MakePatches.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import os\n",
    "import numpy as np\n",
    "from helper_functions import extract_patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/advo/PycharmProjects/ML_ND3_CapstoneProject/Dataset_small/09_10_11_12/\n"
     ]
    }
   ],
   "source": [
    "hwsize = 32\n",
    "\n",
    "# Arrays of rotations to be used for patch creation. \n",
    "pos_rots = [0, 45, 90]\n",
    "neg_rots = [0, 45]\n",
    "\n",
    "# Lists for the datapoints with no centers and no blue ratio data respectively\n",
    "no_centers = []\n",
    "no_br = []\n",
    "\n",
    "# Dictionaries where the labels for each patch will be stored. Needed for training and testing\n",
    "dict_labels_pos = {}\n",
    "dict_labels_neg = {}\n",
    "\n",
    "# Original dataset location\n",
    "orig_dataset_dir = '/home/advo/PycharmProjects/ML_ND3_CapstoneProject/Dataset/mitosis'\n",
    "patches_dataset_dir = '/home/advo/PycharmProjects/ML_ND3_CapstoneProject/Dataset_small/patches/'\n",
    "os.chdir(orig_dataset_dir)\n",
    "root = os.path.splitext(orig_dataset_dir)[0]\n",
    "print(root)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## !! Caution: Takes a long time to execute"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ram usage: 82.3\n",
      "no GT 09_10_pc.png\n",
      "class0 number of pix after edge filtering= 162803\n",
      " final number of pix to be extracted:  326\n",
      "ram usage: 82.3\n",
      "no GT 10_02_pc.png\n",
      "class0 number of pix after edge filtering= 62489\n",
      " final number of pix to be extracted:  125\n",
      "ram usage: 82.5\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 259659\n",
      " final number of pix to be extracted:  519\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 134\n",
      " final number of pix to be extracted:  134\n",
      "class0 number of pix after edge filtering= 166165\n",
      " final number of pix to be extracted:  332\n",
      "ram usage: 81.8\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 158445\n",
      " final number of pix to be extracted:  317\n",
      "ram usage: 82.1\n",
      "no GT 09_01_pc.png\n",
      "class0 number of pix after edge filtering= 149671\n",
      " final number of pix to be extracted:  299\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 116234\n",
      " final number of pix to be extracted:  232\n",
      "ram usage: 82.9\n",
      "class1 number of pix after edge filtering= 134\n",
      " final number of pix to be extracted:  134\n",
      "class0 number of pix after edge filtering= 194050\n",
      " final number of pix to be extracted:  388\n",
      "ram usage: 82.5\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 190202\n",
      " final number of pix to be extracted:  380\n",
      "ram usage: 82.7\n",
      "no GT 10_06_pc.png\n",
      "class0 number of pix after edge filtering= 132793\n",
      " final number of pix to be extracted:  266\n",
      "ram usage: 82.5\n",
      "no GT 09_08_pc.png\n",
      "class0 number of pix after edge filtering= 170337\n",
      " final number of pix to be extracted:  341\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 162791\n",
      " final number of pix to be extracted:  326\n",
      "ram usage: 81.9\n",
      "no GT 10_05_pc.png\n",
      "class0 number of pix after edge filtering= 99218\n",
      " final number of pix to be extracted:  198\n",
      "ram usage: 82.3\n",
      "no GT 10_03_pc.png\n",
      "class0 number of pix after edge filtering= 78688\n",
      " final number of pix to be extracted:  157\n",
      "ram usage: 82.1\n",
      "no GT 10_08_pc.png\n",
      "class0 number of pix after edge filtering= 163907\n",
      " final number of pix to be extracted:  328\n",
      "ram usage: 82.3\n",
      "no GT 12_09_pc.png\n",
      "class0 number of pix after edge filtering= 240697\n",
      " final number of pix to be extracted:  481\n",
      "ram usage: 82.5\n",
      "no GT 10_04_pc.png\n",
      "class0 number of pix after edge filtering= 88483\n",
      " final number of pix to be extracted:  177\n",
      "ram usage: 82.3\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 153832\n",
      " final number of pix to be extracted:  308\n",
      "ram usage: 82.7\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 197243\n",
      " final number of pix to be extracted:  394\n",
      "ram usage: 82.3\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 195217\n",
      " final number of pix to be extracted:  390\n",
      "ram usage: 82.8\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 157862\n",
      " final number of pix to be extracted:  316\n",
      "ram usage: 82.5\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 145166\n",
      " final number of pix to be extracted:  290\n",
      "ram usage: 82.8\n",
      "no GT 12_05_pc.png\n",
      "class0 number of pix after edge filtering= 167795\n",
      " final number of pix to be extracted:  336\n",
      "ram usage: 82.8\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 159025\n",
      " final number of pix to be extracted:  318\n",
      "ram usage: 82.2\n",
      "no GT 10_07_pc.png\n",
      "class0 number of pix after edge filtering= 123698\n",
      " final number of pix to be extracted:  247\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 162905\n",
      " final number of pix to be extracted:  326\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 158257\n",
      " final number of pix to be extracted:  317\n",
      "ram usage: 82.5\n",
      "no GT 10_09_pc.png\n",
      "class0 number of pix after edge filtering= 74810\n",
      " final number of pix to be extracted:  150\n",
      "ram usage: 82.5\n",
      "no GT 09_07_pc.png\n",
      "class0 number of pix after edge filtering= 151737\n",
      " final number of pix to be extracted:  303\n",
      "ram usage: 82.4\n",
      "no GT 09_09_pc.png\n",
      "class0 number of pix after edge filtering= 125359\n",
      " final number of pix to be extracted:  251\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 123375\n",
      " final number of pix to be extracted:  247\n",
      "ram usage: 82.5\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 179299\n",
      " final number of pix to be extracted:  359\n",
      "ram usage: 81.9\n",
      "no GT 09_03_pc.png\n",
      "class0 number of pix after edge filtering= 190803\n",
      " final number of pix to be extracted:  382\n",
      "ram usage: 82.7\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 209964\n",
      " final number of pix to be extracted:  420\n",
      "ram usage: 82.1\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 173469\n",
      " final number of pix to be extracted:  347\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 165834\n",
      " final number of pix to be extracted:  332\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 67\n",
      " final number of pix to be extracted:  67\n",
      "class0 number of pix after edge filtering= 221897\n",
      " final number of pix to be extracted:  444\n",
      "ram usage: 82.4\n",
      "no GT 09_05_pc.png\n",
      "class0 number of pix after edge filtering= 165833\n",
      " final number of pix to be extracted:  332\n",
      "ram usage: 82.4\n",
      "class1 number of pix after edge filtering= 0\n",
      " final number of pix to be extracted:  0\n",
      "class0 number of pix after edge filtering= 183846\n",
      " final number of pix to be extracted:  368\n",
      "ram usage: 82.2\n",
      "no GT 09_04_pc.png\n",
      "class0 number of pix after edge filtering= 157139\n",
      " final number of pix to be extracted:  314\n",
      "ram usage: 82.5\n",
      "no GT 10_10_pc.png\n",
      "class0 number of pix after edge filtering= 66588\n",
      " final number of pix to be extracted:  133\n",
      "ram usage: 82.7\n",
      "no GT 10_01_pc.png\n",
      "class0 number of pix after edge filtering= 129399\n",
      " final number of pix to be extracted:  259\n",
      "ram usage: 81.8\n",
      "no GT 12_10_pc.png\n",
      "class0 number of pix after edge filtering= 228093\n",
      " final number of pix to be extracted:  456\n",
      "\n",
      "The following 22 datapoints contain no Ground Truth info: \n",
      "['09_10', '09_10', '10_02', '09_01', '10_06', '09_08', '10_05', '10_03', '10_08', '12_09', '10_04', '12_05', '10_07', '10_09', '09_07', '09_09', '09_03', '09_05', '09_04', '10_10', '10_01', '12_10']\n",
      "\n",
      "The following 0 datapoints contain no Blue Ration Segmentation info: \n",
      "[]\n"
     ]
    }
   ],
   "source": [
    "for file in os.listdir(orig_dataset_dir):\n",
    "\n",
    "    if file.endswith(\".tif\"):\n",
    "        \n",
    "        # print(file)\n",
    "        print(\"ram usage: \" + str(psutil.virtual_memory().percent))\n",
    "        basename = os.path.splitext(file)[0]\n",
    "        filename = os.path.join(root + file)\n",
    "        no_gt = False\n",
    "        \n",
    "                \n",
    "        image = cv2.imread(file, cv2.IMREAD_UNCHANGED)\n",
    "        \n",
    "        if type(image) == type(None):\n",
    "        \n",
    "            print(\"Error: file \" + file + \" not read\")\n",
    "        \n",
    "        else:\n",
    "            \n",
    "            # Resize the image to 20x magnification (initially it was 2000x2000)\n",
    "            # The reason behind this is to ensure that the entire mitotic center is contained in a patch\n",
    "            image = cv2.resize(src=image, dsize=(1000,1000))\n",
    "            \n",
    "            # Form the name of the Ground Truth file\n",
    "            # Will later check if it actually exists\n",
    "            gt_maskname = os.path.join(root + basename + \"_pc.png\")\n",
    "\n",
    "            # Read the Ground Truth mask\n",
    "            gt_mask = cv2.imread(gt_maskname, cv2.IMREAD_UNCHANGED)\n",
    "\n",
    "            # Check if the image has a ground truth associated\n",
    "            if type(gt_mask) is type(None):\n",
    "\n",
    "                # Add to a list of missing ground truths\n",
    "                no_centers.append(basename)\n",
    "                print(\"no GT \" + os.path.split(gt_maskname)[1])\n",
    "                no_gt = True\n",
    "\n",
    "            else:\n",
    "\n",
    "                # Resize the image to 20x magnification\n",
    "                # The reason behind this is to ensure that the entire mitotic center is contained in a patch\n",
    "                gt_mask = cv2.resize(src=gt_mask, dsize=(1000,1000))\n",
    "                \n",
    "                # Extract the coordinates of the centers \n",
    "                gt_result = np.where(gt_mask == 255)\n",
    "                # print(\"number of GT pos pix in \" + gt_maskname + \"= \" + str(len(gt_result[0])))\n",
    "                \n",
    "                # Extract the patches surrounding the positive pixels\n",
    "                extract_patches(image, gt_result, 'class1', pos_rots, hwsize, dict_labels_pos, patches_dataset_dir, basename)\n",
    "\n",
    "\n",
    "\n",
    "            # Form the name of the Blue Ratio Segmentation mask file\n",
    "            br_maskname = os.path.join(root + basename + \"_cmask.png\")\n",
    "\n",
    "            # Read the Blue Ratio Segmentation mask\n",
    "            br_mask = cv2.imread(br_maskname, cv2.IMREAD_UNCHANGED)\n",
    "\n",
    "            # Check if the image has a BR mask associated\n",
    "            if type(br_mask) == type(None):\n",
    "\n",
    "                # Add to a list of missing BR masks\n",
    "                no_br.append(basename)\n",
    "                print(\"no BR \" + str(br_maskname))\n",
    "\n",
    "            else:\n",
    "\n",
    "                # Resize the image to 20x magnification\n",
    "                # The reason behind this is to ensure that the entire mitotic center is contained in a patch\n",
    "                br_mask = cv2.resize(src=br_mask, dsize=(1000,1000))        \n",
    "        \n",
    "                # Extract the coordinates of the centers \n",
    "                br_result = np.where(br_mask == 255)\n",
    "                # print(\"Size of BR mask before removing pos pix: \" + str(len(br_result[0])) + \", \" +\n",
    "                # str(len(br_result[1])))\n",
    "\n",
    "                # Check if the Ground Truth mask exists \n",
    "                # (ie. if there are any positive centers in the current image)\n",
    "                if not no_gt:\n",
    "\n",
    "                    # Remove the positive centers from the BR mask\n",
    "                    for id1, id2 in zip(gt_result[0], gt_result[1]):\n",
    "                        br_mask[id1][id2] = 0\n",
    "\n",
    "                    # Extract the coordinates of the centers \n",
    "                    br_result = np.where(br_mask == 255)\n",
    "                    # print(\"Size of BR mask after removing pos pix: \" + str(len(br_result[0])) + \", \" +\n",
    "                    # str(len(br_result[1])))\n",
    "\n",
    "                # Extract the patches whether we had positive centers or not\n",
    "                # These ones will be used as negative samples\n",
    "                extract_patches(image, br_result, 'class0', neg_rots, hwsize, dict_labels_neg, patches_dataset_dir, basename)\n",
    "\n",
    "         \n",
    "    # break \"file\" loop\n",
    "    # break\n",
    "    \n",
    "print(\"\\nThe following \"+ str(len(no_centers)) +\" datapoints contain no Ground Truth info: \")\n",
    "print(no_centers)\n",
    "\n",
    "print(\"\\nThe following \"+ str(len(no_br)) +\" datapoints contain no Blue Ration Segmentation info: \")\n",
    "print(no_br)\n",
    "            \n",
    "# Free up the space\n",
    "no_centers = None\n",
    "no_br = None"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Load an image and visualise/resize with cv2 (for testing purposes)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2000, 2000)\n"
     ]
    }
   ],
   "source": [
    "mask = cv2.imread('/home/advo/PycharmProjects/ML_ND3_CapstoneProject/Dataset/mitosis/01_01_cmask.png', cv2.IMREAD_UNCHANGED)\n",
    "print(mask.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv2.imshow(\"image\", mask)\n",
    "cv2.waitKey(0)\n",
    "cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "image = cv2.resize(src=image, dsize=(1000,1000))\n",
    "cv2.imshow(\"image\", image)\n",
    "cv2.waitKey(0)\n",
    "cv2.destroyAllWindows()"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}