Full MNIST Single Image.ipynb

full-mnist-single-image.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Full MNIST Single Image.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyMMfOdJiWTOatfX9WlIbuV0",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/dgrahn/e370bfec4e1b957bc378f42a965c912a/full-mnist-single-image.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SdHLOb3njb95",
        "colab_type": "text"
      },
      "source": [
        "# MNIST Single Image\n",
        "This code combines MNIST training data into a single image.\n",
        "\n",
        "First import appropriate libraries and download MNIST."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZfvbgHJJ-zAA",
        "colab_type": "code",
        "outputId": "eb5b7233-0802-455a-c47f-86af2fa7998c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "%tensorflow_version 2.x\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "(train_images, train_labels), (_, _) = tf.keras.datasets.mnist.load_data()\n",
        "print(train_images.shape, train_labels.shape)\n",
        "\n",
        "mnist = pd.DataFrame()#{ 'x': train_images, 'y': train_labels })\n",
        "mnist['y'] = train_labels\n",
        "mnist['x'] = train_images.tolist()\n",
        "mnist.x = mnist.x.map(np.array)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TensorFlow 2.x selected.\n",
            "(60000, 28, 28) (60000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lkRdTOvVmLvm",
        "colab_type": "text"
      },
      "source": [
        "MNIST takes up about 59 columns and 1125 rows. This will be a tall image."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BpbDccqrBAKC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from PIL import Image\n",
        "\n",
        "COLS = 59\n",
        "ROWS = 1125\n",
        "SIZE = 28\n",
        "\n",
        "full = Image.new('L', (COLS * SIZE, ROWS * SIZE), (0,))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UDpy4WJSn9Ds",
        "colab_type": "text"
      },
      "source": [
        "Now composite each digit into their own columns."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gvo4uzJLE_uf",
        "colab_type": "code",
        "outputId": "5309e57e-8b11-464b-ed63-b4a6b6f00bfc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        }
      },
      "source": [
        "pixels = full.load()\n",
        "col = -1\n",
        "\n",
        "for i in range(10):\n",
        "  digits = mnist[mnist.y == i].x\n",
        "  print('Compositing', i, '->', digits.shape[0])\n",
        "\n",
        "  for j in range(digits.shape[0]):\n",
        "    row    = j % ROWS\n",
        "    if row == 0: col += 1\n",
        "    top    = row * SIZE\n",
        "    left   = col * SIZE\n",
        "    bottom = top + SIZE\n",
        "    right  = left + SIZE\n",
        "\n",
        "    for i2, x in enumerate(range(top, bottom)):\n",
        "      for j2, y in enumerate(range(left, right)):\n",
        "          pixels[y, x] = int(digits.iloc[j][i2, j2]),"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Compositing 0 -> 5923\n",
            "Compositing 1 -> 6742\n",
            "Compositing 2 -> 5958\n",
            "Compositing 3 -> 6131\n",
            "Compositing 4 -> 5842\n",
            "Compositing 5 -> 5421\n",
            "Compositing 6 -> 5918\n",
            "Compositing 7 -> 6265\n",
            "Compositing 8 -> 5851\n",
            "Compositing 9 -> 5949\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4MW2lgtXowcS",
        "colab_type": "text"
      },
      "source": [
        "Finally, put it all in an image. You can download this using a right-click > save."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "s2bOoc3rLJMs",
        "colab_type": "code",
        "outputId": "55820f5d-94f8-4b05-ccf7-949bc49f5e97",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "dpi  = SIZE * 4\n",
        "w, h = full.size\n",
        "\n",
        "plt.figure(figsize=(w / dpi, h / dpi))\n",
        "plt.axis('off')\n",
        "plt.imshow(full, aspect='auto', cmap='gray', vmin=0, vmax=255)\n",
        "plt.show()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {