notebooks/annotation_manipulation_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook will indicate:\n",
    "1. how to load annotations from HALO/ASAP format annotation files;\n",
    "2. how to serialize annotations to ASAP format annotation files;\n",
    "3. how to transform annotations to masks and vice versa."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "from PIL import Image\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append(\"../\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Dataset preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n",
      "                                 Dload  Upload   Total   Spent    Left  Speed\n",
      "100  434k  100  434k    0     0  1119k      0 --:--:-- --:--:-- --:--:--     0-:-- 1116k\n",
      "slide_001.asap.regen.xml  slide_001.asap.xml  slide_001.halo.xml\n"
     ]
    }
   ],
   "source": [
    "!mkdir -p /tmp/annotations\n",
    "!curl https://storage.googleapis.com/kidney_dataset/annotations/slide_001.halo.xml -o /tmp/annotations/slide_001.halo.xml\n",
    "!ls /tmp/annotations/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "working_dir = Path(\"/tmp/annotations/\")\n",
    "halo_annotation_file = working_dir / \"slide_001.halo.xml\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. HALO annotation to ASAP annotation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from annotations import load_annotations_from_halo_xml, create_asap_annotation_file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "annotations = load_annotations_from_halo_xml(halo_annotation_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "asap_annotation_file = working_dir / \"slide_001.asap.xml\"\n",
    "create_asap_annotation_file(annotations, asap_annotation_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "slide_001.asap.regen.xml  slide_001.asap.xml  slide_001.halo.xml\n"
     ]
    }
   ],
   "source": [
    "!ls /tmp/annotations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The generated slide_001.asap.xml can now be used in ASAP to view the annotations."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Read directly from ASAP annotations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This functionality is needed, as later on, when models generate annotations, it will only be provided in ASAP format. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from annotations import load_annotations_from_asap_xml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "annotations = load_annotations_from_asap_xml(asap_annotation_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label_name</th>\n",
       "      <th>color</th>\n",
       "      <th>z-axis</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Biopsy_core</td>\n",
       "      <td>#00ffff</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Background</td>\n",
       "      <td>#000000</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Glomerulus</td>\n",
       "      <td>#ff0000</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Glomerulus_GS</td>\n",
       "      <td>#0000ff</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Artery</td>\n",
       "      <td>#28614f</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Arteriole</td>\n",
       "      <td>#59bb9b</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      label_name    color  z-axis  label\n",
       "1    Biopsy_core  #00ffff       0      1\n",
       "0     Background  #000000       1      0\n",
       "2     Glomerulus  #ff0000       1      2\n",
       "3  Glomerulus_GS  #0000ff       1      3\n",
       "4         Artery  #28614f       1      4\n",
       "5      Arteriole  #59bb9b       1      5"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the following dataframe is useful for rendering \n",
    "# different groups of annotations on ASAP and mask;\n",
    "# can be modified according to the true annotation;\n",
    "# must contain four columns: label, label_name, color and z-axis\n",
    "# here it's directly decoded from HALO annotations.\n",
    "\n",
    "label_info = pd.DataFrame({\n",
    "    \"label_name\": [\"Background\"] + [a.group_name for a in annotations],\n",
    "    \"color\": [\"#000000\"] + [a.render_color for a in annotations],\n",
    "    \"z-axis\": 1\n",
    "})\n",
    "label_info = label_info.drop_duplicates().reset_index(drop=True)\n",
    "label_info[\"label\"] = label_info.index\n",
    "label_info.loc[label_info[\"label_name\"] == \"Biopsy_core\", \"z-axis\"] = 0\n",
    "label_info = label_info.sort_values(\"z-axis\")\n",
    "label_info"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Transformation between mask and annotation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "from annotations import annotation_to_mask, mask_to_annotation, generate_colorful_mask"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1562 1562\n"
     ]
    }
   ],
   "source": [
    "upper_left = (5000, 43000) # in level-0 coordinate system\n",
    "level = 2\n",
    "width, height = (30000 - upper_left[0]) // 4**level, (68000 - upper_left[1]) // 4**level\n",
    "print(width, height)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The annotation to mask transformation is useful for training segmentation models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "mask = annotation_to_mask(annotations, label_info, upper_left, (width, height), level=level)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1562, 1562, 3), array([0, 1, 2, 3, 4, 5], dtype=uint8))"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask.shape, np.unique(mask)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f4c7632da90>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "colorful_mask = generate_colorful_mask(mask, label_info)\n",
    "plt.imshow(colorful_mask)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The inverse transformation is useful in rendering model inference results using polygons on ASAP."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "regen_annotations = mask_to_annotation(mask, label_info, upper_left, level)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "create_asap_annotation_file(regen_annotations, working_dir / \"slide_001.asap.regen.xml\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorch",
   "language": "python",
   "name": "pytorch"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}