image_labeling_missing_data_viirs.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "import random\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "from PIL import Image\n",
    "import cv2\n",
    "\n",
    "import scipy\n",
    "from scipy import ndimage\n",
    "\n",
    "from features import rgb2gray\n",
    "\n",
    "# This is a bit of magic to make matplotlib figures appear inline in the notebook\n",
    "# rather than in a new window.\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (30.0, 16.0) # set default size of plots\n",
    "plt.rcParams['image.interpolation'] = 'nearest'\n",
    "plt.rcParams['image.cmap'] = 'gray'\n",
    "\n",
    "# Some more magic so that the notebook will reload external python modules;\n",
    "# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_labels(label_arr):\n",
    "    \"\"\"    \n",
    "    None = Normal\n",
    "    1 = missing data (expected)\n",
    "    2 = missing data (unexpected)\n",
    "    3 = miscoloration\n",
    "    4 = edge warping\n",
    "    5 = eclipse (missing data)\n",
    "    6 = eclipse (only miscoloration)\n",
    "    \"\"\"\n",
    "    label = 0\n",
    "    if len(label_arr) == 0:\n",
    "        label = 0\n",
    "    else:\n",
    "        if 1 in label_arr or 2 in label_arr or 5 in label_arr: # missing data labels!\n",
    "            label = 1\n",
    "        else:\n",
    "            label = 0\n",
    "    return label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "layer_name = 'VIIRS_SNPP_CorrectedReflectance_TrueColor'\n",
    "img_extension = \".jpg\"\n",
    "data_dir = 'data/4326/'\n",
    "labels_file = os.path.join(data_dir, layer_name + \".txt\")\n",
    "img_size = (4096, 2048)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2015-11-25: Detected LARGE hole (0.2586461305618286 missing data)\n",
      "2015-12-11: Detected SMALL hole (3 missing data holes)\n",
      "2015-12-21: Detected SMALL hole (1 missing data holes)\n",
      "2015-12-23: Detected SMALL hole (1 missing data holes)\n",
      "2016-01-09: Detected SMALL hole (3 missing data holes)\n",
      "2016-01-11: Detected SMALL hole (1 missing data holes)\n",
      "2016-01-14: Detected SMALL hole (1 missing data holes)\n",
      "2016-01-18: Detected SMALL hole (1 missing data holes)\n",
      "2016-01-19: Detected SMALL hole (1 missing data holes)\n",
      "2016-01-30: Detected SMALL hole (1 missing data holes)\n",
      "2016-02-06: Detected SMALL hole (1 missing data holes)\n",
      "2016-02-07: Algorithm MISSED the hole detection (TN).\n",
      "2016-02-18: Detected SMALL hole (1 missing data holes)\n",
      "2016-03-09: Detected SMALL hole (2 missing data holes)\n",
      "2016-03-18: Detected SMALL hole (1 missing data holes)\n",
      "2016-03-31: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-01: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-02: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-03: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-05: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-06: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-07: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-09: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-10: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-12: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-13: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-14: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-16: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-17: Detected SMALL hole (3 missing data holes)\n",
      "2016-04-18: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-19: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-21: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-22: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-23: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-25: Detected SMALL hole (2 missing data holes)\n",
      "2016-04-28: Detected SMALL hole (1 missing data holes)\n",
      "2016-04-29: Detected SMALL hole (4 missing data holes)\n",
      "2016-04-30: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-02: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-03: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-04: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-07: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-09: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-10: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-11: Detected SMALL hole (3 missing data holes)\n",
      "2016-05-12: Detected SMALL hole (2 missing data holes)\n",
      "2016-05-14: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-15: Detected SMALL hole (2 missing data holes)\n",
      "2016-05-16: Detected SMALL hole (1 missing data holes)\n",
      "2016-05-17: Detected SMALL hole (1 missing data holes)\n",
      "2016-06-15: Detected SMALL hole (1 missing data holes)\n",
      "2016-07-13: Detected SMALL hole (1 missing data holes)\n",
      "2016-08-13: Detected SMALL hole (1 missing data holes)\n",
      "2016-08-18: Detected SMALL hole (3 missing data holes)\n",
      "2016-09-01: Detected SMALL hole (4 missing data holes)\n",
      "2016-10-19: Detected SMALL hole (1 missing data holes)\n",
      "2016-11-09: Detected SMALL hole (1 missing data holes)\n",
      "2016-11-10: Detected SMALL hole (1 missing data holes)\n",
      "2016-11-29: Detected SMALL hole (3 missing data holes)\n",
      "2016-12-10: Detected SMALL hole (1 missing data holes)\n",
      "2016-12-14: Detected SMALL hole (1 missing data holes)\n",
      "2016-12-20: Detected LARGE hole (0.22177481651306152 missing data)\n",
      "2016-12-29: Detected SMALL hole (3 missing data holes)\n",
      "2017-01-08: Detected SMALL hole (1 missing data holes)\n",
      "2017-01-28: Detected SMALL hole (3 missing data holes)\n",
      "2017-02-06: Detected SMALL hole (1 missing data holes)\n",
      "2017-02-26: Detected SMALL hole (3 missing data holes)\n",
      "2017-03-08: Detected SMALL hole (1 missing data holes)\n",
      "2017-03-15: Detected SMALL hole (1 missing data holes)\n",
      "2017-03-19: Detected SMALL hole (1 missing data holes)\n",
      "2017-03-26: Detected SMALL hole (1 missing data holes)\n",
      "2017-03-29: Detected SMALL hole (2 missing data holes)\n",
      "2017-03-30: Detected LARGE hole (0.978532075881958 missing data)\n",
      "2017-03-31: Detected LARGE hole (1.0 missing data)\n",
      "2017-04-01: Detected LARGE hole (0.904449462890625 missing data)\n",
      "2017-04-05: Detected SMALL hole (1 missing data holes)\n",
      "2017-04-06: Detected SMALL hole (1 missing data holes)\n",
      "2017-05-06: Detected SMALL hole (1 missing data holes)\n",
      "2017-06-04: Detected SMALL hole (1 missing data holes)\n",
      "2017-06-30: Detected SMALL hole (1 missing data holes)\n",
      "2017-07-03: Detected SMALL hole (1 missing data holes)\n",
      "2017-07-26: Detected SMALL hole (1 missing data holes)\n",
      "2017-08-21: Detected SMALL hole (7 missing data holes)\n",
      "2017-08-27: Detected SMALL hole (2 missing data holes)\n",
      "2017-09-02: Detected SMALL hole (1 missing data holes)\n",
      "2017-09-29: Detected LARGE hole (0.446364164352417 missing data)\n",
      "2017-10-31: Detected SMALL hole (1 missing data holes)\n",
      "2017-11-18: Detected SMALL hole (3 missing data holes)\n",
      "2017-11-25: Detected SMALL hole (2 missing data holes)\n",
      "2017-11-29: Detected SMALL hole (1 missing data holes)\n",
      "2017-11-30: Detected LARGE hole (0.17970287799835205 missing data)\n",
      "2017-12-18: Detected SMALL hole (3 missing data holes)\n",
      "2017-12-29: Detected SMALL hole (1 missing data holes)\n",
      "2017-12-30: Detected LARGE hole (0.2704439163208008 missing data)\n",
      "2017-12-31: Detected SMALL hole (2 missing data holes)\n",
      "2018-01-02: Detected SMALL hole (1 missing data holes)\n",
      "2018-01-05: Detected SMALL hole (1 missing data holes)\n",
      "2018-01-16: Detected SMALL hole (3 missing data holes)\n",
      "2018-01-27: Detected SMALL hole (1 missing data holes)\n",
      "2018-01-29: Detected LARGE hole (0.1728205680847168 missing data)\n",
      "2018-02-13: Detected SMALL hole (1 missing data holes)\n",
      "2018-02-17: Detected SMALL hole (1 missing data holes)\n",
      "2018-02-26: Detected SMALL hole (1 missing data holes)\n",
      "2018-03-10: Detected SMALL hole (1 missing data holes)\n",
      "2018-03-27: Detected SMALL hole (2 missing data holes)\n",
      "2018-04-19: Detected SMALL hole (1 missing data holes)\n",
      "2018-04-25: Detected SMALL hole (1 missing data holes)\n",
      "2018-05-25: Detected SMALL hole (1 missing data holes)\n",
      "2018-06-21: Detected SMALL hole (2 missing data holes)\n",
      "2018-06-23: Detected SMALL hole (1 missing data holes)\n"
     ]
    }
   ],
   "source": [
    "# Check the labels file exists!\n",
    "assert os.path.exists(labels_file), \"Cannot find the {} file\".format(layer_name + \".txt\")\n",
    "        \n",
    "# Read in the file line by line\n",
    "with open(labels_file) as f:\n",
    "    file_lines = f.read().splitlines()\n",
    "    num_total_img = len(file_lines)\n",
    "    for line in file_lines:\n",
    "        line_list = line.split()  \n",
    "        split = line_list[0]\n",
    "        datestring = line_list[1]\n",
    "        label_arr = [int(item) for item in line_list[2:]]\n",
    "        label = parse_labels(label_arr)\n",
    "        \n",
    "        # Construct and resize the image\n",
    "        filename = os.path.join(data_dir, datestring, layer_name + img_extension)\n",
    "        image = np.asarray(Image.open(filename).resize(img_size, Image.BILINEAR))\n",
    "        \n",
    "        # Grayscale the image\n",
    "        imgray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)\n",
    "\n",
    "        # Threshold the image to black and white\n",
    "        _, im_thresh = cv2.threshold(imgray, 0, 255, cv2.THRESH_BINARY_INV)\n",
    "\n",
    "        # Copy the thresholded image\n",
    "        im_floodfill = im_thresh.copy()\n",
    "\n",
    "        # Fill in very small holes\n",
    "        small_kernel = np.ones((4,4),np.uint8)\n",
    "        im_floodfill = cv2.erode(im_floodfill, small_kernel, iterations = 1)\n",
    "\n",
    "        # Expand larger holes\n",
    "        large_kernel = np.ones((30,30),np.uint8)\n",
    "        im_floodfill = cv2.dilate(im_floodfill, large_kernel, iterations = 1)\n",
    "\n",
    "        # Mask used to flood filling.\n",
    "        # Notice the size needs to be 2 pixels than the image.\n",
    "        h, w = im_floodfill.shape\n",
    "        mask = np.zeros((h+2, w+2), np.uint8)\n",
    "\n",
    "        # Floodfill from top and bottom\n",
    "        for w_idx in range(w):\n",
    "            cv2.floodFill(im_floodfill, mask, (w_idx,0), 0);   # flood fill top\n",
    "            cv2.floodFill(im_floodfill, mask, (w_idx,h-1), 0); # flood fill bottom\n",
    "\n",
    "        # Count the number of holes\n",
    "        _, nr_objects = ndimage.label(im_floodfill)\n",
    "        pre_ratio_missing = 1.0 - (im_thresh == 0).sum() / im_thresh.size\n",
    "        post_ratio_missing = 1.0 - (im_floodfill == 0).sum() / im_thresh.size\n",
    "        missing_data_flag = False\n",
    "        if nr_objects == 0 and pre_ratio_missing > 0.165:\n",
    "            missing_data_flag = True\n",
    "            print(\"{}: Detected LARGE hole ({} missing data)\".format(datestring, pre_ratio_missing))\n",
    "        if nr_objects != 0:\n",
    "            missing_data_flag = True\n",
    "            print(\"{}: Detected SMALL hole ({} missing data holes)\".format(datestring, nr_objects))\n",
    "        \n",
    "        if label == 0 and missing_data_flag: # False Positive\n",
    "            print(\"^^^ Algorithm INCORRECTLY detected missing data (FP)\")\n",
    "        elif label == 1 and not missing_data_flag: # True Negative\n",
    "            print(\"{}: Algorithm MISSED the hole detection (TN).\".format(datestring))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [default]",
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}