5.3-using-a-pretrained-convnet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "5.3-using-a-pretrained-convnet.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.5.2"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/onietosi/deep-learning-with-python-notebooks/blob/master/5.3-using-a-pretrained-convnet.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o7bqtrFjQ5yL",
        "colab_type": "text"
      },
      "source": [
        "# Preliminaries"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dhbpioVQQYtG",
        "colab_type": "text"
      },
      "source": [
        "## Check runtime is running on GPU and import keras"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hkOYGyKMQWlu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "9f9ed327-3607-42dd-e3e4-e392c6a427d3"
      },
      "source": [
        "import tensorflow as tf\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KxdoHp7CQjBu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        },
        "outputId": "5d711f8c-7bd4-4789-b0b0-7bf7b2d630d0"
      },
      "source": [
        "import keras\n",
        "keras.__version__"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'2.2.4'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qFp2H8iKP5Kt",
        "colab_type": "text"
      },
      "source": [
        "## Download Kaggle dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xYQGUyTWQCcq",
        "colab_type": "text"
      },
      "source": [
        "Upload Kaggle access token kaggle.json."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qwwyTNMvPien",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "mv ./kaggle.json /root/.kaggle/kaggle.json"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r17y1v_mPvnw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 73
        },
        "outputId": "98af7d4b-ce9b-4dee-edf6-c3f86325bb8b"
      },
      "source": [
        "!kaggle -v"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Warning: Your Kaggle API key is readable by other users on this system! To fix this, you can run 'chmod 600 /root/.kaggle/kaggle.json'\n",
            "Kaggle API 1.5.4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8PXBBIWWP1Cb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!chmod 600 /root/.kaggle/kaggle.json"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DMkDZ52IQJMm",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 181
        },
        "outputId": "e59644e9-c776-48da-c482-f992a8676a87"
      },
      "source": [
        "!kaggle competitions download  -c dogs-vs-cats -p /content/kaggle/dogscatsb"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading sampleSubmission.csv to /content/kaggle/dogscats\n",
            "\r  0% 0.00/86.8k [00:00<?, ?B/s]\n",
            "100% 86.8k/86.8k [00:00<00:00, 33.9MB/s]\n",
            "Downloading test1.zip to /content/kaggle/dogscats\n",
            "100% 271M/271M [00:02<00:00, 117MB/s] \n",
            "100% 271M/271M [00:02<00:00, 109MB/s]\n",
            "Downloading train.zip to /content/kaggle/dogscats\n",
            " 97% 528M/543M [00:14<00:00, 33.0MB/s]\n",
            "100% 543M/543M [00:14<00:00, 39.5MB/s]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nLzTYmGOQvwB",
        "colab_type": "text"
      },
      "source": [
        "## Extract data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VD5w3OnkQ9ob",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os, shutil"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vlZu8eN3Q-Ya",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!unzip -q /content/kaggle/dogscats/train.zip -d /content/kaggle/dogscats"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TC8XH4zWRDOh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# The path to the directory where the original\n",
        "# dataset was uncompressed\n",
        "original_dataset_dir = '/content/kaggle/dogscats/train'\n",
        "\n",
        "# The directory where we will\n",
        "# store our smaller dataset\n",
        "base_dir = '/content/kaggle/dogscats/subset'\n",
        "if not os.path.exists(base_dir):\n",
        "  os.mkdir(base_dir)\n",
        "\n",
        "# Directories for our training,\n",
        "# validation and test splits\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "if not os.path.exists(train_dir):\n",
        "  os.mkdir(train_dir)\n",
        "\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "if not os.path.exists(validation_dir):\n",
        "  os.mkdir(validation_dir)\n",
        "\n",
        "test_dir = os.path.join(base_dir, 'test')\n",
        "if not os.path.exists(test_dir):\n",
        "  os.mkdir(test_dir)\n",
        "\n",
        "# Directory with our training cat pictures\n",
        "train_cats_dir = os.path.join(train_dir, 'cats')\n",
        "if not os.path.exists(train_cats_dir):\n",
        "  os.mkdir(train_cats_dir)\n",
        "\n",
        "# Directory with our training dog pictures\n",
        "train_dogs_dir = os.path.join(train_dir, 'dogs')\n",
        "if not os.path.exists(train_dogs_dir):\n",
        "  os.mkdir(train_dogs_dir)\n",
        "\n",
        "# Directory with our validation cat pictures\n",
        "validation_cats_dir = os.path.join(validation_dir, 'cats')\n",
        "if not os.path.exists(validation_cats_dir):\n",
        "  os.mkdir(validation_cats_dir)\n",
        "\n",
        "# Directory with our validation dog pictures\n",
        "validation_dogs_dir = os.path.join(validation_dir, 'dogs')\n",
        "if not os.path.exists(validation_dogs_dir):\n",
        "  os.mkdir(validation_dogs_dir)\n",
        "\n",
        "# Directory with our validation cat pictures\n",
        "test_cats_dir = os.path.join(test_dir, 'cats')\n",
        "if not os.path.exists(test_cats_dir):\n",
        "  os.mkdir(test_cats_dir)\n",
        "\n",
        "# Directory with our validation dog pictures\n",
        "test_dogs_dir = os.path.join(test_dir, 'dogs')\n",
        "if not os.path.exists(test_dogs_dir):\n",
        "  os.mkdir(test_dogs_dir)\n",
        "\n",
        "# Copy first 1000 cat images to train_cats_dir\n",
        "fnames = ['cat.{}.jpg'.format(i) for i in range(1000)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(train_cats_dir, fname)\n",
        "    shutil.copyfile(src, dst)\n",
        "\n",
        "# Copy next 500 cat images to validation_cats_dir\n",
        "fnames = ['cat.{}.jpg'.format(i) for i in range(1000, 1500)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(validation_cats_dir, fname)\n",
        "    shutil.copyfile(src, dst)\n",
        "    \n",
        "# Copy next 500 cat images to test_cats_dir\n",
        "fnames = ['cat.{}.jpg'.format(i) for i in range(1500, 2000)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(test_cats_dir, fname)\n",
        "    shutil.copyfile(src, dst)\n",
        "    \n",
        "# Copy first 1000 dog images to train_dogs_dir\n",
        "fnames = ['dog.{}.jpg'.format(i) for i in range(1000)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(train_dogs_dir, fname)\n",
        "    shutil.copyfile(src, dst)\n",
        "    \n",
        "# Copy next 500 dog images to validation_dogs_dir\n",
        "fnames = ['dog.{}.jpg'.format(i) for i in range(1000, 1500)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(validation_dogs_dir, fname)\n",
        "    shutil.copyfile(src, dst)\n",
        "    \n",
        "# Copy next 500 dog images to test_dogs_dir\n",
        "fnames = ['dog.{}.jpg'.format(i) for i in range(1500, 2000)]\n",
        "for fname in fnames:\n",
        "    src = os.path.join(original_dataset_dir, fname)\n",
        "    dst = os.path.join(test_dogs_dir, fname)\n",
        "    shutil.copyfile(src, dst)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5TxYUZkKGjjF",
        "colab_type": "code",
        "outputId": "b792670d-401e-4ff4-c48f-88e570c3a2a6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total training cat images:', len(os.listdir(train_cats_dir)))"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total training cat images: 1000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NpXWNRBHGjjM",
        "colab_type": "code",
        "outputId": "cdb516a4-a36b-443f-c7b8-ff28dcd334c9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total training dog images:', len(os.listdir(train_dogs_dir)))"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total training dog images: 1000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CZTVoWYXGjjS",
        "colab_type": "code",
        "outputId": "abeacdfd-017c-47a8-bc59-39cec57593f9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total validation cat images:', len(os.listdir(validation_cats_dir)))"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total validation cat images: 500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-3pUgx3CGjja",
        "colab_type": "code",
        "outputId": "e250697e-a138-4500-90c6-6f10949ad92c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total validation dog images:', len(os.listdir(validation_dogs_dir)))"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total validation dog images: 500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4XBnzuJSGjjg",
        "colab_type": "code",
        "outputId": "78c9ff34-755c-4536-d364-b6e8fafe4d54",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total test cat images:', len(os.listdir(test_cats_dir)))"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total test cat images: 500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vUGtNI3UGjjn",
        "colab_type": "code",
        "outputId": "ab0b3395-0180-4377-d117-f548aeda4542",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print('total test dog images:', len(os.listdir(test_dogs_dir)))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total test dog images: 500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9GJHM9UEOwpy",
        "colab_type": "text"
      },
      "source": [
        "# Using a pre-trained convnet\n",
        "\n",
        "This notebook contains the code sample found in Chapter 5, Section 3 of [Deep Learning with Python](https://www.manning.com/books/deep-learning-with-python?a_aid=keras&a_bid=76564dff). Note that the original text features far more content, in particular further explanations and figures: in this notebook, you will only find source code and related comments.\n",
        "\n",
        "----\n",
        "\n",
        "A common and highly effective approach to deep learning on small image datasets is to leverage a pre-trained network. A pre-trained network \n",
        "is simply a saved network previously trained on a large dataset, typically on a large-scale image classification task. If this original \n",
        "dataset is large enough and general enough, then the spatial feature hierarchy learned by the pre-trained network can effectively act as a \n",
        "generic model of our visual world, and hence its features can prove useful for many different computer vision problems, even though these \n",
        "new problems might involve completely different classes from those of the original task. For instance, one might train a network on \n",
        "ImageNet (where classes are mostly animals and everyday objects) and then re-purpose this trained network for something as remote as \n",
        "identifying furniture items in images. Such portability of learned features across different problems is a key advantage of deep learning \n",
        "compared to many older shallow learning approaches, and it makes deep learning very effective for small-data problems.\n",
        "\n",
        "In our case, we will consider a large convnet trained on the ImageNet dataset (1.4 million labeled images and 1000 different classes). \n",
        "ImageNet contains many animal classes, including different species of cats and dogs, and we can thus expect to perform very well on our cat \n",
        "vs. dog classification problem.\n",
        "\n",
        "We will use the VGG16 architecture, developed by Karen Simonyan and Andrew Zisserman in 2014, a simple and widely used convnet architecture \n",
        "for ImageNet. Although it is a bit of an older model, far from the current state of the art and somewhat heavier than many other recent \n",
        "models, we chose it because its architecture is similar to what you are already familiar with, and easy to understand without introducing \n",
        "any new concepts. This may be your first encounter with one of these cutesie model names -- VGG, ResNet, Inception, Inception-ResNet, \n",
        "Xception... you will get used to them, as they will come up frequently if you keep doing deep learning for computer vision.\n",
        "\n",
        "There are two ways to leverage a pre-trained network: *feature extraction* and *fine-tuning*. We will cover both of them. Let's start with \n",
        "feature extraction."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZebgxYSbOwp0",
        "colab_type": "text"
      },
      "source": [
        "## Feature extraction\n",
        "\n",
        "Feature extraction consists of using the representations learned by a previous network to extract interesting features from new samples. \n",
        "These features are then run through a new classifier, which is trained from scratch.\n",
        "\n",
        "As we saw previously, convnets used for image classification comprise two parts: they start with a series of pooling and convolution \n",
        "layers, and they end with a densely-connected classifier. The first part is called the \"convolutional base\" of the model. In the case of \n",
        "convnets, \"feature extraction\" will simply consist of taking the convolutional base of a previously-trained network, running the new data \n",
        "through it, and training a new classifier on top of the output.\n",
        "\n",
        "![swapping FC classifiers](https://s3.amazonaws.com/book.keras.io/img/ch5/swapping_fc_classifier.png)\n",
        "\n",
        "Why only reuse the convolutional base? Could we reuse the densely-connected classifier as well? In general, it should be avoided. The \n",
        "reason is simply that the representations learned by the convolutional base are likely to be more generic and therefore more reusable: the \n",
        "feature maps of a convnet are presence maps of generic concepts over a picture, which is likely to be useful regardless of the computer \n",
        "vision problem at hand. On the other end, the representations learned by the classifier will necessarily be very specific to the set of \n",
        "classes that the model was trained on -- they will only contain information about the presence probability of this or that class in the \n",
        "entire picture. Additionally, representations found in densely-connected layers no longer contain any information about _where_ objects are \n",
        "located in the input image: these layers get rid of the notion of space, whereas the object location is still described by convolutional \n",
        "feature maps. For problems where object location matters, densely-connected features would be largely useless.\n",
        "\n",
        "Note that the level of generality (and therefore reusability) of the representations extracted by specific convolution layers depends on \n",
        "the depth of the layer in the model. Layers that come earlier in the model extract local, highly generic feature maps (such as visual \n",
        "edges, colors, and textures), while layers higher-up extract more abstract concepts (such as \"cat ear\" or \"dog eye\"). So if your new \n",
        "dataset differs a lot from the dataset that the original model was trained on, you may be better off using only the first few layers of the \n",
        "model to do feature extraction, rather than using the entire convolutional base.\n",
        "\n",
        "In our case, since the ImageNet class set did contain multiple dog and cat classes, it is likely that it would be beneficial to reuse the \n",
        "information contained in the densely-connected layers of the original model. However, we will chose not to, in order to cover the more \n",
        "general case where the class set of the new problem does not overlap with the class set of the original model."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ueyJRoszOwp2",
        "colab_type": "text"
      },
      "source": [
        "Let's put this in practice by using the convolutional base of the VGG16 network, trained on ImageNet, to extract interesting features from \n",
        "our cat and dog images, and then training a cat vs. dog classifier on top of these features.\n",
        "\n",
        "The VGG16 model, among others, comes pre-packaged with Keras. You can import it from the `keras.applications` module. Here's the list of \n",
        "image classification models (all pre-trained on the ImageNet dataset) that are available as part of `keras.applications`:\n",
        "\n",
        "* Xception\n",
        "* InceptionV3\n",
        "* ResNet50\n",
        "* VGG16\n",
        "* VGG19\n",
        "* MobileNet\n",
        "\n",
        "Let's instantiate the VGG16 model:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QH3XHJFOOwp4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 310
        },
        "outputId": "cc7a1db5-a286-494b-c29c-1c15395bc650"
      },
      "source": [
        "from keras.applications import VGG16\n",
        "\n",
        "conv_base = VGG16(weights='imagenet',\n",
        "                  include_top=False,\n",
        "                  input_shape=(150, 150, 3))"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0628 08:48:04.130437 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "W0628 08:48:04.135722 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "W0628 08:48:04.144317 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "W0628 08:48:04.181787 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3976: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.1/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
            "58892288/58889256 [==============================] - 1s 0us/step\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0628 08:48:05.497823 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:174: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n",
            "\n",
            "W0628 08:48:05.498902 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:181: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n",
            "\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_ogG9nN_OwqA",
        "colab_type": "text"
      },
      "source": [
        "We passed three arguments to the constructor:\n",
        "\n",
        "* `weights`, to specify which weight checkpoint to initialize the model from\n",
        "* `include_top`, which refers to including or not the densely-connected classifier on top of the network. By default, this \n",
        "densely-connected classifier would correspond to the 1000 classes from ImageNet. Since we intend to use our own densely-connected \n",
        "classifier (with only two classes, cat and dog), we don't need to include it.\n",
        "* `input_shape`, the shape of the image tensors that we will feed to the network. This argument is purely optional: if we don't pass it, \n",
        "then the network will be able to process inputs of any size.\n",
        "\n",
        "Here's the detail of the architecture of the VGG16 convolutional base: it's very similar to the simple convnets that you are already \n",
        "familiar with."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eMUzDvEmOwqC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 835
        },
        "outputId": "2d778dee-186b-41e0-8c30-c50f1a219a7a"
      },
      "source": [
        "conv_base.summary()"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         (None, 150, 150, 3)       0         \n",
            "_________________________________________________________________\n",
            "block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792      \n",
            "_________________________________________________________________\n",
            "block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928     \n",
            "_________________________________________________________________\n",
            "block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0         \n",
            "_________________________________________________________________\n",
            "block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856     \n",
            "_________________________________________________________________\n",
            "block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584    \n",
            "_________________________________________________________________\n",
            "block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0         \n",
            "_________________________________________________________________\n",
            "block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168    \n",
            "_________________________________________________________________\n",
            "block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0         \n",
            "_________________________________________________________________\n",
            "block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160   \n",
            "_________________________________________________________________\n",
            "block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0         \n",
            "_________________________________________________________________\n",
            "block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
            "=================================================================\n",
            "Total params: 14,714,688\n",
            "Trainable params: 14,714,688\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U9aWHlIZOwqK",
        "colab_type": "text"
      },
      "source": [
        "The final feature map has shape `(4, 4, 512)`. That's the feature on top of which we will stick a densely-connected classifier.\n",
        "\n",
        "At this point, there are two ways we could proceed: \n",
        "\n",
        "* Running the convolutional base over our dataset, recording its output to a Numpy array on disk, then using this data as input to a \n",
        "standalone densely-connected classifier similar to those you have seen in the first chapters of this book. This solution is very fast and \n",
        "cheap to run, because it only requires running the convolutional base once for every input image, and the convolutional base is by far the \n",
        "most expensive part of the pipeline. However, for the exact same reason, this technique would not allow us to leverage data augmentation at \n",
        "all.\n",
        "* Extending the model we have (`conv_base`) by adding `Dense` layers on top, and running the whole thing end-to-end on the input data. This \n",
        "allows us to use data augmentation, because every input image is going through the convolutional base every time it is seen by the model. \n",
        "However, for this same reason, this technique is far more expensive than the first one.\n",
        "\n",
        "We will cover both techniques. Let's walk through the code required to set-up the first one: recording the output of `conv_base` on our \n",
        "data and using these outputs as inputs to a new model.\n",
        "\n",
        "We will start by simply running instances of the previously-introduced `ImageDataGenerator` to extract images as Numpy arrays as well as \n",
        "their labels. We will extract features from these images simply by calling the `predict` method of the `conv_base` model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0J-Cb3hUOwqL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        },
        "outputId": "9204ec6f-96e7-4491-f7b2-88e6a346a7b6"
      },
      "source": [
        "import os\n",
        "import numpy as np\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "base_dir = '/content/kaggle/dogscats/subset'\n",
        "\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "test_dir = os.path.join(base_dir, 'test')\n",
        "\n",
        "datagen = ImageDataGenerator(rescale=1./255)\n",
        "batch_size = 20\n",
        "\n",
        "def extract_features(directory, sample_count):\n",
        "    features = np.zeros(shape=(sample_count, 4, 4, 512))\n",
        "    labels = np.zeros(shape=(sample_count))\n",
        "    generator = datagen.flow_from_directory(\n",
        "        directory,\n",
        "        target_size=(150, 150),\n",
        "        batch_size=batch_size,\n",
        "        class_mode='binary')\n",
        "    i = 0\n",
        "    for inputs_batch, labels_batch in generator:\n",
        "        features_batch = conv_base.predict(inputs_batch)\n",
        "        features[i * batch_size : (i + 1) * batch_size] = features_batch\n",
        "        labels[i * batch_size : (i + 1) * batch_size] = labels_batch\n",
        "        i += 1\n",
        "        if i * batch_size >= sample_count:\n",
        "            # Note that since generators yield data indefinitely in a loop,\n",
        "            # we must `break` after every image has been seen once.\n",
        "            break\n",
        "    return features, labels\n",
        "\n",
        "train_features, train_labels = extract_features(train_dir, 2000)\n",
        "validation_features, validation_labels = extract_features(validation_dir, 1000)\n",
        "test_features, test_labels = extract_features(test_dir, 1000)"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c7tmVD24OwqZ",
        "colab_type": "text"
      },
      "source": [
        "The extracted features are currently of shape `(samples, 4, 4, 512)`. We will feed them to a densely-connected classifier, so first we must \n",
        "flatten them to `(samples, 8192)`:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MIHZ2CX5Owqc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_features = np.reshape(train_features, (2000, 4 * 4 * 512))\n",
        "validation_features = np.reshape(validation_features, (1000, 4 * 4 * 512))\n",
        "test_features = np.reshape(test_features, (1000, 4 * 4 * 512))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HvDLUxkROwqo",
        "colab_type": "text"
      },
      "source": [
        "At this point, we can define our densely-connected classifier (note the use of dropout for regularization), and train it on the data and \n",
        "labels that we just recorded:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DgwRKuGbOwqq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "24b416f2-cadf-4274-9e68-5ce27eae76c1"
      },
      "source": [
        "from keras import models\n",
        "from keras import layers\n",
        "from keras import optimizers\n",
        "\n",
        "model = models.Sequential()\n",
        "model.add(layers.Dense(256, activation='relu', input_dim=4 * 4 * 512))\n",
        "model.add(layers.Dropout(0.5))\n",
        "model.add(layers.Dense(1, activation='sigmoid'))\n",
        "\n",
        "model.compile(optimizer=optimizers.RMSprop(lr=2e-5),\n",
        "              loss='binary_crossentropy',\n",
        "              metrics=['acc'])\n",
        "\n",
        "history = model.fit(train_features, train_labels,\n",
        "                    epochs=30,\n",
        "                    batch_size=20,\n",
        "                    validation_data=(validation_features, validation_labels))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "W0628 08:58:11.603560 140577419925376 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "W0628 08:58:11.638642 140577419925376 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n",
            "W0628 08:58:11.647429 140577419925376 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/nn_impl.py:180: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Train on 2000 samples, validate on 1000 samples\n",
            "Epoch 1/30\n",
            "2000/2000 [==============================] - 2s 810us/step - loss: 0.6088 - acc: 0.6595 - val_loss: 0.4535 - val_acc: 0.8220\n",
            "Epoch 2/30\n",
            "2000/2000 [==============================] - 1s 284us/step - loss: 0.4413 - acc: 0.8025 - val_loss: 0.3673 - val_acc: 0.8610\n",
            "Epoch 3/30\n",
            "2000/2000 [==============================] - 1s 279us/step - loss: 0.3668 - acc: 0.8420 - val_loss: 0.3314 - val_acc: 0.8700\n",
            "Epoch 4/30\n",
            "2000/2000 [==============================] - 1s 274us/step - loss: 0.3154 - acc: 0.8785 - val_loss: 0.3047 - val_acc: 0.8840\n",
            "Epoch 5/30\n",
            "2000/2000 [==============================] - 1s 283us/step - loss: 0.2912 - acc: 0.8845 - val_loss: 0.2872 - val_acc: 0.8840\n",
            "Epoch 6/30\n",
            "2000/2000 [==============================] - 1s 279us/step - loss: 0.2736 - acc: 0.8955 - val_loss: 0.2758 - val_acc: 0.8890\n",
            "Epoch 7/30\n",
            "2000/2000 [==============================] - 1s 286us/step - loss: 0.2516 - acc: 0.8985 - val_loss: 0.2802 - val_acc: 0.8850\n",
            "Epoch 8/30\n",
            "2000/2000 [==============================] - 1s 280us/step - loss: 0.2310 - acc: 0.9110 - val_loss: 0.2625 - val_acc: 0.8960\n",
            "Epoch 9/30\n",
            "2000/2000 [==============================] - 1s 292us/step - loss: 0.2176 - acc: 0.9155 - val_loss: 0.2555 - val_acc: 0.8950\n",
            "Epoch 10/30\n",
            "2000/2000 [==============================] - 1s 285us/step - loss: 0.2051 - acc: 0.9250 - val_loss: 0.2669 - val_acc: 0.8870\n",
            "Epoch 11/30\n",
            "2000/2000 [==============================] - 1s 292us/step - loss: 0.1929 - acc: 0.9295 - val_loss: 0.2484 - val_acc: 0.8990\n",
            "Epoch 12/30\n",
            "2000/2000 [==============================] - 1s 284us/step - loss: 0.1876 - acc: 0.9335 - val_loss: 0.2453 - val_acc: 0.8990\n",
            "Epoch 13/30\n",
            "2000/2000 [==============================] - 1s 278us/step - loss: 0.1799 - acc: 0.9345 - val_loss: 0.2452 - val_acc: 0.9010\n",
            "Epoch 14/30\n",
            "2000/2000 [==============================] - 1s 284us/step - loss: 0.1689 - acc: 0.9405 - val_loss: 0.2420 - val_acc: 0.8990\n",
            "Epoch 15/30\n",
            "2000/2000 [==============================] - 1s 279us/step - loss: 0.1629 - acc: 0.9455 - val_loss: 0.2449 - val_acc: 0.9030\n",
            "Epoch 16/30\n",
            "2000/2000 [==============================] - 1s 289us/step - loss: 0.1571 - acc: 0.9400 - val_loss: 0.2397 - val_acc: 0.9040\n",
            "Epoch 17/30\n",
            "2000/2000 [==============================] - 1s 281us/step - loss: 0.1485 - acc: 0.9460 - val_loss: 0.2391 - val_acc: 0.9000\n",
            "Epoch 18/30\n",
            "2000/2000 [==============================] - 1s 280us/step - loss: 0.1440 - acc: 0.9545 - val_loss: 0.2407 - val_acc: 0.9040\n",
            "Epoch 19/30\n",
            "2000/2000 [==============================] - 1s 286us/step - loss: 0.1367 - acc: 0.9570 - val_loss: 0.2405 - val_acc: 0.9050\n",
            "Epoch 20/30\n",
            "2000/2000 [==============================] - 1s 291us/step - loss: 0.1307 - acc: 0.9540 - val_loss: 0.2377 - val_acc: 0.9020\n",
            "Epoch 21/30\n",
            "2000/2000 [==============================] - 1s 286us/step - loss: 0.1290 - acc: 0.9595 - val_loss: 0.2494 - val_acc: 0.9020\n",
            "Epoch 22/30\n",
            "2000/2000 [==============================] - 1s 280us/step - loss: 0.1202 - acc: 0.9585 - val_loss: 0.2372 - val_acc: 0.9020\n",
            "Epoch 23/30\n",
            "2000/2000 [==============================] - 1s 287us/step - loss: 0.1139 - acc: 0.9625 - val_loss: 0.2411 - val_acc: 0.9060\n",
            "Epoch 24/30\n",
            "2000/2000 [==============================] - 1s 283us/step - loss: 0.1169 - acc: 0.9585 - val_loss: 0.2418 - val_acc: 0.9080\n",
            "Epoch 25/30\n",
            "2000/2000 [==============================] - 1s 291us/step - loss: 0.1094 - acc: 0.9650 - val_loss: 0.2374 - val_acc: 0.9010\n",
            "Epoch 26/30\n",
            "2000/2000 [==============================] - 1s 284us/step - loss: 0.1015 - acc: 0.9680 - val_loss: 0.2463 - val_acc: 0.9020\n",
            "Epoch 27/30\n",
            "2000/2000 [==============================] - 1s 279us/step - loss: 0.1017 - acc: 0.9685 - val_loss: 0.2383 - val_acc: 0.9010\n",
            "Epoch 28/30\n",
            "2000/2000 [==============================] - 1s 285us/step - loss: 0.1015 - acc: 0.9695 - val_loss: 0.2425 - val_acc: 0.9020\n",
            "Epoch 29/30\n",
            "2000/2000 [==============================] - 1s 293us/step - loss: 0.0918 - acc: 0.9730 - val_loss: 0.2446 - val_acc: 0.9040\n",
            "Epoch 30/30\n",
            "2000/2000 [==============================] - 1s 292us/step - loss: 0.0908 - acc: 0.9700 - val_loss: 0.2421 - val_acc: 0.9020\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WNdEMlfLOwqz",
        "colab_type": "text"
      },
      "source": [
        "Training is very fast, since we only have to deal with two `Dense` layers -- an epoch takes less than one second even on CPU.\n",
        "\n",
        "Let's take a look at the loss and accuracy curves during training:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NeofRckZOwq2",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "outputId": "d16c3e45-4e08-4633-a49b-634ba91e6eac"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(len(acc))\n",
        "\n",
        "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\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": 25,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEICAYAAACktLTqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIhlIwV1EJAMpuIuIZKD/D+Bx7U9/XdIGAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "93GYmHFcOwq9",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "We reach a validation accuracy of about 90%, much better than what we could achieve in the previous section with our small model trained from \n",
        "scratch. However, our plots also indicate that we are overfitting almost from the start -- despite using dropout with a fairly large rate. \n",
        "This is because this technique does not leverage data augmentation, which is essential to preventing overfitting with small image datasets.\n",
        "\n",
        "Now, let's review the second technique we mentioned for doing feature extraction, which is much slower and more expensive, but which allows \n",
        "us to leverage data augmentation during training: extending the `conv_base` model and running it end-to-end on the inputs. Note that this \n",
        "technique is in fact so expensive that you should only attempt it if you have access to a GPU: it is absolutely intractable on CPU. If you \n",
        "cannot run your code on GPU, then the previous technique is the way to go.\n",
        "\n",
        "Because models behave just like layers, you can add a model (like our `conv_base`) to a `Sequential` model just like you would add a layer. \n",
        "So you can do the following:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4GadeAokOwq9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from keras import models\n",
        "from keras import layers\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": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qsp9QErfOwrC",
        "colab_type": "text"
      },
      "source": [
        "This is what our model looks like now:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C3LsOTIOOwrD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 290
        },
        "outputId": "e5e9bbd8-f617-4fc1-a016-abe4368df314"
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "vgg16 (Model)                (None, 4, 4, 512)         14714688  \n",
            "_________________________________________________________________\n",
            "flatten_2 (Flatten)          (None, 8192)              0         \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 256)               2097408   \n",
            "_________________________________________________________________\n",
            "dense_6 (Dense)              (None, 1)                 257       \n",
            "=================================================================\n",
            "Total params: 16,812,353\n",
            "Trainable params: 2,097,665\n",
            "Non-trainable params: 14,714,688\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HiiR0KSaOwrJ",
        "colab_type": "text"
      },
      "source": [
        "As you can see, the convolutional base of VGG16 has 14,714,688 parameters, which is very large. The classifier we are adding on top has 2 \n",
        "million parameters.\n",
        "\n",
        "Before we compile and train our model, a very important thing to do is to freeze the convolutional base. \"Freezing\" a layer or set of \n",
        "layers means preventing their weights from getting updated during training. If we don't do this, then the representations that were \n",
        "previously learned by the convolutional base would get modified during training. Since the `Dense` layers on top are randomly initialized, \n",
        "very large weight updates would be propagated through the network, effectively destroying the representations previously learned.\n",
        "\n",
        "In Keras, freezing a network is done by setting its `trainable` attribute to `False`:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pZWs5VvEOwrL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 55
        },
        "outputId": "203e702f-17ac-474b-e131-23efc9177523"
      },
      "source": [
        "print('This is the number of trainable weights '\n",
        "      'before freezing the conv base:', len(model.trainable_weights))"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "This is the number of trainable weights before freezing the conv base: 4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i4MMyygdOwrR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "conv_base.trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jm-XPI7cOwrX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "2c5ad90f-a5bb-4faf-9853-2d1e6fc59448"
      },
      "source": [
        "print('This is the number of trainable weights '\n",
        "      'after freezing the conv base:', len(model.trainable_weights))"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "This is the number of trainable weights after freezing the conv base: 4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CJ9ELCd3Owrc",
        "colab_type": "text"
      },
      "source": [
        "With this setup, only the weights from the two `Dense` layers that we added will be trained. That's a total of four weight tensors: two per \n",
        "layer (the main weight matrix and the bias vector). Note that in order for these changes to take effect, we must first compile the model. \n",
        "If you ever modify weight trainability after compilation, you should then re-compile the model, or these changes would be ignored.\n",
        "\n",
        "Now we can start training our model, with the same data augmentation configuration that we used in our previous example:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bUq843lnOwre",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "97da7875-32d5-49c6-ff49-134785d716ee"
      },
      "source": [
        "from keras.preprocessing.image import ImageDataGenerator\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",
        "# Note that the validation data should not be augmented!\n",
        "test_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "        # This is the target directory\n",
        "        train_dir,\n",
        "        # All images will be resized to 150x150\n",
        "        target_size=(150, 150),\n",
        "        batch_size=20,\n",
        "        # Since we use binary_crossentropy loss, we need binary labels\n",
        "        class_mode='binary')\n",
        "\n",
        "validation_generator = test_datagen.flow_from_directory(\n",
        "        validation_dir,\n",
        "        target_size=(150, 150),\n",
        "        batch_size=20,\n",
        "        class_mode='binary')\n",
        "\n",
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer=optimizers.RMSprop(lr=2e-5),\n",
        "              metrics=['acc'])\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",
        "      verbose=2)"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n",
            "Epoch 1/30\n",
            " - 19s - loss: 0.6059 - acc: 0.6750 - val_loss: 0.4557 - val_acc: 0.8370\n",
            "Epoch 2/30\n",
            " - 17s - loss: 0.4886 - acc: 0.7905 - val_loss: 0.3750 - val_acc: 0.8740\n",
            "Epoch 3/30\n",
            " - 17s - loss: 0.4387 - acc: 0.8125 - val_loss: 0.3311 - val_acc: 0.8800\n",
            "Epoch 4/30\n",
            " - 17s - loss: 0.4065 - acc: 0.8285 - val_loss: 0.3063 - val_acc: 0.8820\n",
            "Epoch 5/30\n",
            " - 17s - loss: 0.3950 - acc: 0.8250 - val_loss: 0.2954 - val_acc: 0.8800\n",
            "Epoch 6/30\n",
            " - 17s - loss: 0.3639 - acc: 0.8395 - val_loss: 0.2826 - val_acc: 0.8920\n",
            "Epoch 7/30\n",
            " - 16s - loss: 0.3761 - acc: 0.8315 - val_loss: 0.2759 - val_acc: 0.8960\n",
            "Epoch 8/30\n",
            " - 16s - loss: 0.3526 - acc: 0.8495 - val_loss: 0.2743 - val_acc: 0.8900\n",
            "Epoch 9/30\n",
            " - 17s - loss: 0.3341 - acc: 0.8490 - val_loss: 0.2698 - val_acc: 0.8930\n",
            "Epoch 10/30\n",
            " - 17s - loss: 0.3350 - acc: 0.8575 - val_loss: 0.2572 - val_acc: 0.8960\n",
            "Epoch 11/30\n",
            " - 17s - loss: 0.3322 - acc: 0.8510 - val_loss: 0.2571 - val_acc: 0.8970\n",
            "Epoch 12/30\n",
            " - 17s - loss: 0.3307 - acc: 0.8530 - val_loss: 0.2643 - val_acc: 0.8900\n",
            "Epoch 13/30\n",
            " - 16s - loss: 0.3453 - acc: 0.8525 - val_loss: 0.2651 - val_acc: 0.8890\n",
            "Epoch 14/30\n",
            " - 16s - loss: 0.3312 - acc: 0.8570 - val_loss: 0.2514 - val_acc: 0.8970\n",
            "Epoch 15/30\n",
            " - 17s - loss: 0.3186 - acc: 0.8625 - val_loss: 0.2516 - val_acc: 0.8980\n",
            "Epoch 16/30\n",
            " - 17s - loss: 0.3154 - acc: 0.8595 - val_loss: 0.2521 - val_acc: 0.8980\n",
            "Epoch 17/30\n",
            " - 16s - loss: 0.3150 - acc: 0.8615 - val_loss: 0.2667 - val_acc: 0.8890\n",
            "Epoch 18/30\n",
            " - 16s - loss: 0.3115 - acc: 0.8680 - val_loss: 0.2446 - val_acc: 0.9010\n",
            "Epoch 19/30\n",
            " - 16s - loss: 0.3103 - acc: 0.8655 - val_loss: 0.2477 - val_acc: 0.9010\n",
            "Epoch 20/30\n",
            " - 17s - loss: 0.3059 - acc: 0.8685 - val_loss: 0.2427 - val_acc: 0.9020\n",
            "Epoch 21/30\n",
            " - 16s - loss: 0.3023 - acc: 0.8595 - val_loss: 0.2443 - val_acc: 0.9010\n",
            "Epoch 22/30\n",
            " - 17s - loss: 0.2912 - acc: 0.8800 - val_loss: 0.2458 - val_acc: 0.9000\n",
            "Epoch 23/30\n",
            " - 16s - loss: 0.2915 - acc: 0.8680 - val_loss: 0.2370 - val_acc: 0.9040\n",
            "Epoch 24/30\n",
            " - 16s - loss: 0.2985 - acc: 0.8720 - val_loss: 0.2362 - val_acc: 0.9060\n",
            "Epoch 25/30\n",
            " - 17s - loss: 0.3020 - acc: 0.8755 - val_loss: 0.2406 - val_acc: 0.9060\n",
            "Epoch 26/30\n",
            " - 16s - loss: 0.2869 - acc: 0.8855 - val_loss: 0.2440 - val_acc: 0.8970\n",
            "Epoch 27/30\n",
            " - 17s - loss: 0.2890 - acc: 0.8745 - val_loss: 0.2459 - val_acc: 0.9020\n",
            "Epoch 28/30\n",
            " - 16s - loss: 0.2910 - acc: 0.8785 - val_loss: 0.2372 - val_acc: 0.9060\n",
            "Epoch 29/30\n",
            " - 16s - loss: 0.2854 - acc: 0.8810 - val_loss: 0.2400 - val_acc: 0.9040\n",
            "Epoch 30/30\n",
            " - 17s - loss: 0.2920 - acc: 0.8670 - val_loss: 0.2415 - val_acc: 0.9030\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TFBFTsi5Owrj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.save('cats_and_dogs_small_3.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iaIOqzw1Owro",
        "colab_type": "text"
      },
      "source": [
        "Let's plot our results again:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PF_avxutOwro",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "outputId": "04083044-2f91-49c1-f152-a908ed98c0f5"
      },
      "source": [
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(len(acc))\n",
        "\n",
        "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\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": 43,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nI/8fvl6IjpGTtyAAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAw3bDUwE16nzAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pttb1NHWOwrv",
        "colab_type": "text"
      },
      "source": [
        "As you can see, we reach a validation accuracy of about 96%. This is much better than our small convnet trained from scratch."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bj0tm2VZOwrw",
        "colab_type": "text"
      },
      "source": [
        "## Fine-tuning\n",
        "\n",
        "Another widely used technique for model reuse, complementary to feature extraction, is _fine-tuning_. \n",
        "Fine-tuning consists in unfreezing a few of the top layers \n",
        "of a frozen model base used for feature extraction, and jointly training both the newly added part of the model (in our case, the \n",
        "fully-connected classifier) and these top layers. This is called \"fine-tuning\" because it slightly adjusts the more abstract \n",
        "representations of the model being reused, in order to make them more relevant for the problem at hand.\n",
        "\n",
        "![fine-tuning VGG16](https://s3.amazonaws.com/book.keras.io/img/ch5/vgg16_fine_tuning.png)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eZXLcsekOwrz",
        "colab_type": "text"
      },
      "source": [
        "We have stated before that it was necessary to freeze the convolution base of VGG16 in order to be able to train a randomly initialized \n",
        "classifier on top. For the same reason, it is only possible to fine-tune the top layers of the convolutional base once the classifier on \n",
        "top has already been trained. If the classified wasn't already trained, then the error signal propagating through the network during \n",
        "training would be too large, and the representations previously learned by the layers being fine-tuned would be destroyed. Thus the steps \n",
        "for fine-tuning a network are as follow:\n",
        "\n",
        "* 1) Add your custom network on top of an already trained base network.\n",
        "* 2) Freeze the base network.\n",
        "* 3) Train the part you added.\n",
        "* 4) Unfreeze some layers in the base network.\n",
        "* 5) Jointly train both these layers and the part you added.\n",
        "\n",
        "We have already completed the first 3 steps when doing feature extraction. Let's proceed with the 4th step: we will unfreeze our `conv_base`, \n",
        "and then freeze individual layers inside of it.\n",
        "\n",
        "As a reminder, this is what our convolutional base looks like:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NGTMKsX-Owr0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 835
        },
        "outputId": "0352b67d-a59d-49ec-c831-b1fd50ad36bc"
      },
      "source": [
        "conv_base.summary()"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         (None, 150, 150, 3)       0         \n",
            "_________________________________________________________________\n",
            "block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792      \n",
            "_________________________________________________________________\n",
            "block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928     \n",
            "_________________________________________________________________\n",
            "block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0         \n",
            "_________________________________________________________________\n",
            "block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856     \n",
            "_________________________________________________________________\n",
            "block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584    \n",
            "_________________________________________________________________\n",
            "block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0         \n",
            "_________________________________________________________________\n",
            "block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168    \n",
            "_________________________________________________________________\n",
            "block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0         \n",
            "_________________________________________________________________\n",
            "block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160   \n",
            "_________________________________________________________________\n",
            "block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0         \n",
            "_________________________________________________________________\n",
            "block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808   \n",
            "_________________________________________________________________\n",
            "block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
            "=================================================================\n",
            "Total params: 14,714,688\n",
            "Trainable params: 0\n",
            "Non-trainable params: 14,714,688\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_AhQ7Fy0OwsA",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "We will fine-tune the last 3 convolutional layers, which means that all layers up until `block4_pool` should be frozen, and the layers \n",
        "`block5_conv1`, `block5_conv2` and `block5_conv3` should be trainable.\n",
        "\n",
        "Why not fine-tune more layers? Why not fine-tune the entire convolutional base? We could. However, we need to consider that:\n",
        "\n",
        "* Earlier layers in the convolutional base encode more generic, reusable features, while layers higher up encode more specialized features. It is \n",
        "more useful to fine-tune the more specialized features, as these are the ones that need to be repurposed on our new problem. There would \n",
        "be fast-decreasing returns in fine-tuning lower layers.\n",
        "* The more parameters we are training, the more we are at risk of overfitting. The convolutional base has 15M parameters, so it would be \n",
        "risky to attempt to train it on our small dataset.\n",
        "\n",
        "Thus, in our situation, it is a good strategy to only fine-tune the top 2 to 3 layers in the convolutional base.\n",
        "\n",
        "Let's set this up, starting from where we left off in the previous example:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lpgqqE2lOwsB",
        "colab_type": "code",
        "colab": {}
      },
      "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": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JWbVE6IhOwsE",
        "colab_type": "text"
      },
      "source": [
        "Now we can start fine-tuning our network. We will do this with the RMSprop optimizer, using a very low learning rate. The reason for using \n",
        "a low learning rate is that we want to limit the magnitude of the modifications we make to the representations of the 3 layers that we are \n",
        "fine-tuning. Updates that are too large may harm these representations.\n",
        "\n",
        "Now let's proceed with fine-tuning:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mojSraeoOwsF",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "1ca4b562-8ea4-40d9-8457-5f71f9bcd252"
      },
      "source": [
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer=optimizers.RMSprop(lr=1e-5),\n",
        "              metrics=['acc'])\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": 48,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "100/100 [==============================] - 20s 196ms/step - loss: 0.2804 - acc: 0.8810 - val_loss: 0.2272 - val_acc: 0.9110\n",
            "Epoch 2/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.2522 - acc: 0.8960 - val_loss: 0.2271 - val_acc: 0.9140\n",
            "Epoch 3/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.2281 - acc: 0.9125 - val_loss: 0.1953 - val_acc: 0.9220\n",
            "Epoch 4/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.2176 - acc: 0.9090 - val_loss: 0.2029 - val_acc: 0.9180\n",
            "Epoch 5/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.2014 - acc: 0.9125 - val_loss: 0.1986 - val_acc: 0.9120\n",
            "Epoch 6/100\n",
            "100/100 [==============================] - 18s 182ms/step - loss: 0.1862 - acc: 0.9255 - val_loss: 0.2293 - val_acc: 0.9000\n",
            "Epoch 7/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.1833 - acc: 0.9255 - val_loss: 0.2013 - val_acc: 0.9270\n",
            "Epoch 8/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.1979 - acc: 0.9290 - val_loss: 0.1813 - val_acc: 0.9310\n",
            "Epoch 9/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.1573 - acc: 0.9385 - val_loss: 0.2079 - val_acc: 0.9230\n",
            "Epoch 10/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.1667 - acc: 0.9350 - val_loss: 0.2125 - val_acc: 0.9140\n",
            "Epoch 11/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.1455 - acc: 0.9400 - val_loss: 0.2135 - val_acc: 0.9210\n",
            "Epoch 12/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.1501 - acc: 0.9440 - val_loss: 0.1864 - val_acc: 0.9340\n",
            "Epoch 13/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.1279 - acc: 0.9515 - val_loss: 0.1666 - val_acc: 0.9350\n",
            "Epoch 14/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.1228 - acc: 0.9495 - val_loss: 0.1700 - val_acc: 0.9360\n",
            "Epoch 15/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.1274 - acc: 0.9475 - val_loss: 0.1950 - val_acc: 0.9300\n",
            "Epoch 16/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.1248 - acc: 0.9545 - val_loss: 0.1798 - val_acc: 0.9330\n",
            "Epoch 17/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.1208 - acc: 0.9530 - val_loss: 0.1740 - val_acc: 0.9390\n",
            "Epoch 18/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0965 - acc: 0.9660 - val_loss: 0.1878 - val_acc: 0.9350\n",
            "Epoch 19/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.1015 - acc: 0.9580 - val_loss: 0.1979 - val_acc: 0.9320\n",
            "Epoch 20/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.1026 - acc: 0.9605 - val_loss: 0.1763 - val_acc: 0.9380\n",
            "Epoch 21/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0936 - acc: 0.9640 - val_loss: 0.1683 - val_acc: 0.9400\n",
            "Epoch 22/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0810 - acc: 0.9705 - val_loss: 0.1889 - val_acc: 0.9350\n",
            "Epoch 23/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0907 - acc: 0.9605 - val_loss: 0.2242 - val_acc: 0.9280\n",
            "Epoch 24/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0846 - acc: 0.9700 - val_loss: 0.1731 - val_acc: 0.9360\n",
            "Epoch 25/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0844 - acc: 0.9675 - val_loss: 0.1756 - val_acc: 0.9420\n",
            "Epoch 26/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0809 - acc: 0.9690 - val_loss: 0.2206 - val_acc: 0.9240\n",
            "Epoch 27/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0850 - acc: 0.9635 - val_loss: 0.1916 - val_acc: 0.9360\n",
            "Epoch 28/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0595 - acc: 0.9785 - val_loss: 0.1855 - val_acc: 0.9310\n",
            "Epoch 29/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0750 - acc: 0.9725 - val_loss: 0.2046 - val_acc: 0.9370\n",
            "Epoch 30/100\n",
            "100/100 [==============================] - 17s 174ms/step - loss: 0.0685 - acc: 0.9775 - val_loss: 0.1846 - val_acc: 0.9360\n",
            "Epoch 31/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0616 - acc: 0.9780 - val_loss: 0.2105 - val_acc: 0.9410\n",
            "Epoch 32/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0743 - acc: 0.9730 - val_loss: 0.2167 - val_acc: 0.9270\n",
            "Epoch 33/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0648 - acc: 0.9735 - val_loss: 0.1819 - val_acc: 0.9340\n",
            "Epoch 34/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0489 - acc: 0.9825 - val_loss: 0.1814 - val_acc: 0.9410\n",
            "Epoch 35/100\n",
            "100/100 [==============================] - 17s 174ms/step - loss: 0.0566 - acc: 0.9815 - val_loss: 0.2070 - val_acc: 0.9380\n",
            "Epoch 36/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0607 - acc: 0.9810 - val_loss: 0.3226 - val_acc: 0.9130\n",
            "Epoch 37/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0490 - acc: 0.9810 - val_loss: 0.2624 - val_acc: 0.9310\n",
            "Epoch 38/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0554 - acc: 0.9780 - val_loss: 0.2016 - val_acc: 0.9380\n",
            "Epoch 39/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0528 - acc: 0.9805 - val_loss: 0.2032 - val_acc: 0.9340\n",
            "Epoch 40/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0590 - acc: 0.9820 - val_loss: 0.1953 - val_acc: 0.9390\n",
            "Epoch 41/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0432 - acc: 0.9865 - val_loss: 0.2054 - val_acc: 0.9330\n",
            "Epoch 42/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0594 - acc: 0.9790 - val_loss: 0.1989 - val_acc: 0.9400\n",
            "Epoch 43/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0425 - acc: 0.9835 - val_loss: 0.2283 - val_acc: 0.9340\n",
            "Epoch 44/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0438 - acc: 0.9830 - val_loss: 0.2367 - val_acc: 0.9290\n",
            "Epoch 45/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0454 - acc: 0.9825 - val_loss: 0.2179 - val_acc: 0.9370\n",
            "Epoch 46/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0407 - acc: 0.9845 - val_loss: 0.2675 - val_acc: 0.9300\n",
            "Epoch 47/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0506 - acc: 0.9830 - val_loss: 0.1838 - val_acc: 0.9400\n",
            "Epoch 48/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0414 - acc: 0.9830 - val_loss: 0.1912 - val_acc: 0.9420\n",
            "Epoch 49/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0385 - acc: 0.9850 - val_loss: 0.2005 - val_acc: 0.9410\n",
            "Epoch 50/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0445 - acc: 0.9845 - val_loss: 0.2375 - val_acc: 0.9390\n",
            "Epoch 51/100\n",
            "100/100 [==============================] - 18s 181ms/step - loss: 0.0350 - acc: 0.9870 - val_loss: 0.2278 - val_acc: 0.9300\n",
            "Epoch 52/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0385 - acc: 0.9885 - val_loss: 0.1990 - val_acc: 0.9470\n",
            "Epoch 53/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0407 - acc: 0.9875 - val_loss: 0.2161 - val_acc: 0.9440\n",
            "Epoch 54/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0312 - acc: 0.9900 - val_loss: 0.2056 - val_acc: 0.9410\n",
            "Epoch 55/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0413 - acc: 0.9830 - val_loss: 0.1956 - val_acc: 0.9400\n",
            "Epoch 56/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0296 - acc: 0.9915 - val_loss: 0.2658 - val_acc: 0.9330\n",
            "Epoch 57/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0353 - acc: 0.9890 - val_loss: 0.4432 - val_acc: 0.9050\n",
            "Epoch 58/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0333 - acc: 0.9885 - val_loss: 0.2139 - val_acc: 0.9430\n",
            "Epoch 59/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0294 - acc: 0.9900 - val_loss: 0.2790 - val_acc: 0.9340\n",
            "Epoch 60/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0313 - acc: 0.9880 - val_loss: 0.2630 - val_acc: 0.9420\n",
            "Epoch 61/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0344 - acc: 0.9860 - val_loss: 0.6541 - val_acc: 0.8910\n",
            "Epoch 62/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0279 - acc: 0.9905 - val_loss: 0.2677 - val_acc: 0.9390\n",
            "Epoch 63/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0250 - acc: 0.9920 - val_loss: 0.2353 - val_acc: 0.9420\n",
            "Epoch 64/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0216 - acc: 0.9905 - val_loss: 0.3011 - val_acc: 0.9290\n",
            "Epoch 65/100\n",
            "100/100 [==============================] - 17s 174ms/step - loss: 0.0303 - acc: 0.9890 - val_loss: 0.2491 - val_acc: 0.9420\n",
            "Epoch 66/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0255 - acc: 0.9910 - val_loss: 0.2297 - val_acc: 0.9330\n",
            "Epoch 67/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0340 - acc: 0.9890 - val_loss: 0.2039 - val_acc: 0.9450\n",
            "Epoch 68/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0234 - acc: 0.9920 - val_loss: 0.2480 - val_acc: 0.9400\n",
            "Epoch 69/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0263 - acc: 0.9890 - val_loss: 0.2438 - val_acc: 0.9410\n",
            "Epoch 70/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0345 - acc: 0.9875 - val_loss: 0.2306 - val_acc: 0.9360\n",
            "Epoch 71/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0311 - acc: 0.9900 - val_loss: 0.2306 - val_acc: 0.9400\n",
            "Epoch 72/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0270 - acc: 0.9905 - val_loss: 0.3882 - val_acc: 0.9220\n",
            "Epoch 73/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0277 - acc: 0.9910 - val_loss: 0.2399 - val_acc: 0.9400\n",
            "Epoch 74/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0238 - acc: 0.9910 - val_loss: 0.2660 - val_acc: 0.9390\n",
            "Epoch 75/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0211 - acc: 0.9925 - val_loss: 0.3789 - val_acc: 0.9300\n",
            "Epoch 76/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0193 - acc: 0.9925 - val_loss: 0.2504 - val_acc: 0.9370\n",
            "Epoch 77/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0208 - acc: 0.9910 - val_loss: 0.3572 - val_acc: 0.9270\n",
            "Epoch 78/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0283 - acc: 0.9920 - val_loss: 0.2532 - val_acc: 0.9430\n",
            "Epoch 79/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0268 - acc: 0.9940 - val_loss: 0.2639 - val_acc: 0.9390\n",
            "Epoch 80/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0177 - acc: 0.9940 - val_loss: 0.3257 - val_acc: 0.9310\n",
            "Epoch 81/100\n",
            "100/100 [==============================] - 17s 174ms/step - loss: 0.0231 - acc: 0.9920 - val_loss: 0.2927 - val_acc: 0.9370\n",
            "Epoch 82/100\n",
            "100/100 [==============================] - 18s 178ms/step - loss: 0.0281 - acc: 0.9885 - val_loss: 0.2619 - val_acc: 0.9380\n",
            "Epoch 83/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0249 - acc: 0.9940 - val_loss: 0.3143 - val_acc: 0.9350\n",
            "Epoch 84/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0204 - acc: 0.9910 - val_loss: 0.2660 - val_acc: 0.9360\n",
            "Epoch 85/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0246 - acc: 0.9920 - val_loss: 0.2232 - val_acc: 0.9430\n",
            "Epoch 86/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0230 - acc: 0.9905 - val_loss: 0.2428 - val_acc: 0.9470\n",
            "Epoch 87/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0209 - acc: 0.9930 - val_loss: 0.2560 - val_acc: 0.9450\n",
            "Epoch 88/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0214 - acc: 0.9920 - val_loss: 0.2629 - val_acc: 0.9390\n",
            "Epoch 89/100\n",
            "100/100 [==============================] - 17s 174ms/step - loss: 0.0126 - acc: 0.9960 - val_loss: 0.2848 - val_acc: 0.9370\n",
            "Epoch 90/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0161 - acc: 0.9940 - val_loss: 0.2915 - val_acc: 0.9400\n",
            "Epoch 91/100\n",
            "100/100 [==============================] - 18s 179ms/step - loss: 0.0256 - acc: 0.9900 - val_loss: 0.2404 - val_acc: 0.9370\n",
            "Epoch 92/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0198 - acc: 0.9935 - val_loss: 0.3244 - val_acc: 0.9310\n",
            "Epoch 93/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0129 - acc: 0.9945 - val_loss: 0.2344 - val_acc: 0.9470\n",
            "Epoch 94/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0350 - acc: 0.9915 - val_loss: 0.2315 - val_acc: 0.9420\n",
            "Epoch 95/100\n",
            "100/100 [==============================] - 17s 175ms/step - loss: 0.0257 - acc: 0.9915 - val_loss: 0.5066 - val_acc: 0.9070\n",
            "Epoch 96/100\n",
            "100/100 [==============================] - 18s 177ms/step - loss: 0.0201 - acc: 0.9920 - val_loss: 0.2769 - val_acc: 0.9410\n",
            "Epoch 97/100\n",
            "100/100 [==============================] - 18s 180ms/step - loss: 0.0247 - acc: 0.9940 - val_loss: 0.3260 - val_acc: 0.9300\n",
            "Epoch 98/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0215 - acc: 0.9900 - val_loss: 0.2563 - val_acc: 0.9440\n",
            "Epoch 99/100\n",
            "100/100 [==============================] - 18s 175ms/step - loss: 0.0172 - acc: 0.9945 - val_loss: 0.2726 - val_acc: 0.9410\n",
            "Epoch 100/100\n",
            "100/100 [==============================] - 18s 176ms/step - loss: 0.0090 - acc: 0.9975 - val_loss: 0.2339 - val_acc: 0.9420\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l5WFnvA1OwsK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.save('cats_and_dogs_small_4.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lthExw6rOwsM",
        "colab_type": "text"
      },
      "source": [
        "Let's plot our results using the same plotting code as before:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YPZLJS-KOwsM",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "outputId": "6bf7790f-8347-4019-9794-083e6943e9e7"
      },
      "source": [
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(len(acc))\n",
        "\n",
        "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\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": 49,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEICAYAAACktLTqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n3BVFURKQ/wcMgTj7jtGcpwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "v-T8nfo1OwsS",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "These curves look very noisy. To make them more readable, we can smooth them by replacing every loss and accuracy with exponential moving \n",
        "averages of these quantities. Here's a trivial utility function to do this:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xKaxGmv8OwsT",
        "colab_type": "code",
        "colab": {},
        "outputId": "aa499349-2ca6-472e-ce97-b41220303da6"
      },
      "source": [
        "def smooth_curve(points, factor=0.8):\n",
        "  smoothed_points = []\n",
        "  for point in points:\n",
        "    if smoothed_points:\n",
        "      previous = smoothed_points[-1]\n",
        "      smoothed_points.append(previous * factor + point * (1 - factor))\n",
        "    else:\n",
        "      smoothed_points.append(point)\n",
        "  return smoothed_points\n",
        "\n",
        "plt.plot(epochs,\n",
        "         smooth_curve(acc), 'bo', label='Smoothed training acc')\n",
        "plt.plot(epochs,\n",
        "         smooth_curve(val_acc), 'b', label='Smoothed validation acc')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "\n",
        "plt.figure()\n",
        "\n",
        "plt.plot(epochs,\n",
        "         smooth_curve(loss), 'bo', label='Smoothed training loss')\n",
        "plt.plot(epochs,\n",
        "         smooth_curve(val_loss), 'b', label='Smoothed validation loss')\n",
        "plt.title('Training and validation loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEICAYAAABfz4NwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fd53cc076d8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEICAYAAACzliQjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ny6VyEc4AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7fd53cbf15c0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UYGgW-ZEOwsZ",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "These curves look much cleaner and more stable. We are seeing a nice 1% absolute improvement.\n",
        "\n",
        "Note that the loss curve does not show any real improvement (in fact, it is deteriorating). You may wonder, how could accuracy improve if the \n",
        "loss isn't decreasing? The answer is simple: what we display is an average of pointwise loss values, but what actually matters for accuracy \n",
        "is the distribution of the loss values, not their average, since accuracy is the result of a binary thresholding of the class probability \n",
        "predicted by the model. The model may still be improving even if this isn't reflected in the average loss.\n",
        "\n",
        "We can now finally evaluate this model on the test data:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mN7GNgkqOwsb",
        "colab_type": "code",
        "colab": {},
        "outputId": "f79996ef-49aa-4d9d-fd6d-85121992a5bc"
      },
      "source": [
        "test_generator = test_datagen.flow_from_directory(\n",
        "        test_dir,\n",
        "        target_size=(150, 150),\n",
        "        batch_size=20,\n",
        "        class_mode='binary')\n",
        "\n",
        "test_loss, test_acc = model.evaluate_generator(test_generator, steps=50)\n",
        "print('test acc:', test_acc)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 1000 images belonging to 2 classes.\n",
            "test acc: 0.967999992371\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5Fa_TEjlOwse",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "Here we get a test accuracy of 97%. In the original Kaggle competition around this dataset, this would have been one of the top results. \n",
        "However, using modern deep learning techniques, we managed to reach this result using only a very small fraction of the training data \n",
        "available (about 10%). There is a huge difference between being able to train on 20,000 samples compared to 2,000 samples!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2uYs66DPOwse",
        "colab_type": "text"
      },
      "source": [
        "## Take-aways: using convnets with small datasets\n",
        "\n",
        "Here's what you should take away from the exercises of these past two sections:\n",
        "\n",
        "* Convnets are the best type of machine learning models for computer vision tasks. It is possible to train one from scratch even on a very \n",
        "small dataset, with decent results.\n",
        "* On a small dataset, overfitting will be the main issue. Data augmentation is a powerful way to fight overfitting when working with image \n",
        "data.\n",
        "* It is easy to reuse an existing convnet on a new dataset, via feature extraction. This is a very valuable technique for working with \n",
        "small image datasets.\n",
        "* As a complement to feature extraction, one may use fine-tuning, which adapts to a new problem some of the representations previously \n",
        "learned by an existing model. This pushes performance a bit further.\n",
        "\n",
        "Now you have a solid set of tools for dealing with image classification problems, in particular with small datasets."
      ]
    }
  ]
}