11-test-height-generalizability.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import cv2\n",
    "import glob\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from deep_shadow import *\n",
    "from utils import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Store tile heights in CSV files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "dates = ['summer', 'winter', 'spring']\n",
    "cities = ['la', 'bos', 'nyc', 'chi', 'sp', 'bue', 'dc',\n",
    "          'joh', 'syd', 'tok', 'par', 'mex', 'sea', 'aus']\n",
    "\n",
    "for city in cities:\n",
    "    t_ij = set()\n",
    "    for date in dates:\n",
    "        df = pd.read_csv('data/evaluation/%s-%s-16.csv'%(city, date))\n",
    "        t_ij = t_ij.union(set(zip(df['i'], df['j'])))\n",
    "    \n",
    "    avg_tile_heights = []\n",
    "    for tile in t_ij:\n",
    "        i,j = tile\n",
    "        img = cv2.imread('data/heights_new/%s/%d/%d/%d.png'%(city, zoom, i, j), cv2.IMREAD_GRAYSCALE)\n",
    "        values = (img / 255.0) * 550\n",
    "        masked = np.ma.masked_where(values == 0, values)\n",
    "        \n",
    "        mean = 0 if np.ma.count(masked) == 0 else np.ma.mean(masked)\n",
    "        avg_tile_heights.append([i, j, mean])\n",
    "\n",
    "    # create a dataframe from the list\n",
    "    df = pd.DataFrame(avg_tile_heights, columns=['i', 'j', 'height'])\n",
    "\n",
    "    # sort the dataframe by i and j\n",
    "    df = df.sort_values(by=['i', 'j'])\n",
    "    \n",
    "    # store the dataframe as csv file\n",
    "    df.to_csv('data/tile_heights/%s-%s.csv'%(city, zoom), index=False)\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate the average height tile quantiles of all **train cities** combined"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.10     5.807628\n",
       "0.25     6.470588\n",
       "0.50     6.470588\n",
       "0.75     7.627339\n",
       "0.95    13.582397\n",
       "Name: height, dtype: float64"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# append all these csv files into one and find the percentiles of the heights\n",
    "train_cities = ['aus', 'par', 'syd', 'sp', 'mex', 'la', 'chi']\n",
    "\n",
    "# files = glob.glob('data/tile_heights/*.csv')\n",
    "df = pd.concat([pd.read_csv('data/tile_heights/%s-16.csv'%(city)) for city in train_cities], ignore_index=True)\n",
    "\n",
    "df.quantile([0.1, 0.25, 0.5, 0.75, 0.95])['height']"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Bin tile heights into groups"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "files = glob.glob('data/evaluation/*-summer-16.csv')\n",
    "results = {}\n",
    "\n",
    "for filepath in files:\n",
    "    city = filepath.split('\\\\')[1].split('-')[0]\n",
    "    zoom = 16\n",
    "\n",
    "    df = pd.read_csv('data/tile_heights/%s-%d.csv' % (city, zoom))\n",
    "\n",
    "    bins = [0, 8, 14, np.inf]\n",
    "    labels = ['[0-8)', '[8-14)', '>=14']\n",
    "    df['height_bin'] = pd.cut(df['height'], bins=bins, labels=labels, include_lowest=True)\n",
    "\n",
    "    counts = df['height_bin'].value_counts()\n",
    "    counts = counts.sort_index()\n",
    "\n",
    "    results[city] = counts\n",
    "\n",
    "df_counts = pd.DataFrame.from_dict(results, orient='index')\n",
    "df_counts.columns = ['[0-8)', '[8-14)', '>=14']\n",
    "\n",
    "# create a DataFrame with percentages as a column for each city\n",
    "df_percentages = df_counts.apply(lambda x: (x / x.sum()) * 100, axis=1)\n",
    "df_percentages.columns = ['[0-8)', '[8-14)', '>=14']\n",
    "\n",
    "# concatenate the two dataframes\n",
    "df = pd.concat([df_counts], keys=['Counts', 'Percentages'])\n",
    "\n",
    "df.to_csv('./data/tile_height_distribution.csv', index=True, index_label=['Metric', 'City'])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Save tile height distribution plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_cities = ['aus', 'par', 'syd', 'sp', 'mex', 'la', 'chi']\n",
    "test_cities = ['dc', 'nyc', 'joh', 'bue', 'bos', 'sea', 'tok']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 4000.0)"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot percentage distributions for each city\n",
    "\n",
    "df = pd.read_csv('./data/tile_height_distribution.csv', index_col=[0, 1])\n",
    "df = df.loc['Counts']\n",
    "\n",
    "df = df.loc[train_cities]\n",
    "\n",
    "colors = ['#3b5a9d', '#ffa633', '#4fb2aa', '#e7298a']\n",
    "\n",
    "# df = df.loc[test_cities]\n",
    "ax = df.plot(kind='barh', stacked=True, color=colors, figsize=(20, 8), linewidth=0.2, edgecolor='black')\n",
    "\n",
    "plt.title('Building Height Distribution by City', fontsize=14)\n",
    "\n",
    "plt.xlabel('Counts', fontsize=16)\n",
    "plt.ylabel('City', fontsize=16)\n",
    "\n",
    "\n",
    "plt.xticks(fontsize=14)\n",
    "plt.yticks(fontsize=14)\n",
    "\n",
    "ax.get_xaxis().set_major_formatter(\n",
    "    matplotlib.ticker.FuncFormatter(lambda x, p: format(int(x), ',')))\n",
    "\n",
    "legend = plt.legend(title='Average Tile Height', fontsize=16)\n",
    "plt.setp(legend.get_title(), fontsize=16)\n",
    "\n",
    "ax.xaxis.grid(True, linestyle='--')\n",
    "ax.set_xlim([0, 4000])\n",
    "\n",
    "# frame1 = plt.gca()\n",
    "# frame1.axes.xaxis.set_ticklabels([])\n",
    "# frame1.axes.yaxis.set_ticklabels([])\n",
    "# # remove xlabel and ylabel\n",
    "# ax.set(xlabel=None)\n",
    "# ax.set(ylabel=None)\n",
    "# ax.get_legend().remove()\n",
    "\n",
    "# ax.spines['top'].set_visible(False)\n",
    "# ax.spines['right'].set_visible(False)\n",
    "# ax.spines['bottom'].set_visible(False)\n",
    "# ax.spines['left'].set_visible(False)\n",
    "# plt.title('', fontsize=14);\n",
    "\n",
    "# plt.savefig('./data/tile_height_distribution.png', dpi=300, bbox_inches='tight')\n",
    "# plt.close()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load DeepShadow model ckpt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.keras.backend.clear_session()\n",
    "down_stack, up_stack = get_generator_arch('resnet9', attn=False)\n",
    "deep_shadow = DeepShadow(512, 512, down_stack, up_stack, latitude=True, date=True, loss_funcs=[ssim_loss, sobel_loss, l1_loss], type='resnet9', attention=False)\n",
    "deep_shadow.restore('training_checkpoints_new/uniform_cities_resnet_wo_attn__ssim_sobel_plus_l1')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sample tiles according to tile height bins and compute results using the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "height_path = 'data/heights_new'\n",
    "shadow_path = 'data/shadows'\n",
    "prediction_path = 'data/shadows_new'\n",
    "\n",
    "files = glob.glob('data/heights_new/*')\n",
    "cities = [filepath.split('\\\\')[1] for filepath in files]\n",
    "dates = ['summer', 'winter', 'spring']\n",
    "zoom = 16\n",
    "\n",
    "def sobel(img): return tf.image.sobel_edges(img)\n",
    "\n",
    "for city in cities:    \n",
    "    df = pd.read_csv('data/tile_heights/%s-%d.csv' % (city, zoom))\n",
    "    results = []\n",
    "\n",
    "    for date in dates:\n",
    "        bins = [0, 8, 14, np.inf]\n",
    "        labels = ['[0-8)', '[8-14)', '>=14']\n",
    "        df['height_bin'] = pd.cut(df['height'], bins=bins, labels=labels, include_lowest=True)\n",
    "        \n",
    "        for label in labels:\n",
    "            df_label = df[df['height_bin'] == label]\n",
    "            \n",
    "            maes, mses, rmses, sobels, ssims = [], [], [], [], []\n",
    "            for i, j in zip(df_label['i'], df_label['j']):\n",
    "\n",
    "                try:\n",
    "                    ip_height = cv2.imread('%s/%s/%d/%d/%d.png'%(height_path, city, zoom, i, j), cv2.IMREAD_GRAYSCALE)\n",
    "                    ip_height = ip_height / 255.0\n",
    "\n",
    "                    gt = cv2.imread('%s/%s-%s/%d/%d/%d.png'%(shadow_path, city, date, zoom, i, j), cv2.IMREAD_GRAYSCALE)\n",
    "                    gt = gt / 255.0\n",
    "\n",
    "                    gt = np.where(ip_height > 0, 0, gt)\n",
    "\n",
    "                    prediction = cv2.imread('%s/%s-%s/%d/%d/%d.png'%(prediction_path, city, date, zoom, i, j), cv2.IMREAD_GRAYSCALE)\n",
    "                    prediction = prediction / 255.0\n",
    "\n",
    "                    # We did not do this before. This is to remove shadows from buildings in the prediction\n",
    "                    prediction = np.where(ip_height > 0, 0, prediction)\n",
    "\n",
    "                    maes.append(np.mean(np.abs(gt-prediction)))\n",
    "                    mses.append(np.mean((prediction - gt) ** 2))\n",
    "                    rmses.append(np.sqrt(np.mean((prediction - gt) ** 2)))\n",
    "\n",
    "                    # convert 256x256 to rank 3 tensor\n",
    "                    gt = np.expand_dims(gt, axis=0)\n",
    "                    gt = np.expand_dims(gt, axis=-1)\n",
    "\n",
    "                    prediction = np.expand_dims(prediction, axis=0)\n",
    "                    prediction = np.expand_dims(prediction, axis=-1)\n",
    "                    \n",
    "                    ssim = tf.reduce_mean(tf.image.ssim(tf.convert_to_tensor(gt),\n",
    "                                  tf.convert_to_tensor(prediction), 1.0))\n",
    "                    ssims.append(ssim.numpy())\n",
    "\n",
    "                    sobel_loss = tf.reduce_mean(tf.square(\n",
    "                        sobel(tf.convert_to_tensor(gt)) - sobel(tf.convert_to_tensor(prediction))))\n",
    "                    sobels.append(sobel_loss.numpy())\n",
    "                \n",
    "                except Exception as e:\n",
    "                    print(e)\n",
    "                    \n",
    "\n",
    "            results.append([date, 16, len(maes), label, sum(maes)/len(maes), sum(mses)/len(mses), sum(rmses)/len(rmses), sum(sobels)/len(sobels), sum(ssims)/len(ssims)])\n",
    "\n",
    "    res = pd.DataFrame(results, columns=['season', 'zoom', 'numer of tiles', 'height bin', 'mae', 'mse', 'rmse', 'sobel', 'ssim'])\n",
    "    res.to_csv('data/evaluation_height/%s-%d.csv' % (city, zoom), index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Average per test city per height bin for all seasons and store in one csv file\n",
    "\n",
    "df_csv = pd.DataFrame(columns=['city', 'height bin', 'mae', 'mse', 'rmse', 'sobel', 'ssim'])\n",
    "for city in test_cities:\n",
    "    df = pd.read_csv('data/evaluation_height/%s-%d.csv' % (city, zoom))[['height bin', 'numer of tiles', 'mae', 'mse', 'rmse', 'sobel', 'ssim']]\n",
    "    df = df.groupby(['height bin']).mean()\n",
    "    df = df.reindex(['[0-8)', '[8-14)', '>=14'])\n",
    "    # rmse, mae, mse, sobel, then ssim\n",
    "    df = df.reindex(['numer of tiles', 'rmse', 'mae', 'mse', 'sobel', 'ssim'], axis=1)\n",
    "    # round to 4 decimal places\n",
    "    df = df.round(4)\n",
    "    df = df.reset_index()\n",
    "    # add city column in the beginning\n",
    "    df.insert(0, 'city', city)\n",
    "    # number of tiles is an integer\n",
    "    df['numer of tiles'] = df['numer of tiles'].astype(int)\n",
    "    \n",
    "    df_csv = pd.read_csv('data/evaluation_height/_res.csv')\n",
    "    df_csv = pd.concat([df_csv, df], ignore_index=True)\n",
    "    df_csv.to_csv('data/evaluation_height/_res.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# total number of tiles in three height bins for all 14 cities\n",
    "\n",
    "bin_1, bin_2, bin_3 = 0, 0, 0\n",
    "for city in test_cities + train_cities:\n",
    "    df = pd.read_csv('data/evaluation_height/%s-%d.csv' % (city, zoom))[['height bin', 'numer of tiles']]\n",
    "    df = df.groupby(['height bin']).mean()\n",
    "    bin_1 += df.loc['[0-8)', 'numer of tiles']\n",
    "    bin_2 += df.loc['[8-14)', 'numer of tiles']\n",
    "    bin_3 += df.loc['>=14', 'numer of tiles']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.9.16"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}