Transfer_Learning_Using_Pretrained_ConvNets.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Transfer Learning Using Pretrained ConvNets.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "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/ActionPace/DeepLearningTests/blob/master/Transfer_Learning_Using_Pretrained_ConvNets.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "77gENRVX40S7"
      },
      "source": [
        "##### Copyright 2019 The TensorFlow Authors."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "colab_type": "code",
        "id": "d8jyt37T42Vf",
        "colab": {}
      },
      "source": [
        "#@title 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": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "form",
        "colab_type": "code",
        "id": "aPxHdjwW5P2j",
        "colab": {}
      },
      "source": [
        "#@title MIT License\n",
        "#\n",
        "# Copyright (c) 2017 François Chollet                                                                                                                    # IGNORE_COPYRIGHT: cleared by OSS licensing\n",
        "#\n",
        "# Permission is hereby granted, free of charge, to any person obtaining a\n",
        "# copy of this software and associated documentation files (the \"Software\"),\n",
        "# to deal in the Software without restriction, including without limitation\n",
        "# the rights to use, copy, modify, merge, publish, distribute, sublicense,\n",
        "# and/or sell copies of the Software, and to permit persons to whom the\n",
        "# Software is furnished to do so, subject to the following conditions:\n",
        "#\n",
        "# The above copyright notice and this permission notice shall be included in\n",
        "# all copies or substantial portions of the Software.\n",
        "#\n",
        "# THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n",
        "# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n",
        "# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL\n",
        "# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n",
        "# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING\n",
        "# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER\n",
        "# DEALINGS IN THE SOFTWARE."
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "hRTa3Ee15WsJ"
      },
      "source": [
        "# Transfer Learning Using Pretrained ConvNets"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "dQHMcypT3vDT"
      },
      "source": [
        "<table class=\"tfo-notebook-buttons\" align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://www.tensorflow.org/tutorials/images/transfer_learning\"><img src=\"https://www.tensorflow.org/images/tf_logo_32px.png\" />View on TensorFlow.org</a>\n",
        "  </td>\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/images/transfer_learning.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n",
        "  </td>\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://github.com/tensorflow/docs/blob/master/site/en/tutorials/images/transfer_learning.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n",
        "  </td>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "2X4KyhORdSeO"
      },
      "source": [
        "In this tutorial we will discuss how to classify cats vs dogs images by using transfer learning from a pre-trained network. This will allows us to get higher accuracies than we saw by training our network from scratch.\n",
        "\n",
        "A **pre-trained model **is a saved network that was previously trained on a large dataset, typically on a large-scale image-classification task. We can either use the pretrained model as it is or transfer learning using the pretrained convents. The intuition behind **transfer learning **is that if this model trained on a large and general enough dataset, this model will effectively serve as a generic model of the visual world. We can leverage these learned feature maps without having to train a large model on a large dataset by using these models as the basis of our own model specific to our task. There are 2 scenarios of transfer learning using a pretrained model:\n",
        "\n",
        "1. **Feature Extraction** - use the representations of learned by a previous network to extract meaningful features from new samples. We simply add a new classifier, which will be trained from scratch, on top of the pretrained model so that we can repurpose the feature maps learned previously for our dataset. **Do we use the entire pretrained model or just the convolutional base?** - We use the feature extraction portion of these pretrained convnets (convolutional base) since they are likely to be generic features and learned concepts over a picture. However, the classification part of the pretrained model is often specific to original classification task, and subsequently specific to the set of classes on which the model was trained.\n",
        "2. **Fine-Tuning** - unfreezing a few of the top layers of a frozen model base used for feature extraction, and jointly training both the newly added classifier layers as well as the last layers of the frozen model. This allows us to \"fine tune\" the higher order feature representations in addition to our final classifier in order to make them more relevant for the specific task involved.\n",
        "\n",
        "**We will follow the general machine learning workflow:**\n",
        "1. Examine and understand data\n",
        "2. Build an input pipeline - using Keras ImageDataGenerator as we did in the image classification tutorial\n",
        "3. Compose our model\n",
        "  * Load in our pretrained model (and pretrained weights)\n",
        "  * Stack our classification layers on top\n",
        "4. Train our model\n",
        "5. Evaluate model\n",
        "\n",
        "We will see an example of using the pre-trained convnet as the feature extraction and then fine-tune to train the last few layers of the base model.\n",
        "\n",
        "**Audience:** This post is geared towards beginners with some Keras API and ML background. To get the most out of this post, you should have some basic ML background, know what CNNs are, and be familiar with the Keras Sequential API.\n",
        "\n",
        "**Time Estimated**: 30 minutes"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "iBMcobPHdD8O",
        "outputId": "fbc9a85e-1589-4d52-e7d0-e1f5633e12c1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "from __future__ import absolute_import, division, print_function, unicode_literals\n",
        "\n",
        "import os\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "print(\"TensorFlow version is \", tf.__version__)\n",
        "\n",
        "import numpy as np\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.image as mpimg"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TensorFlow version is  1.13.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "v77rlkCKW0IJ"
      },
      "source": [
        "## Data preprocessing"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "aXzwKdouXf1h"
      },
      "source": [
        "### Download data - cats_and_dogs_filtered.zip\n",
        "We will download a filtered version of Kaggle's [Dogs vs Cats](https://www.kaggle.com/c/dogs-vs-cats/data) dataset. Then store the downloaded zip file to the \"/tmp/\" directory."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "nRnO59Kr6enO",
        "outputId": "a25fa3ac-574f-413f-8637-efda78c3cfdb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 74
        }
      },
      "source": [
        "zip_file = tf.keras.utils.get_file(origin=\"https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip\",\n",
        "                                   fname=\"cats_and_dogs_filtered.zip\", extract=True)\n",
        "base_dir, _ = os.path.splitext(zip_file)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip\n",
            "68608000/68606236 [==============================] - 1s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "9_6h-c5EXN91"
      },
      "source": [
        "### Prepare training and validation cats and dogs datasets\n",
        "Create the training and validation directories for cats datasets and dog datasets."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "RWcldM4TXLen",
        "outputId": "22a10030-eed5-4b51-f984-54d6f833e387",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 91
        }
      },
      "source": [
        "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",
        "print ('Total training cat images:', len(os.listdir(train_cats_dir)))\n",
        "\n",
        "# Directory with our training dog pictures\n",
        "train_dogs_dir = os.path.join(train_dir, 'dogs')\n",
        "print ('Total training dog images:', len(os.listdir(train_dogs_dir)))\n",
        "\n",
        "# Directory with our validation cat pictures\n",
        "validation_cats_dir = os.path.join(validation_dir, 'cats')\n",
        "print ('Total validation cat images:', len(os.listdir(validation_cats_dir)))\n",
        "\n",
        "# Directory with our validation dog pictures\n",
        "validation_dogs_dir = os.path.join(validation_dir, 'dogs')\n",
        "print ('Total validation dog images:', len(os.listdir(validation_dogs_dir)))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total training cat images: 1000\n",
            "Total training dog images: 1000\n",
            "Total validation cat images: 500\n",
            "Total validation dog images: 500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "wvidPx6jeFzf"
      },
      "source": [
        "### Create Image Data Generator with Image Augmentation\n",
        "\n",
        "We will use ImageDataGenerator to rescale the images.\n",
        "\n",
        "Create the train generator and specify where the train dataset directory, image size, batch size and it's a binary classification.\n",
        "\n",
        "Create the validation generator with similar approach as the train generator with the flow_from_directory() method."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "y3PM6GVHcC31",
        "outputId": "15de6956-10b6-459c-ad46-ba42cfbfcbb2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "image_size = 160 # All images will be resized to 160x160\n",
        "batch_size = 32\n",
        "\n",
        "# Rescale all images by 1./255 and apply image augmentation\n",
        "train_datagen = keras.preprocessing.image.ImageDataGenerator(\n",
        "                rescale=1./255)\n",
        "\n",
        "validation_datagen = keras.preprocessing.image.ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "# Flow training images in batches of 20 using train_datagen generator\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "                train_dir,  # Source directory for the training images\n",
        "                target_size=(image_size, image_size),\n",
        "                batch_size=batch_size,\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 = validation_datagen.flow_from_directory(\n",
        "                validation_dir, # Source directory for the validation images\n",
        "                target_size=(image_size, image_size),\n",
        "                batch_size=batch_size,\n",
        "                class_mode='binary')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 1000 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "OkH-kazQecHB"
      },
      "source": [
        "## Create the base model from the pre-trained convnets\n",
        "We will create the base model from the **MobileNet V2** model developed at Google, and pre-trained on the ImageNet dataset, a large dataset of 1.4M images and 1000 classes of web images. 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 MobileNet V2 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 to 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 MobileNet V2 model pre-loaded with weights trained on ImageNet. By specifying the **include_top=False** argument, we load a network that doesn't include the classification layers at the top, which is ideal for feature extraction."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "19IQ2gqneqmS",
        "outputId": "0c6a2b86-0404-438c-f5d2-d56775645aa1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 130
        }
      },
      "source": [
        "IMG_SHAPE = (image_size, image_size, 3)\n",
        "\n",
        "# Create the base model from the pre-trained model MobileNet V2\n",
        "base_model = tf.keras.applications.MobileNetV2(input_shape=IMG_SHAPE,\n",
        "                                               include_top=False,\n",
        "                                               weights='imagenet')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n",
            "Downloading data from https://github.com/JonathanCMitchell/mobilenet_v2_keras/releases/download/v1.1/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_160_no_top.h5\n",
            "9412608/9406464 [==============================] - 1s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "rlx56nQtfe8Y"
      },
      "source": [
        "## Feature extraction\n",
        "We will freeze the convolutional base created from the previous step and use that as a feature extractor, add a classifier on top of it and train the top-level classifier."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "CnMLieHBCwil"
      },
      "source": [
        "### Freeze the convolutional base\n",
        "It's important to freeze the convolutional based before we compile and train the model. By freezing (or setting `layer.trainable = False`), we prevent the weights in these layers from being updated during training."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "OTCJH4bphOeo",
        "colab": {}
      },
      "source": [
        "base_model.trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "KpbzSmPkDa-N",
        "outputId": "38b23c87-3966-4680-fe89-bba4823e5857",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 6097
        }
      },
      "source": [
        "# Let's take a look at the base model architecture\n",
        "base_model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            (None, 160, 160, 3)  0                                            \n",
            "__________________________________________________________________________________________________\n",
            "Conv1_pad (ZeroPadding2D)       (None, 161, 161, 3)  0           input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "Conv1 (Conv2D)                  (None, 80, 80, 32)   864         Conv1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "bn_Conv1 (BatchNormalizationV1) (None, 80, 80, 32)   128         Conv1[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "Conv1_relu (ReLU)               (None, 80, 80, 32)   0           bn_Conv1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "expanded_conv_depthwise (Depthw (None, 80, 80, 32)   288         Conv1_relu[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "expanded_conv_depthwise_BN (Bat (None, 80, 80, 32)   128         expanded_conv_depthwise[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "expanded_conv_depthwise_relu (R (None, 80, 80, 32)   0           expanded_conv_depthwise_BN[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "expanded_conv_project (Conv2D)  (None, 80, 80, 16)   512         expanded_conv_depthwise_relu[0][0\n",
            "__________________________________________________________________________________________________\n",
            "expanded_conv_project_BN (Batch (None, 80, 80, 16)   64          expanded_conv_project[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_1_expand (Conv2D)         (None, 80, 80, 96)   1536        expanded_conv_project_BN[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "block_1_expand_BN (BatchNormali (None, 80, 80, 96)   384         block_1_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_1_expand_relu (ReLU)      (None, 80, 80, 96)   0           block_1_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_1_pad (ZeroPadding2D)     (None, 81, 81, 96)   0           block_1_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_1_depthwise (DepthwiseCon (None, 40, 40, 96)   864         block_1_pad[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_1_depthwise_BN (BatchNorm (None, 40, 40, 96)   384         block_1_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_1_depthwise_relu (ReLU)   (None, 40, 40, 96)   0           block_1_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_1_project (Conv2D)        (None, 40, 40, 24)   2304        block_1_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_1_project_BN (BatchNormal (None, 40, 40, 24)   96          block_1_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_2_expand (Conv2D)         (None, 40, 40, 144)  3456        block_1_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_2_expand_BN (BatchNormali (None, 40, 40, 144)  576         block_2_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_2_expand_relu (ReLU)      (None, 40, 40, 144)  0           block_2_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_2_depthwise (DepthwiseCon (None, 40, 40, 144)  1296        block_2_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_2_depthwise_BN (BatchNorm (None, 40, 40, 144)  576         block_2_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_2_depthwise_relu (ReLU)   (None, 40, 40, 144)  0           block_2_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_2_project (Conv2D)        (None, 40, 40, 24)   3456        block_2_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_2_project_BN (BatchNormal (None, 40, 40, 24)   96          block_2_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_2_add (Add)               (None, 40, 40, 24)   0           block_1_project_BN[0][0]         \n",
            "                                                                 block_2_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_3_expand (Conv2D)         (None, 40, 40, 144)  3456        block_2_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_3_expand_BN (BatchNormali (None, 40, 40, 144)  576         block_3_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_3_expand_relu (ReLU)      (None, 40, 40, 144)  0           block_3_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_3_pad (ZeroPadding2D)     (None, 41, 41, 144)  0           block_3_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_3_depthwise (DepthwiseCon (None, 20, 20, 144)  1296        block_3_pad[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_3_depthwise_BN (BatchNorm (None, 20, 20, 144)  576         block_3_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_3_depthwise_relu (ReLU)   (None, 20, 20, 144)  0           block_3_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_3_project (Conv2D)        (None, 20, 20, 32)   4608        block_3_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_3_project_BN (BatchNormal (None, 20, 20, 32)   128         block_3_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_4_expand (Conv2D)         (None, 20, 20, 192)  6144        block_3_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_4_expand_BN (BatchNormali (None, 20, 20, 192)  768         block_4_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_4_expand_relu (ReLU)      (None, 20, 20, 192)  0           block_4_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_4_depthwise (DepthwiseCon (None, 20, 20, 192)  1728        block_4_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_4_depthwise_BN (BatchNorm (None, 20, 20, 192)  768         block_4_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_4_depthwise_relu (ReLU)   (None, 20, 20, 192)  0           block_4_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_4_project (Conv2D)        (None, 20, 20, 32)   6144        block_4_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_4_project_BN (BatchNormal (None, 20, 20, 32)   128         block_4_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_4_add (Add)               (None, 20, 20, 32)   0           block_3_project_BN[0][0]         \n",
            "                                                                 block_4_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_5_expand (Conv2D)         (None, 20, 20, 192)  6144        block_4_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_5_expand_BN (BatchNormali (None, 20, 20, 192)  768         block_5_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_5_expand_relu (ReLU)      (None, 20, 20, 192)  0           block_5_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_5_depthwise (DepthwiseCon (None, 20, 20, 192)  1728        block_5_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_5_depthwise_BN (BatchNorm (None, 20, 20, 192)  768         block_5_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_5_depthwise_relu (ReLU)   (None, 20, 20, 192)  0           block_5_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_5_project (Conv2D)        (None, 20, 20, 32)   6144        block_5_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_5_project_BN (BatchNormal (None, 20, 20, 32)   128         block_5_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_5_add (Add)               (None, 20, 20, 32)   0           block_4_add[0][0]                \n",
            "                                                                 block_5_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_6_expand (Conv2D)         (None, 20, 20, 192)  6144        block_5_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_6_expand_BN (BatchNormali (None, 20, 20, 192)  768         block_6_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_6_expand_relu (ReLU)      (None, 20, 20, 192)  0           block_6_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_6_pad (ZeroPadding2D)     (None, 21, 21, 192)  0           block_6_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_6_depthwise (DepthwiseCon (None, 10, 10, 192)  1728        block_6_pad[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_6_depthwise_BN (BatchNorm (None, 10, 10, 192)  768         block_6_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_6_depthwise_relu (ReLU)   (None, 10, 10, 192)  0           block_6_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_6_project (Conv2D)        (None, 10, 10, 64)   12288       block_6_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_6_project_BN (BatchNormal (None, 10, 10, 64)   256         block_6_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_7_expand (Conv2D)         (None, 10, 10, 384)  24576       block_6_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_7_expand_BN (BatchNormali (None, 10, 10, 384)  1536        block_7_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_7_expand_relu (ReLU)      (None, 10, 10, 384)  0           block_7_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_7_depthwise (DepthwiseCon (None, 10, 10, 384)  3456        block_7_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_7_depthwise_BN (BatchNorm (None, 10, 10, 384)  1536        block_7_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_7_depthwise_relu (ReLU)   (None, 10, 10, 384)  0           block_7_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_7_project (Conv2D)        (None, 10, 10, 64)   24576       block_7_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_7_project_BN (BatchNormal (None, 10, 10, 64)   256         block_7_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_7_add (Add)               (None, 10, 10, 64)   0           block_6_project_BN[0][0]         \n",
            "                                                                 block_7_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_8_expand (Conv2D)         (None, 10, 10, 384)  24576       block_7_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_8_expand_BN (BatchNormali (None, 10, 10, 384)  1536        block_8_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_8_expand_relu (ReLU)      (None, 10, 10, 384)  0           block_8_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_8_depthwise (DepthwiseCon (None, 10, 10, 384)  3456        block_8_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_8_depthwise_BN (BatchNorm (None, 10, 10, 384)  1536        block_8_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_8_depthwise_relu (ReLU)   (None, 10, 10, 384)  0           block_8_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_8_project (Conv2D)        (None, 10, 10, 64)   24576       block_8_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_8_project_BN (BatchNormal (None, 10, 10, 64)   256         block_8_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_8_add (Add)               (None, 10, 10, 64)   0           block_7_add[0][0]                \n",
            "                                                                 block_8_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_9_expand (Conv2D)         (None, 10, 10, 384)  24576       block_8_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_9_expand_BN (BatchNormali (None, 10, 10, 384)  1536        block_9_expand[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block_9_expand_relu (ReLU)      (None, 10, 10, 384)  0           block_9_expand_BN[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_9_depthwise (DepthwiseCon (None, 10, 10, 384)  3456        block_9_expand_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_9_depthwise_BN (BatchNorm (None, 10, 10, 384)  1536        block_9_depthwise[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "block_9_depthwise_relu (ReLU)   (None, 10, 10, 384)  0           block_9_depthwise_BN[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_9_project (Conv2D)        (None, 10, 10, 64)   24576       block_9_depthwise_relu[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "block_9_project_BN (BatchNormal (None, 10, 10, 64)   256         block_9_project[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_9_add (Add)               (None, 10, 10, 64)   0           block_8_add[0][0]                \n",
            "                                                                 block_9_project_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_10_expand (Conv2D)        (None, 10, 10, 384)  24576       block_9_add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "block_10_expand_BN (BatchNormal (None, 10, 10, 384)  1536        block_10_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_10_expand_relu (ReLU)     (None, 10, 10, 384)  0           block_10_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_10_depthwise (DepthwiseCo (None, 10, 10, 384)  3456        block_10_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_10_depthwise_BN (BatchNor (None, 10, 10, 384)  1536        block_10_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_10_depthwise_relu (ReLU)  (None, 10, 10, 384)  0           block_10_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_10_project (Conv2D)       (None, 10, 10, 96)   36864       block_10_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_10_project_BN (BatchNorma (None, 10, 10, 96)   384         block_10_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_11_expand (Conv2D)        (None, 10, 10, 576)  55296       block_10_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_11_expand_BN (BatchNormal (None, 10, 10, 576)  2304        block_11_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_11_expand_relu (ReLU)     (None, 10, 10, 576)  0           block_11_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_11_depthwise (DepthwiseCo (None, 10, 10, 576)  5184        block_11_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_11_depthwise_BN (BatchNor (None, 10, 10, 576)  2304        block_11_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_11_depthwise_relu (ReLU)  (None, 10, 10, 576)  0           block_11_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_11_project (Conv2D)       (None, 10, 10, 96)   55296       block_11_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_11_project_BN (BatchNorma (None, 10, 10, 96)   384         block_11_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_11_add (Add)              (None, 10, 10, 96)   0           block_10_project_BN[0][0]        \n",
            "                                                                 block_11_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_12_expand (Conv2D)        (None, 10, 10, 576)  55296       block_11_add[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block_12_expand_BN (BatchNormal (None, 10, 10, 576)  2304        block_12_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_12_expand_relu (ReLU)     (None, 10, 10, 576)  0           block_12_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_12_depthwise (DepthwiseCo (None, 10, 10, 576)  5184        block_12_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_12_depthwise_BN (BatchNor (None, 10, 10, 576)  2304        block_12_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_12_depthwise_relu (ReLU)  (None, 10, 10, 576)  0           block_12_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_12_project (Conv2D)       (None, 10, 10, 96)   55296       block_12_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_12_project_BN (BatchNorma (None, 10, 10, 96)   384         block_12_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_12_add (Add)              (None, 10, 10, 96)   0           block_11_add[0][0]               \n",
            "                                                                 block_12_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_13_expand (Conv2D)        (None, 10, 10, 576)  55296       block_12_add[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block_13_expand_BN (BatchNormal (None, 10, 10, 576)  2304        block_13_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_13_expand_relu (ReLU)     (None, 10, 10, 576)  0           block_13_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_13_pad (ZeroPadding2D)    (None, 11, 11, 576)  0           block_13_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_13_depthwise (DepthwiseCo (None, 5, 5, 576)    5184        block_13_pad[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block_13_depthwise_BN (BatchNor (None, 5, 5, 576)    2304        block_13_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_13_depthwise_relu (ReLU)  (None, 5, 5, 576)    0           block_13_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_13_project (Conv2D)       (None, 5, 5, 160)    92160       block_13_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_13_project_BN (BatchNorma (None, 5, 5, 160)    640         block_13_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_14_expand (Conv2D)        (None, 5, 5, 960)    153600      block_13_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_14_expand_BN (BatchNormal (None, 5, 5, 960)    3840        block_14_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_14_expand_relu (ReLU)     (None, 5, 5, 960)    0           block_14_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_14_depthwise (DepthwiseCo (None, 5, 5, 960)    8640        block_14_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_14_depthwise_BN (BatchNor (None, 5, 5, 960)    3840        block_14_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_14_depthwise_relu (ReLU)  (None, 5, 5, 960)    0           block_14_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_14_project (Conv2D)       (None, 5, 5, 160)    153600      block_14_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_14_project_BN (BatchNorma (None, 5, 5, 160)    640         block_14_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_14_add (Add)              (None, 5, 5, 160)    0           block_13_project_BN[0][0]        \n",
            "                                                                 block_14_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_15_expand (Conv2D)        (None, 5, 5, 960)    153600      block_14_add[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block_15_expand_BN (BatchNormal (None, 5, 5, 960)    3840        block_15_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_15_expand_relu (ReLU)     (None, 5, 5, 960)    0           block_15_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_15_depthwise (DepthwiseCo (None, 5, 5, 960)    8640        block_15_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_15_depthwise_BN (BatchNor (None, 5, 5, 960)    3840        block_15_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_15_depthwise_relu (ReLU)  (None, 5, 5, 960)    0           block_15_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_15_project (Conv2D)       (None, 5, 5, 160)    153600      block_15_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_15_project_BN (BatchNorma (None, 5, 5, 160)    640         block_15_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block_15_add (Add)              (None, 5, 5, 160)    0           block_14_add[0][0]               \n",
            "                                                                 block_15_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "block_16_expand (Conv2D)        (None, 5, 5, 960)    153600      block_15_add[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "block_16_expand_BN (BatchNormal (None, 5, 5, 960)    3840        block_16_expand[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block_16_expand_relu (ReLU)     (None, 5, 5, 960)    0           block_16_expand_BN[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_16_depthwise (DepthwiseCo (None, 5, 5, 960)    8640        block_16_expand_relu[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "block_16_depthwise_BN (BatchNor (None, 5, 5, 960)    3840        block_16_depthwise[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "block_16_depthwise_relu (ReLU)  (None, 5, 5, 960)    0           block_16_depthwise_BN[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "block_16_project (Conv2D)       (None, 5, 5, 320)    307200      block_16_depthwise_relu[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block_16_project_BN (BatchNorma (None, 5, 5, 320)    1280        block_16_project[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "Conv_1 (Conv2D)                 (None, 5, 5, 1280)   409600      block_16_project_BN[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "Conv_1_bn (BatchNormalizationV1 (None, 5, 5, 1280)   5120        Conv_1[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "out_relu (ReLU)                 (None, 5, 5, 1280)   0           Conv_1_bn[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 2,257,984\n",
            "Trainable params: 0\n",
            "Non-trainable params: 2,257,984\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "wdMRM8YModbk"
      },
      "source": [
        "#### Add a classification head"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "0iqnBeZrfoIc"
      },
      "source": [
        "Now let's add a few layers on top of the base model:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "eApvroIyn1K0",
        "colab": {}
      },
      "source": [
        "model = tf.keras.Sequential([\n",
        "  base_model,\n",
        "  keras.layers.GlobalAveragePooling2D(),\n",
        "  keras.layers.Dense(1, activation='sigmoid')\n",
        "])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "g0ylJXE_kRLi"
      },
      "source": [
        "### Compile the model\n",
        "\n",
        "You must compile the model before training it."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "RpR8HdyMhukJ",
        "colab": {}
      },
      "source": [
        "model.compile(optimizer=tf.keras.optimizers.RMSprop(lr=0.0001),\n",
        "              loss='binary_crossentropy',\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "I8ARiyMFsgbH",
        "outputId": "c8d91d0e-f34e-40c7-c6fb-91505ba4cf9f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 258
        }
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "mobilenetv2_1.00_160 (Model) (None, 5, 5, 1280)        2257984   \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d (Gl (None, 1280)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 1)                 1281      \n",
            "=================================================================\n",
            "Total params: 2,259,265\n",
            "Trainable params: 1,281\n",
            "Non-trainable params: 2,257,984\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "lxOcmVr0ydFZ"
      },
      "source": [
        "These 1.2K trainable parameters are divided among 2 TensorFlow `Variable` objects, the weights and biases of the two dense layers:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "krvBumovycVA",
        "outputId": "c660fa80-f898-4fc1-e130-524f68acee4e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "len(model.trainable_variables)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "RxvgOYTDSWTx"
      },
      "source": [
        "### Train the model\n",
        "\n",
        "After training for 10 epochs, we are able to get ~94% accuracy.\n",
        "\n",
        "If you have more time, train it to convergence (50 epochs, ~96% accuracy)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Om4O3EESkab1",
        "outputId": "a85ac9b2-2b8b-4678-d924-c9c157f09c38",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 649
        }
      },
      "source": [
        "epochs = 10\n",
        "steps_per_epoch = train_generator.n // batch_size\n",
        "validation_steps = validation_generator.n // batch_size\n",
        "\n",
        "history = model.fit_generator(train_generator,\n",
        "                              steps_per_epoch = steps_per_epoch,\n",
        "                              epochs=epochs,\n",
        "                              workers=4,\n",
        "                              validation_data=validation_generator,\n",
        "                              validation_steps=validation_steps)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "Epoch 1/10\n",
            "32/32 [==============================] - 4s 127ms/step - loss: 0.4935 - acc: 0.7720\n",
            "63/63 [==============================] - 14s 220ms/step - loss: 0.6214 - acc: 0.6560 - val_loss: 0.4935 - val_acc: 0.7720\n",
            "Epoch 2/10\n",
            "32/32 [==============================] - 4s 111ms/step - loss: 0.3849 - acc: 0.8500\n",
            "63/63 [==============================] - 9s 139ms/step - loss: 0.5144 - acc: 0.7570 - val_loss: 0.3849 - val_acc: 0.8500\n",
            "Epoch 3/10\n",
            "32/32 [==============================] - 4s 113ms/step - loss: 0.3241 - acc: 0.8830\n",
            "63/63 [==============================] - 9s 136ms/step - loss: 0.4389 - acc: 0.8180 - val_loss: 0.3241 - val_acc: 0.8830\n",
            "Epoch 4/10\n",
            "32/32 [==============================] - 4s 123ms/step - loss: 0.2805 - acc: 0.9030\n",
            "63/63 [==============================] - 9s 149ms/step - loss: 0.3754 - acc: 0.8560 - val_loss: 0.2805 - val_acc: 0.9030\n",
            "Epoch 5/10\n",
            "32/32 [==============================] - 4s 110ms/step - loss: 0.2472 - acc: 0.9130\n",
            "63/63 [==============================] - 9s 140ms/step - loss: 0.3412 - acc: 0.8705 - val_loss: 0.2472 - val_acc: 0.9130\n",
            "Epoch 6/10\n",
            "32/32 [==============================] - 4s 110ms/step - loss: 0.2262 - acc: 0.9200\n",
            "63/63 [==============================] - 9s 135ms/step - loss: 0.3061 - acc: 0.8810 - val_loss: 0.2262 - val_acc: 0.9200\n",
            "Epoch 7/10\n",
            "32/32 [==============================] - 4s 110ms/step - loss: 0.2087 - acc: 0.9240\n",
            "63/63 [==============================] - 9s 135ms/step - loss: 0.2829 - acc: 0.8995 - val_loss: 0.2087 - val_acc: 0.9240\n",
            "Epoch 8/10\n",
            "32/32 [==============================] - 3s 109ms/step - loss: 0.2049 - acc: 0.9220\n",
            "63/63 [==============================] - 9s 135ms/step - loss: 0.2696 - acc: 0.9040 - val_loss: 0.2049 - val_acc: 0.9220\n",
            "Epoch 9/10\n",
            "32/32 [==============================] - 4s 110ms/step - loss: 0.1867 - acc: 0.9330\n",
            "63/63 [==============================] - 9s 136ms/step - loss: 0.2516 - acc: 0.9065 - val_loss: 0.1867 - val_acc: 0.9330\n",
            "Epoch 10/10\n",
            "32/32 [==============================] - 4s 113ms/step - loss: 0.1736 - acc: 0.9410\n",
            "63/63 [==============================] - 9s 146ms/step - loss: 0.2252 - acc: 0.9295 - val_loss: 0.1736 - val_acc: 0.9410\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Hd94CKImf8vi"
      },
      "source": [
        "### Learning curves\n",
        "\n",
        "Let's take a look at the learning curves of the training and validation accuracy / loss, when using the MobileNet V2 base model as a fixed feature extractor."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "l7HOsQTPNgO9"
      },
      "source": [
        "If you train to convergence (`epochs=50`) the resulting graph should look like this:\n",
        "\n",
        "![Before fine tuning, the model reaches 96% accuracy](https://github.com/tensorflow/docs/blob/master/site/en/tutorials/images/images/before_fine_tuning.pnd?raw=1)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "53OTCh3jnbwV",
        "outputId": "d90e44ed-f66f-4a74-b956-a72f74c95707",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 499
        }
      },
      "source": [
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "plt.figure(figsize=(8, 8))\n",
        "plt.subplot(2, 1, 1)\n",
        "plt.plot(acc, label='Training Accuracy')\n",
        "plt.plot(val_acc, label='Validation Accuracy')\n",
        "plt.legend(loc='lower right')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([min(plt.ylim()),1])\n",
        "plt.title('Training and Validation Accuracy')\n",
        "\n",
        "plt.subplot(2, 1, 2)\n",
        "plt.plot(loss, label='Training Loss')\n",
        "plt.plot(val_loss, label='Validation Loss')\n",
        "plt.legend(loc='upper right')\n",
        "plt.ylabel('Cross Entropy')\n",
        "plt.ylim([0,max(plt.ylim())])\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfsAAAHiCAYAAADvUmWBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njDEmwFmyN8YYYwKcJXtjjDEmwP1/J22++pBCklMAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 576x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "CqwV-CRdS6Nv"
      },
      "source": [
        "## Fine tuning\n",
        "In our feature extraction experiment, we were only training a few layers on top of an MobileNet V2 base model. The weights of the pre-trained 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 pre-trained model alongside the training of the top-level classifier. The training process will force the weights to be tuned from generic features maps to features associated specifically to our dataset.\n",
        "\n",
        "Note: this should only be attempted after you have trained the top-level classifier with the pre-trained model set to non-trainable. If you add a randomly initialized classifier on top of a pre-trained 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 pre-trained model will just forget everything it has learned.\n",
        "\n",
        "Additionally, the reasoning behind fine-tuning the top layers of the pre-trained model rather than all layers of the pre-trained model is the following: in a convnet, the higher up a layer is, the more specialized it is. The first few layers in a convnet learned very simple and generic features, which generalize to almost all types of images. But as you go higher up, the features are increasingly more specific to the dataset that the model was trained on. The goal of fine-tuning is to adapt these specialized features to work with the new dataset."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "CPXnzUK0QonF"
      },
      "source": [
        "### Un-freeze the top layers of the model\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "rfxv_ifotQak"
      },
      "source": [
        "All we need to do is unfreeze the `base_model`, and set the bottom layers be un-trainable. Then, recompile the model (necessary for these changes to take effect), and resume training."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "4nzcagVitLQm",
        "colab": {}
      },
      "source": [
        "base_model.trainable = True"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "-4HgVAacRs5v",
        "outputId": "6f03dbf6-fa9d-4367-cfad-35bbe5a398b2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "# Let's take a look to see how many layers are in the base model\n",
        "print(\"Number of layers in the base model: \", len(base_model.layers))\n",
        "\n",
        "# Fine tune from this layer onwards\n",
        "fine_tune_at = 100\n",
        "\n",
        "# Freeze all the layers before the `fine_tune_at` layer\n",
        "for layer in base_model.layers[:fine_tune_at]:\n",
        "  layer.trainable =  False"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of layers in the base model:  155\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "4Uk1dgsxT0IS"
      },
      "source": [
        "### Compile the model\n",
        "\n",
        "Compile the model using a much-lower training rate."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "NtUnaz0WUDva",
        "colab": {}
      },
      "source": [
        "\n",
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer = tf.keras.optimizers.RMSprop(lr=2e-5),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "WwBWy7J2kZvA",
        "outputId": "43b85fe4-777d-4e22-90cd-8be351310cad",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 258
        }
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "mobilenetv2_1.00_160 (Model) (None, 5, 5, 1280)        2257984   \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d (Gl (None, 1280)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 1)                 1281      \n",
            "=================================================================\n",
            "Total params: 2,259,265\n",
            "Trainable params: 1,863,873\n",
            "Non-trainable params: 395,392\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "bNXelbMQtonr",
        "outputId": "42c17701-6f97-4aa1-ba11-7fe0f8e7ef5a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 36
        }
      },
      "source": [
        "len(model.trainable_variables)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "58"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "4G5O4jd6TuAG"
      },
      "source": [
        "### Continue Train the model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "0foWUN-yDLo_"
      },
      "source": [
        "If you trained to convergence earlier, this will get you a few percent more accuracy."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ECQLkAsFTlun",
        "outputId": "89272842-e0e8-478d-9678-8da768d273ed",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 537
        }
      },
      "source": [
        "history_fine = model.fit_generator(train_generator,\n",
        "                                   steps_per_epoch = steps_per_epoch,\n",
        "                                   epochs=epochs,\n",
        "                                   workers=4,\n",
        "                                   validation_data=validation_generator,\n",
        "                                   validation_steps=validation_steps)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "32/32 [==============================] - 4s 120ms/step - loss: 0.0963 - acc: 0.9640\n",
            "63/63 [==============================] - 11s 176ms/step - loss: 0.1651 - acc: 0.9425 - val_loss: 0.0963 - val_acc: 0.9640\n",
            "Epoch 2/10\n",
            "32/32 [==============================] - 4s 121ms/step - loss: 0.0842 - acc: 0.9700\n",
            "63/63 [==============================] - 9s 146ms/step - loss: 0.0890 - acc: 0.9785 - val_loss: 0.0842 - val_acc: 0.9700\n",
            "Epoch 3/10\n",
            "32/32 [==============================] - 3s 106ms/step - loss: 0.0764 - acc: 0.9720\n",
            "63/63 [==============================] - 9s 146ms/step - loss: 0.0567 - acc: 0.9845 - val_loss: 0.0764 - val_acc: 0.9720\n",
            "Epoch 4/10\n",
            "32/32 [==============================] - 3s 105ms/step - loss: 0.0698 - acc: 0.9780\n",
            "63/63 [==============================] - 9s 136ms/step - loss: 0.0455 - acc: 0.9880 - val_loss: 0.0698 - val_acc: 0.9780\n",
            "Epoch 5/10\n",
            "32/32 [==============================] - 3s 108ms/step - loss: 0.0696 - acc: 0.9740\n",
            "63/63 [==============================] - 9s 138ms/step - loss: 0.0317 - acc: 0.9915 - val_loss: 0.0696 - val_acc: 0.9740\n",
            "Epoch 6/10\n",
            "32/32 [==============================] - 3s 106ms/step - loss: 0.0760 - acc: 0.9730\n",
            "63/63 [==============================] - 9s 137ms/step - loss: 0.0250 - acc: 0.9915 - val_loss: 0.0760 - val_acc: 0.9730\n",
            "Epoch 7/10\n",
            "32/32 [==============================] - 4s 123ms/step - loss: 0.0732 - acc: 0.9770\n",
            "63/63 [==============================] - 9s 144ms/step - loss: 0.0185 - acc: 0.9955 - val_loss: 0.0732 - val_acc: 0.9770\n",
            "Epoch 8/10\n",
            "32/32 [==============================] - 3s 106ms/step - loss: 0.0713 - acc: 0.9770\n",
            "63/63 [==============================] - 9s 148ms/step - loss: 0.0156 - acc: 0.9960 - val_loss: 0.0713 - val_acc: 0.9770\n",
            "Epoch 9/10\n",
            "32/32 [==============================] - 3s 107ms/step - loss: 0.0796 - acc: 0.9740\n",
            "63/63 [==============================] - 9s 137ms/step - loss: 0.0094 - acc: 0.9995 - val_loss: 0.0796 - val_acc: 0.9740\n",
            "Epoch 10/10\n",
            "32/32 [==============================] - 3s 105ms/step - loss: 0.0856 - acc: 0.9760\n",
            "63/63 [==============================] - 9s 138ms/step - loss: 0.0072 - acc: 0.9985 - val_loss: 0.0856 - val_acc: 0.9760\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "TfXEmsxQf6eP"
      },
      "source": [
        "### Learning curves\n",
        "\n",
        "Let's take a look at the learning curves of the training and validation accuracy / loss, when fine tuning the last few layers of the MobileNet V2 base model, as well as the classifier on top of it. Note the validation loss much higher than the training loss which means there maybe some overfitting.\n",
        "\n",
        "**Note**: the training dataset is fairly small, and is similar to the original datasets that MobileNet V2 was trained on, so fine-tuning may result in overfitting.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "DNtfNZKlInGT"
      },
      "source": [
        "If you train to convergence (`epochs=50`) the resulting graph should look like this:\n",
        "\n",
        "![After fine tuning the model nearly reaches 98% accuracy](https://github.com/tensorflow/docs/blob/master/site/en/tutorials/images/images/fine_tuning.pnd?raw=1)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "PpA8PlpQKygw",
        "colab": {}
      },
      "source": [
        "acc += history_fine.history['acc']\n",
        "val_acc += history_fine.history['val_acc']\n",
        "\n",
        "loss += history_fine.history['loss']\n",
        "val_loss += history_fine.history['val_loss']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "chW103JUItdk",
        "outputId": "139662e5-7a72-4a5f-bcab-5e13267e7e64",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 499
        }
      },
      "source": [
        "plt.figure(figsize=(8, 8))\n",
        "plt.subplot(2, 1, 1)\n",
        "plt.plot(acc, label='Training Accuracy')\n",
        "plt.plot(val_acc, label='Validation Accuracy')\n",
        "plt.ylim([0.9, 1])\n",
        "plt.plot([epochs-1,epochs-1], plt.ylim(), label='Start Fine Tuning')\n",
        "plt.legend(loc='lower right')\n",
        "plt.title('Training and Validation Accuracy')\n",
        "\n",
        "plt.subplot(2, 1, 2)\n",
        "plt.plot(loss, label='Training Loss')\n",
        "plt.plot(val_loss, label='Validation Loss')\n",
        "plt.ylim([0, 0.2])\n",
        "plt.plot([epochs-1,epochs-1], plt.ylim(), label='Start Fine Tuning')\n",
        "plt.legend(loc='upper right')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfMAAAHiCAYAAAD8hSV1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niESckrmIiEjEKZmLiIhE3P8BSaoCUxU6WHYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 576x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "_TZTwG7nhm0C"
      },
      "source": [
        "# Key takeaways\n",
        "In summary here is what we covered in this tutorial on how to do transfer learning using a pre-trained model to improve accuracy:\n",
        "* Using a pre-trained model for **feature extraction** - when working with a small dataset, it is common to leverage the features learned by a model trained on a larger dataset in the same domain. This is done by instantiating the pre-trained model and adding a fully connected classifier on top. The pre-trained model is \"frozen\" and only the weights of the classifier are updated during training.\n",
        "In this case, the convolutional base extracts all the features associated with each image and we train a classifier that determines, given these set of features to which class it belongs.\n",
        "* **Fine-tuning** a pre-trained model - to further improve performance, one might want to repurpose the top-level layers of the pre-trained models to the new dataset via fine-tuning.\n",
        "In this case, we tune our weights such that we learn highly specified and high level features specific to our dataset. This only make sense when the training dataset is large and very similar to the original dataset that the pre-trained model was trained on.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DTwapxkd9Llw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}