image_classification_part3.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "image_classification_part3.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [
        "jTEzoMx6CasV"
      ]
    },
    "kernelspec": {
      "display_name": "Python 2",
      "name": "python2"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "[View in Colaboratory](https://colab.research.google.com/github/FaboNo/CNN-learning/blob/master/image_classification_part3.ipynb)"
      ]
    },
    {
      "metadata": {
        "id": "jTEzoMx6CasV",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#### Copyright 2018 Google LLC."
      ]
    },
    {
      "metadata": {
        "id": "IhmPj1VVCfWb",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Licensed under the Apache License, Version 2.0 (the \"License\");\n",
        "# you may not use this file except in compliance with the License.\n",
        "# You may obtain a copy of the License at\n",
        "#\n",
        "# https://www.apache.org/licenses/LICENSE-2.0\n",
        "#\n",
        "# Unless required by applicable law or agreed to in writing, software\n",
        "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
        "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
        "# See the License for the specific language governing permissions and\n",
        "# limitations under the License."
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "YHK6DyunSbs4",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Cat vs. Dog Image Classification\n",
        "## Exercise 3: Feature Extraction and Fine-Tuning\n",
        "**_Estimated completion time: 30 minutes_**\n",
        "\n",
        "In Exercise 1, we built a convnet from scratch, and were able to achieve an accuracy of about 70%. With the addition of data augmentation and dropout in Exercise 2, we were able to increase accuracy to about 80%. That seems decent, but 20% is still too high of an error rate. Maybe we just don't have enough training data available to properly solve the problem. What other approaches can we try?\n",
        "\n",
        "In this exercise, we'll look at two techniques for repurposing feature data generated from image models that have already been trained on large sets of data, **feature extraction** and **fine tuning**, and use them to improve the accuracy of our cat vs. dog classification model."
      ]
    },
    {
      "metadata": {
        "id": "dI5rmt4UBwXs",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Feature Extraction Using a Pretrained Model\n",
        "\n",
        "One thing that is commonly done in computer vision is to take a model trained on a very large dataset, run it on your own, smaller dataset, and extract the intermediate representations (features) that the model generates. These representations are frequently informative for your own computer vision task, even though the task may be quite different from the problem that the original model was trained on. This versatility and repurposability of convnets is one of the most interesting aspects of deep learning.\n",
        "\n",
        "In our case, we will use the [Inception V3 model](https://arxiv.org/abs/1512.00567) developed at Google, and pre-trained on [ImageNet](http://image-net.org/), a large dataset of web images (1.4M images and 1000 classes). This is a powerful model; let's see what the features that it has learned can do for our cat vs. dog problem.\n",
        "\n",
        "First, we need to pick which intermediate layer of Inception V3 we will use for feature extraction. A common practice is to use the output of the very last layer before the `Flatten` operation, the so-called \"bottleneck layer.\" The reasoning here is that the following fully connected layers will be too specialized for the task the network was trained on, and thus the features learned by these layers won't be very useful for a new task. The bottleneck features, however, retain much generality.\n",
        "\n",
        "Let's instantiate an Inception V3 model preloaded with weights trained on ImageNet:\n"
      ]
    },
    {
      "metadata": {
        "id": "1xJZ5glPPCRz",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3f647382-18b0-4e5c-dd1b-5d914ba96a13"
      },
      "cell_type": "code",
      "source": [
        "import os\n",
        "import tensorflow as tf\n",
        "\n",
        "from keras.applications.inception_v3 import InceptionV3\n",
        "from keras import layers\n",
        "from keras.models import Model\n",
        "from keras.optimizers import RMSprop\n",
        "from keras import backend as K\n",
        "\n",
        "# Configure the TF backend session\n",
        "tf_config = tf.ConfigProto(\n",
        "    gpu_options=tf.GPUOptions(allow_growth=True))\n",
        "K.set_session(tf.Session(config=tf_config))"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "VaXLMtYiF0t9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Now let's download the weights:"
      ]
    },
    {
      "metadata": {
        "id": "KMrbllgAFipZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "f64bfa86-ef21-45c5-807d-6cfecc747248"
      },
      "cell_type": "code",
      "source": [
        "!wget --no-check-certificate \\\n",
        "    https://storage.googleapis.com/mledu-datasets/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 \\\n",
        "    -O /tmp/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2018-05-31 21:05:09--  https://storage.googleapis.com/mledu-datasets/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5\r\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 74.125.141.128, 2607:f8b0:400c:c06::80\r\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|74.125.141.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 87910968 (84M) [application/x-hdf]\n",
            "Saving to: ‘/tmp/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5’\n",
            "\n",
            "/tmp/inception_v3_w 100%[===================>]  83.84M  92.4MB/s    in 0.9s    \n",
            "\n",
            "2018-05-31 21:05:10 (92.4 MB/s) - ‘/tmp/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5’ saved [87910968/87910968]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "UnRiGBfOF8rq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "7d71a14b-2845-4691-ae8d-38abb8d73df5"
      },
      "cell_type": "code",
      "source": [
        "local_weights_file = '/tmp/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5'\n",
        "pre_trained_model = InceptionV3(\n",
        "    input_shape=(150, 150, 3), include_top=False, weights=None)\n",
        "pre_trained_model.load_weights(local_weights_file)"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "metadata": {
        "id": "IcYZPBS3bTAj",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "By specifying the `include_top=False` argument, we load a network that doesn't include the classification layers at the top—ideal for feature extraction."
      ]
    },
    {
      "metadata": {
        "id": "CFxrqTuJee5m",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Let's make the model non-trainable, since we will only use it for feature extraction; we won't update the weights of the pretrained model during training."
      ]
    },
    {
      "metadata": {
        "id": "a38rB3lyedcB",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "f3e8bd9b-ec47-4765-ac14-42b4c73bad2f"
      },
      "cell_type": "code",
      "source": [
        "for layer in pre_trained_model.layers:\n",
        "  layer.trainable = False"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "metadata": {
        "id": "XGBGDiOAepnO",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "The layer we will use for feature extraction in Inception v3 is called `mixed7`. It is not the bottleneck of the network, but we are using it to keep a sufficiently large feature map (7x7 in this case). (Using the bottleneck layer would have resulting in a 3x3 feature map, which is a bit small.) Let's get the output from `mixed7`:"
      ]
    },
    {
      "metadata": {
        "id": "Cj4rXshqbQlS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "415e6a20-7498-4f5f-84ee-423387edb93b"
      },
      "cell_type": "code",
      "source": [
        "last_layer = pre_trained_model.get_layer('mixed7')\n",
        "print 'last layer output shape:', last_layer.output_shape\n",
        "last_output = last_layer.output"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "last layer output shape: (None, 7, 7, 768)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "XxHk6XQLeUWh",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Now let's stick a fully connected classifier on top of `last_output`:"
      ]
    },
    {
      "metadata": {
        "id": "BMXb913pbvFg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "8647319f-21f6-4766-a532-9979170328a4"
      },
      "cell_type": "code",
      "source": [
        "# Flatten the output layer to 1 dimension\n",
        "x = layers.Flatten()(last_output)\n",
        "# Add a fully connected layer with 1,024 hidden units and ReLU activation\n",
        "x = layers.Dense(1024, activation='relu')(x)\n",
        "# Add a dropout rate of 0.2\n",
        "x = layers.Dropout(0.2)(x)\n",
        "# Add a final sigmoid layer for classification\n",
        "x = layers.Dense(1, activation='sigmoid')(x)\n",
        "\n",
        "# Configure and compile the model\n",
        "model = Model(pre_trained_model.input, x)\n",
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer=RMSprop(lr=0.0001),\n",
        "              metrics=['acc'])"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_6ECjowwV5Ug",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "For examples and data preprocessing, let's use the same files and `train_generator` as we did in Exercise 2."
      ]
    },
    {
      "metadata": {
        "id": "Cl-IqOTjZVw_",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "**NOTE:** The 2,000 images used in this exercise are excerpted from the [\"Dogs vs. Cats\" dataset](https://www.kaggle.com/c/dogs-vs-cats/data) available on Kaggle, which contains 25,000 images. Here, we use a subset of the full dataset to decrease training time for educational purposes."
      ]
    },
    {
      "metadata": {
        "id": "O4s8HckqGlnb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "9c1f7288-e593-418f-bec1-bea75ccf7db2"
      },
      "cell_type": "code",
      "source": [
        "!wget --no-check-certificate \\\n",
        "   https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip -O \\\n",
        "   /tmp/cats_and_dogs_filtered.zip"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2018-05-31 21:07:49--  https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip\r\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 74.125.141.128, 2607:f8b0:400c:c06::80\r\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|74.125.141.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 68606236 (65M) [application/zip]\n",
            "Saving to: ‘/tmp/cats_and_dogs_filtered.zip’\n",
            "\n",
            "/tmp/cats_and_dogs_ 100%[===================>]  65.43M   105MB/s    in 0.6s    \n",
            "\n",
            "2018-05-31 21:07:50 (105 MB/s) - ‘/tmp/cats_and_dogs_filtered.zip’ saved [68606236/68606236]\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Fl9XXARuV_eg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "fa2e4259-d86f-452b-b71f-8d7532dc5d5c"
      },
      "cell_type": "code",
      "source": [
        "import os\n",
        "import zipfile\n",
        "\n",
        "local_zip = '/tmp/cats_and_dogs_filtered.zip'\n",
        "zip_ref = zipfile.ZipFile(local_zip, 'r')\n",
        "zip_ref.extractall('/tmp')\n",
        "zip_ref.close()\n",
        "\n",
        "# Define our example directories and files\n",
        "base_dir = '/tmp/cats_and_dogs_filtered'\n",
        "train_dir = os.path.join(base_dir, 'train')\n",
        "validation_dir = os.path.join(base_dir, 'validation')\n",
        "\n",
        "# Directory with our training cat pictures\n",
        "train_cats_dir = os.path.join(train_dir, 'cats')\n",
        "\n",
        "# Directory with our training dog pictures\n",
        "train_dogs_dir = os.path.join(train_dir, 'dogs')\n",
        "\n",
        "# Directory with our validation cat pictures\n",
        "validation_cats_dir = os.path.join(validation_dir, 'cats')\n",
        "\n",
        "# Directory with our validation dog pictures\n",
        "validation_dogs_dir = os.path.join(validation_dir, 'dogs')\n",
        "\n",
        "train_cat_fnames = os.listdir(train_cats_dir)\n",
        "train_dog_fnames = os.listdir(train_dogs_dir)\n",
        "\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "# Add our data-augmentation parameters to ImageDataGenerator\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",
        "\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",
        "        train_dir, # This is the source directory for training images\n",
        "        target_size=(150, 150),  # All images will be resized to 150x150\n",
        "        batch_size=20,\n",
        "        # Since we use binary_crossentropy loss, we need binary labels\n",
        "        class_mode='binary')\n",
        "\n",
        "# Flow validation images in batches of 20 using test_datagen generator\n",
        "validation_generator = test_datagen.flow_from_directory(\n",
        "        validation_dir,\n",
        "        target_size=(150, 150),\n",
        "        batch_size=20,\n",
        "        class_mode='binary')"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "qEC1AL7iVRLz",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Finally, let's train the model using the features we extracted. We'll train on all 2000 images available, for 2 epochs, and validate on all 1,000 test images."
      ]
    },
    {
      "metadata": {
        "id": "Blhq2MAUeyGA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 153
        },
        "outputId": "4d18117e-59c1-4c94-b890-30c3b077a1ef"
      },
      "cell_type": "code",
      "source": [
        "history = model.fit_generator(\n",
        "      train_generator,\n",
        "      steps_per_epoch=100,\n",
        "      epochs=4,\n",
        "      validation_data=validation_generator,\n",
        "      validation_steps=50,\n",
        "      verbose=2)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/4\n",
            " - 25s - loss: 0.3398 - acc: 0.8535 - val_loss: 0.2230 - val_acc: 0.9560\n",
            "Epoch 2/4\n",
            " - 23s - loss: 0.3146 - acc: 0.8590 - val_loss: 0.3290 - val_acc: 0.9420\n",
            "Epoch 3/4\n",
            " - 23s - loss: 0.3202 - acc: 0.8610 - val_loss: 0.5268 - val_acc: 0.9170\n",
            "Epoch 4/4\n",
            " - 23s - loss: 0.3164 - acc: 0.8725 - val_loss: 0.3813 - val_acc: 0.9390\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "lRjyAkE62aOG",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "You can see that we reach a validation accuracy of 88–90% very quickly. This is much better than the small model we trained from scratch."
      ]
    },
    {
      "metadata": {
        "id": "tt15y6IS2pBo",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Further Improving Accuracy with Fine-Tuning\n",
        "\n",
        "In our feature-extraction experiment, we only tried adding two classification layers on top of an Inception V3 layer. The weights of the pretrained network were not updated during training. One way to increase performance even further is to \"fine-tune\" the weights of the top layers of the pretrained model alongside the training of the top-level classifier. A couple of important notes on fine-tuning:\n",
        "\n",
        "- **Fine-tuning should only be attempted *after* you have trained the top-level classifier with the pretrained model set to non-trainable**. If you add a randomly initialized classifier on top of a pretrained model and attempt to train all layers jointly, the magnitude of the gradient updates will be too large (due to the random weights from the classifier), and your pretrained model will just forget everything it has learned.\n",
        "- Additionally, we **fine-tune only the *top layers* of the pre-trained model** rather than all layers of the pretrained model because, in a convnet, the higher up a layer is, the more specialized it is. The first few layers in a convnet learn very simple and generic features, which generalize to almost all types of images. But as you go higher up, the features are increasingly specific to the dataset that the model is trained on. The goal of fine-tuning is to adapt these specialized features to work with the new dataset.\n",
        "\n",
        "All we need to do to implement fine-tuning is to set the top layers of Inception V3 to be trainable, recompile the model (necessary for these changes to take effect), and resume training. Let's unfreeze all layers belonging to the `mixed7` module—i.e., all layers found after `mixed6`—and recompile the model:"
      ]
    },
    {
      "metadata": {
        "id": "_l_J4S0Z2rgg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "63346036-e577-4b6f-8ab3-2af7fda5f2ec"
      },
      "cell_type": "code",
      "source": [
        "unfreeze = False\n",
        "\n",
        "# Unfreeze all models after \"mixed6\"\n",
        "for layer in pre_trained_model.layers:\n",
        "  if unfreeze:\n",
        "    layer.trainable = True\n",
        "  if layer.name == 'mixed6':\n",
        "    unfreeze = True\n",
        "\n",
        "from keras.optimizers import SGD\n",
        "\n",
        "# As an optimizer, here we will use SGD \n",
        "# with a very low learning rate (0.00001)\n",
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer=SGD(lr=0.00001, momentum=0.9),\n",
        "              metrics=['acc'])"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "metadata": {
        "id": "zE37ARlqY9da",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Now let's retrain the model. We'll train on all 2000 images available, for 50 epochs, and validate on all 1,000 test images. (This may take 15-20 minutes to run.)"
      ]
    },
    {
      "metadata": {
        "id": "o_GgDGG4Y_hJ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1717
        },
        "outputId": "9d735460-c4bb-42a2-87d7-d7e1a97e642a"
      },
      "cell_type": "code",
      "source": [
        "history = model.fit_generator(\n",
        "      train_generator,\n",
        "      steps_per_epoch=100,\n",
        "      epochs=50,\n",
        "      validation_data=validation_generator,\n",
        "      validation_steps=50,\n",
        "      verbose=2)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            " - 28s - loss: 0.2495 - acc: 0.8915 - val_loss: 0.3911 - val_acc: 0.9420\n",
            "Epoch 2/50\n",
            " - 23s - loss: 0.2504 - acc: 0.8900 - val_loss: 0.3703 - val_acc: 0.9440\n",
            "Epoch 3/50\n",
            " - 24s - loss: 0.2547 - acc: 0.8960 - val_loss: 0.3662 - val_acc: 0.9440\n",
            "Epoch 4/50\n",
            " - 23s - loss: 0.2555 - acc: 0.8895 - val_loss: 0.3569 - val_acc: 0.9450\n",
            "Epoch 5/50\n",
            " - 24s - loss: 0.2462 - acc: 0.8960 - val_loss: 0.3538 - val_acc: 0.9450\n",
            "Epoch 6/50\n",
            " - 23s - loss: 0.2329 - acc: 0.8975 - val_loss: 0.3543 - val_acc: 0.9450\n",
            "Epoch 7/50\n",
            " - 24s - loss: 0.2416 - acc: 0.9025 - val_loss: 0.3488 - val_acc: 0.9450\n",
            "Epoch 8/50\n",
            " - 23s - loss: 0.2487 - acc: 0.8905 - val_loss: 0.3486 - val_acc: 0.9450\n",
            "Epoch 9/50\n",
            " - 24s - loss: 0.2509 - acc: 0.8925 - val_loss: 0.3450 - val_acc: 0.9450\n",
            "Epoch 10/50\n",
            " - 23s - loss: 0.2513 - acc: 0.8930 - val_loss: 0.3438 - val_acc: 0.9450\n",
            "Epoch 11/50\n",
            " - 24s - loss: 0.2527 - acc: 0.8885 - val_loss: 0.3433 - val_acc: 0.9450\n",
            "Epoch 12/50\n",
            " - 23s - loss: 0.2465 - acc: 0.8965 - val_loss: 0.3375 - val_acc: 0.9460\n",
            "Epoch 13/50\n",
            " - 23s - loss: 0.2446 - acc: 0.9035 - val_loss: 0.3387 - val_acc: 0.9460\n",
            "Epoch 14/50\n",
            " - 23s - loss: 0.2477 - acc: 0.8855 - val_loss: 0.3375 - val_acc: 0.9460\n",
            "Epoch 15/50\n",
            " - 23s - loss: 0.2811 - acc: 0.8800 - val_loss: 0.3348 - val_acc: 0.9470\n",
            "Epoch 16/50\n",
            " - 23s - loss: 0.2358 - acc: 0.9000 - val_loss: 0.3363 - val_acc: 0.9470\n",
            "Epoch 17/50\n",
            " - 23s - loss: 0.2411 - acc: 0.8975 - val_loss: 0.3361 - val_acc: 0.9470\n",
            "Epoch 18/50\n",
            " - 23s - loss: 0.2322 - acc: 0.9030 - val_loss: 0.3365 - val_acc: 0.9460\n",
            "Epoch 19/50\n",
            " - 23s - loss: 0.2273 - acc: 0.9045 - val_loss: 0.3296 - val_acc: 0.9480\n",
            "Epoch 20/50\n",
            " - 23s - loss: 0.2259 - acc: 0.9045 - val_loss: 0.3313 - val_acc: 0.9470\n",
            "Epoch 21/50\n",
            " - 23s - loss: 0.2535 - acc: 0.8890 - val_loss: 0.3359 - val_acc: 0.9460\n",
            "Epoch 22/50\n",
            " - 23s - loss: 0.2282 - acc: 0.9050 - val_loss: 0.3352 - val_acc: 0.9470\n",
            "Epoch 23/50\n",
            " - 23s - loss: 0.2358 - acc: 0.9035 - val_loss: 0.3304 - val_acc: 0.9480\n",
            "Epoch 24/50\n",
            " - 23s - loss: 0.2419 - acc: 0.8915 - val_loss: 0.3314 - val_acc: 0.9480\n",
            "Epoch 25/50\n",
            " - 23s - loss: 0.2517 - acc: 0.8935 - val_loss: 0.3320 - val_acc: 0.9480\n",
            "Epoch 26/50\n",
            " - 23s - loss: 0.2422 - acc: 0.8915 - val_loss: 0.3337 - val_acc: 0.9480\n",
            "Epoch 27/50\n",
            " - 23s - loss: 0.2184 - acc: 0.9110 - val_loss: 0.3264 - val_acc: 0.9480\n",
            "Epoch 28/50\n",
            " - 23s - loss: 0.2349 - acc: 0.8930 - val_loss: 0.3272 - val_acc: 0.9490\n",
            "Epoch 29/50\n",
            " - 23s - loss: 0.2237 - acc: 0.9105 - val_loss: 0.3299 - val_acc: 0.9480\n",
            "Epoch 30/50\n",
            " - 23s - loss: 0.2516 - acc: 0.8885 - val_loss: 0.3353 - val_acc: 0.9470\n",
            "Epoch 31/50\n",
            " - 23s - loss: 0.2328 - acc: 0.8950 - val_loss: 0.3360 - val_acc: 0.9470\n",
            "Epoch 32/50\n",
            " - 23s - loss: 0.2453 - acc: 0.8970 - val_loss: 0.3343 - val_acc: 0.9480\n",
            "Epoch 33/50\n",
            " - 23s - loss: 0.2494 - acc: 0.8900 - val_loss: 0.3344 - val_acc: 0.9480\n",
            "Epoch 34/50\n",
            " - 23s - loss: 0.2365 - acc: 0.8990 - val_loss: 0.3340 - val_acc: 0.9470\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 35/50\n",
            " - 23s - loss: 0.2521 - acc: 0.8925 - val_loss: 0.3324 - val_acc: 0.9470\n",
            "Epoch 36/50\n",
            " - 23s - loss: 0.2421 - acc: 0.8945 - val_loss: 0.3342 - val_acc: 0.9470\n",
            "Epoch 37/50\n",
            " - 23s - loss: 0.2342 - acc: 0.8975 - val_loss: 0.3353 - val_acc: 0.9470\n",
            "Epoch 38/50\n",
            " - 23s - loss: 0.2450 - acc: 0.8985 - val_loss: 0.3306 - val_acc: 0.9490\n",
            "Epoch 39/50\n",
            " - 23s - loss: 0.2299 - acc: 0.9030 - val_loss: 0.3291 - val_acc: 0.9490\n",
            "Epoch 40/50\n",
            " - 23s - loss: 0.2502 - acc: 0.8930 - val_loss: 0.3315 - val_acc: 0.9480\n",
            "Epoch 41/50\n",
            " - 23s - loss: 0.2314 - acc: 0.9035 - val_loss: 0.3317 - val_acc: 0.9480\n",
            "Epoch 42/50\n",
            " - 23s - loss: 0.2403 - acc: 0.8985 - val_loss: 0.3293 - val_acc: 0.9490\n",
            "Epoch 43/50\n",
            " - 23s - loss: 0.2444 - acc: 0.8925 - val_loss: 0.3241 - val_acc: 0.9490\n",
            "Epoch 44/50\n",
            " - 23s - loss: 0.2378 - acc: 0.8950 - val_loss: 0.3271 - val_acc: 0.9490\n",
            "Epoch 45/50\n",
            " - 23s - loss: 0.2417 - acc: 0.8980 - val_loss: 0.3256 - val_acc: 0.9490\n",
            "Epoch 46/50\n",
            " - 23s - loss: 0.2414 - acc: 0.8990 - val_loss: 0.3228 - val_acc: 0.9490\n",
            "Epoch 47/50\n",
            " - 23s - loss: 0.2216 - acc: 0.9025 - val_loss: 0.3266 - val_acc: 0.9490\n",
            "Epoch 48/50\n",
            " - 23s - loss: 0.2317 - acc: 0.8970 - val_loss: 0.3289 - val_acc: 0.9480\n",
            "Epoch 49/50\n",
            " - 23s - loss: 0.2424 - acc: 0.8880 - val_loss: 0.3221 - val_acc: 0.9490\n",
            "Epoch 50/50\n",
            " - 23s - loss: 0.2307 - acc: 0.9015 - val_loss: 0.3266 - val_acc: 0.9490\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "3EPGn58ofwq5",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "We are seeing a nice improvement, with the validation loss going from ~1.7 down to ~1.2, and accuracy going from 88% to 92%. That's a 4.5% relative improvement in accuracy.\n",
        "\n",
        "Let's plot the training and validation loss and accuracy to show it conclusively:"
      ]
    },
    {
      "metadata": {
        "id": "1FtxcKjJfxL9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 560
        },
        "outputId": "e62d990c-4df8-4848-f0aa-b216a3c28aee"
      },
      "cell_type": "code",
      "source": [
        "%matplotlib inline\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.image as mpimg\n",
        "\n",
        "# Retrieve a list of accuracy results on training and test data\n",
        "# sets for each training epoch\n",
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "\n",
        "# Retrieve a list of list results on training and test data\n",
        "# sets for each training epoch\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "# Get number of epochs\n",
        "epochs = range(len(acc))\n",
        "\n",
        "# Plot training and validation accuracy per epoch\n",
        "plt.plot(epochs, acc)\n",
        "plt.plot(epochs, val_acc)\n",
        "plt.title('Training and validation accuracy')\n",
        "\n",
        "plt.figure()\n",
        "\n",
        "# Plot training and validation loss per epoch\n",
        "plt.plot(epochs, loss)\n",
        "plt.plot(epochs, val_loss)\n",
        "plt.title('Training and validation loss')"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5,1,u'Training and validation loss')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEHCAYAAABV4gY/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f6866bf53d0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAEHCAYAAABCwJb2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nlvSNMcZDLOkbY4yH/D8q9Xf6d4eBtwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f6823a72750>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "X-fUIeizakjE",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Congratulations! Using feature extraction and fine-tuning, you've built an image classification model that can identify cats vs. dogs in images with over 90% accuracy."
      ]
    },
    {
      "metadata": {
        "id": "x_ANwJCnx7w-",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Clean Up\n",
        "\n",
        "Run the following cell to terminate the kernel and free memory resources:"
      ]
    },
    {
      "metadata": {
        "id": "-hUmyohAyBzh",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import os, signal\n",
        "os.kill(os.getpid(), signal.SIGKILL)"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}