{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "visualizing_examples.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "-ZF4GR4A_L1c"
      },
      "source": [
        "This example was done using [Colab](https://colab.research.google.com/\n",
        "). You can read and use the Python code in other environments if you have never used Colab before. For example, we used a path to a file in gs://, but you can also just copy that file to your environment and use a local path instead.\n",
        "\n",
        "This example is for **advanced** users who want to understand the data representation. **You do NOT need to understand this in order to use DeepVariant.**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "hbneXvSWAL8i"
      },
      "source": [
        "<table class=\"tfo-notebook-buttons\" align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/google/deepvariant/blob/r1.6.1/docs/visualizing_examples.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n",
        "  </td>\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://github.com/google/deepvariant/blob/r1.6.1/docs/visualizing_examples.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n",
        "  </td>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AV5gBKsCgF-A",
        "colab_type": "text"
      },
      "source": [
        "Note: This notebook was updated in the 1.0 release to use the new `nucleus.util.vis` module that makes visualizing DeepVariant examples easier. The old notebook can be found at https://github.com/google/deepvariant/blob/r0.10/docs/visualizing_examples.ipynb"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "AS8XlRzD_L1k"
      },
      "source": [
        "# Visualizing DeepVariant examples\n",
        "\n",
        "DeepVariant represents putative variants as pileup image tensors and classifies them using a convolutional neural network.\n",
        "\n",
        "In this notebook, we will inspect some DeepVariant examples and visualize the pileup images.\n",
        "\n",
        "We (the DeepVariant team) also wrote a blog post that explains what these pileup images are and how they are used in DeepVariant:\n",
        "https://google.github.io/deepvariant/posts/2020-02-20-looking-through-deepvariants-eyes/"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y2Z9eNK7FT6q",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Nucleus is the only dependency we need to install. We need at least\n",
        "# nucleus 0.5.0, since this was the first release of Nucleus that includes the\n",
        "# visualization utilities we are using here.\n",
        "%%capture\n",
        "! pip install google-nucleus==0.5.1\n",
        "# Nucleus may show an error about failing to build a wheel, but it still\n",
        "# installs successfully."
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8R_lXnKWFQIw",
        "colab_type": "code",
        "outputId": "61cdf33b-9711-490f-a770-e2e73de90771",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "import tensorflow as tf\n",
        "from nucleus.util import vis\n",
        "print(tf.__version__)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.2.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9Ntg7xw0JSSd",
        "colab_type": "text"
      },
      "source": [
        "Download a file of DeepVariant examples from Google Cloud Storage to the local file storage of this notebook. DeepVariant examples are stored in TFRecord format."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "PYEQP7B-Ae2t",
        "colab": {}
      },
      "source": [
        "!gsutil -q cp gs://deepvariant/datalab-testdata/make_examples_datalab.tfrecord.gz /tmp/make_examples_colab.tfrecord.gz"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nOjW_uniUAxj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "examples_path = '/tmp/make_examples_colab.tfrecord.gz'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "caRpnLLgIAte",
        "colab_type": "text"
      },
      "source": [
        "# Show one DeepVariant example\n",
        "We will use the nucleus.util.vis module to inspect and visualize a single DeepVariant example."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LI11fJoHF_Qp",
        "colab_type": "code",
        "outputId": "f04807a8-d28f-4168-cf58-ed191d581e51",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        }
      },
      "source": [
        "dataset = tf.data.TFRecordDataset(examples_path, compression_type=\"GZIP\")\n",
        "\n",
        "# Take a single example and show the deepvariant pileup image for it.\n",
        "for e in dataset.take(1):\n",
        "  example = tf.train.Example()\n",
        "  example.ParseFromString(e.numpy())\n",
        "\n",
        "  # For example, in chr20:19503712_T_C, T is the reference allele and C is the\n",
        "  # alternate allele that is proposed in this pileup image.\n",
        "  print(\"Locus ID with alt:\", vis.locus_id_with_alt(example))\n",
        "  print(\"Draw the channels:\")\n",
        "  vis.draw_deepvariant_pileup(example)\n",
        "\n",
        "  print(\"Truth label:\", vis.label_from_example(example))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Locus ID with alt: chr20:19503712_T_C\n",
            "Draw the channels:\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Truth label: 1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6GMIhtT1JyRX",
        "colab_type": "text"
      },
      "source": [
        "# You can also save the images as .png files"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iHWZiV82Jwwy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for e in dataset.take(10):\n",
        "  example = tf.train.Example()\n",
        "  example.ParseFromString(e.numpy())\n",
        "\n",
        "  # For example, in chr20:19503712_T_C, T is the reference allele and C is the\n",
        "  # alternate allele that is proposed in this pileup image.\n",
        "  filename = 'pileup_{}_truth={}.png'.format(vis.locus_id_with_alt(example), vis.label_from_example(example))\n",
        "  vis.draw_deepvariant_pileup(example, path=filename, show=False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eP7cgpvYefXF",
        "colab_type": "text"
      },
      "source": [
        "This creates 10 PNG images labeled with locus IDs and truth labels."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_sxWsvFQZki9",
        "colab_type": "code",
        "outputId": "eecb38d6-dedb-49f3-ec9d-9716ae37fd0f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        }
      },
      "source": [
        "! ls"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "'pileup_chr20:12534952_A_G_truth=1.png'\n",
            "'pileup_chr20:19503712_T_C_truth=1.png'\n",
            "'pileup_chr20:24448381_T_A_truth=2.png'\n",
            "'pileup_chr20:32765978_A_G_truth=1.png'\n",
            "'pileup_chr20:37886261_G_A_truth=1.png'\n",
            "'pileup_chr20:55019799_C_T_truth=1.png'\n",
            "'pileup_chr20:59581048_T_G_truth=0.png'\n",
            "'pileup_chr20:64063271_A_G_truth=0.png'\n",
            "'pileup_chr20:7912729_G_A_truth=1.png'\n",
            "'pileup_chr20:845848_G_C_truth=2.png'\n",
            " sample_data\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TnYFOI03ZzIK",
        "colab_type": "text"
      },
      "source": [
        "## Flexible, modular functions\n",
        "The nucleus.util.vis library also has smaller building blocks you can use if you need more flexibility."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Sy4vkwgsaBEo",
        "colab_type": "code",
        "outputId": "f567592e-3c67-463f-bb16-c3c6e60e436e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 982
        }
      },
      "source": [
        "for e in dataset.take(1):\n",
        "  example = tf.train.Example()\n",
        "  example.ParseFromString(e.numpy())\n",
        "  print(\"Extracting the variant:\")\n",
        "  print(vis.variant_from_example(example))\n",
        "\n",
        "  # Get the pileup image as an array.\n",
        "  arr = vis.get_image_array_from_example(example)\n",
        "  print(\"\\nShape of the pileup numpy array:\", arr.shape)\n",
        "  # Split that array into channels.\n",
        "  channels = vis.split_3d_array_into_channels(arr)\n",
        "  print(\"\\nNumber of channels:\", len(channels))\n",
        "\n",
        "  # Visualize the numpy array as an image.\n",
        "  print(\"\\nShowing just the first channel:\")\n",
        "  vis.array_to_png(channels[0])"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Extracting the variant:\n",
            "reference_bases: \"T\"\n",
            "alternate_bases: \"C\"\n",
            "calls {\n",
            "  info {\n",
            "    key: \"AD\"\n",
            "    value {\n",
            "      values {\n",
            "        number_value: 31.0\n",
            "      }\n",
            "      values {\n",
            "        number_value: 24.0\n",
            "      }\n",
            "    }\n",
            "  }\n",
            "  info {\n",
            "    key: \"DP\"\n",
            "    value {\n",
            "      values {\n",
            "        number_value: 55.0\n",
            "      }\n",
            "    }\n",
            "  }\n",
            "  info {\n",
            "    key: \"VAF\"\n",
            "    value {\n",
            "      values {\n",
            "        number_value: 0.43636363636363634\n",
            "      }\n",
            "    }\n",
            "  }\n",
            "  genotype: -1\n",
            "  genotype: -1\n",
            "  call_set_name: \"training\"\n",
            "}\n",
            "end: 19503713\n",
            "reference_name: \"chr20\"\n",
            "start: 19503712\n",
            "\n",
            "\n",
            "Shape of the pileup numpy array: (100, 221, 6)\n",
            "\n",
            "Number of channels: 6\n",
            "\n",
            "Showing just the first channel:\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAboAAADICAAAAAC7AXsnAAAV90lEQVR4nO2de2xb133Hv+d1L3mvJOtpO7ETJ438UmLlMadpljZpu3XYUKDYurbo1rUDgg1Yhg0olm3BgKVN0DSPAhsKpECLDvAfHdIV6YpiBbpubRrPsV0vSxzXsi1ZsmTLlm09SMqkRN7L+zpnf5B2bEeiKIkieZXz+UfifZxzyC8Pz7nn9zjE+uiLewYGhgaO3fPSvTj6ZfadW/e9Nkz77t176dsbX9783eMmB2XpwTSjkFF3X7uz6wvbkBucwaZ+DKQYBcDNwVeORx9/afeR8AEykDFn3sbeLfxk6djbl7Bl79BTr+G3vt4PADj+1Mn7+3f39UXvqP52zA+feLvnT+9EYfT1fbd+R/559HL/0C9emX2QvqP6O5x7Hse+AZMDAChLD04WP/r1/msl0OH9r0zd/bHP34aJH/xyoP+lezG2jz/Br90hw8176bETg+i7d+/0vqkHusrlwet/XH439Yg4dmKQ9m/CzOAMOns//vjlv4i++Rs4+mX6cv/QL16Zuvt3Hp/86/Rf4VvdL3d/Sz7B9520rgygv8Pp+1L3W3hw6l/oE7xcGvYNmAabfGvPNzd+9x3c/zj53n8P3P/Nzm/5j98F4PhTA/2f+ELnW1F/O3CtrYXRA+XWA9kB9mD6e8ex5wvbAGQHxMMnv5x+YsvbxwfZA+GR3S/di7F9ePwuAOdfOW1lB9OMAhSamKKliy1autiipYstWrrYoqWLLbzRDVgcQiqdZGyB06J0jJa/kJRxyigBZZwTQillBhiloMyIGOOcUUIoAyUEhDJDECIAQYiAYIRQZghKOAgVBiNCMEopAaWUCUPAEIwKUS4NjFLKKKHCYBTMIOX/mQAAQQllwhDUAADBKKECMEp1CAAwBDMEo7j6SkBQUm5f6WrBIMp3U0oIIc0s3cLilFHEMG4+q4AkUwAIFwAASUSCGpyCG0mTES6EacPkApFpR4ZICINTygU4JaDMTFgMSSDJqAXLZISZCYtRE5RZtiFtYQrBKbgQho0kYBvUFuXSYHIhOCfMSpocpk1MRphlG7AAwGKEmTaSkQUAtskIswBRqsMGACvJYBscRvmVgMWomTQ4pRyEWYBVLss2hOC0uaUj/qzCYtoR5gyzW+wbThPunR3JdhNQf0oBAJT0kvTsTAEzY2Ym604Xi8n9GJkSCEb2q9HsAD87U6DTAWbzHgqp4QOpK/LNHE5nyZHU+HCumB4+kM7Mn8I8O9w+Jg+wkcuJ+QKmnWLiQPsg8pmz5AArl4aRKSFoupg+3DEyhZH9ZDjjpQ+3jQWHtgI4k/EyIwfaBuWVVgDTIxkvcyQN9+JwrpgeeWMTgPlRms+OTmH04GYA86PcHcvMD6bHZgpsKixmjqRx8SwObQUwNXo5MZ/3KG1e6QgrDHXezoha8CxNzLzed/eGGwZqZqX++diW2xnl+YE5AAqUGQfhuEXMjewP08E7XA4cRNqhkCcOqnTwf8Rxi8gJhG4I58r4/uC8et5GYYJ8zXLnZgN3Yn9wIfxXzEw+w2fwFkm7NHSR4/L4QV6AHaTIW6RcGtIOJSQonPyKSDkYPIhMzj35FT6j3jYBOBfdUxP7eUFZDIA/O+eees6CLM7PBu6F/zEARA6xw7SLU4fKr1rcifD7wnGKyCHvPGfBS+NtE4CfKdLIDQmaV7rVQ4jvlv8tZgHAA3CxdGD2IoCZ0rnSkaCYnQAwBAAYBAD4uQkA4wBOArh8tSwAl8r/X76utPIt777O5QBMlk8Ec9euAYBg8NoNNxy/VjIAnL9an1+6ehI30qzSqcju7UC0cKeDLG79bF9PKK8/Fjlb/mbkdSdC2LJXASBi+th8HVraMJpVOijR2TG/iHJQkbVzp6VuOK1Cc3f3O4GCNDoFACrYWH6xAtYDTSsdVLRYnwNAlO+rmyYxBHAjBkCFAMAUWsziWrawwTSvdFCVuoyKopulAxCUbpESABTrppF0CvloTZrXcJpYulUilbl19w46Mjgw1+imrA3rVzoo1rKln4VpK7DkfNjo1tSedSwdlAyK1FN2cndwdLbRjak98ZWOCFppkRMAoEIaKpa8pbgeV9ljK51iDKzy3L+0CqpCEvrrcaYZV+mIyL5NPVHRuBDMmaGhRCdTid2pdGBxmezxz7oVbokVcZWOWeM/op+zKjSfUmc0v9eU1i7COvamfpLZtaG4/dNTT4/Ur5FrS1ylqwISZkRICSDDlrtTbzibup3tj87sSFNIaz30vrhKFzl3PE29Y0vO+QlzxrGN+F6kwiDw5ll/JBDsWA+9L67SqaC9H0NBxQUX0d0tFAlmsQVRqJSUMvRIz+YEitsfnenNttB4r7TEVTqQKKJRpVmKVNauzVYkCbveYqvCMEDozZO72+4VI4PHY2xbiI107/FUUQGtzi5w441SSsjQQxfpN8JMqxPfbheXZ9WKnioLQplz+rTDFnmDKgqKrocN9uqb1ihi0utIkKOty9SOBGkeLHqPjEISKm6bBKEbgiVae7zxwmrbWU/iIR2ludPknuX2u6VRgX1nJ2bPzWLD1g/+/uSzp2pdwVoSD+lWRGmGWfkSKTbdAiNVgN2z88OXO1pus+FeytanfaslHtJJZe2iVm5Z7gpXZ5iVr1JhgFAqyMgruqz3yV0Y+/bBJe5pEuIhHZTopFi2o4msQgMZQSoFJUM/IO0PbUfPDzpvlaMxWK+OywyzoqfKglBWGBwsLDbDXKQSF5jL3fXcM7cur66GEJNet4SnSs0qCQDfb3vY/I9gxqtDfasiLr1u+cjI7uuzlxrrFoAEs/af/GH3GjSptqxf6fBuxM8ykS7t29VmdrXXtjW1Zv1KV3k1pSIEbmBtfOi+5v5w+KLRNLGn8mpKZULJE92WuNzM6yu0YgDi+xgV5G7/2lfvaHQzKsAjn0HwFf2uNDmKd3Xxlc9Lpdf20G1d1kbM8Nv5DDY1nV2d+3kDlmnUfn2w0Uhl9S69mlIBEhWL4rbP4dX2v2t/FZ9PNZtdncv0r7KjZ86lg6W9GmOGEl1dS6xhVkb6AWu9G61dj2w8isfSOzI0KnrNY9/juPii7bhFN6zoGPd+RUkPMnIKvsyxfpnIj081TwADh1OK5BQha3BTag3x50RgrLIQFULJwI+kJ3putbIpO+nnmmR1+tr0JJw3sPwF3iaGMmd0dGXPdQuiwiDw/dYHH0jWqsRVcm0NU6V+FW23GBHr5oeThBkR1vDNSCllJDaqyUu52hW6Ct79Ul548Z8KPQazkkt48r+/kUXrwf7meJJ613LgDLk/N7LR4VvnzOZo2mqpwkq+gkJD2sX4XAEtzrkGP+ddb/SZfIaOq2c3924wgoa1p3ZUZyVfPqGz9Y/sQbZn+JkztS56eVwvnXcSwMnByHIIYZyXMl5RGt8l6mqs5Msv1Gt92G6hd2/8QKGxKyzvNbXKY8zZQBJJkzIraXLBQWQcRz9Ks4Oz93fV/ntHIsd2Aeza1tgVlgWs5FlAzIwZ0/PicNfIlGAwOwQlhICUstsJwBAcwlgghcL7AhX4AXx0Jh5Ll+KGukuJpvx8XSPWF3ZwKJw4+7qfmn5GpBxK0PWhJCtJR0t56ZBMJpDkS0WVNhap7L7NK7GSV4mKirlS3FDvp/CTUSH49K+za1XZQiwsXZTLAaXcWADciWTeoxRS5qcKRePgZuTPzWJiNllIEgDggjdn/1uhlbxKJDzRszmB4o6P4XQ+YQgkNtwmUg6sbmRyrnUbn1HdJoDCeWct6q/Grch9h7oBAaAKs0weP2QgckNw647P7vQiADxhNqHhgdLc6ex9nWuqXjluqAAvDIiK+D1Pdvz4NHZ+ivz0f8/sfLLt1eBTWwEMvnB6LSqvRrroytX/AhfXZY8b2+YUIwDMOJsNQQVEU41+q7GSV4uUEjLyEUkpI0m7H9l4KosdHyOjZy50P9J91P9wN4AN3SLZ1lVKqpg/69eq7tU48+V/+DOpABBSzHjCsJEktHlsIg1ARY7rhfAKxItU5BR83wEAP7Fh+0O/2zYoe1tx4ulLS5VSLauRLro+lyNNHb7jRHKHXWuj7bJdZ6+xSiv5SpCBH0lEPomkkoEfRQEAhNzs2vFYoiW6D7jl4NHZvOfX4hteMxfa8NfPJvJ7HmttydXS2ZXIwE8skop2CVZtJa8dSkVeIeFGjoWevx/7+a8vT9bCW6lm0slsFij+cvdQYVuC1+rzUpF5VzHnr8z1adVW8tqhZBT4QeRbMHZtGZ9Id99lIH9ulf7VtXVcn3zedrb+Wafp1+hHUxa7Ptn3k1FzPVmBC56X3/aXvTj51QurK6i20nlDQOro5gtcGYwTgAoI1qifrJpYyWtPEEV+20N3YtN/4cqqItlrHy6S/Tcjv+3RNpiMEGYhaRZX/qNFE/mfnsolVtSHS9mKks24dk4iF2j74zt+ObSaMa/20gUXAJxQ6kw2lKlf3TGU27jyOgjzx4bbzJVpT8IrZi2t5DVEBYDZlzqE9pbOq+nxnfPL1HFtgrRmX/1P5WaK6sQzycKmz9yz8oIUEcbKJphYSdaHeqL8+Smv75HfLm1RIe4589zgkvfcwNpI50+U/mazQM5PQtCb05dUzWo+/qZ2yldRGIqu7Y+2AkCH0dfVuswC1j40MpFIwmJk/YalrBAleYdPvEIrAMeNUFjus8LaS+ePHN42norgNccjVhNBEybCAAD8AIps7ESyy0DhQpVj3tpLl/rRawn3UoC5dZg5e5Vct+6kwtb7gN7f24yhF4aqu3vtpfMmSgNfUP/n6rrEn6+C66VTYtNm9H5kA1qtKu+u32MPq7ZJNWP5WR8aiQrD0C8A+XyVN9Qvg8Pcm/IDdV3bULyjo86Wg1UhJaIAsD5IqlvdrF+vu/DCN1JA/eIaZGT19jaH5WBZ3P4PT91S1YX163WFobmP7EHLnXWLaYhbrytj77Z/mqhmbaWeKW9S32jBnmcrZUmvLQtv2dT82J/+ZDVrK/WUzj8NuI4dx0+zrhjbrb7uDUteVu9EU8kqs/7WguZew1wc6RFUYQ6qt3SGUbexrrS7SJ0qqyUqChAsPb2qtzmLWXadfKZrHdVaZ5b+kOrd67KHjVyd+l0T2+uWhi7tTFVv6c49La601KnSuI51gAJvX9Kltd7S5U/CTlLG6xKl0NT2usoQ3tYRTVb0lG7AUFDMwUyYdZxpxpHIveVLX+ypeEkDstBG89PnFLYm1t7JvdktBxVQodXHDjjFCn7SjUgg7JwYSzz0IdtYk9Cl64iX5eAmSDTfcz+fmFk8O1IjpItyOajX7i9Y5ppWE9M1zKtIn22anrUTwWK7Ozcqbffk87v/4DeTa2lHaKKYg5VBVBh4rdu9o1cWPt8o6byhiVuNya67rLVL5xjzXgcoKaXY7CXab5dnFrDgNS5ZvvPj/eZ9z29qNVfsZ7kUcbUcXI/yI97/lPu34+891Tjp5KVLmPlo35kznmkAhmCECkmlgIjrg/RaoKSkXQ/zT7x5JXvzLomN3aJi5gXb5b39rYCwTMrsiCkLVq0i0+O7mnIDykv1Pjn22qG3bjreWOn8IYCbYwc3wb04nM8clLPRkdT4cC4ijIBSygwwWsrDsIKtJvw5I1jbSWx9iBy685axS4YHwM8UaeSGBI2WDgDC05OvG5DF+XTqHzGOr1nu3OwGJpjkQpg2TC44SgEWy4NS53R2rxVXy8H1EGDuypYvbgYw9bPxxPy5LKXNIB3m5soRCu4JAIMAYEwaBTntFJP7MTIlGADRsh76z8rxiq0P9QDInfGtKymXNYd0CzB/lDoqx+XAQaQdSgC039eyzInoWmXmaxAkcgGg4AdBKJVSzSpdeAVAEcC1HBFByOUyf/uU6OxskljyWlAaMoJISqmaWLr3IoSQyzUUrYfnugrERbowRVU2WOY2NjG2HFRBXKSbP2YqYgpvHfei5RIX6cIsgE4e+XFp8NoTq09ijkYFsb52Y1gFsZIuBFJH5HZrPWXAWTlxW2wo7cag+x1i1utQ2o3BmjMY48JYcroZY4+wKoibdMDkM8Zsa8Iwk3Sp3NPrxHKwGPGTzjsJW6TOqsKMHSQq/d6vG8vBIsRPOgDFS+mhRDLR+8WuCqPeOrIcLEwspYscBwDG77KdJs31Xg9iKV2Z3PcNp3XRXrXOLAfvJc7SRefR3rF4r1tnloP3EPOhoGLkQqy9n5cm3tLxyk9u69tyEG/p1vlDd2XiLZ0K3sf5/uItnSxkw3i/g1UQ7zeuQh9MAILRGvxyvndaU7uJTpWLcst5F3F+OAAAyk0bsE22+lGPKP8moYjyfVWTL7ciBqsitoIsx5wVd+nCzMgbmzA9kvFCDgjGhFhhvg0VWTsj6+YjXZZTg35HmDPc6ZhLWjqC2Q3JqtWLu3SFUxOvG/Bn59yiAVgGt8UKQxak1/PZaOONRzZ+JrNxuAbrMTQx88Oubb1LfKcIc4bwAbvaQuMuXTQ3V/bVTB/J4eJZdoCNZLrM0hMdZZwyumZBYA0m7tK9y4UXbXhp8hZJJx/sCiUASUSCGpxW122omfphtL3nhiMz/57+nFmDwU4WN36i87y3RENUZO2+M1F1oetHOqeU7foy0HIpCiUAFXlJenamIAjjvLS7swFGKSgzIsY4Z5QQykAJAaEiEYxEPiAIERCMECoS6ZG0Z1AqDEaEYJTSUvyRMAQMwagQ5dLAKKWMlnZyBzNI+X8mAEBQpdr2JLMlbzbBKKGiFFFIaOkKQzBDMIrE5hYAhhAQlJTbV7paMIjy3ZQSQsh6ku5d3ONDCgAU+EE4brGdJJImI9cihxCZdmSIhDA4pVyAUwLKjISdjFqAJKOlGD9mJGwrkSSMWbYhbWEKwSlKe0QnAdugtiiXBpMLwTlhVtLkMG1iMsIs24AFABajXHAkmA2UZsPMAoRlMsJMGwCsJINtcBhm6ZWAxaiZNDilHIRZgFUuyzaE4HTdShdlb3ztzIwZ6aw7XSxHDiEY2a9GswP87EyBTgeYzXsopIYPpObkmzmczpIjqfHhXDE9fCCddQYxRw+3j8kDbORyYr6AaaeYONA+iHzmLDnAyqVhZEoImi6mD3eMTGFkPxnOeOnDbWPBoa0AzmS8zPDrbYMy2wpgeiTjZY6k4V4czhXTI2+UNoah+ezoFM68sRnA/Ch3xzLzg+mxmQKbCouZI2lcPItDWwFMjV5OzOc9uvLtWmIFa7HtIB1YXCa7kXYopNWt0kELcdwikgKhG4IlWruD8+o2G4UJcpvlzs0Gybbu4EK4ETP8dj6DHpJ2aegiyWWimxdgBynSQ8qllbaUDwrJBTe3di6617bZBfzZOTe51YIszs8Gydar2zHZYdpFsuvq5kzuRNgjHKcIC/nEVgteGt0mbgiN1Gg0Go1Go9FoNBqNRqPRaDQajUaj0Wg0Go1Go9FoNBqNRqPRaDQajUaj0Wg0Go1Go9FoNBqNRqPRaDQajUaj0Wg0Go1Go9FoNBqNRqPRaDQajUaj0Wg0Go1Go9FoNBqNRqPRaDQajUaj0Wg0Gs2a8f9mMQ8Sge+w6QAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}