2_4_2_sign_language_multiclass_classification.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0l5n9ToXGbRC"
      },
      "source": [
        "# Week 4: Multi-class Classification\n",
        "\n",
        "Welcome to this assignment! In this exercise, you will get a chance to work on a multi-class classification problem. You will be using the [Sign Language MNIST](https://www.kaggle.com/datamunge/sign-language-mnist) dataset, which contains 28x28 images of hands depicting the 26 letters of the english alphabet. \n",
        "\n",
        "You will need to pre-process the data so that it can be fed into your convolutional neural network to correctly classify each image as the letter it represents.\n",
        "\n",
        "\n",
        "Let's get started!"
      ],
      "id": "0l5n9ToXGbRC"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wYtuKeK0dImp",
        "tags": [
          "graded"
        ]
      },
      "outputs": [],
      "source": [
        "import csv\n",
        "import string\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator, array_to_img"
      ],
      "id": "wYtuKeK0dImp"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-zwAPcriH40C"
      },
      "source": [
        "Download the training and test sets (the test set will actually be used as a validation set):"
      ],
      "id": "-zwAPcriH40C"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lydsaiTvaRhI",
        "outputId": "57d2f877-fa30-4699-8f5d-8ecf4764a494",
        "tags": []
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/gdown/cli.py:131: FutureWarning: Option `--id` was deprecated in version 4.3.1 and will be removed in 5.0. You don't need to pass it anymore to use a file ID.\n",
            "  category=FutureWarning,\n",
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1z0DkA9BytlLxO1C0BAWzknLyQmZAp0HR\n",
            "To: /content/sign_mnist_train.csv\n",
            "100% 83.3M/83.3M [00:00<00:00, 135MB/s]\n",
            "/usr/local/lib/python3.7/dist-packages/gdown/cli.py:131: FutureWarning: Option `--id` was deprecated in version 4.3.1 and will be removed in 5.0. You don't need to pass it anymore to use a file ID.\n",
            "  category=FutureWarning,\n",
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1z1BIj4qmri59GWBG4ivMNFtpZ4AXIbzg\n",
            "To: /content/sign_mnist_test.csv\n",
            "100% 21.8M/21.8M [00:00<00:00, 101MB/s] \n"
          ]
        }
      ],
      "source": [
        "# sign_mnist_train.csv\n",
        "!gdown --id 1z0DkA9BytlLxO1C0BAWzknLyQmZAp0HR\n",
        "# sign_mnist_test.csv\n",
        "!gdown --id 1z1BIj4qmri59GWBG4ivMNFtpZ4AXIbzg"
      ],
      "id": "lydsaiTvaRhI"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tnVJZI96IMh0"
      },
      "source": [
        "Define some globals with the path to both files you just downloaded:"
      ],
      "id": "tnVJZI96IMh0"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "N3fQlI99JPw9",
        "tags": [
          "graded"
        ]
      },
      "outputs": [],
      "source": [
        "TRAINING_FILE = './sign_mnist_train.csv'\n",
        "VALIDATION_FILE = './sign_mnist_test.csv'"
      ],
      "id": "N3fQlI99JPw9"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iybvOtoHISHV"
      },
      "source": [
        "Unlike previous assignments, you will not have the actual images provided, instead you will have the data serialized as `csv` files.\n",
        "\n",
        "Take a look at how the data looks like within the `csv` file:"
      ],
      "id": "iybvOtoHISHV"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ftfeq3JYagks",
        "outputId": "e35bfdac-4454-4af5-deac-9fac91f94f3a",
        "tags": [
          "graded"
        ]
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "First line (header) looks like this:\n",
            "label,pixel1,pixel2,pixel3,pixel4,pixel5,pixel6,pixel7,pixel8,pixel9,pixel10,pixel11,pixel12,pixel13,pixel14,pixel15,pixel16,pixel17,pixel18,pixel19,pixel20,pixel21,pixel22,pixel23,pixel24,pixel25,pixel26,pixel27,pixel28,pixel29,pixel30,pixel31,pixel32,pixel33,pixel34,pixel35,pixel36,pixel37,pixel38,pixel39,pixel40,pixel41,pixel42,pixel43,pixel44,pixel45,pixel46,pixel47,pixel48,pixel49,pixel50,pixel51,pixel52,pixel53,pixel54,pixel55,pixel56,pixel57,pixel58,pixel59,pixel60,pixel61,pixel62,pixel63,pixel64,pixel65,pixel66,pixel67,pixel68,pixel69,pixel70,pixel71,pixel72,pixel73,pixel74,pixel75,pixel76,pixel77,pixel78,pixel79,pixel80,pixel81,pixel82,pixel83,pixel84,pixel85,pixel86,pixel87,pixel88,pixel89,pixel90,pixel91,pixel92,pixel93,pixel94,pixel95,pixel96,pixel97,pixel98,pixel99,pixel100,pixel101,pixel102,pixel103,pixel104,pixel105,pixel106,pixel107,pixel108,pixel109,pixel110,pixel111,pixel112,pixel113,pixel114,pixel115,pixel116,pixel117,pixel118,pixel119,pixel120,pixel121,pixel122,pixel123,pixel124,pixel125,pixel126,pixel127,pixel128,pixel129,pixel130,pixel131,pixel132,pixel133,pixel134,pixel135,pixel136,pixel137,pixel138,pixel139,pixel140,pixel141,pixel142,pixel143,pixel144,pixel145,pixel146,pixel147,pixel148,pixel149,pixel150,pixel151,pixel152,pixel153,pixel154,pixel155,pixel156,pixel157,pixel158,pixel159,pixel160,pixel161,pixel162,pixel163,pixel164,pixel165,pixel166,pixel167,pixel168,pixel169,pixel170,pixel171,pixel172,pixel173,pixel174,pixel175,pixel176,pixel177,pixel178,pixel179,pixel180,pixel181,pixel182,pixel183,pixel184,pixel185,pixel186,pixel187,pixel188,pixel189,pixel190,pixel191,pixel192,pixel193,pixel194,pixel195,pixel196,pixel197,pixel198,pixel199,pixel200,pixel201,pixel202,pixel203,pixel204,pixel205,pixel206,pixel207,pixel208,pixel209,pixel210,pixel211,pixel212,pixel213,pixel214,pixel215,pixel216,pixel217,pixel218,pixel219,pixel220,pixel221,pixel222,pixel223,pixel224,pixel225,pixel226,pixel227,pixel228,pixel229,pixel230,pixel231,pixel232,pixel233,pixel234,pixel235,pixel236,pixel237,pixel238,pixel239,pixel240,pixel241,pixel242,pixel243,pixel244,pixel245,pixel246,pixel247,pixel248,pixel249,pixel250,pixel251,pixel252,pixel253,pixel254,pixel255,pixel256,pixel257,pixel258,pixel259,pixel260,pixel261,pixel262,pixel263,pixel264,pixel265,pixel266,pixel267,pixel268,pixel269,pixel270,pixel271,pixel272,pixel273,pixel274,pixel275,pixel276,pixel277,pixel278,pixel279,pixel280,pixel281,pixel282,pixel283,pixel284,pixel285,pixel286,pixel287,pixel288,pixel289,pixel290,pixel291,pixel292,pixel293,pixel294,pixel295,pixel296,pixel297,pixel298,pixel299,pixel300,pixel301,pixel302,pixel303,pixel304,pixel305,pixel306,pixel307,pixel308,pixel309,pixel310,pixel311,pixel312,pixel313,pixel314,pixel315,pixel316,pixel317,pixel318,pixel319,pixel320,pixel321,pixel322,pixel323,pixel324,pixel325,pixel326,pixel327,pixel328,pixel329,pixel330,pixel331,pixel332,pixel333,pixel334,pixel335,pixel336,pixel337,pixel338,pixel339,pixel340,pixel341,pixel342,pixel343,pixel344,pixel345,pixel346,pixel347,pixel348,pixel349,pixel350,pixel351,pixel352,pixel353,pixel354,pixel355,pixel356,pixel357,pixel358,pixel359,pixel360,pixel361,pixel362,pixel363,pixel364,pixel365,pixel366,pixel367,pixel368,pixel369,pixel370,pixel371,pixel372,pixel373,pixel374,pixel375,pixel376,pixel377,pixel378,pixel379,pixel380,pixel381,pixel382,pixel383,pixel384,pixel385,pixel386,pixel387,pixel388,pixel389,pixel390,pixel391,pixel392,pixel393,pixel394,pixel395,pixel396,pixel397,pixel398,pixel399,pixel400,pixel401,pixel402,pixel403,pixel404,pixel405,pixel406,pixel407,pixel408,pixel409,pixel410,pixel411,pixel412,pixel413,pixel414,pixel415,pixel416,pixel417,pixel418,pixel419,pixel420,pixel421,pixel422,pixel423,pixel424,pixel425,pixel426,pixel427,pixel428,pixel429,pixel430,pixel431,pixel432,pixel433,pixel434,pixel435,pixel436,pixel437,pixel438,pixel439,pixel440,pixel441,pixel442,pixel443,pixel444,pixel445,pixel446,pixel447,pixel448,pixel449,pixel450,pixel451,pixel452,pixel453,pixel454,pixel455,pixel456,pixel457,pixel458,pixel459,pixel460,pixel461,pixel462,pixel463,pixel464,pixel465,pixel466,pixel467,pixel468,pixel469,pixel470,pixel471,pixel472,pixel473,pixel474,pixel475,pixel476,pixel477,pixel478,pixel479,pixel480,pixel481,pixel482,pixel483,pixel484,pixel485,pixel486,pixel487,pixel488,pixel489,pixel490,pixel491,pixel492,pixel493,pixel494,pixel495,pixel496,pixel497,pixel498,pixel499,pixel500,pixel501,pixel502,pixel503,pixel504,pixel505,pixel506,pixel507,pixel508,pixel509,pixel510,pixel511,pixel512,pixel513,pixel514,pixel515,pixel516,pixel517,pixel518,pixel519,pixel520,pixel521,pixel522,pixel523,pixel524,pixel525,pixel526,pixel527,pixel528,pixel529,pixel530,pixel531,pixel532,pixel533,pixel534,pixel535,pixel536,pixel537,pixel538,pixel539,pixel540,pixel541,pixel542,pixel543,pixel544,pixel545,pixel546,pixel547,pixel548,pixel549,pixel550,pixel551,pixel552,pixel553,pixel554,pixel555,pixel556,pixel557,pixel558,pixel559,pixel560,pixel561,pixel562,pixel563,pixel564,pixel565,pixel566,pixel567,pixel568,pixel569,pixel570,pixel571,pixel572,pixel573,pixel574,pixel575,pixel576,pixel577,pixel578,pixel579,pixel580,pixel581,pixel582,pixel583,pixel584,pixel585,pixel586,pixel587,pixel588,pixel589,pixel590,pixel591,pixel592,pixel593,pixel594,pixel595,pixel596,pixel597,pixel598,pixel599,pixel600,pixel601,pixel602,pixel603,pixel604,pixel605,pixel606,pixel607,pixel608,pixel609,pixel610,pixel611,pixel612,pixel613,pixel614,pixel615,pixel616,pixel617,pixel618,pixel619,pixel620,pixel621,pixel622,pixel623,pixel624,pixel625,pixel626,pixel627,pixel628,pixel629,pixel630,pixel631,pixel632,pixel633,pixel634,pixel635,pixel636,pixel637,pixel638,pixel639,pixel640,pixel641,pixel642,pixel643,pixel644,pixel645,pixel646,pixel647,pixel648,pixel649,pixel650,pixel651,pixel652,pixel653,pixel654,pixel655,pixel656,pixel657,pixel658,pixel659,pixel660,pixel661,pixel662,pixel663,pixel664,pixel665,pixel666,pixel667,pixel668,pixel669,pixel670,pixel671,pixel672,pixel673,pixel674,pixel675,pixel676,pixel677,pixel678,pixel679,pixel680,pixel681,pixel682,pixel683,pixel684,pixel685,pixel686,pixel687,pixel688,pixel689,pixel690,pixel691,pixel692,pixel693,pixel694,pixel695,pixel696,pixel697,pixel698,pixel699,pixel700,pixel701,pixel702,pixel703,pixel704,pixel705,pixel706,pixel707,pixel708,pixel709,pixel710,pixel711,pixel712,pixel713,pixel714,pixel715,pixel716,pixel717,pixel718,pixel719,pixel720,pixel721,pixel722,pixel723,pixel724,pixel725,pixel726,pixel727,pixel728,pixel729,pixel730,pixel731,pixel732,pixel733,pixel734,pixel735,pixel736,pixel737,pixel738,pixel739,pixel740,pixel741,pixel742,pixel743,pixel744,pixel745,pixel746,pixel747,pixel748,pixel749,pixel750,pixel751,pixel752,pixel753,pixel754,pixel755,pixel756,pixel757,pixel758,pixel759,pixel760,pixel761,pixel762,pixel763,pixel764,pixel765,pixel766,pixel767,pixel768,pixel769,pixel770,pixel771,pixel772,pixel773,pixel774,pixel775,pixel776,pixel777,pixel778,pixel779,pixel780,pixel781,pixel782,pixel783,pixel784\n",
            "\n",
            "Each subsequent line (data points) look like this:\n",
            "3,107,118,127,134,139,143,146,150,153,156,158,160,163,165,159,166,168,170,170,171,171,171,172,171,171,170,170,169,111,121,129,135,141,144,148,151,154,157,160,163,164,170,119,152,171,171,170,171,172,172,172,172,172,171,171,170,113,123,131,137,142,145,150,152,155,158,161,163,164,172,105,142,170,171,171,171,172,172,173,173,172,171,171,171,116,125,133,139,143,146,151,153,156,159,162,163,167,167,95,144,171,172,172,172,172,172,173,173,173,172,172,171,117,126,134,140,145,149,153,156,158,161,163,164,175,156,87,154,172,173,173,173,173,173,174,174,174,173,172,172,119,128,136,142,146,150,153,156,159,163,165,164,184,148,89,164,172,174,174,174,174,175,175,174,175,174,173,173,122,130,138,143,147,150,154,158,162,165,166,172,181,128,94,170,173,175,174,175,176,177,177,177,177,175,175,174,122,132,139,145,149,152,156,160,163,165,166,181,172,103,113,175,176,178,178,179,179,179,179,178,179,177,175,174,125,134,141,147,150,153,157,161,164,167,168,184,179,116,126,165,176,179,180,180,181,180,180,180,179,178,177,176,128,135,142,148,152,154,158,162,165,168,170,187,180,156,161,124,143,179,178,178,181,182,181,180,181,180,179,179,129,136,144,150,153,155,159,163,166,169,172,187,184,153,102,117,110,175,169,154,182,183,183,182,182,181,181,179,131,138,145,150,155,157,161,165,168,174,190,189,175,146,94,97,113,151,158,129,184,184,184,184,183,183,182,180,131,139,146,151,155,159,163,167,175,182,179,171,159,114,102,89,121,136,136,96,172,186,186,185,185,184,182,181,131,140,147,154,157,160,164,179,186,191,187,180,157,100,88,84,108,111,126,90,120,186,187,187,186,185,184,182,133,141,149,155,158,160,174,201,189,165,151,143,146,120,87,78,87,76,108,98,96,181,188,187,186,186,185,183,133,141,150,156,160,161,179,197,174,135,99,72,95,134,97,72,74,68,116,105,108,187,189,187,187,186,186,185,134,143,151,156,161,163,179,194,156,110,74,42,52,139,94,67,75,75,118,106,129,189,191,190,188,188,187,186,135,144,152,158,163,163,177,193,161,122,84,43,71,134,81,57,71,88,112,98,157,193,193,192,190,190,189,188,136,144,152,158,162,163,176,192,164,128,98,62,60,100,71,76,96,101,105,95,174,195,194,194,194,193,191,190,137,145,152,159,164,165,178,191,164,135,113,82,59,87,98,111,120,108,97,108,190,196,195,195,194,193,193,192,139,146,154,160,164,165,175,186,163,139,112,85,67,102,126,133,126,105,104,176,197,198,197,196,195,195,194,193,138,147,155,161,165,167,172,186,163,137,107,87,76,106,122,125,117,96,156,199,199,200,198,196,196,195,195,194,139,148,156,163,166,168,172,180,158,131,108,99,86,108,118,116,103,107,191,202,201,200,200,200,199,197,198,196,140,149,157,164,168,167,177,178,155,131,118,105,87,100,106,100,96,164,202,202,202,202,202,201,200,199,199,198,140,150,157,165,167,170,181,175,152,130,115,98,82,85,90,99,165,202,203,204,203,203,202,202,201,201,200,200,142,150,159,165,170,191,173,157,144,119,97,84,79,79,91,172,202,203,203,205,204,204,204,203,202,202,201,200,142,151,160,165,188,190,187,150,119,109,85,79,79,78,137,203,205,206,206,207,207,206,206,204,205,204,203,202,142,151,160,172,196,188,188,190,135,96,86,77,77,79,176,205,207,207,207,207,207,207,206,206,206,204,203,202\n",
            "\n"
          ]
        }
      ],
      "source": [
        "with open(TRAINING_FILE) as training_file:\n",
        "  line = training_file.readline()\n",
        "  print(f\"First line (header) looks like this:\\n{line}\")\n",
        "  line = training_file.readline()\n",
        "  print(f\"Each subsequent line (data points) look like this:\\n{line}\")"
      ],
      "id": "ftfeq3JYagks"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RH23LS7RHA-U"
      },
      "source": [
        "As you can see, each file includes a header (the first line) and each subsequent data point is represented as a line that contains 785 values. \n",
        "\n",
        "The first value is the label (the numeric representation of each letter) and the other 784 values are the value of each pixel of the image. Remember that the original images have a resolution of 28x28, which sums up to 784 pixels."
      ],
      "id": "RH23LS7RHA-U"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_Slnchp3JkCk"
      },
      "source": [
        " ## Parsing the dataset\n",
        " \n",
        " Now complete the `parse_data_from_input` below.\n",
        "\n",
        " This function should be able to read a file passed as input and return 2 numpy arrays, one containing the labels and one containing the 28x28 representation of each image within the file. These numpy arrays should have type `float64`.\n",
        "\n",
        " A couple of things to keep in mind:\n",
        " \n",
        "- The first line contains the column headers, so you should ignore it.\n",
        "\n",
        "- Each successive line contains 785 comma-separated values between 0 and 255\n",
        "  - The first value is the label\n",
        "\n",
        "  - The rest are the pixel values for that picture\n",
        "\n",
        "  \n",
        "**Hint**:\n",
        "\n",
        "You have two options to solve this function. \n",
        "  \n",
        "   - 1. One is to use `csv.reader` and create a for loop that reads from it, if you take this approach take this into consideration:\n",
        "\n",
        "        - `csv.reader` returns an iterable that returns a row of the csv file in each iteration.\n",
        "    Following this convention, row[0] has the label and row[1:] has the 784 pixel values.\n",
        "\n",
        "        - To reshape the arrays (going from 784 to 28x28), you can use functions such as [`np.array_split`](https://numpy.org/doc/stable/reference/generated/numpy.array_split.html) or [`np.reshape`](https://numpy.org/doc/stable/reference/generated/numpy.reshape.html).\n",
        "\n",
        "        - For type conversion of the numpy arrays, use the method [`np.ndarray.astype`](https://numpy.org/doc/stable/reference/generated/numpy.ndarray.astype.html).\n",
        "\n",
        "\n",
        "   - 2. The other one is to use `np.loadtxt`. You can find the documentation [here](https://numpy.org/doc/stable/reference/generated/numpy.loadtxt.html).\n",
        "   \n",
        "   \n",
        "Regardless of the method you chose, your function should finish its execution in under 1 minute. If you see that your function is taking a long time to run, try changing your implementation."
      ],
      "id": "_Slnchp3JkCk"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "cellView": "code",
        "id": "EO3u5NIhdNgE",
        "lines_to_next_cell": 2,
        "tags": [
          "graded"
        ]
      },
      "outputs": [],
      "source": [
        "# GRADED FUNCTION: parse_data_from_input\n",
        "def parse_data_from_input(filename):\n",
        "  \"\"\"\n",
        "  Parses the images and labels from a CSV file\n",
        "  \n",
        "  Args:\n",
        "    filename (string): path to the CSV file\n",
        "    \n",
        "  Returns:\n",
        "    images, labels: tuple of numpy arrays containing the images and labels\n",
        "  \"\"\"\n",
        "  with open(filename) as file:\n",
        "    ### START CODE HERE\n",
        "\n",
        "    # Use csv.reader, passing in the appropriate delimiter\n",
        "    # Remember that csv.reader can be iterated and returns one line in each iteration\n",
        "    csv_reader = csv.reader(file, delimiter=',')\n",
        "    first_line=True\n",
        "    temp_labels=[]\n",
        "    temp_images=[]\n",
        "    for i in csv_reader:\n",
        "      if first_line:\n",
        "        first_line=False\n",
        "      else:\n",
        "        temp_labels.append(i[0])\n",
        "        img_data=i[1:785]\n",
        "        img_data_arr=np.array_split(img_data, 28)\n",
        "        temp_images.append(img_data_arr)\n",
        "    \n",
        "    labels = np.array(temp_labels).astype('float')\n",
        "    images = np.array(temp_images).astype('float')\n",
        "\n",
        "    \n",
        "    ### END CODE HERE\n",
        "\n",
        "    return images, labels"
      ],
      "id": "EO3u5NIhdNgE"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HeSMXBTUsnhl",
        "outputId": "194d5d58-585e-4b57-f5fe-fbeb3e74ec27",
        "tags": [
          "graded"
        ]
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Training images has shape: (27455, 28, 28) and dtype: float64\n",
            "Training labels has shape: (27455,) and dtype: float64\n",
            "Validation images has shape: (7172, 28, 28) and dtype: float64\n",
            "Validation labels has shape: (7172,) and dtype: float64\n"
          ]
        }
      ],
      "source": [
        "# Test your function\n",
        "training_images, training_labels = parse_data_from_input(TRAINING_FILE)\n",
        "validation_images, validation_labels = parse_data_from_input(VALIDATION_FILE)\n",
        "\n",
        "print(f\"Training images has shape: {training_images.shape} and dtype: {training_images.dtype}\")\n",
        "print(f\"Training labels has shape: {training_labels.shape} and dtype: {training_labels.dtype}\")\n",
        "print(f\"Validation images has shape: {validation_images.shape} and dtype: {validation_images.dtype}\")\n",
        "print(f\"Validation labels has shape: {validation_labels.shape} and dtype: {validation_labels.dtype}\")"
      ],
      "id": "HeSMXBTUsnhl"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xwqJykm1JqvK"
      },
      "source": [
        "**Expected Output:**\n",
        "```\n",
        "Training images has shape: (27455, 28, 28) and dtype: float64\n",
        "Training labels has shape: (27455,) and dtype: float64\n",
        "Validation images has shape: (7172, 28, 28) and dtype: float64\n",
        "Validation labels has shape: (7172,) and dtype: float64\n",
        "```"
      ],
      "id": "xwqJykm1JqvK"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tURDCvUKsXIw"
      },
      "source": [
        "## Visualizing the numpy arrays\n",
        "\n",
        "Now that you have converted the initial csv data into a format that is compatible with computer vision tasks, take a moment to actually see how the images of the dataset look like:"
      ],
      "id": "tURDCvUKsXIw"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 150
        },
        "id": "BsAH3kVWB1ye",
        "outputId": "f471e14c-d836-475b-9a3b-5296a66c74c5",
        "tags": []
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1152x1080 with 10 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "# Plot a sample of 10 images from the training set\n",
        "def plot_categories(training_images, training_labels):\n",
        "  fig, axes = plt.subplots(1, 10, figsize=(16, 15))\n",
        "  axes = axes.flatten()\n",
        "  letters = list(string.ascii_lowercase)\n",
        "\n",
        "  for k in range(10):\n",
        "    img = training_images[k]\n",
        "    img = np.expand_dims(img, axis=-1)\n",
        "    img = array_to_img(img)\n",
        "    ax = axes[k]\n",
        "    ax.imshow(img, cmap=\"Greys_r\")\n",
        "    ax.set_title(f\"{letters[int(training_labels[k])]}\")\n",
        "    ax.set_axis_off()\n",
        "\n",
        "  plt.tight_layout()\n",
        "  plt.show()\n",
        "\n",
        "plot_categories(training_images, training_labels)"
      ],
      "id": "BsAH3kVWB1ye"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0QNkjIRCN5Kg"
      },
      "source": [
        "## Creating the generators for the CNN\n",
        "\n",
        "Now that you have successfully organized the data in a way that can be easily fed to Keras' `ImageDataGenerator`, it is time for you to code the generators that will yield batches of images, both for training and validation. For this complete the `train_val_generators` function below.\n",
        "\n",
        "Some important notes:\n",
        "\n",
        "- The images in this dataset come in the same resolution so you don't need to set a custom `target_size` in this case. In fact, you can't even do so because this time you will not be using the `flow_from_directory` method (as in previous assignments). Instead you will use the [`flow`](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator#flow) method.\n",
        "- You need to add the \"color\" dimension to the numpy arrays that encode the images. These are black and white images, so this new dimension should have a size of 1 (instead of 3, which is used when dealing with colored images). Take a look at the function [`np.expand_dims`](https://numpy.org/doc/stable/reference/generated/numpy.expand_dims.html) for this."
      ],
      "id": "0QNkjIRCN5Kg"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "cellView": "code",
        "id": "9ZDIvKGtNISO",
        "lines_to_next_cell": 2,
        "tags": [
          "graded"
        ]
      },
      "outputs": [],
      "source": [
        "# GRADED FUNCTION: train_val_generators\n",
        "def train_val_generators(training_images, training_labels, validation_images, validation_labels):\n",
        "  \"\"\"\n",
        "  Creates the training and validation data generators\n",
        "  \n",
        "  Args:\n",
        "    training_images (array): parsed images from the train CSV file\n",
        "    training_labels (array): parsed labels from the train CSV file\n",
        "    validation_images (array): parsed images from the test CSV file\n",
        "    validation_labels (array): parsed labels from the test CSV file\n",
        "    \n",
        "  Returns:\n",
        "    train_generator, validation_generator - tuple containing the generators\n",
        "  \"\"\"\n",
        "  ### START CODE HERE\n",
        "\n",
        "  # In this section you will have to add another dimension to the data\n",
        "  # So, for example, if your array is (10000, 28, 28)\n",
        "  # You will need to make it (10000, 28, 28, 1)\n",
        "  # Hint: np.expand_dims\n",
        "  training_images = np.expand_dims(training_images, axis=3)\n",
        "  validation_images = np.expand_dims(validation_images, axis=3)\n",
        "\n",
        "  # Instantiate the ImageDataGenerator class \n",
        "  # Don't forget to normalize pixel values \n",
        "  # and set arguments to augment the images (if desired)\n",
        "  train_datagen = ImageDataGenerator(\n",
        "    rescale=1. / 255,\n",
        "    rotation_range=40,\n",
        "    width_shift_range=0.2,\n",
        "    height_shift_range=0.2,\n",
        "    shear_range=0.2,\n",
        "    zoom_range=0.2,\n",
        "    horizontal_flip=True,\n",
        "    fill_mode='nearest')\n",
        "\n",
        "\n",
        "  # Pass in the appropriate arguments to the flow method\n",
        "  train_generator = train_datagen.flow(x=training_images,\n",
        "                                       y=training_labels,\n",
        "                                       batch_size=32) \n",
        "\n",
        "  \n",
        "  # Instantiate the ImageDataGenerator class (don't forget to set the rescale argument)\n",
        "  # Remember that validation data should not be augmented\n",
        "  validation_datagen = ImageDataGenerator(\n",
        "    rescale=1. / 255)\n",
        "\n",
        "  # Pass in the appropriate arguments to the flow method\n",
        "  validation_generator = validation_datagen.flow(x=validation_images,\n",
        "                                                 y=validation_labels,\n",
        "                                                 batch_size=32) \n",
        "\n",
        "  ### END CODE HERE\n",
        "\n",
        "  return train_generator, validation_generator"
      ],
      "id": "9ZDIvKGtNISO"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2oxxBZDwPozr",
        "outputId": "937614d7-a3a5-43dd-e88f-3ceeb68be82c",
        "tags": [
          "graded"
        ]
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Images of training generator have shape: (27455, 28, 28, 1)\n",
            "Labels of training generator have shape: (27455,)\n",
            "Images of validation generator have shape: (7172, 28, 28, 1)\n",
            "Labels of validation generator have shape: (7172,)\n"
          ]
        }
      ],
      "source": [
        "# Test your generators\n",
        "train_generator, validation_generator = train_val_generators(training_images, training_labels, validation_images, validation_labels)\n",
        "\n",
        "print(f\"Images of training generator have shape: {train_generator.x.shape}\")\n",
        "print(f\"Labels of training generator have shape: {train_generator.y.shape}\")\n",
        "print(f\"Images of validation generator have shape: {validation_generator.x.shape}\")\n",
        "print(f\"Labels of validation generator have shape: {validation_generator.y.shape}\")"
      ],
      "id": "2oxxBZDwPozr"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cok5oQa5Rknv"
      },
      "source": [
        "**Expected Output:**\n",
        "```\n",
        "Images of training generator have shape: (27455, 28, 28, 1)\n",
        "Labels of training generator have shape: (27455,)\n",
        "Images of validation generator have shape: (7172, 28, 28, 1)\n",
        "Labels of validation generator have shape: (7172,)\n",
        "```"
      ],
      "id": "cok5oQa5Rknv"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CkHUj4PsP_jT"
      },
      "source": [
        "## Coding the CNN\n",
        "\n",
        "One last step before training is to define the architecture of the model that will be trained.\n",
        "\n",
        "Complete the `create_model` function below. This function should return a Keras' model that uses the `Sequential` or the `Functional` API.\n",
        "\n",
        "The last layer of your model should have a number of units that corresponds to the number of possible categories, as well as the correct activation function.\n",
        "\n",
        "Aside from defining the architecture of the model, you should also compile it so make sure to use a `loss` function that is suitable for multi-class classification.\n",
        "\n",
        "**Note that you should use no more than 2 Conv2D and 2 MaxPooling2D layers to achieve the desired performance.**"
      ],
      "id": "CkHUj4PsP_jT"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "cellView": "code",
        "id": "Rmb7S32cgRqS",
        "tags": [
          "graded"
        ]
      },
      "outputs": [],
      "source": [
        "def create_model():\n",
        "\n",
        "  ### START CODE HERE       \n",
        "\n",
        "  # Define the model\n",
        "  # Use no more than 2 Conv2D and 2 MaxPooling2D\n",
        "  model = tf.keras.models.Sequential([\n",
        "    tf.keras.layers.Conv2D(64, (3, 3), activation='relu', input_shape=(28, 28, 1)),\n",
        "    tf.keras.layers.MaxPooling2D(2, 2),\n",
        "    tf.keras.layers.Conv2D(64, (3, 3), activation='relu'),\n",
        "    tf.keras.layers.MaxPooling2D(2, 2),\n",
        "    tf.keras.layers.Flatten(),\n",
        "    tf.keras.layers.Dense(128, activation=tf.nn.relu),\n",
        "    tf.keras.layers.Dense(26, activation=tf.nn.softmax)])\n",
        "  \n",
        "\n",
        "  model.compile(optimizer='adam',loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=['accuracy'])\n",
        "\n",
        "  ### END CODE HERE       \n",
        "  \n",
        "  return model"
      ],
      "id": "Rmb7S32cgRqS"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-e7ywtgZS5pU",
        "lines_to_next_cell": 2,
        "outputId": "749a60d9-33c5-4e0a-e3a9-004d3ea7ad4e",
        "tags": []
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/15\n",
            "858/858 [==============================] - 22s 13ms/step - loss: 3.2319 - accuracy: 0.0764 - val_loss: 3.1941 - val_accuracy: 0.1122\n",
            "Epoch 2/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 3.1874 - accuracy: 0.1295 - val_loss: 3.0975 - val_accuracy: 0.2310\n",
            "Epoch 3/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 3.1474 - accuracy: 0.1734 - val_loss: 3.0493 - val_accuracy: 0.2821\n",
            "Epoch 4/15\n",
            "858/858 [==============================] - 11s 13ms/step - loss: 3.1027 - accuracy: 0.2218 - val_loss: 3.0193 - val_accuracy: 0.3074\n",
            "Epoch 5/15\n",
            "858/858 [==============================] - 11s 13ms/step - loss: 3.0517 - accuracy: 0.2757 - val_loss: 2.9776 - val_accuracy: 0.3477\n",
            "Epoch 6/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 3.0087 - accuracy: 0.3171 - val_loss: 2.9244 - val_accuracy: 0.4060\n",
            "Epoch 7/15\n",
            "858/858 [==============================] - 11s 13ms/step - loss: 2.9762 - accuracy: 0.3503 - val_loss: 2.8638 - val_accuracy: 0.4626\n",
            "Epoch 8/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.9331 - accuracy: 0.3958 - val_loss: 2.8391 - val_accuracy: 0.4895\n",
            "Epoch 9/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.9069 - accuracy: 0.4225 - val_loss: 2.8280 - val_accuracy: 0.5004\n",
            "Epoch 10/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.8915 - accuracy: 0.4371 - val_loss: 2.8076 - val_accuracy: 0.5187\n",
            "Epoch 11/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.8628 - accuracy: 0.4667 - val_loss: 2.7772 - val_accuracy: 0.5501\n",
            "Epoch 12/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.8486 - accuracy: 0.4781 - val_loss: 2.8122 - val_accuracy: 0.5130\n",
            "Epoch 13/15\n",
            "858/858 [==============================] - 11s 13ms/step - loss: 2.8380 - accuracy: 0.4892 - val_loss: 2.8146 - val_accuracy: 0.5074\n",
            "Epoch 14/15\n",
            "858/858 [==============================] - 11s 12ms/step - loss: 2.8233 - accuracy: 0.5047 - val_loss: 2.7797 - val_accuracy: 0.5446\n",
            "Epoch 15/15\n",
            "858/858 [==============================] - 11s 13ms/step - loss: 2.8167 - accuracy: 0.5100 - val_loss: 2.7700 - val_accuracy: 0.5581\n"
          ]
        }
      ],
      "source": [
        "# Save your model\n",
        "model = create_model()\n",
        "\n",
        "# Train your model\n",
        "history = model.fit(train_generator,\n",
        "                    epochs=15,\n",
        "                    validation_data=validation_generator)"
      ],
      "id": "-e7ywtgZS5pU"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mmpadXR_WGbK"
      },
      "source": [
        "Now take a look at your training history:"
      ],
      "id": "mmpadXR_WGbK"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 545
        },
        "id": "_Q3Zpr46dsij",
        "outputId": "ae00ecd8-d0bd-4826-c063-3c793bcde9ac",
        "tags": []
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "# Plot the chart for accuracy and loss on both training and validation\n",
        "acc = history.history['accuracy']\n",
        "val_acc = history.history['val_accuracy']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "\n",
        "epochs = range(len(acc))\n",
        "\n",
        "plt.plot(epochs, acc, 'r', label='Training accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='Validation accuracy')\n",
        "plt.title('Training and validation accuracy')\n",
        "plt.legend()\n",
        "plt.figure()\n",
        "\n",
        "plt.plot(epochs, loss, 'r', label='Training Loss')\n",
        "plt.plot(epochs, val_loss, 'b', label='Validation Loss')\n",
        "plt.title('Training and validation loss')\n",
        "plt.legend()\n",
        "\n",
        "plt.show()"
      ],
      "id": "_Q3Zpr46dsij"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PdWizvXnXAGz"
      },
      "source": [
        "You will not be graded based on the accuracy of your model but try making it as high as possible for both training and validation, as an optional exercise, **after submitting your notebook for grading**.\n",
        "\n",
        "A reasonable benchmark is to achieve over 99% accuracy for training and over 95% accuracy for validation within 15 epochs. Try tweaking your model's architecture or the augmentation techniques to see if you can achieve these levels of accuracy."
      ],
      "id": "PdWizvXnXAGz"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uTikUFbYTqBo"
      },
      "source": [
        "You need to submit this notebook for grading. To download it, click on the `File` tab in the upper left corner of the screen then click on `Download` -> `Download .ipynb`. You can name it anything you want as long as it is a valid `.ipynb` (jupyter notebook) file."
      ],
      "id": "uTikUFbYTqBo"
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DmHC02GaTuDg"
      },
      "source": [
        "**Congratulations on finishing this week's assignment!**\n",
        "\n",
        "You have successfully implemented a convolutional neural network that is able to perform multi-class classification tasks! Nice job!\n",
        "\n",
        "**Keep it up!**"
      ],
      "id": "DmHC02GaTuDg"
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.4"
    },
    "colab": {
      "name": "sign_language_multiclass_classification.ipynb",
      "provenance": []
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}