computer-vision/fine-tuning-densenet121-for-image-classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Fine-Tuning DenseNet121 for Image Classification",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cgvnrCfBbuNt"
      },
      "source": [
        "# Fine-Tuning DenseNet121 for Image Classification\n",
        "*By Carlos Santiago Bañón*\n",
        "\n",
        "**Year:** 2020\n",
        "\n",
        "**Technologies:** Python, TensorFlow, Keras, Matplotlib, NumPy, TeX, Jupyter Notebook\n",
        "\n",
        "**Discipline(s):** Computer Vision\n",
        "\n",
        "**Keywords:** `computer-vision`, `cnn`, `convolutional-neural-network`, `deep-learning`, `image-classification`, `densenet121`"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AYamXeN0wOVK"
      },
      "source": [
        "This notebook shows the use of a pretrained *Convolutional Neural Network (CNN)* on a large data set from Google, containing 2,000 JPG images of cats and dogs.\n",
        "\n",
        "In *Deep Learning with Python*, François Chollet defines pretrained networks as \"a saved network that was previously trained on a large dataset, typically on a large-scale image-classification task.\"\n",
        "\n",
        "Here, we employ the [DenseNet121](https://keras.io/applications/#densenet) convolutional base for feature extraction. These are then run through a new classifier trained from scratch. With it, we then use data augmentation, fine-tuning, and no dropout."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fnJ_JIpBb5ob"
      },
      "source": [
        "## 1. Get the Data\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7SVDE3nnztYy"
      },
      "source": [
        "### 1.1. Download and Extract the Data from Google"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IZ2ubzgDzNdT"
      },
      "source": [
        "First, we download the data from Google, which containes 2,000 JPG images of cats and dogs. These are extracted locally in `/tmp`."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lCDU6140cA2v",
        "outputId": "1d8a5680-46dd-48a9-ad2f-2f7232e2db15",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        }
      },
      "source": [
        "# Download the data from Google.\n",
        "\n",
        "!wget --no-check-certificate \\\n",
        "    https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip \\\n",
        "    -O /tmp/cats_and_dogs_filtered.zip"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2020-04-22 01:05:57--  https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 74.125.195.128, 2607:f8b0:400e:c09::80\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|74.125.195.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 68606236 (65M) [application/zip]\n",
            "Saving to: ‘/tmp/cats_and_dogs_filtered.zip’\n",
            "\n",
            "\r          /tmp/cats   0%[                    ]       0  --.-KB/s               \r         /tmp/cats_  86%[================>   ]  56.29M   281MB/s               \r/tmp/cats_and_dogs_ 100%[===================>]  65.43M   297MB/s    in 0.2s    \n",
            "\n",
            "2020-04-22 01:05:57 (297 MB/s) - ‘/tmp/cats_and_dogs_filtered.zip’ saved [68606236/68606236]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "viU0x8YYcI7v"
      },
      "source": [
        "import os\n",
        "import zipfile\n",
        "\n",
        "# Unzip the file and extract the data.\n",
        "local_zip = '/tmp/cats_and_dogs_filtered.zip'\n",
        "zip_ref = zipfile.ZipFile(local_zip, 'r')\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5bwSilqPzw4J"
      },
      "source": [
        "### 1.2. Split the Data into Training and Validation Sets"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sC4rtapmzl2l"
      },
      "source": [
        "We then take this data and split it into training and validation sets. It must be noted that the data provided by Google does not have a test set."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HzlF1HLFcKd0"
      },
      "source": [
        "# Split the data into training and validation sets.\n",
        "\n",
        "base_dir = '/tmp/cats_and_dogs_filtered'\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "\n",
        "# Directory with our training cat pictures.\n",
        "train_cats_dir = os.path.join(train_dir, 'cats')\n",
        "\n",
        "# Directory with our training dog pictures.\n",
        "train_dogs_dir = os.path.join(train_dir, 'dogs')\n",
        "\n",
        "# Directory with our validation cat pictures.\n",
        "validation_cats_dir = os.path.join(validation_dir, 'cats')\n",
        "\n",
        "# Directory with our validation dog pictures.\n",
        "validation_dogs_dir = os.path.join(validation_dir, 'dogs')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WcGwpQhxcQyr"
      },
      "source": [
        "## 2. Build the New Model\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k1NISnSZ0W2v"
      },
      "source": [
        "Once we have the data ready, we can proceed to build the new model using the pretrained DenseNet convolutional base. This will then be concatenated with a newly-trained custom set of densely-connected layers."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "v85m_lDMcjgn"
      },
      "source": [
        "### 2.1. Load the Convolutional Base"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EpAuGUM90pcI"
      },
      "source": [
        "First, load the pretrained DenseNet convolutional base."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G0YPYH1EcnmM",
        "outputId": "e6705420-7286-4393-ff9b-7692d26efaf9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "from keras.applications import DenseNet121\n",
        "\n",
        "conv_base = DenseNet121(\n",
        "    weights='imagenet', \n",
        "    include_top=False, \n",
        "    input_shape=(150, 150, 3))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KkWkDCqpeAz1"
      },
      "source": [
        "### 2.2. Freeze the Convolutional Base"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pgYukGqN01Wy"
      },
      "source": [
        "Once the pretrained convolutional base is ready, freeze it by setting its `trainable` attribute to `False`."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "76FfZbLaeC45"
      },
      "source": [
        "conv_base.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EneyLvQQeI1H"
      },
      "source": [
        "### 2.3. Add a Densely-Connected Classifier on Top of the Convolutional Base"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dcCBCP9A1MRf"
      },
      "source": [
        "After freezing the pretrained model, we can then proceed to add a densely-connected classifier on top of the convolutional base. This will provide more specific results for our problem."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hr0tVGmpeLnM"
      },
      "source": [
        "from keras import layers\n",
        "from keras import models\n",
        "from keras import optimizers\n",
        "\n",
        "model = models.Sequential()\n",
        "model.add(conv_base)\n",
        "model.add(layers.Flatten())\n",
        "model.add(layers.Dense(256, activation='relu'))\n",
        "model.add(layers.Dense(1, activation='sigmoid'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PgeL3w65eZSF",
        "outputId": "d888e29a-b194-44a5-de01-b7320aeedcab",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 289
        }
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "densenet121 (Model)          (None, 4, 4, 1024)        7037504   \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 16384)             0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 256)               4194560   \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 1)                 257       \n",
            "=================================================================\n",
            "Total params: 11,232,321\n",
            "Trainable params: 4,194,817\n",
            "Non-trainable params: 7,037,504\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3w2zI_7Mek7l"
      },
      "source": [
        "### 2.4. Train the Model End to End with a Frozen Convolutional Base"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nzAtKnmtEWmK"
      },
      "source": [
        "After concatenating the pretrained model and the densely-connected layers, the next step is to train the model end to end with the frozen convolutional base.\n",
        "\n",
        "First, we use data augmentation."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jGbjUN0DemEi"
      },
      "source": [
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras import optimizers"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G2WYu7VLenzH",
        "outputId": "d0ec87e7-5107-412d-b652-024223838aba",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "# Use data augmentation.\n",
        "\n",
        "train_datagen = ImageDataGenerator(\n",
        "    rescale=1./255, \n",
        "    rotation_range=40,\n",
        "    width_shift_range=0.2,\n",
        "    height_shift_range=0.2,\n",
        "    shear_range=0.2,\n",
        "    zoom_range=0.2,\n",
        "    horizontal_flip=True,\n",
        "    fill_mode='nearest'\n",
        ")\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "    train_dir,\n",
        "    target_size=(150, 150),\n",
        "    batch_size=20,\n",
        "    class_mode='binary')\n",
        "\n",
        "validation_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "validation_generator = validation_datagen.flow_from_directory(\n",
        "    validation_dir,\n",
        "    target_size=(150, 150),\n",
        "    batch_size=20,\n",
        "    class_mode='binary')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DNH9csF3En33"
      },
      "source": [
        "Then, we compile and train the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Pwa4_mYSerhI"
      },
      "source": [
        "# Compile the model.\n",
        "\n",
        "model.compile(\n",
        "    loss='binary_crossentropy', \n",
        "    optimizer=optimizers.RMSprop(lr=2e-5), \n",
        "    metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CZmZqzCRetA6",
        "outputId": "4bd8dc02-3a2d-4f90-ef44-64a5b917e9fe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# Train the model.\n",
        "\n",
        "history = model.fit_generator(\n",
        "    train_generator,\n",
        "    steps_per_epoch=100,\n",
        "    epochs=30,\n",
        "    validation_data=validation_generator,\n",
        "    validation_steps=50\n",
        ")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "100/100 [==============================] - 26s 260ms/step - loss: 0.3753 - acc: 0.8230 - val_loss: 0.3794 - val_acc: 0.9390\n",
            "Epoch 2/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.2499 - acc: 0.8855 - val_loss: 0.0079 - val_acc: 0.9430\n",
            "Epoch 3/30\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.2458 - acc: 0.8925 - val_loss: 0.0091 - val_acc: 0.9440\n",
            "Epoch 4/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.2317 - acc: 0.9035 - val_loss: 0.1025 - val_acc: 0.9590\n",
            "Epoch 5/30\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.2086 - acc: 0.9115 - val_loss: 0.0980 - val_acc: 0.9540\n",
            "Epoch 6/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1905 - acc: 0.9160 - val_loss: 0.0282 - val_acc: 0.9370\n",
            "Epoch 7/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.2275 - acc: 0.9060 - val_loss: 0.1580 - val_acc: 0.9490\n",
            "Epoch 8/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.2045 - acc: 0.9075 - val_loss: 0.2591 - val_acc: 0.9610\n",
            "Epoch 9/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1939 - acc: 0.9210 - val_loss: 0.0971 - val_acc: 0.9430\n",
            "Epoch 10/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1873 - acc: 0.9210 - val_loss: 0.0160 - val_acc: 0.9660\n",
            "Epoch 11/30\n",
            "100/100 [==============================] - 18s 184ms/step - loss: 0.1959 - acc: 0.9180 - val_loss: 0.0014 - val_acc: 0.9380\n",
            "Epoch 12/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1718 - acc: 0.9270 - val_loss: 0.0184 - val_acc: 0.9520\n",
            "Epoch 13/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1599 - acc: 0.9310 - val_loss: 0.0081 - val_acc: 0.9540\n",
            "Epoch 14/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1650 - acc: 0.9340 - val_loss: 0.2921 - val_acc: 0.9720\n",
            "Epoch 15/30\n",
            "100/100 [==============================] - 18s 185ms/step - loss: 0.1767 - acc: 0.9235 - val_loss: 0.9676 - val_acc: 0.9430\n",
            "Epoch 16/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1711 - acc: 0.9335 - val_loss: 0.0165 - val_acc: 0.9560\n",
            "Epoch 17/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1777 - acc: 0.9245 - val_loss: 0.3038 - val_acc: 0.9170\n",
            "Epoch 18/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1750 - acc: 0.9240 - val_loss: 0.0013 - val_acc: 0.9480\n",
            "Epoch 19/30\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.1559 - acc: 0.9390 - val_loss: 3.5386e-05 - val_acc: 0.9700\n",
            "Epoch 20/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1671 - acc: 0.9345 - val_loss: 0.0205 - val_acc: 0.9540\n",
            "Epoch 21/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1580 - acc: 0.9340 - val_loss: 0.6773 - val_acc: 0.9580\n",
            "Epoch 22/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1822 - acc: 0.9170 - val_loss: 0.4594 - val_acc: 0.9520\n",
            "Epoch 23/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1672 - acc: 0.9265 - val_loss: 0.0187 - val_acc: 0.9450\n",
            "Epoch 24/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1489 - acc: 0.9420 - val_loss: 0.1047 - val_acc: 0.9530\n",
            "Epoch 25/30\n",
            "100/100 [==============================] - 18s 184ms/step - loss: 0.1437 - acc: 0.9425 - val_loss: 0.0048 - val_acc: 0.9540\n",
            "Epoch 26/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1631 - acc: 0.9330 - val_loss: 0.7208 - val_acc: 0.9470\n",
            "Epoch 27/30\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1540 - acc: 0.9350 - val_loss: 0.0017 - val_acc: 0.9720\n",
            "Epoch 28/30\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.1372 - acc: 0.9420 - val_loss: 0.0079 - val_acc: 0.9700\n",
            "Epoch 29/30\n",
            "100/100 [==============================] - 18s 184ms/step - loss: 0.1179 - acc: 0.9540 - val_loss: 0.0031 - val_acc: 0.9710\n",
            "Epoch 30/30\n",
            "100/100 [==============================] - 18s 183ms/step - loss: 0.1480 - acc: 0.9360 - val_loss: 0.6581 - val_acc: 0.9440\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ft3V0Yejg1_Q"
      },
      "source": [
        "### 2.5. Show Loss and Accuracy Curves During Training"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wMiLaS9DFAEP"
      },
      "source": [
        "After the training phase is complete, we can then visualize the training and validation accuracy and loss using *Matplotlib*."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZmYVtydPg2jK",
        "outputId": "2eaa22c7-d12c-4240-db43-50408dd98d93",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        }
      },
      "source": [
        "%matplotlib inline\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\n",
        "\n",
        "# Training and Validation Accuracy\n",
        "\n",
        "plt.plot(epochs, acc, 'bo', label='Training Accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation Accuracy')\n",
        "plt.title('Training and Validation Accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\n",
        "\n",
        "# Training and Validation Loss\n",
        "\n",
        "plt.plot(epochs, loss, 'bo', label='Training Loss')\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation Loss')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qD5p0LMHg7ZM"
      },
      "source": [
        "## 3. Perform Fine-Tuning\n",
        "\n",
        "---"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AG_6ikl_Fpbu"
      },
      "source": [
        "After the initial training phase, we can use *fine-tuning* to get a higher accuracy."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IKA5PcvAhFcD"
      },
      "source": [
        "### 3.1. Freeze All Layers Up to a Specific One"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-VR5aEPyGnuY"
      },
      "source": [
        "First, we must freeze all the layers up to a specific one."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_vOKcKYhhGeW"
      },
      "source": [
        "conv_base.trainable = True\n",
        "\n",
        "set_trainable = False\n",
        "for layer in conv_base.layers:\n",
        "  if layer.name == 'block5_conv1':\n",
        "    set_trainable = True\n",
        "  if set_trainable:\n",
        "    layer.trainable = True\n",
        "  else:\n",
        "    layer.trainable = False"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FgcYAKwEhLg8"
      },
      "source": [
        "### 3.2. Recompile and Train the Model with Fine-Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "J34G6YebGrQv"
      },
      "source": [
        "Then, we recompile the model and train it once more with the fine-tuning."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-stZICYbhMAh"
      },
      "source": [
        "# Recompile the model.\n",
        "\n",
        "model.compile(\n",
        "    loss='binary_crossentropy',\n",
        "    optimizer=optimizers.RMSprop(lr=1e-5), \n",
        "    metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0PBOOuNnhN9u",
        "outputId": "4464a239-a433-46e1-f146-df66314cb1b8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# Train the model.\n",
        "\n",
        "history = model.fit_generator(\n",
        "    train_generator,\n",
        "    steps_per_epoch=100,\n",
        "    epochs=100,\n",
        "    validation_data=validation_generator,\n",
        "    validation_steps=50)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "100/100 [==============================] - 28s 282ms/step - loss: 0.1485 - acc: 0.9390 - val_loss: 0.1357 - val_acc: 0.9670\n",
            "Epoch 2/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1278 - acc: 0.9485 - val_loss: 0.1359 - val_acc: 0.9690\n",
            "Epoch 3/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1308 - acc: 0.9435 - val_loss: 0.2101 - val_acc: 0.9640\n",
            "Epoch 4/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1418 - acc: 0.9395 - val_loss: 0.0171 - val_acc: 0.9590\n",
            "Epoch 5/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1675 - acc: 0.9340 - val_loss: 0.0088 - val_acc: 0.9660\n",
            "Epoch 6/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1298 - acc: 0.9505 - val_loss: 0.0945 - val_acc: 0.9590\n",
            "Epoch 7/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.1212 - acc: 0.9500 - val_loss: 0.0825 - val_acc: 0.9650\n",
            "Epoch 8/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1174 - acc: 0.9520 - val_loss: 0.0074 - val_acc: 0.9620\n",
            "Epoch 9/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1182 - acc: 0.9505 - val_loss: 0.0301 - val_acc: 0.9720\n",
            "Epoch 10/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.1303 - acc: 0.9505 - val_loss: 0.0060 - val_acc: 0.9620\n",
            "Epoch 11/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1367 - acc: 0.9455 - val_loss: 0.0189 - val_acc: 0.9630\n",
            "Epoch 12/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1223 - acc: 0.9530 - val_loss: 0.1227 - val_acc: 0.9680\n",
            "Epoch 13/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1175 - acc: 0.9490 - val_loss: 0.0431 - val_acc: 0.9660\n",
            "Epoch 14/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1296 - acc: 0.9535 - val_loss: 0.1828 - val_acc: 0.9710\n",
            "Epoch 15/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.1314 - acc: 0.9470 - val_loss: 0.0057 - val_acc: 0.9660\n",
            "Epoch 16/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1217 - acc: 0.9435 - val_loss: 0.0649 - val_acc: 0.9630\n",
            "Epoch 17/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1313 - acc: 0.9420 - val_loss: 0.0035 - val_acc: 0.9620\n",
            "Epoch 18/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1292 - acc: 0.9485 - val_loss: 0.1491 - val_acc: 0.9620\n",
            "Epoch 19/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1288 - acc: 0.9440 - val_loss: 0.1169 - val_acc: 0.9540\n",
            "Epoch 20/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1345 - acc: 0.9450 - val_loss: 0.0060 - val_acc: 0.9740\n",
            "Epoch 21/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0986 - acc: 0.9595 - val_loss: 0.0746 - val_acc: 0.9570\n",
            "Epoch 22/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1311 - acc: 0.9470 - val_loss: 0.0014 - val_acc: 0.9690\n",
            "Epoch 23/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1113 - acc: 0.9540 - val_loss: 0.0762 - val_acc: 0.9590\n",
            "Epoch 24/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1266 - acc: 0.9440 - val_loss: 0.1415 - val_acc: 0.9640\n",
            "Epoch 25/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1128 - acc: 0.9560 - val_loss: 0.0347 - val_acc: 0.9780\n",
            "Epoch 26/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1251 - acc: 0.9520 - val_loss: 0.0220 - val_acc: 0.9690\n",
            "Epoch 27/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1127 - acc: 0.9570 - val_loss: 0.0316 - val_acc: 0.9700\n",
            "Epoch 28/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1203 - acc: 0.9515 - val_loss: 0.1840 - val_acc: 0.9610\n",
            "Epoch 29/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0931 - acc: 0.9620 - val_loss: 0.0218 - val_acc: 0.9690\n",
            "Epoch 30/100\n",
            "100/100 [==============================] - 19s 195ms/step - loss: 0.1128 - acc: 0.9555 - val_loss: 0.0058 - val_acc: 0.9660\n",
            "Epoch 31/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1056 - acc: 0.9595 - val_loss: 0.1267 - val_acc: 0.9720\n",
            "Epoch 32/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1001 - acc: 0.9605 - val_loss: 0.0529 - val_acc: 0.9710\n",
            "Epoch 33/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1244 - acc: 0.9485 - val_loss: 0.0354 - val_acc: 0.9620\n",
            "Epoch 34/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0956 - acc: 0.9615 - val_loss: 0.0217 - val_acc: 0.9680\n",
            "Epoch 35/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1088 - acc: 0.9590 - val_loss: 4.8882e-04 - val_acc: 0.9690\n",
            "Epoch 36/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1084 - acc: 0.9580 - val_loss: 0.0785 - val_acc: 0.9630\n",
            "Epoch 37/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1031 - acc: 0.9605 - val_loss: 0.1764 - val_acc: 0.9650\n",
            "Epoch 38/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0998 - acc: 0.9610 - val_loss: 0.1280 - val_acc: 0.9610\n",
            "Epoch 39/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1034 - acc: 0.9615 - val_loss: 0.3110 - val_acc: 0.9630\n",
            "Epoch 40/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0939 - acc: 0.9600 - val_loss: 0.0861 - val_acc: 0.9560\n",
            "Epoch 41/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1081 - acc: 0.9575 - val_loss: 0.0629 - val_acc: 0.9650\n",
            "Epoch 42/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1032 - acc: 0.9545 - val_loss: 0.0673 - val_acc: 0.9600\n",
            "Epoch 43/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0934 - acc: 0.9630 - val_loss: 0.0559 - val_acc: 0.9600\n",
            "Epoch 44/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1245 - acc: 0.9510 - val_loss: 0.0157 - val_acc: 0.9770\n",
            "Epoch 45/100\n",
            "100/100 [==============================] - 20s 196ms/step - loss: 0.1048 - acc: 0.9605 - val_loss: 0.0062 - val_acc: 0.9620\n",
            "Epoch 46/100\n",
            "100/100 [==============================] - 20s 196ms/step - loss: 0.0958 - acc: 0.9620 - val_loss: 0.0382 - val_acc: 0.9670\n",
            "Epoch 47/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1073 - acc: 0.9620 - val_loss: 0.2550 - val_acc: 0.9680\n",
            "Epoch 48/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.1126 - acc: 0.9530 - val_loss: 0.0505 - val_acc: 0.9610\n",
            "Epoch 49/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1083 - acc: 0.9585 - val_loss: 0.0178 - val_acc: 0.9670\n",
            "Epoch 50/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1125 - acc: 0.9530 - val_loss: 0.1640 - val_acc: 0.9640\n",
            "Epoch 51/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0996 - acc: 0.9615 - val_loss: 0.0171 - val_acc: 0.9720\n",
            "Epoch 52/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1215 - acc: 0.9520 - val_loss: 0.0692 - val_acc: 0.9600\n",
            "Epoch 53/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1037 - acc: 0.9580 - val_loss: 8.7379e-04 - val_acc: 0.9640\n",
            "Epoch 54/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.1057 - acc: 0.9590 - val_loss: 0.0056 - val_acc: 0.9690\n",
            "Epoch 55/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0977 - acc: 0.9625 - val_loss: 0.1610 - val_acc: 0.9610\n",
            "Epoch 56/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0976 - acc: 0.9625 - val_loss: 0.0421 - val_acc: 0.9570\n",
            "Epoch 57/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0973 - acc: 0.9645 - val_loss: 0.0548 - val_acc: 0.9630\n",
            "Epoch 58/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0986 - acc: 0.9590 - val_loss: 0.0557 - val_acc: 0.9650\n",
            "Epoch 59/100\n",
            "100/100 [==============================] - 20s 202ms/step - loss: 0.0852 - acc: 0.9675 - val_loss: 0.0065 - val_acc: 0.9600\n",
            "Epoch 60/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0886 - acc: 0.9630 - val_loss: 0.0131 - val_acc: 0.9650\n",
            "Epoch 61/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0831 - acc: 0.9645 - val_loss: 0.1726 - val_acc: 0.9710\n",
            "Epoch 62/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.1197 - acc: 0.9490 - val_loss: 0.1715 - val_acc: 0.9600\n",
            "Epoch 63/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0979 - acc: 0.9620 - val_loss: 0.0034 - val_acc: 0.9670\n",
            "Epoch 64/100\n",
            "100/100 [==============================] - 20s 203ms/step - loss: 0.1033 - acc: 0.9580 - val_loss: 0.5870 - val_acc: 0.9700\n",
            "Epoch 65/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0961 - acc: 0.9620 - val_loss: 0.0027 - val_acc: 0.9590\n",
            "Epoch 66/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0978 - acc: 0.9630 - val_loss: 0.0652 - val_acc: 0.9630\n",
            "Epoch 67/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.1069 - acc: 0.9590 - val_loss: 0.0763 - val_acc: 0.9690\n",
            "Epoch 68/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0950 - acc: 0.9585 - val_loss: 0.0496 - val_acc: 0.9640\n",
            "Epoch 69/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0918 - acc: 0.9630 - val_loss: 0.0754 - val_acc: 0.9570\n",
            "Epoch 70/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0804 - acc: 0.9715 - val_loss: 0.0425 - val_acc: 0.9650\n",
            "Epoch 71/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0907 - acc: 0.9595 - val_loss: 0.0339 - val_acc: 0.9630\n",
            "Epoch 72/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0862 - acc: 0.9660 - val_loss: 0.1673 - val_acc: 0.9720\n",
            "Epoch 73/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0967 - acc: 0.9640 - val_loss: 0.0462 - val_acc: 0.9600\n",
            "Epoch 74/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0879 - acc: 0.9675 - val_loss: 0.0413 - val_acc: 0.9720\n",
            "Epoch 75/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0952 - acc: 0.9640 - val_loss: 0.1698 - val_acc: 0.9660\n",
            "Epoch 76/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0905 - acc: 0.9660 - val_loss: 0.0285 - val_acc: 0.9660\n",
            "Epoch 77/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0843 - acc: 0.9635 - val_loss: 0.0679 - val_acc: 0.9690\n",
            "Epoch 78/100\n",
            "100/100 [==============================] - 20s 197ms/step - loss: 0.0838 - acc: 0.9675 - val_loss: 0.0931 - val_acc: 0.9690\n",
            "Epoch 79/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0831 - acc: 0.9715 - val_loss: 9.2556e-04 - val_acc: 0.9640\n",
            "Epoch 80/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0976 - acc: 0.9610 - val_loss: 0.0028 - val_acc: 0.9630\n",
            "Epoch 81/100\n",
            "100/100 [==============================] - 20s 196ms/step - loss: 0.0958 - acc: 0.9610 - val_loss: 0.0156 - val_acc: 0.9680\n",
            "Epoch 82/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1062 - acc: 0.9580 - val_loss: 0.0589 - val_acc: 0.9550\n",
            "Epoch 83/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0967 - acc: 0.9630 - val_loss: 0.0827 - val_acc: 0.9660\n",
            "Epoch 84/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0841 - acc: 0.9670 - val_loss: 0.0721 - val_acc: 0.9680\n",
            "Epoch 85/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0976 - acc: 0.9620 - val_loss: 0.0688 - val_acc: 0.9540\n",
            "Epoch 86/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0764 - acc: 0.9730 - val_loss: 0.0131 - val_acc: 0.9640\n",
            "Epoch 87/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0842 - acc: 0.9660 - val_loss: 0.1083 - val_acc: 0.9530\n",
            "Epoch 88/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0852 - acc: 0.9665 - val_loss: 0.0987 - val_acc: 0.9620\n",
            "Epoch 89/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0878 - acc: 0.9640 - val_loss: 0.0340 - val_acc: 0.9620\n",
            "Epoch 90/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0929 - acc: 0.9645 - val_loss: 0.2563 - val_acc: 0.9670\n",
            "Epoch 91/100\n",
            "100/100 [==============================] - 21s 207ms/step - loss: 0.0858 - acc: 0.9650 - val_loss: 0.1806 - val_acc: 0.9650\n",
            "Epoch 92/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.1011 - acc: 0.9560 - val_loss: 0.0615 - val_acc: 0.9680\n",
            "Epoch 93/100\n",
            "100/100 [==============================] - 20s 200ms/step - loss: 0.0872 - acc: 0.9665 - val_loss: 0.0254 - val_acc: 0.9700\n",
            "Epoch 94/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0841 - acc: 0.9680 - val_loss: 0.0016 - val_acc: 0.9670\n",
            "Epoch 95/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0871 - acc: 0.9690 - val_loss: 0.0266 - val_acc: 0.9670\n",
            "Epoch 96/100\n",
            "100/100 [==============================] - 20s 201ms/step - loss: 0.0754 - acc: 0.9700 - val_loss: 0.0018 - val_acc: 0.9670\n",
            "Epoch 97/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0877 - acc: 0.9630 - val_loss: 0.1574 - val_acc: 0.9620\n",
            "Epoch 98/100\n",
            "100/100 [==============================] - 20s 198ms/step - loss: 0.0910 - acc: 0.9665 - val_loss: 0.0231 - val_acc: 0.9680\n",
            "Epoch 99/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0962 - acc: 0.9625 - val_loss: 0.0010 - val_acc: 0.9620\n",
            "Epoch 100/100\n",
            "100/100 [==============================] - 20s 199ms/step - loss: 0.0935 - acc: 0.9680 - val_loss: 0.0269 - val_acc: 0.9630\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RgC7U_17G-SU"
      },
      "source": [
        "### 3.3. Show Loss and Accuracy Curves During Training Using Fine-Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UlQObIkiG_3k"
      },
      "source": [
        "Once the new training phase is done, we can then use *Matplotlib* once more to show the loss and accuracy curves."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "M2_mPptmHISp",
        "outputId": "26c60416-0304-4f7c-8341-d6862f3d122a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        }
      },
      "source": [
        "%matplotlib inline\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(1, len(acc) + 1)\n",
        "\n",
        "# Training and Validation Accuracy\n",
        "\n",
        "plt.plot(epochs, acc, 'bo', label='Training Accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation Accuracy')\n",
        "plt.title('Training and Validation Accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\n",
        "\n",
        "# Training and Validation Loss\n",
        "\n",
        "plt.plot(epochs, loss, 'bo', label='Training Loss')\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation Loss')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UGSIfNH6HUEy"
      },
      "source": [
        "### 3.4. Display the Validation Loss and Accuracy"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FGuUJC3bHWxI"
      },
      "source": [
        "Finally, we can simply print out the values for the validation loss and accuracy."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "to1OJqmkHcYR"
      },
      "source": [
        "val_loss, val_acc = model.evaluate_generator(validation_generator, steps=50)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fJrA5HQvHdVK",
        "outputId": "e87bf44b-2251-4eaa-c6fe-f4508e063aee",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print(\"Validation loss:\", val_loss)\n",
        "print(\"Validation accuracy:\", val_acc)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Validation loss: 0.010449341498315334\n",
            "Validation accuracy: 0.9629999995231628\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}