demo_01_segment_patches.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# demo_01_segment_patches"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A demonstration of running the ADP tuning set through HistoSegNet and evaluating the results qualitatively and quantitatively."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import hsn_v1\n",
    "import pandas as pd\n",
    "import matplotlib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from hsn_v1.adp import Atlas\n",
    "from hsn_v1.utilities import *\n",
    "from hsn_v1.histonet import HistoNet\n",
    "from hsn_v1.gradcam import GradCAM\n",
    "from hsn_v1.densecrf import DenseCRF"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Initialize HistoSegNet"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* `input_name` $\\in$ {'01_tuning_patch', '02_glas_full'}: the segmentation set to be solved\n",
    "* `input_size` $\\in$ [int > 0, int > 0]: the resized size of the input image\n",
    "* `input_mode` $\\in$ {'patch', 'wsi'}: the type of input image, either patch ('patch') or slide ('wsi'); only patches supported in public code release\n",
    "* `down_fac` $\\in$ 0 $\\leq$ float $\\leq$ 1: the scalar by which to downsample the input image to ensure equivalent pixel resolution (1.2143 um/px / input pixel resolution)\n",
    "* `batch_size` $\\in$ int > 0: batch size of input images\n",
    "* `htt_mode` $\\in$ {'both', 'morph', 'func', 'glas'}: the type of classes to segment from the images: both morphological and functional types ('both'), only morphological types ('morph'), only functional types ('func'), and glandular/non-glandular ('glas')\n",
    "* `gt_mode` $\\in$ {'on', 'off'}: whether to assess the segmentation against the ground-truth annotations, either 'on' or 'off'\n",
    "* `run_level` $\\in$ {1, 2, 3}: the final stage to run in HistoSegNet: the first stage/CNN confidence scores (1), the third stage/modified Grad-CAMs (2), or the fourth stage/dense CRF segmentation masks (3)\n",
    "* `save_types` $\\in$ [{0, 1}, {0, 1}, {0, 1}, {0, 1}]: the types of files to save for debugging, as a list of four binary values (0/1)\n",
    "  1. HTT confidence scores: save (1), do not save (0)\n",
    "  2. Continuous Grad-CAMs: save (1), do not save (0)\n",
    "  3. Discrete segmentation masks: save (1), do not save (0)\n",
    "  4. Summary images: save (1), do not save (0)\n",
    "* `verbosity` $\\in$ {'NORMAL', 'QUIET'}: the verbosity of debug messages, either on ('NORMAL') or off ('QUIET')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "hsn = hsn_v1.HistoSegNetV1(params={'input_name': '01_tuning_patch', 'input_size': [224, 224], \n",
    "                                   'input_mode': 'patch', 'down_fac': 1, 'batch_size': 16, \n",
    "                                   'htt_mode': 'both', 'gt_mode': 'on', 'run_level': 3, \n",
    "                                   'save_types': [0, 0, 1, 0], 'verbosity': 'QUIET'})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load images"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load images from file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'008.png_crop_3788.png'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hsn.find_img()\n",
    "hsn.input_files_all[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Obtain logarithmic inverse class frequencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "hsn.analyze_img()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Display logarithmic inverse class frequencies for the morphological types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>morph</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Background</th>\n",
       "      <td>0.028387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.S</th>\n",
       "      <td>0.037301</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.U</th>\n",
       "      <td>0.030246</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.O</th>\n",
       "      <td>0.031373</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.S</th>\n",
       "      <td>0.032309</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.U</th>\n",
       "      <td>0.031002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.O</th>\n",
       "      <td>0.034507</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.P</th>\n",
       "      <td>0.034125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.D.I</th>\n",
       "      <td>0.028471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.D.R</th>\n",
       "      <td>0.033361</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.L</th>\n",
       "      <td>0.029153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.E</th>\n",
       "      <td>0.032533</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.K</th>\n",
       "      <td>0.037720</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.Y</th>\n",
       "      <td>0.033824</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.M.C</th>\n",
       "      <td>0.032883</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.M.S</th>\n",
       "      <td>0.035775</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.E</th>\n",
       "      <td>0.040472</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.C.H</th>\n",
       "      <td>0.034068</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.R</th>\n",
       "      <td>0.033321</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.W</th>\n",
       "      <td>0.032159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.B</th>\n",
       "      <td>0.037867</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.M</th>\n",
       "      <td>0.039213</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>M.M</th>\n",
       "      <td>0.030903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>M.K</th>\n",
       "      <td>0.035579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.P</th>\n",
       "      <td>0.029396</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.R.B</th>\n",
       "      <td>0.033176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.R.A</th>\n",
       "      <td>0.033845</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.M</th>\n",
       "      <td>0.044829</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.W</th>\n",
       "      <td>0.052202</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               morph\n",
       "Background  0.028387\n",
       "E.M.S       0.037301\n",
       "E.M.U       0.030246\n",
       "E.M.O       0.031373\n",
       "E.T.S       0.032309\n",
       "E.T.U       0.031002\n",
       "E.T.O       0.034507\n",
       "E.P         0.034125\n",
       "C.D.I       0.028471\n",
       "C.D.R       0.033361\n",
       "C.L         0.029153\n",
       "H.E         0.032533\n",
       "H.K         0.037720\n",
       "H.Y         0.033824\n",
       "S.M.C       0.032883\n",
       "S.M.S       0.035775\n",
       "S.E         0.040472\n",
       "S.C.H       0.034068\n",
       "S.R         0.033321\n",
       "A.W         0.032159\n",
       "A.B         0.037867\n",
       "A.M         0.039213\n",
       "M.M         0.030903\n",
       "M.K         0.035579\n",
       "N.P         0.029396\n",
       "N.R.B       0.033176\n",
       "N.R.A       0.033845\n",
       "N.G.M       0.044829\n",
       "N.G.W       0.052202"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(data=hsn.httclass_loginvfreq[0], index=hsn.httclass_valid_classes[0], columns=np.array([hsn.htt_classes[0]]))\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Display logarithmic inverse class frequencies for the functional types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>func</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Background</th>\n",
       "      <td>0.200454</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other</th>\n",
       "      <td>0.180285</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G.O</th>\n",
       "      <td>0.196248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G.N</th>\n",
       "      <td>0.211562</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>T</th>\n",
       "      <td>0.211451</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                func\n",
       "Background  0.200454\n",
       "Other       0.180285\n",
       "G.O         0.196248\n",
       "G.N         0.211562\n",
       "T           0.211451"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(data=hsn.httclass_loginvfreq[1], index=hsn.httclass_valid_classes[1], columns=np.array([hsn.htt_classes[1]]))\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\chanlynd\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From C:\\Users\\chanlynd\\Anaconda3\\lib\\site-packages\\keras\\backend\\tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 224, 224, 64)      1792      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 224, 224, 64)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 224, 224, 64)      256       \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 224, 224, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 224, 224, 64)      36928     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 224, 224, 64)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 224, 224, 64)      256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 112, 112, 128)     73856     \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 112, 112, 128)     0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 112, 112, 128)     512       \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 112, 112, 128)     0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 112, 112, 128)     147584    \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 112, 112, 128)     0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 112, 112, 128)     512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 56, 56, 256)       295168    \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 56, 56, 256)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 56, 56, 256)       1024      \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 56, 56, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "activation_6 (Activation)    (None, 56, 56, 256)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 56, 56, 256)       1024      \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 56, 56, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "activation_7 (Activation)    (None, 56, 56, 256)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_7 (Batch (None, 56, 56, 256)       1024      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 1, 1, 256)         0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 51)                13107     \n",
      "_________________________________________________________________\n",
      "activation_8 (Activation)    (None, 51)                0         \n",
      "=================================================================\n",
      "Total params: 1,753,203\n",
      "Trainable params: 1,750,899\n",
      "Non-trainable params: 2,304\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "hsn.load_histonet(params={'model_name': 'histonet_X1.7_clrdecay_5'})\n",
    "print(hsn.hn.model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Print the loaded class score thresholds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>E</th>\n",
       "      <td>0.720844</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M</th>\n",
       "      <td>0.656404</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.S</th>\n",
       "      <td>0.420312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.U</th>\n",
       "      <td>0.419097</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.M.O</th>\n",
       "      <td>0.428349</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T</th>\n",
       "      <td>0.603345</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.S</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.U</th>\n",
       "      <td>0.489809</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.T.O</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>E.P</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C</th>\n",
       "      <td>0.756199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.D</th>\n",
       "      <td>0.340588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.D.I</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.D.R</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C.L</th>\n",
       "      <td>0.551754</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H</th>\n",
       "      <td>0.577365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.E</th>\n",
       "      <td>0.487684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.K</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>H.Y</th>\n",
       "      <td>0.455780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S</th>\n",
       "      <td>0.423960</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.M</th>\n",
       "      <td>0.357728</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.M.C</th>\n",
       "      <td>0.453480</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.M.S</th>\n",
       "      <td>0.377504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.E</th>\n",
       "      <td>0.728794</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.C</th>\n",
       "      <td>0.738026</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.C.H</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>S.R</th>\n",
       "      <td>0.964739</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A</th>\n",
       "      <td>0.339476</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.W</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.B</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A.M</th>\n",
       "      <td>0.963768</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>M</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>M.M</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>M.K</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.P</th>\n",
       "      <td>0.887729</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.R</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.R.B</th>\n",
       "      <td>0.540835</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.R.A</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G</th>\n",
       "      <td>0.881226</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.M</th>\n",
       "      <td>0.387146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.A</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.O</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.E</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.R</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.W</th>\n",
       "      <td>0.937337</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>N.G.T</th>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G</th>\n",
       "      <td>0.728526</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G.O</th>\n",
       "      <td>0.777848</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G.N</th>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>T</th>\n",
       "      <td>0.553801</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              0\n",
       "E      0.720844\n",
       "E.M    0.656404\n",
       "E.M.S  0.420312\n",
       "E.M.U  0.419097\n",
       "E.M.O  0.428349\n",
       "E.T    0.603345\n",
       "E.T.S  0.333333\n",
       "E.T.U  0.489809\n",
       "E.T.O  0.333333\n",
       "E.P    0.333333\n",
       "C      0.756199\n",
       "C.D    0.340588\n",
       "C.D.I  0.333333\n",
       "C.D.R  0.333333\n",
       "C.L    0.551754\n",
       "H      0.577365\n",
       "H.E    0.487684\n",
       "H.K    0.333333\n",
       "H.Y    0.455780\n",
       "S      0.423960\n",
       "S.M    0.357728\n",
       "S.M.C  0.453480\n",
       "S.M.S  0.377504\n",
       "S.E    0.728794\n",
       "S.C    0.738026\n",
       "S.C.H  0.333333\n",
       "S.R    0.964739\n",
       "A      0.339476\n",
       "A.W    0.333333\n",
       "A.B    0.500000\n",
       "A.M    0.963768\n",
       "M      0.333333\n",
       "M.M    0.333333\n",
       "M.K    0.333333\n",
       "N      0.333333\n",
       "N.P    0.887729\n",
       "N.R    0.333333\n",
       "N.R.B  0.540835\n",
       "N.R.A  0.333333\n",
       "N.G    0.881226\n",
       "N.G.M  0.387146\n",
       "N.G.A  0.500000\n",
       "N.G.O  0.500000\n",
       "N.G.E  0.500000\n",
       "N.G.R  0.500000\n",
       "N.G.W  0.937337\n",
       "N.G.T  0.500000\n",
       "G      0.728526\n",
       "G.O    0.777848\n",
       "G.N    0.333333\n",
       "T      0.553801"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(data=hsn.hn.thresholds[0], index=hsn.atlas.level5)\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run all batches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\chanlynd\\Documents\\Grad Research\\hsn_v1\\hsn_v1\\utilities.py:367: MatplotlibDeprecationWarning: \n",
      "The tick1On function was deprecated in Matplotlib 3.1 and will be removed in 3.3. Use Tick.tick1line.set_visible instead.\n",
      "  t.tick1On = False\n",
      "C:\\Users\\chanlynd\\Documents\\Grad Research\\hsn_v1\\hsn_v1\\utilities.py:368: MatplotlibDeprecationWarning: \n",
      "The tick2On function was deprecated in Matplotlib 3.1 and will be removed in 3.3. Use Tick.tick2line.set_visible instead.\n",
      "  t.tick2On = False\n",
      "C:\\Users\\chanlynd\\Documents\\Grad Research\\hsn_v1\\hsn_v1\\utilities.py:370: MatplotlibDeprecationWarning: \n",
      "The tick1On function was deprecated in Matplotlib 3.1 and will be removed in 3.3. Use Tick.tick1line.set_visible instead.\n",
      "  t.tick1On = False\n",
      "C:\\Users\\chanlynd\\Documents\\Grad Research\\hsn_v1\\hsn_v1\\utilities.py:371: MatplotlibDeprecationWarning: \n",
      "The tick2On function was deprecated in Matplotlib 3.1 and will be removed in 3.3. Use Tick.tick2line.set_visible instead.\n",
      "  t.tick2On = False\n"
     ]
    }
   ],
   "source": [
    "confscores = np.zeros((len(hsn.input_files_all), len(hsn.hn.class_names)))\n",
    "iou = {}\n",
    "fiou = {}\n",
    "miou = {}\n",
    "num_batches = (len(hsn.input_files_all) + hsn.batch_size - 1) // hsn.batch_size\n",
    "for iter_batch in range(num_batches):\n",
    "    start = iter_batch * hsn.batch_size\n",
    "    end = min((iter_batch + 1) * hsn.batch_size, len(hsn.input_files_all))\n",
    "    hsn.input_files_batch = hsn.input_files_all[start:end]\n",
    "    # Normalize image batch\n",
    "    hsn.load_norm_imgs()\n",
    "    # Load ground-truth annotations\n",
    "    hsn.load_gt()\n",
    "    # Segment image\n",
    "    hsn.segment_img()\n",
    "    # Evaluate segmentation\n",
    "    iou['GradCAM'], fiou['GradCAM'], miou['GradCAM'] = hsn.eval_segmentation(hsn.intersect_counts['GradCAM'], hsn.union_counts['GradCAM'],\n",
    "                                                                             hsn.confusion_matrix['GradCAM'], hsn.gt_counts['GradCAM'],\n",
    "                                                                             httclass_pred_segmasks=hsn.ablative_segmasks['GradCAM'], tag_name='GradCAM')\n",
    "    iou['Adjust'], fiou['Adjust'], miou['Adjust'] = hsn.eval_segmentation(hsn.intersect_counts['Adjust'], hsn.union_counts['Adjust'],\n",
    "                                                                          hsn.confusion_matrix['Adjust'], hsn.gt_counts['Adjust'],\n",
    "                                                                          httclass_pred_segmasks=hsn.ablative_segmasks['Adjust'], tag_name='Adjust')\n",
    "    iou['CRF'], fiou['CRF'], miou['CRF'] = hsn.eval_segmentation(hsn.intersect_counts['CRF'], hsn.union_counts['CRF'],\n",
    "                                                                 hsn.confusion_matrix['CRF'], hsn.gt_counts['CRF'],\n",
    "                                                                 httclass_pred_segmasks=hsn.ablative_segmasks['CRF'], tag_name='CRF')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Qualitative performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Original images and Ground-truth annotations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Display original image, morphological ground truth, and functional ground truth annotations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Functional\\n ground truth')"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure\n",
    "plt.subplot(131)\n",
    "plt.imshow(hsn.input_images[0].astype('uint8'))\n",
    "plt.title('Original image')\n",
    "plt.subplot(132)\n",
    "plt.imshow(hsn.httclass_gt_segmasks[0][0].astype('uint8'))\n",
    "plt.title('Morphological\\n ground truth')\n",
    "plt.subplot(133)\n",
    "plt.imshow(hsn.httclass_gt_segmasks[1][0].astype('uint8'))\n",
    "plt.title('Functional\\n ground truth')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predicted segmentations, stage-by-stage"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Display the predicted segmentations after Stages 2, 3, and 4 for a sample image, for the morphological types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, '(4) Segmentation\\n Post-Processing\\n(morphological)')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure\n",
    "plt.subplot(131)\n",
    "plt.imshow(hsn.ablative_segmasks['GradCAM'][0][0])\n",
    "plt.title('(2) Pixel-level\\n Segmentation\\n(morphological)')\n",
    "plt.subplot(132)\n",
    "plt.imshow(hsn.ablative_segmasks['Adjust'][0][0])\n",
    "plt.title('(3) Inter-HTT\\n Adjustments\\n(morphological)')\n",
    "plt.subplot(133)\n",
    "plt.imshow(hsn.ablative_segmasks['CRF'][0][0])\n",
    "plt.title('(4) Segmentation\\n Post-Processing\\n(morphological)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Display the predicted segmentations after Stages 2, 3, and 4 for a sample image, for the functional types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, '(4) Segmentation\\n Post-Processing\\n(functional)')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure\n",
    "plt.subplot(131)\n",
    "plt.imshow(hsn.ablative_segmasks['GradCAM'][1][0])\n",
    "plt.title('(2) Pixel-level\\n Segmentation\\n(functional)')\n",
    "plt.subplot(132)\n",
    "plt.imshow(hsn.ablative_segmasks['Adjust'][1][0])\n",
    "plt.title('(3) Inter-HTT\\n Adjustments\\n(functional)')\n",
    "plt.subplot(133)\n",
    "plt.imshow(hsn.ablative_segmasks['CRF'][1][0])\n",
    "plt.title('(4) Segmentation\\n Post-Processing\\n(functional)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Quantitative performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Morphological types"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Class IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2ac33037908>"
      ]
     },
     "execution_count": 13,
     "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": [
    "df = pd.DataFrame(data=iou['CRF'][0], index=hsn.httclass_valid_classes[0], columns=np.array(['IoU']))\n",
    "df.plot.bar(y='IoU')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Frequency-weighted IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.20567515021442848"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fiou['CRF'][0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mean IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.22060561066883094"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "miou['CRF'][0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Functional types"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Class IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2ac38244a58>"
      ]
     },
     "execution_count": 16,
     "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": [
    "df = pd.DataFrame(data=iou['CRF'][1], index=hsn.httclass_valid_classes[1], columns=np.array(['IoU']))\n",
    "df.plot.bar(y='IoU')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Frequency-weighted IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5421715749664008"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fiou['CRF'][1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mean IoU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5505286801193577"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "miou['CRF'][1]"
   ]
  }
 ],
 "metadata": {
  "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}