06_Transfer_Learning_Scaling_Up.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "06_Transfer_Learning_Scaling_Up.ipynb",
      "provenance": [],
      "toc_visible": true,
      "mount_file_id": "1CMdDaNTxXDFMulY8axMJEhB_IA0StMpY",
      "authorship_tag": "ABX9TyP1JGQlsp4vQW1zrpjstT5M",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Tower5954/TensorFlow/blob/main/06_Transfer_Learning_Scaling_Up.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Transfer Learning part.3\n",
        "\n",
        "What we are going to cover:\n",
        "* Downloading and preparing 10% of all Food101 classes (7500+ training images)\n",
        "* Training a transfer learning feature extraction model\n",
        "* Fine-tuning a feature extraction model to beat the original Food101 paper with only 10% of the data.\n",
        "* Evaluating \"Food Visions mini\" predictions.\n",
        "  * Finding the most wrong predictions (on the test data)\n",
        "* Making predictions with Food Vision mini on our own custom data.\n",
        "\n",
        "We have now seen the power of transfer learning feature extraction and fine-tuning, now it is time to scale up to all of the classes in food101 (101 total classes of food).\n",
        "\n",
        "The goal is to beat the original food 101 paper with only 10% of the training (leveraging the power of deep learning).\n",
        "\n",
        " [Original Food101 paper.](https://data.vision.ee.ethz.ch/cvl/datasets_extra/food-101/)\n",
        "\n",
        " The baseline to beat is 50.76% accuracy across 101 classes."
      ],
      "metadata": {
        "id": "58J6Oc1fDnmC"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Set up GPU \n",
        "# Check to see if running a GPU\n",
        "\n",
        "!nvidia-smi"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5NWy2x_DIgdz",
        "outputId": "6737a992-cda5-4b95-f044-d87ca9b6768e"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mon Jan 31 20:22:29 2022       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 495.46       Driver Version: 460.32.03    CUDA Version: 11.2     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla T4            Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   41C    P8     9W /  70W |      0MiB / 15109MiB |      0%      Default |\n",
            "|                               |                      |                  N/A |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Creating helper functions\n",
        "\n",
        "In previous notebooks, we have created a series of helper functions."
      ],
      "metadata": {
        "id": "cIJEhdYfI2Gr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!wget https://raw.githubusercontent.com/mrdbourke/tensorflow-deep-learning/main/extras/helper_functions.py"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FQG7Xnt3OFQI",
        "outputId": "6325b855-d5e8-4cc9-dc5f-ffc4d064b315"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-01-31 20:22:30--  https://raw.githubusercontent.com/mrdbourke/tensorflow-deep-learning/main/extras/helper_functions.py\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 10246 (10K) [text/plain]\n",
            "Saving to: ‘helper_functions.py’\n",
            "\n",
            "\rhelper_functions.py   0%[                    ]       0  --.-KB/s               \rhelper_functions.py 100%[===================>]  10.01K  --.-KB/s    in 0s      \n",
            "\n",
            "2022-01-31 20:22:30 (82.6 MB/s) - ‘helper_functions.py’ saved [10246/10246]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Import series of helper functions for the notebook\n",
        "from helper_functions import create_tensorboard_callback, plot_loss_curves, unzip_data, compare_historys, walk_through_dir"
      ],
      "metadata": {
        "id": "3IZ6Mf2TOG6y"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 101 Food Classes: working with less data.\n",
        "\n",
        "Our goal is to beat the original Food101 paper with 10% of the training data, so let's download it.\n",
        "\n",
        "> The data that we are downloading comes from the original Food101 dataset but has been preprocessed using the image_data_modification notebook - https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/extras/image_data_modification.ipynb"
      ],
      "metadata": {
        "id": "POGc5DlmOrRo"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Download data from Google Storage (already preformatted)\n",
        "!wget https://storage.googleapis.com/ztm_tf_course/food_vision/101_food_classes_10_percent.zip \n",
        "\n",
        "unzip_data(\"101_food_classes_10_percent.zip\")\n",
        "\n",
        "train_dir = \"101_food_classes_10_percent/train/\"\n",
        "test_dir = \"101_food_classes_10_percent/test/\""
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Nz3ya-dZPJVB",
        "outputId": "4199c73d-13be-4e1b-aee6-2dc8d239e242"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-01-31 20:22:34--  https://storage.googleapis.com/ztm_tf_course/food_vision/101_food_classes_10_percent.zip\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 74.125.20.128, 74.125.142.128, 74.125.195.128, ...\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|74.125.20.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 1625420029 (1.5G) [application/zip]\n",
            "Saving to: ‘101_food_classes_10_percent.zip’\n",
            "\n",
            "101_food_classes_10 100%[===================>]   1.51G   163MB/s    in 9.1s    \n",
            "\n",
            "2022-01-31 20:22:44 (171 MB/s) - ‘101_food_classes_10_percent.zip’ saved [1625420029/1625420029]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Checking the data"
      ],
      "metadata": {
        "id": "F_vDux56QNMQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# How many images/classes are there?\n",
        "walk_through_dir(\"101_food_classes_10_percent\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jdmmm77GQqP1",
        "outputId": "a801180a-9af1-40dc-adc3-2fbae8820294"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "There are 2 directories and 0 images in '101_food_classes_10_percent'.\n",
            "There are 101 directories and 0 images in '101_food_classes_10_percent/test'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/beef_tartare'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/clam_chowder'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/poutine'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/steak'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/french_onion_soup'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/miso_soup'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/edamame'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/peking_duck'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pancakes'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/spaghetti_carbonara'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/chocolate_cake'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/bread_pudding'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/churros'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/filet_mignon'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/chocolate_mousse'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/cup_cakes'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/huevos_rancheros'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/nachos'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/macaroni_and_cheese'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/onion_rings'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/caprese_salad'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/ravioli'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/frozen_yogurt'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/grilled_salmon'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/bibimbap'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/scallops'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/prime_rib'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/chicken_wings'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/donuts'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/greek_salad'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/cheese_plate'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/breakfast_burrito'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/lasagna'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/sashimi'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/takoyaki'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/cheesecake'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/foie_gras'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/gyoza'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/paella'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/beef_carpaccio'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/panna_cotta'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/macarons'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/fried_calamari'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pho'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/croque_madame'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/escargots'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/eggs_benedict'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/omelette'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/oysters'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/dumplings'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/carrot_cake'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/tacos'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/seaweed_salad'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/crab_cakes'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/guacamole'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pad_thai'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/grilled_cheese_sandwich'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/french_fries'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/ramen'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/creme_brulee'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/gnocchi'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/hamburger'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/chicken_quesadilla'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/sushi'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/spring_rolls'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/apple_pie'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pizza'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/chicken_curry'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/fried_rice'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/strawberry_shortcake'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/ice_cream'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/hot_and_sour_soup'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/lobster_bisque'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/tuna_tartare'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/red_velvet_cake'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pulled_pork_sandwich'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/baby_back_ribs'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/pork_chop'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/risotto'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/shrimp_and_grits'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/hummus'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/french_toast'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/bruschetta'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/hot_dog'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/mussels'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/deviled_eggs'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/fish_and_chips'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/club_sandwich'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/spaghetti_bolognese'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/cannoli'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/garlic_bread'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/tiramisu'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/samosa'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/ceviche'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/baklava'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/waffles'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/falafel'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/lobster_roll_sandwich'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/beet_salad'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/beignets'.\n",
            "There are 0 directories and 250 images in '101_food_classes_10_percent/test/caesar_salad'.\n",
            "There are 101 directories and 0 images in '101_food_classes_10_percent/train'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/beef_tartare'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/clam_chowder'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/poutine'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/steak'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/french_onion_soup'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/miso_soup'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/edamame'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/peking_duck'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pancakes'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/spaghetti_carbonara'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/chocolate_cake'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/bread_pudding'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/churros'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/filet_mignon'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/chocolate_mousse'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/cup_cakes'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/huevos_rancheros'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/nachos'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/macaroni_and_cheese'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/onion_rings'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/caprese_salad'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/ravioli'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/frozen_yogurt'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/grilled_salmon'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/bibimbap'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/scallops'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/prime_rib'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/chicken_wings'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/donuts'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/greek_salad'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/cheese_plate'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/breakfast_burrito'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/lasagna'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/sashimi'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/takoyaki'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/cheesecake'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/foie_gras'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/gyoza'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/paella'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/beef_carpaccio'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/panna_cotta'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/macarons'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/fried_calamari'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pho'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/croque_madame'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/escargots'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/eggs_benedict'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/omelette'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/oysters'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/dumplings'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/carrot_cake'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/tacos'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/seaweed_salad'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/crab_cakes'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/guacamole'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pad_thai'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/grilled_cheese_sandwich'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/french_fries'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/ramen'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/creme_brulee'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/gnocchi'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/hamburger'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/chicken_quesadilla'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/sushi'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/spring_rolls'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/apple_pie'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pizza'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/chicken_curry'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/fried_rice'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/strawberry_shortcake'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/ice_cream'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/hot_and_sour_soup'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/lobster_bisque'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/tuna_tartare'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/red_velvet_cake'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pulled_pork_sandwich'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/baby_back_ribs'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/pork_chop'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/risotto'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/shrimp_and_grits'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/hummus'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/french_toast'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/bruschetta'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/hot_dog'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/mussels'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/deviled_eggs'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/fish_and_chips'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/club_sandwich'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/spaghetti_bolognese'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/cannoli'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/garlic_bread'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/tiramisu'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/samosa'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/ceviche'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/baklava'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/waffles'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/falafel'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/lobster_roll_sandwich'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/beet_salad'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/beignets'.\n",
            "There are 0 directories and 75 images in '101_food_classes_10_percent/train/caesar_salad'.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Setup data inputs\n",
        "import tensorflow as tf\n",
        "IMG_SIZE = (224, 224)\n",
        "train_data_all_10_percent = tf.keras.preprocessing.image_dataset_from_directory(train_dir,\n",
        "                                                                                label_mode='categorical',\n",
        "                                                                                image_size=IMG_SIZE)\n",
        "\n",
        "test_data = tf.keras.preprocessing.image_dataset_from_directory(test_dir,\n",
        "                                                                label_mode='categorical',\n",
        "                                                                image_size=IMG_SIZE,\n",
        "                                                                shuffle=False) # don't shuffle test data for prediction analysis "
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tIipFNauQ-9d",
        "outputId": "f1b9763c-e03e-420f-f163-6f0786225a0f"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 7575 files belonging to 101 classes.\n",
            "Found 25250 files belonging to 101 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Train a big model with transfer learning on 10% of the 101 Food Classes\n",
        "\n",
        "The steps that we will take:\n",
        "* Create a ModelCheckpoint callback\n",
        "* Create a data augmentation layer to build data augmentation right into the model\n",
        "* Build a headless (no top layers) Functional EfficientNetB0 backboned-model (we will create our output layer)\n",
        "* Compile the model\n",
        "* Feature extract for 5 full passes (5 epochs on the train dataset and validate on 15% of the test data, to save epoch time)\n",
        " "
      ],
      "metadata": {
        "id": "RWxY-mT8SS4_"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Create checkpoint callback"
      ],
      "metadata": {
        "id": "tPdvSb_ug2IK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "checkpoint_path = \"101_classes_10_percent_model_checkpoint\"\n",
        "checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(checkpoint_path,\n",
        "                                                         save_weights_only=True,\n",
        "                                                         monitor='val_accuracy',\n",
        "                                                         save_best_only=True) "
      ],
      "metadata": {
        "id": "YQ5H1jIXejqX"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Create data augmentation"
      ],
      "metadata": {
        "id": "9hb6AZS8gvd3"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.layers.experimental import preprocessing\n",
        "\n",
        "from tensorflow.keras.models import Sequential \n",
        "\n",
        "# Setup data augmentation\n",
        "data_augmentation = Sequential([\n",
        "                      preprocessing.RandomFlip(\"horizontal\"),\n",
        "                      preprocessing.RandomRotation(0.2),\n",
        "                      preprocessing.RandomHeight(0.2),\n",
        "                      preprocessing.RandomWidth(0.2),\n",
        "                      preprocessing.RandomZoom(0.2),\n",
        "                     # preprocessing.Rescaling(1/255.) # rescale inputs of images to between 0 & 1, required for models like ResNet50\n",
        "                     \n",
        "                      ],name=\"data-augmentation\")"
      ],
      "metadata": {
        "id": "Lr6QcB32g_tk"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Set up the base model and freeze it's layers (this will extract features)"
      ],
      "metadata": {
        "id": "nJDEPSy_kLy9"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "base_model = tf.keras.applications.EfficientNetB0(include_top=False)\n",
        "base_model.trainable = False\n",
        "\n",
        "# Setup model architecture with trainable top layers\n",
        "\n",
        "inputs = layers.Input(shape=(224,224,3), name='input_layer')\n",
        "x = data_augmentation(inputs) # augment images (only happens during training phase)\n",
        "x = base_model(x, training=False) # put the base model in inference mode so weights which need to stay frozen, stay frozen.\n",
        "x = layers.GlobalAveragePooling2D(name='global_avg_pool_layer')(x)\n",
        "outputs = layers.Dense(len(train_data_all_10_percent.class_names), activation='softmax', name='output_layer')(x)\n",
        "model = tf.keras.Model(inputs, outputs)"
      ],
      "metadata": {
        "id": "ru1hHQaJBrgA",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ea00e75d-ef90-49db-8321-9c4dc7eebbb9"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/keras-applications/efficientnetb0_notop.h5\n",
            "16711680/16705208 [==============================] - 0s 0us/step\n",
            "16719872/16705208 [==============================] - 0s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Get a summary of the model we have created\n",
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Uu4a6fzXCn9H",
        "outputId": "89ef640e-d4e8-4cdd-b3b3-8d1aba570f87"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_layer (InputLayer)    [(None, 224, 224, 3)]     0         \n",
            "                                                                 \n",
            " data-augmentation (Sequenti  (None, None, None, 3)    0         \n",
            " al)                                                             \n",
            "                                                                 \n",
            " efficientnetb0 (Functional)  (None, None, None, 1280)  4049571  \n",
            "                                                                 \n",
            " global_avg_pool_layer (Glob  (None, 1280)             0         \n",
            " alAveragePooling2D)                                             \n",
            "                                                                 \n",
            " output_layer (Dense)        (None, 101)               129381    \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 4,178,952\n",
            "Trainable params: 129,381\n",
            "Non-trainable params: 4,049,571\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Compile the Model"
      ],
      "metadata": {
        "id": "d0UhloTJDUCB"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer=tf.keras.optimizers.Adam(),\n",
        "              metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "kJJwAUWOECvD"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Fit the model"
      ],
      "metadata": {
        "id": "nb0A30f_EipB"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "history_101_classes_10_percent_model = model.fit(train_data_all_10_percent,\n",
        "                                                 epochs=5,\n",
        "                                                 validation_data=test_data,\n",
        "                                                 validation_steps=int(0.15*len(test_data)),\n",
        "                                                 callbacks=[checkpoint_callback])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZK_ETMHYEmIN",
        "outputId": "5cd611cc-63ff-485e-f93b-35fbc5cec696"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "237/237 [==============================] - 88s 309ms/step - loss: 3.4646 - accuracy: 0.2463 - val_loss: 2.6860 - val_accuracy: 0.4092\n",
            "Epoch 2/5\n",
            "237/237 [==============================] - 58s 240ms/step - loss: 2.3587 - accuracy: 0.4595 - val_loss: 2.2486 - val_accuracy: 0.4746\n",
            "Epoch 3/5\n",
            "237/237 [==============================] - 53s 223ms/step - loss: 1.9744 - accuracy: 0.5304 - val_loss: 2.1549 - val_accuracy: 0.4680\n",
            "Epoch 4/5\n",
            "237/237 [==============================] - 50s 210ms/step - loss: 1.7641 - accuracy: 0.5706 - val_loss: 2.0162 - val_accuracy: 0.4960\n",
            "Epoch 5/5\n",
            "237/237 [==============================] - 47s 196ms/step - loss: 1.6031 - accuracy: 0.6049 - val_loss: 2.0077 - val_accuracy: 0.4865\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Evaluate the model "
      ],
      "metadata": {
        "id": "Nj-hfwedLimi"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "feature_extraction_results = model.evaluate(test_data)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2hN8gfL5F0RP",
        "outputId": "37731ea9-4259-4121-d2ea-b7c5bff1a0ce"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "790/790 [==============================] - 59s 75ms/step - loss: 1.7191 - accuracy: 0.5520\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "feature_extraction_results"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EmDtTEn5KWmR",
        "outputId": "8643d847-9641-4174-8e01-42b7902443ae"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.7190937995910645, 0.5519604086875916]"
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plot_loss_curves(history_101_classes_10_percent_model)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "id": "eERzKH19LLlG",
        "outputId": "617ca7db-2746-4a4b-e0f0-4c53b2670dcc"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Fine-tuning\n",
        "\n",
        "Unfreeze all of the layers in the base model."
      ],
      "metadata": {
        "id": "jvBtUHZFLwDg"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "base_model.trainable = True\n",
        "\n",
        "# Refreeze every layer except the last 5\n",
        "for layer in base_model.layers[:-5]:\n",
        "  layer.trainable = False\n"
      ],
      "metadata": {
        "id": "YzCQM5Tbo3Wi"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Recompile model with a lower learning rate\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), #learning rate lowered by 10x \n",
        "              metrics=['accuracy'])"
      ],
      "metadata": {
        "id": "EpH0N7HxpoeK"
      },
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# What layers in the model are trainable?\n",
        "for layer in model.layers:\n",
        "  print(layer.name, layer.trainable)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8A2uZ7_ZstSO",
        "outputId": "7d565102-6e02-4f99-c6fe-182a1da8ff16"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "input_layer True\n",
            "data-augmentation True\n",
            "efficientnetb0 True\n",
            "global_avg_pool_layer True\n",
            "output_layer True\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Check which layers are trainable in the base_model\n",
        "for layer_number, layer in enumerate(base_model.layers):\n",
        "  print(layer_number, layer.name, layer.trainable)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fljtoh6Ts_g6",
        "outputId": "ba9ed9d5-ddfb-4e5f-9a48-8060e1b87003"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0 input_1 False\n",
            "1 rescaling False\n",
            "2 normalization False\n",
            "3 stem_conv_pad False\n",
            "4 stem_conv False\n",
            "5 stem_bn False\n",
            "6 stem_activation False\n",
            "7 block1a_dwconv False\n",
            "8 block1a_bn False\n",
            "9 block1a_activation False\n",
            "10 block1a_se_squeeze False\n",
            "11 block1a_se_reshape False\n",
            "12 block1a_se_reduce False\n",
            "13 block1a_se_expand False\n",
            "14 block1a_se_excite False\n",
            "15 block1a_project_conv False\n",
            "16 block1a_project_bn False\n",
            "17 block2a_expand_conv False\n",
            "18 block2a_expand_bn False\n",
            "19 block2a_expand_activation False\n",
            "20 block2a_dwconv_pad False\n",
            "21 block2a_dwconv False\n",
            "22 block2a_bn False\n",
            "23 block2a_activation False\n",
            "24 block2a_se_squeeze False\n",
            "25 block2a_se_reshape False\n",
            "26 block2a_se_reduce False\n",
            "27 block2a_se_expand False\n",
            "28 block2a_se_excite False\n",
            "29 block2a_project_conv False\n",
            "30 block2a_project_bn False\n",
            "31 block2b_expand_conv False\n",
            "32 block2b_expand_bn False\n",
            "33 block2b_expand_activation False\n",
            "34 block2b_dwconv False\n",
            "35 block2b_bn False\n",
            "36 block2b_activation False\n",
            "37 block2b_se_squeeze False\n",
            "38 block2b_se_reshape False\n",
            "39 block2b_se_reduce False\n",
            "40 block2b_se_expand False\n",
            "41 block2b_se_excite False\n",
            "42 block2b_project_conv False\n",
            "43 block2b_project_bn False\n",
            "44 block2b_drop False\n",
            "45 block2b_add False\n",
            "46 block3a_expand_conv False\n",
            "47 block3a_expand_bn False\n",
            "48 block3a_expand_activation False\n",
            "49 block3a_dwconv_pad False\n",
            "50 block3a_dwconv False\n",
            "51 block3a_bn False\n",
            "52 block3a_activation False\n",
            "53 block3a_se_squeeze False\n",
            "54 block3a_se_reshape False\n",
            "55 block3a_se_reduce False\n",
            "56 block3a_se_expand False\n",
            "57 block3a_se_excite False\n",
            "58 block3a_project_conv False\n",
            "59 block3a_project_bn False\n",
            "60 block3b_expand_conv False\n",
            "61 block3b_expand_bn False\n",
            "62 block3b_expand_activation False\n",
            "63 block3b_dwconv False\n",
            "64 block3b_bn False\n",
            "65 block3b_activation False\n",
            "66 block3b_se_squeeze False\n",
            "67 block3b_se_reshape False\n",
            "68 block3b_se_reduce False\n",
            "69 block3b_se_expand False\n",
            "70 block3b_se_excite False\n",
            "71 block3b_project_conv False\n",
            "72 block3b_project_bn False\n",
            "73 block3b_drop False\n",
            "74 block3b_add False\n",
            "75 block4a_expand_conv False\n",
            "76 block4a_expand_bn False\n",
            "77 block4a_expand_activation False\n",
            "78 block4a_dwconv_pad False\n",
            "79 block4a_dwconv False\n",
            "80 block4a_bn False\n",
            "81 block4a_activation False\n",
            "82 block4a_se_squeeze False\n",
            "83 block4a_se_reshape False\n",
            "84 block4a_se_reduce False\n",
            "85 block4a_se_expand False\n",
            "86 block4a_se_excite False\n",
            "87 block4a_project_conv False\n",
            "88 block4a_project_bn False\n",
            "89 block4b_expand_conv False\n",
            "90 block4b_expand_bn False\n",
            "91 block4b_expand_activation False\n",
            "92 block4b_dwconv False\n",
            "93 block4b_bn False\n",
            "94 block4b_activation False\n",
            "95 block4b_se_squeeze False\n",
            "96 block4b_se_reshape False\n",
            "97 block4b_se_reduce False\n",
            "98 block4b_se_expand False\n",
            "99 block4b_se_excite False\n",
            "100 block4b_project_conv False\n",
            "101 block4b_project_bn False\n",
            "102 block4b_drop False\n",
            "103 block4b_add False\n",
            "104 block4c_expand_conv False\n",
            "105 block4c_expand_bn False\n",
            "106 block4c_expand_activation False\n",
            "107 block4c_dwconv False\n",
            "108 block4c_bn False\n",
            "109 block4c_activation False\n",
            "110 block4c_se_squeeze False\n",
            "111 block4c_se_reshape False\n",
            "112 block4c_se_reduce False\n",
            "113 block4c_se_expand False\n",
            "114 block4c_se_excite False\n",
            "115 block4c_project_conv False\n",
            "116 block4c_project_bn False\n",
            "117 block4c_drop False\n",
            "118 block4c_add False\n",
            "119 block5a_expand_conv False\n",
            "120 block5a_expand_bn False\n",
            "121 block5a_expand_activation False\n",
            "122 block5a_dwconv False\n",
            "123 block5a_bn False\n",
            "124 block5a_activation False\n",
            "125 block5a_se_squeeze False\n",
            "126 block5a_se_reshape False\n",
            "127 block5a_se_reduce False\n",
            "128 block5a_se_expand False\n",
            "129 block5a_se_excite False\n",
            "130 block5a_project_conv False\n",
            "131 block5a_project_bn False\n",
            "132 block5b_expand_conv False\n",
            "133 block5b_expand_bn False\n",
            "134 block5b_expand_activation False\n",
            "135 block5b_dwconv False\n",
            "136 block5b_bn False\n",
            "137 block5b_activation False\n",
            "138 block5b_se_squeeze False\n",
            "139 block5b_se_reshape False\n",
            "140 block5b_se_reduce False\n",
            "141 block5b_se_expand False\n",
            "142 block5b_se_excite False\n",
            "143 block5b_project_conv False\n",
            "144 block5b_project_bn False\n",
            "145 block5b_drop False\n",
            "146 block5b_add False\n",
            "147 block5c_expand_conv False\n",
            "148 block5c_expand_bn False\n",
            "149 block5c_expand_activation False\n",
            "150 block5c_dwconv False\n",
            "151 block5c_bn False\n",
            "152 block5c_activation False\n",
            "153 block5c_se_squeeze False\n",
            "154 block5c_se_reshape False\n",
            "155 block5c_se_reduce False\n",
            "156 block5c_se_expand False\n",
            "157 block5c_se_excite False\n",
            "158 block5c_project_conv False\n",
            "159 block5c_project_bn False\n",
            "160 block5c_drop False\n",
            "161 block5c_add False\n",
            "162 block6a_expand_conv False\n",
            "163 block6a_expand_bn False\n",
            "164 block6a_expand_activation False\n",
            "165 block6a_dwconv_pad False\n",
            "166 block6a_dwconv False\n",
            "167 block6a_bn False\n",
            "168 block6a_activation False\n",
            "169 block6a_se_squeeze False\n",
            "170 block6a_se_reshape False\n",
            "171 block6a_se_reduce False\n",
            "172 block6a_se_expand False\n",
            "173 block6a_se_excite False\n",
            "174 block6a_project_conv False\n",
            "175 block6a_project_bn False\n",
            "176 block6b_expand_conv False\n",
            "177 block6b_expand_bn False\n",
            "178 block6b_expand_activation False\n",
            "179 block6b_dwconv False\n",
            "180 block6b_bn False\n",
            "181 block6b_activation False\n",
            "182 block6b_se_squeeze False\n",
            "183 block6b_se_reshape False\n",
            "184 block6b_se_reduce False\n",
            "185 block6b_se_expand False\n",
            "186 block6b_se_excite False\n",
            "187 block6b_project_conv False\n",
            "188 block6b_project_bn False\n",
            "189 block6b_drop False\n",
            "190 block6b_add False\n",
            "191 block6c_expand_conv False\n",
            "192 block6c_expand_bn False\n",
            "193 block6c_expand_activation False\n",
            "194 block6c_dwconv False\n",
            "195 block6c_bn False\n",
            "196 block6c_activation False\n",
            "197 block6c_se_squeeze False\n",
            "198 block6c_se_reshape False\n",
            "199 block6c_se_reduce False\n",
            "200 block6c_se_expand False\n",
            "201 block6c_se_excite False\n",
            "202 block6c_project_conv False\n",
            "203 block6c_project_bn False\n",
            "204 block6c_drop False\n",
            "205 block6c_add False\n",
            "206 block6d_expand_conv False\n",
            "207 block6d_expand_bn False\n",
            "208 block6d_expand_activation False\n",
            "209 block6d_dwconv False\n",
            "210 block6d_bn False\n",
            "211 block6d_activation False\n",
            "212 block6d_se_squeeze False\n",
            "213 block6d_se_reshape False\n",
            "214 block6d_se_reduce False\n",
            "215 block6d_se_expand False\n",
            "216 block6d_se_excite False\n",
            "217 block6d_project_conv False\n",
            "218 block6d_project_bn False\n",
            "219 block6d_drop False\n",
            "220 block6d_add False\n",
            "221 block7a_expand_conv False\n",
            "222 block7a_expand_bn False\n",
            "223 block7a_expand_activation False\n",
            "224 block7a_dwconv False\n",
            "225 block7a_bn False\n",
            "226 block7a_activation False\n",
            "227 block7a_se_squeeze False\n",
            "228 block7a_se_reshape False\n",
            "229 block7a_se_reduce False\n",
            "230 block7a_se_expand False\n",
            "231 block7a_se_excite False\n",
            "232 block7a_project_conv True\n",
            "233 block7a_project_bn True\n",
            "234 top_conv True\n",
            "235 top_bn True\n",
            "236 top_activation True\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Fine tune for 5 more epochs\n",
        "fine_tune_epochs = 10 # model has already done 5 epochs (feature extraction), this is the total number of epochs we are after\n",
        "\n",
        "# Fine-tune the model \n",
        "history_all_classes_10_percent_fine_tune = model.fit(train_data_all_10_percent,\n",
        "                                                     epochs = fine_tune_epochs,\n",
        "                                                     validation_data = test_data,\n",
        "                                                     validation_steps = int(0.15* len(test_data)),\n",
        "                                                     initial_epoch=history_101_classes_10_percent_model.epoch[-1])"
      ],
      "metadata": {
        "id": "jkHktEYpt9zh",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d06dd7b9-b07a-4d31-df23-a83620ad9a47"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 5/10\n",
            "237/237 [==============================] - 53s 197ms/step - loss: 1.3582 - accuracy: 0.6547 - val_loss: 1.9657 - val_accuracy: 0.4971\n",
            "Epoch 6/10\n",
            "237/237 [==============================] - 44s 184ms/step - loss: 1.2240 - accuracy: 0.6756 - val_loss: 1.9650 - val_accuracy: 0.5000\n",
            "Epoch 7/10\n",
            "237/237 [==============================] - 42s 174ms/step - loss: 1.1652 - accuracy: 0.6903 - val_loss: 1.9390 - val_accuracy: 0.5079\n",
            "Epoch 8/10\n",
            "237/237 [==============================] - 42s 175ms/step - loss: 1.0794 - accuracy: 0.7154 - val_loss: 1.9251 - val_accuracy: 0.5098\n",
            "Epoch 9/10\n",
            "237/237 [==============================] - 44s 184ms/step - loss: 1.0351 - accuracy: 0.7284 - val_loss: 1.9501 - val_accuracy: 0.5074\n",
            "Epoch 10/10\n",
            "237/237 [==============================] - 41s 172ms/step - loss: 0.9571 - accuracy: 0.7440 - val_loss: 1.9181 - val_accuracy: 0.5082\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plot_loss_curves(history_all_classes_10_percent_fine_tune)"
      ],
      "metadata": {
        "id": "8juP8JYV_4ow",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "outputId": "bcaac943-6f11-418b-a1e8-82800a0fbd42"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate on the whole test set\n",
        "all_classes_10_percent_fine_tune_results = model.evaluate(test_data)\n",
        "all_classes_10_percent_fine_tune_results"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ScLdWRLqABH6",
        "outputId": "6d132d4e-f8ff-44d1-fff0-a7fd11ed4e58"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "790/790 [==============================] - 62s 78ms/step - loss: 1.6151 - accuracy: 0.5771\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.6150940656661987, 0.5770692825317383]"
            ]
          },
          "metadata": {},
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "compare_historys(original_history=history_101_classes_10_percent_model,\n",
        "                 new_history=history_all_classes_10_percent_fine_tune,\n",
        "                 initial_epochs=5)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 513
        },
        "id": "ExY5SRMeAgWi",
        "outputId": "c074fd65-94eb-4b2f-c2ad-1eb1d1fe04e9"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Saving and Loading a trained model\n",
        "\n",
        "To use a model in an external application, we will need to save it and export it somewhere."
      ],
      "metadata": {
        "id": "Uoenc5KkBFDi"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Save the fine-tuned model\n"
      ],
      "metadata": {
        "id": "Q1XySmIGQX4k"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model.save(\"drive/MyDrive/tensorflow_course/101_food_classes_10_percent_saved_big_model\", save_format='h5')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LwemcdcBQkcQ",
        "outputId": "d2c0992e-ee5a-40cf-ea15-80f5522e67f3"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/keras/engine/functional.py:1410: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
            "  layer_config = serialize_layer_fn(layer)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Load and evaluate a saved model"
      ],
      "metadata": {
        "id": "cem3rZ9bRrqE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "loaded_model = tf.keras.models.load_model(\"drive/MyDrive/tensorflow_course/101_food_classes_10_percent_saved_big_model\")"
      ],
      "metadata": {
        "id": "2YVWhAy1SyB-"
      },
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate loaded model and compare performance to pre-saved model\n",
        "loaded_model_results = loaded_model.evaluate(test_data)\n",
        "loaded_model_results"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JkR9jBIRVwDd",
        "outputId": "e9f12de6-a3e3-473e-e36f-a9d6baad568f"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "790/790 [==============================] - 60s 74ms/step - loss: 2.2101 - accuracy: 0.4487\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[2.2100837230682373, 0.44871288537979126]"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# The results(above) should be very similar to the results below:\n",
        "all_classes_10_percent_fine_tune_results = model.evaluate(test_data)\n",
        "all_classes_10_percent_fine_tune_results"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LPCElpK1WSF4",
        "outputId": "ea01a42d-15fc-472d-8a97-10650029c988"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "790/790 [==============================] - 58s 74ms/step - loss: 1.6151 - accuracy: 0.5771\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.6150940656661987, 0.5770692825317383]"
            ]
          },
          "metadata": {},
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Evaluating the performance of the big model across all different classes\n",
        "\n",
        "Making some predictions and then visualising them and then later find out which predictions were the \"most\" wrong."
      ],
      "metadata": {
        "id": "Klo-8Mn2Wz1z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "PavsKmFp2cvR"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}