"Training accuracy\")\n",
        "plt.plot(epochs, val_acc, \"b\", label=\"Validation accuracy\")\n",
        "plt.title(\"Training and validation accuracy\")\n",
        "plt.legend()\n",
        "plt.figure()\n",
        "plt.plot(epochs, loss, \"bo\", label=\"Training loss\")\n",
        "plt.plot(epochs, val_loss, \"b\", label=\"Validation loss\")\n",
        "plt.title(\"Training and validation loss\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "9hTGauNA8OYo",
        "outputId": "7efcf985-4ad9-43d6-c923-bdd84cfcb579"
      },
      "execution_count": null,
      "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"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 2nd Method to use pretrained model "
      ],
      "metadata": {
        "id": "ciuc_Iq0hSkx"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "conv_base = keras.applications.vgg16.VGG16(weights=\"imagenet\",include_top=False,input_shape=(180,180,3))"
      ],
      "metadata": {
        "id": "EpX0mct3i-XO"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "conv_base.trainable = False"
      ],
      "metadata": {
        "id": "os_ePOqQndE3"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "data_augmentation = keras.Sequential([\n",
        "                                        layers.RandomFlip(\"horizontal\"),\n",
        "                                        layers.RandomRotation(0.1),\n",
        "                                        layers.RandomZoom(0.2)\n",
        "]) "
      ],
      "metadata": {
        "id": "uw5iM0Zw86JQ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "inputs = keras.Input(shape=(180,180,3))\n",
        "x  = data_augmentation(inputs)\n",
        "x  = keras.applications.vgg16.preprocess_input(x)\n",
        "x  = conv_base(x)\n",
        "x  = layers.Flatten()(x)\n",
        "x  = layers.Dense(256,activation='relu')(x)\n",
        "x  = layers.Dropout(0.5)(x)\n",
        "outputs  = layers.Dense(1,activation=\"sigmoid\")(x)\n",
        "model = keras.Model(inputs=inputs,outputs=outputs)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6vqxdZQqhx3P",
        "outputId": "84db8637-3794-4fd6-d120-284b678e4d80"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "z7QvjJBOlisp",
        "outputId": "95693766-2d63-4bf9-9a24-a2c635663302"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_11 (InputLayer)       [(None, 180, 180, 3)]     0         \n",
            "                                                                 \n",
            " sequential_2 (Sequential)   (None, 180, 180, 3)       0         \n",
            "                                                                 \n",
            " tf.__operators__.getitem_3   (None, 180, 180, 3)      0         \n",
            " (SlicingOpLambda)                                               \n",
            "                                                                 \n",
            " tf.nn.bias_add_3 (TFOpLambd  (None, 180, 180, 3)      0         \n",
            " a)                                                              \n",
            "                                                                 \n",
            " vgg16 (Functional)          (None, 5, 5, 512)         14714688  \n",
            "                                                                 \n",
            " flatten_3 (Flatten)         (None, 12800)             0         \n",
            "                                                                 \n",
            " dense_6 (Dense)             (None, 256)               3277056   \n",
            "                                                                 \n",
            " dropout_3 (Dropout)         (None, 256)               0         \n",
            "                                                                 \n",
            " dense_7 (Dense)             (None, 1)                 257       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 17,992,001\n",
            "Trainable params: 3,277,313\n",
            "Non-trainable params: 14,714,688\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(optimizer=\"rmsprop\",loss=\"binary_crossentropy\",metrics=[\"accuracy\"])"
      ],
      "metadata": {
        "id": "zeoezOu8q8ZE"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "callbacks = [keras.callbacks.ModelCheckpoint(save_best_only=True,filepath = \"keras_vgg16_aug.keras\",monitor=\"val_loss\")]"
      ],
      "metadata": {
        "id": "lW2BkwnxnQs2"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(train_dataset,epochs=50,validation_data= validation_dataset,callbacks=callbacks)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bAgwAcM8ogJH",
        "outputId": "ffd56e6e-23d7-46d9-9ada-fca7212055a3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.\n",
            "WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "63/63 [==============================] - 36s 332ms/step - loss: 7.7353 - accuracy: 0.8835 - val_loss: 0.8204 - val_accuracy: 0.9710\n",
            "Epoch 2/50\n",
            "63/63 [==============================] - 18s 285ms/step - loss: 1.0802 - accuracy: 0.9475 - val_loss: 0.7355 - val_accuracy: 0.9680\n",
            "Epoch 3/50\n",
            "63/63 [==============================] - 19s 293ms/step - loss: 0.8021 - accuracy: 0.9560 - val_loss: 0.7193 - val_accuracy: 0.9630\n",
            "Epoch 4/50\n",
            "63/63 [==============================] - 18s 280ms/step - loss: 0.4949 - accuracy: 0.9630 - val_loss: 0.5699 - val_accuracy: 0.9730\n",
            "Epoch 5/50\n",
            "63/63 [==============================] - 18s 277ms/step - loss: 0.3984 - accuracy: 0.9665 - val_loss: 0.5900 - val_accuracy: 0.9660\n",
            "Epoch 6/50\n",
            "63/63 [==============================] - 18s 279ms/step - loss: 0.2913 - accuracy: 0.9695 - val_loss: 0.4935 - val_accuracy: 0.9700\n",
            "Epoch 7/50\n",
            "63/63 [==============================] - 20s 313ms/step - loss: 0.2633 - accuracy: 0.9720 - val_loss: 0.2434 - val_accuracy: 0.9750\n",
            "Epoch 8/50\n",
            "63/63 [==============================] - 19s 290ms/step - loss: 0.1828 - accuracy: 0.9740 - val_loss: 0.3937 - val_accuracy: 0.9710\n",
            "Epoch 9/50\n",
            "63/63 [==============================] - 18s 274ms/step - loss: 0.1996 - accuracy: 0.9775 - val_loss: 0.2956 - val_accuracy: 0.9800\n",
            "Epoch 10/50\n",
            "63/63 [==============================] - 18s 276ms/step - loss: 0.1791 - accuracy: 0.9770 - val_loss: 0.5081 - val_accuracy: 0.9780\n",
            "Epoch 11/50\n",
            "63/63 [==============================] - 20s 307ms/step - loss: 0.1786 - accuracy: 0.9825 - val_loss: 0.4753 - val_accuracy: 0.9700\n",
            "Epoch 12/50\n",
            "63/63 [==============================] - 18s 277ms/step - loss: 0.1817 - accuracy: 0.9820 - val_loss: 0.5060 - val_accuracy: 0.9750\n",
            "Epoch 13/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.1199 - accuracy: 0.9885 - val_loss: 0.4107 - val_accuracy: 0.9840\n",
            "Epoch 14/50\n",
            "63/63 [==============================] - 18s 274ms/step - loss: 0.1968 - accuracy: 0.9850 - val_loss: 0.5336 - val_accuracy: 0.9790\n",
            "Epoch 15/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.1223 - accuracy: 0.9815 - val_loss: 0.5102 - val_accuracy: 0.9790\n",
            "Epoch 16/50\n",
            "63/63 [==============================] - 18s 283ms/step - loss: 0.1173 - accuracy: 0.9865 - val_loss: 0.4080 - val_accuracy: 0.9820\n",
            "Epoch 17/50\n",
            "63/63 [==============================] - 18s 277ms/step - loss: 0.1755 - accuracy: 0.9845 - val_loss: 0.4817 - val_accuracy: 0.9700\n",
            "Epoch 18/50\n",
            "63/63 [==============================] - 18s 280ms/step - loss: 0.1433 - accuracy: 0.9880 - val_loss: 0.3840 - val_accuracy: 0.9800\n",
            "Epoch 19/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.1435 - accuracy: 0.9875 - val_loss: 0.3605 - val_accuracy: 0.9820\n",
            "Epoch 20/50\n",
            "63/63 [==============================] - 18s 276ms/step - loss: 0.1455 - accuracy: 0.9865 - val_loss: 0.2556 - val_accuracy: 0.9850\n",
            "Epoch 21/50\n",
            "63/63 [==============================] - 20s 310ms/step - loss: 0.1404 - accuracy: 0.9910 - val_loss: 0.3013 - val_accuracy: 0.9830\n",
            "Epoch 22/50\n",
            "63/63 [==============================] - 20s 308ms/step - loss: 0.0705 - accuracy: 0.9930 - val_loss: 0.3375 - val_accuracy: 0.9820\n",
            "Epoch 23/50\n",
            "63/63 [==============================] - 18s 285ms/step - loss: 0.0936 - accuracy: 0.9905 - val_loss: 0.4417 - val_accuracy: 0.9760\n",
            "Epoch 24/50\n",
            "63/63 [==============================] - 17s 273ms/step - loss: 0.0778 - accuracy: 0.9910 - val_loss: 0.3098 - val_accuracy: 0.9820\n",
            "Epoch 25/50\n",
            "63/63 [==============================] - 18s 274ms/step - loss: 0.0539 - accuracy: 0.9925 - val_loss: 0.3011 - val_accuracy: 0.9820\n",
            "Epoch 26/50\n",
            "63/63 [==============================] - 17s 273ms/step - loss: 0.0854 - accuracy: 0.9870 - val_loss: 0.3146 - val_accuracy: 0.9830\n",
            "Epoch 27/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.1207 - accuracy: 0.9900 - val_loss: 0.3919 - val_accuracy: 0.9740\n",
            "Epoch 28/50\n",
            "63/63 [==============================] - 18s 283ms/step - loss: 0.0986 - accuracy: 0.9890 - val_loss: 0.6377 - val_accuracy: 0.9720\n",
            "Epoch 29/50\n",
            "63/63 [==============================] - 17s 273ms/step - loss: 0.0596 - accuracy: 0.9925 - val_loss: 0.3870 - val_accuracy: 0.9820\n",
            "Epoch 30/50\n",
            "63/63 [==============================] - 18s 284ms/step - loss: 0.0373 - accuracy: 0.9930 - val_loss: 0.4204 - val_accuracy: 0.9820\n",
            "Epoch 31/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.0343 - accuracy: 0.9935 - val_loss: 0.4572 - val_accuracy: 0.9810\n",
            "Epoch 32/50\n",
            "63/63 [==============================] - 20s 308ms/step - loss: 0.0487 - accuracy: 0.9920 - val_loss: 0.3873 - val_accuracy: 0.9810\n",
            "Epoch 33/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.0516 - accuracy: 0.9940 - val_loss: 0.4083 - val_accuracy: 0.9820\n",
            "Epoch 34/50\n",
            "63/63 [==============================] - 20s 307ms/step - loss: 0.0456 - accuracy: 0.9955 - val_loss: 0.3420 - val_accuracy: 0.9830\n",
            "Epoch 35/50\n",
            "63/63 [==============================] - 18s 278ms/step - loss: 0.0586 - accuracy: 0.9935 - val_loss: 0.3740 - val_accuracy: 0.9810\n",
            "Epoch 36/50\n",
            "63/63 [==============================] - 18s 282ms/step - loss: 0.0265 - accuracy: 0.9970 - val_loss: 0.3868 - val_accuracy: 0.9840\n",
            "Epoch 37/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.0750 - accuracy: 0.9945 - val_loss: 0.5313 - val_accuracy: 0.9810\n",
            "Epoch 38/50\n",
            "63/63 [==============================] - 18s 276ms/step - loss: 0.0568 - accuracy: 0.9935 - val_loss: 0.3778 - val_accuracy: 0.9820\n",
            "Epoch 39/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.0453 - accuracy: 0.9965 - val_loss: 0.4861 - val_accuracy: 0.9780\n",
            "Epoch 40/50\n",
            "63/63 [==============================] - 17s 273ms/step - loss: 0.0532 - accuracy: 0.9965 - val_loss: 0.5659 - val_accuracy: 0.9810\n",
            "Epoch 41/50\n",
            "63/63 [==============================] - 18s 283ms/step - loss: 0.0201 - accuracy: 0.9970 - val_loss: 0.3612 - val_accuracy: 0.9830\n",
            "Epoch 42/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.0197 - accuracy: 0.9965 - val_loss: 0.4048 - val_accuracy: 0.9780\n",
            "Epoch 43/50\n",
            "63/63 [==============================] - 18s 286ms/step - loss: 0.0486 - accuracy: 0.9935 - val_loss: 0.3936 - val_accuracy: 0.9790\n",
            "Epoch 44/50\n",
            "63/63 [==============================] - 17s 274ms/step - loss: 0.0367 - accuracy: 0.9955 - val_loss: 0.3792 - val_accuracy: 0.9810\n",
            "Epoch 45/50\n",
            "63/63 [==============================] - 17s 273ms/step - loss: 0.0445 - accuracy: 0.9955 - val_loss: 0.4130 - val_accuracy: 0.9810\n",
            "Epoch 46/50\n",
            "63/63 [==============================] - 18s 289ms/step - loss: 0.0290 - accuracy: 0.9980 - val_loss: 0.4244 - val_accuracy: 0.9780\n",
            "Epoch 47/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.0261 - accuracy: 0.9965 - val_loss: 0.5046 - val_accuracy: 0.9810\n",
            "Epoch 48/50\n",
            "63/63 [==============================] - 18s 275ms/step - loss: 0.0293 - accuracy: 0.9955 - val_loss: 0.4958 - val_accuracy: 0.9770\n",
            "Epoch 49/50\n",
            "63/63 [==============================] - 18s 276ms/step - loss: 0.0653 - accuracy: 0.9925 - val_loss: 0.3720 - val_accuracy: 0.9870\n",
            "Epoch 50/50\n",
            "63/63 [==============================] - 18s 288ms/step - loss: 0.0560 - accuracy: 0.9940 - val_loss: 0.3458 - val_accuracy: 0.9850\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Fine Tuning the model"
      ],
      "metadata": {
        "id": "Ih-5iRADQDwp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "conv_base.trainable = True\n",
        "for layer in conv_base.layers[:-4]:\n",
        "    layer.trainable = False"
      ],
      "metadata": {
        "id": "fin38KiSof-H"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(loss=\"binary_crossentropy\",\n",
        "              optimizer=keras.optimizers.RMSprop(learning_rate=1e-5),\n",
        "              metrics=[\"accuracy\"])\n",
        "  \n",
        "callbacks = [\n",
        "    keras.callbacks.ModelCheckpoint(\n",
        "        filepath=\"fine_tuning.keras\",\n",
        "        save_best_only=True,\n",
        "        monitor=\"val_loss\")\n",
        "]\n",
        "history = model.fit(\n",
        "    train_dataset,\n",
        "    epochs=30,\n",
        "    validation_data=validation_dataset,\n",
        "    callbacks=callbacks)"
      ],
      "metadata": {
        "id": "KQ1h-mHIRLJ-"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyPdRt7BxcVYJdeohD6HlPYC",
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}