{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC Curves related to StyleGAN3 generated image detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# download and unzip pkls with results\n",
    "!wget https://www.dropbox.com/s/h19xrb3v77llxoe/results.zip?dl=0\n",
    "!unzip results.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.metrics import roc_curve, roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib as mpl\n",
    "mpl.use(\"pgf\")\n",
    "pgf_with_rc_fonts = {\n",
    "    \"font.family\": \"serif\",\n",
    "    \"font.serif\": [],                   # use latex default serif font\n",
    "    \"font.sans-serif\": [\"DejaVu Sans\"], # use a specific sans-serif font\n",
    "}\n",
    "mpl.rcParams.update(pgf_with_rc_fonts)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "fontsize = 18\n",
    "plt.rc('text', usetex=True)\n",
    "plt.rc('font', family='sans',size=fontsize)\n",
    "plt.rc('legend',fontsize=fontsize)\n",
    "plt.rc('xtick',labelsize=fontsize)\n",
    "plt.rc('ytick',labelsize=fontsize)\n",
    "plt.rc('axes',labelsize=fontsize,titlesize=fontsize)\n",
    "plt.rc('lines',linewidth=2)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def scores_voting(row):\n",
    "    scores_vot = row['scores'][:, row['maj_voting']]\n",
    "    return scores_vot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def max_scores_voting(row):\n",
    "    max_score_vot = np.nanmax(row['scores_maj_voting']) if row['maj_voting'] == 1 else -np.nanmax(row['scores_maj_voting'])\n",
    "    return max_score_vot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def scores_fusion(row):\n",
    "    fused_score = np.mean([row['image_score_A'], row['image_score_B'], row['image_score_C'], row['image_score_D'], row['image_score_E']])\n",
    "    return fused_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_roc_auc(y_true, y_pred):\n",
    "    auc = roc_auc_score(y_true, y_pred)\n",
    "    print(f'AUC: {auc}')\n",
    "    fpr, tpr, thr = roc_curve(y_true, y_pred)\n",
    "    \n",
    "    plt.figure(figsize=(6,6))\n",
    "    ax = plt.gca()\n",
    "    ax.plot(fpr, tpr, label=f'AUC: {auc:.6f}')\n",
    "    plt.xlabel('$\\mathrm{FPR}$')\n",
    "    plt.ylabel('$\\mathrm{TPR}$')\n",
    "    plt.xlim(-0.01, 1.01);\n",
    "    plt.ylim(-0.01, 1.01);\n",
    "    plt.legend(loc='lower right')\n",
    "    plt.grid();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train method A"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_A = pd.read_pickle('results/method_A.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_A['hard_predictions'] = df_cnn_images_A['scores'].apply(np.argmax, axis=1)\n",
    "df_cnn_images_A['maj_voting'] = df_cnn_images_A['hard_predictions'].map(np.any).astype(int)\n",
    "df_cnn_images_A['scores_maj_voting'] = df_cnn_images_A.apply(scores_voting, axis=1)\n",
    "df_cnn_images_A['image_score'] = df_cnn_images_A.apply(max_scores_voting, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train method B"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_B = pd.read_pickle('results/method_B.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_B['hard_predictions'] = df_cnn_images_B['scores'].apply(np.argmax, axis=1)\n",
    "df_cnn_images_B['maj_voting'] = df_cnn_images_B['hard_predictions'].map(np.any).astype(int)\n",
    "df_cnn_images_B['scores_maj_voting'] = df_cnn_images_B.apply(scores_voting, axis=1)\n",
    "df_cnn_images_B['image_score'] = df_cnn_images_B.apply(max_scores_voting, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train method C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_C = pd.read_pickle('results/method_C.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_C['hard_predictions'] = df_cnn_images_C['scores'].apply(np.argmax, axis=1)\n",
    "df_cnn_images_C['maj_voting'] = df_cnn_images_C['hard_predictions'].map(np.any).astype(int)\n",
    "df_cnn_images_C['scores_maj_voting'] = df_cnn_images_C.apply(scores_voting, axis=1)\n",
    "df_cnn_images_C['image_score'] = df_cnn_images_C.apply(max_scores_voting, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train method D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_D = pd.read_pickle('results/method_D.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_D['hard_predictions'] = df_cnn_images_D['scores'].apply(np.argmax, axis=1)\n",
    "df_cnn_images_D['maj_voting'] = df_cnn_images_D['hard_predictions'].map(np.any).astype(int)\n",
    "df_cnn_images_D['scores_maj_voting'] = df_cnn_images_D.apply(scores_voting, axis=1)\n",
    "df_cnn_images_D['image_score'] = df_cnn_images_D.apply(max_scores_voting, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train method E"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_E = pd.read_pickle('results/method_E.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_E['hard_predictions'] = df_cnn_images_E['scores'].apply(np.argmax, axis=1)\n",
    "df_cnn_images_E['maj_voting'] = df_cnn_images_E['hard_predictions'].map(np.any).astype(int)\n",
    "df_cnn_images_E['scores_maj_voting'] = df_cnn_images_E.apply(scores_voting, axis=1)\n",
    "df_cnn_images_E['image_score'] = df_cnn_images_E.apply(max_scores_voting, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fusion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_fus = df_cnn_images_A.rename(columns={'image_score': 'image_score_A'})\n",
    "df_cnn_images_fus['image_score_B'] = df_cnn_images_B['image_score']\n",
    "df_cnn_images_fus['image_score_C'] = df_cnn_images_C['image_score']\n",
    "df_cnn_images_fus['image_score_D'] = df_cnn_images_D['image_score']\n",
    "df_cnn_images_fus['image_score_E'] = df_cnn_images_E['image_score']\n",
    "\n",
    "df_cnn_images_fus['fused_scores'] = df_cnn_images_fus.apply(scores_fusion, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 1) afhq-v2-test vs alias_free_r_afhqv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'alias-free-r-afhqv2-512x512'\n",
    "test_real_folder = 'afhq-v2-test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9991032719836401\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 2) afhq-v2-test vs alias_free_t_afhqv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'alias-free-t-afhqv2-512x512'\n",
    "test_real_folder = 'afhq-v2-test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]\n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9999360599863667\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 3) Metfaces-in-the-wild-test vs alias_free_r_metfacesu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'alias-free-r-metfacesu-1024x1024'\n",
    "test_real_folder = 'Metfaces-in-the-wild-test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9919059375000001\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 4) Metfaces-in-the-wild-test vs alias_free_t_metfacesu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'alias-free-t-metfacesu-1024x1024'\n",
    "test_real_folder = 'Metfaces-in-the-wild-test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9964493749999999\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 5) FFHQ-in-the-wild-20k vs alias_free_r_no_compression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'generated-ours-r-ffhq-u-8324-2-24400'\n",
    "test_real_folder = 'FFHQ-in-the-wild-20k'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.99999998\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZQAAAGCCAYAAADHSjUmAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAat0lEQVR4nO3dz28c553n8c93IltGAsWtlgLCl6zQ2kuOoeg9EzEJLDC3WUqe6wJj9vwBO+Ro/gGHnD3ktsveBfaQy8Yk5mbAADvYPm+onqMxBzU8QW6R2z2RktiJhe8cqopqtYr989tVXez3CxDgqqeq++Fjdn9Yz/PUU+buAgBgUX9RdgUAANcDgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQtwo+g3NrCbpkaRdd3845Tn7kvrpZsPdj5dUPQDAnAoNFDPblNRQEg6NKc/Zl9R397N0u2FmJ+7eXF5NAQCzKjRQ3L0rqZsGy7Sa7v5g6DV6ZrY1zYl37971e/fuXW7//ve/1/e+970Z3vr6oQ0StEOCdqANMsPt8OTJk2fu/oNZX6PwLq9ZpN1jeVcyfTPbcff2uPPv3buni4uLy+1Op6Pt7e3QOlYNbZCgHRK0A22QGW4HM/vXeV5jpQNFr7rHRg00ZZdZ5r/+n/+v//cvv5c++zSiXtVGGyRohwTtQBukvthe7PxVD5T6mLJa3s50zGVfkjY2NtTpdCQpCRMAwJWy78t5rXqgzMzdW5JakrS1teWXl7LpXyBf/PQvS6rZauDyPkE7JGgH2iAT0Q5VvQ+lVnYFAACvW/VAuVB+t1ddUrfgugAAxljpQHH3gZIZXbWRotqkGV4AgGKVFSi5g+3pTYunIwFypHSQPT1mUxJhAgArpug75RuS9iTtSto0syNJT9OBdCmZCryjJHAGUjLIbmb7ZraXHcNd8gCweoq+U74n6Tj9l1felnQ7Z38r53AAwApZ6TEUAEB1ECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgxI0y3tTM9iX1082Gux9PeU6mJqnl7oP42gEA5lF4oGRh4u5n6XbDzE7cvTnmnAONBIiZnUi68hwAQLHK6PJqZmEiSe7ek7Q14Zz3c65GBmZWC64bAGBOhQZKGgCNnKK+me2MObWRU16jywsAVkfRVygNvRo7GTZQftBkDiWdm9mRJJnZnqST8NoBAOZW9BhKfUxZ7aoCd2+b2QNJT9LxlF137+Ydm47R7EvSxsaGOp3Oa+Wj2+vmxYsXa98GEu2QoR1og0xEO5Qyy2tWZtaQ9KGk25IeK7laabp7a/TYdF9Lkra2tnx7ezsp+OxTSdLl9prqdDpr3wYS7ZChHWiDTEQ7rEqg1CaUHw7NAjs0s19I+qWZ9dy9vdyqAQCmUfQYyoXyu73qkq7qwtqRdD68L+3u+kjSbnQFAQDzKTRQ0llZ/ZzpvrU5rjS6kr6MqBcAYHFl3IdypHTQXJLMbFNSe2i7YWanWeikQfNhzuvsKR0rAQCUr/AxFHdvmdl+OvVXSpZeGb7jvSFpR0k32CDd91E6Zfhpul2TdMZ9KACwOkoZlM+bnTVU1lYym2t430DJvSgAgBXFasMAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhxo4w3NbN9Sf10s+Hux1OedyBpkJ3r7mdLqSAAYGaFB0oWJlkYmFnDzE7cvTnhvHNJD919kG5/ZWbtbBsAUK4yrlCa7v4g23D3npltjTshvTI5HQmPB4QJAKyOQsdQzKwmqZFT1DeznTGnPpb0yfAOd+8FVg0AsKCir1AaejV2Mmyg/KDJQqgmqZ6GzkDSpqQWVygAsDqKDpT6mLLaFfu3lIRIbWjc5ULSqaTd0YPTMZp9SdrY2FCn03mtfHR73bx48WLt20CiHTK0A22QiWiHUmZ5zaEm6bKLy90HZlY3s0137w4f6O4tSS1J2tra8u3t7aTgs08lSZfba6rT6ax9G0i0Q4Z2oA0yEe2wKveh1MaUZVOEBzn7x427AAAKVHSgXCi/26suqZuzXxq6MskxWLRCAIAYhQZKepXRTwfah9XcvT3mnK6ZjQ7aN5QEFABgBZTR5XWkdNBcksxsU1J7aLthZqcjofOxpMORc3qj4ycAgPIUPijv7i0z2zezvXRXY+Qu+YaSsZG60i4tdz8zs+wGR0m64+5vzPACAJSnlFle6Uysq8rakm7n7GfdLgBYYasyywsAUHEECgAgBIECAAhBoAAAQhAoAIAQBAoAIERYoJjZvajXAgBUT+QVym0z+6vA1wMAVMjUgWJm3x9X7u7/vHh1AABVNfFO+fSq439LetfMzt39Pw+V/UTJUin3lTxFcSDpn5ZTVQDAKhsbKGb2Y0n/IOkjJWGxa2b/Q8lCjV0l623107ILDS3gCABYL5OuUPbdfWto+5dm9neSHkt6SDcXACAzaQwl7+FWLUl9wgQAMGxSoPgbO9z/TeOfoggAWEPzThv+Km+nmf23BeoCAKiwSWMoTTO7I+nLkf3vp09NHHZHyYwvAMAamhQod5RMCb6fU/afcva90UUGAFgPkwLlY3f/x2lfzMx+umB9AAAVNXYMZZYwSY//+8WqAwCoqqkG5c3snpn9FQtAAgCuMjFQzOx/KpkmfCbpqZn936XXCgBQOWMDxcw+Sv/ztrv/hZJB+i/M7G+WXjMAQKVMukLZdPe/TW9mlLsP0nGS/7j8qgEAqmRSoPzbFfufRlcEAFBtkwJl9IbGTD+6IgCAapt5La9x+1l6BQDW1yJLr7yfc/yepP8eUjMAQKUssvRK3j4AwJqKXnrl7xasDwCgoqKXXpnpeADA9THpmfL3JNWUdG+5pLa7/2751QIAVM2kWV5dSTtKguSfCBMAwFUmjaG03Z1ZWwCAiSZdocz07Hgz+8kCdQEAVNikQHk24+uNPhYYALAmJnV5/bWZzXK/ySNxYyMArKVJgSJJNuVr1RepCACg2iYFyi+4sREAMI1JYyizXnV0560IAKDaJgVKY5YXc/dfLlAXAECFTQqUXR73CwCYxqQxlA8k1c3sv4ilVwAAY4wNFHf/56IqAgCotkldXgAATIVAAQCEIFAAACEIFABACAIFABCCQAEAhCBQAAAhCBQAQIhplq8PZ2b7kvrpZsPdj2c8/9TdH8bXDAAwr8IDJQsTdz9LtxtmduLuzSnP35S0t8w6AgBmV8YVStPdH2Qb7t4zs60Zzp9pBWQAQDEKHUMxs5ryA6FvZjtTnL+XXdkAAFZL0YPyDb0aOxk20IQrDzNrSOotoU4AgABFd3mNewJkbcK5m9NcnaRjNPuStLGxoU6n81r56Pa6efHixdq3gUQ7ZGgH2iAT0Q6lzPKaVdod1p7mWHdvSWpJ0tbWlm9vbycFn30qSbrcXlOdTmft20CiHTK0A22QiWiHVbkPpXZVQTruIncfFFQXAMAcir5CuVB+t1ddUveKc/aly+nCl8zsQNIgvSIBAJSs0EBx94GZ9c2sNnLFUXP33C6tvJsezexo1pshAQDLVUaX15HSqw7p8sqjPbTdMLPTrKsLAFANhQdK2kU1MLM9M9uTtDNyl3xD0o5yusbMbMfMTtL/Ppnm3hUAQDFKmeU1btwj7fq6PaasLWmqZVoAAMVZlVleAICKI1AAACEIFABACAIFABCCQAEAhCBQAAAhCBQAQAgCBQAQgkABAIQgUAAAIQgUAEAIAgUAEIJAAQCEIFAAACEIFABACAIFABCCQAEAhCBQAAAhCBQAQAgCBQAQgkABAIQgUAAAIQgUAEAIAgUAEIJAAQCEIFAAACEIFABACAIFABCCQAEAhCBQAAAhCBQAQAgCBQAQgkABAIQgUAAAIQgUAEAIAgUAEIJAAQCEIFAAACEIFABACAIFABCCQAEAhCBQAAAhCBQAQAgCBQAQgkABAIQgUAAAIQgUAEAIAgUAEIJAAQCEIFAAACFulPGmZrYvqZ9uNtz9eMpzJOm+pIakj9x9sJwaAgBmVXigZGHi7mfpdsPMTty9Oe4cd28Nbe9JeqIkXAAAK6CMLq9mFiaS5O49SVtXHWxmDY0ER3p+PQ0WAMAKKDRQzKympLtqVN/Mdsacup+zry+pHlEvAMDiiu7yaujV2MmwgfKDJruCuX3Fa12E1QwAsJCiu7zGXVHUpn2RdBym7e7dhWsEAAhRyiyvRaRjKk13f3BF+b7SLrKNjQ11Op3Xyke3182LFy/Wvg0k2iFDO9AGmYh2WJVAqc1w7JGkD64qTGeDtSRpa2vLt7e3k4LPPpUkXW6vqU6ns/ZtINEOGdqBNshEtEPRXV4Xyu/2qkua2H1lZkeSDrn/BABWT6GBkgZBP53tNazm7u1x56ZdWSfpIH22b9zMMABAgcq4D+VIQ9OAzWxTUntou2Fmp8OhkwbHxUiYbBZTXQDANAofQ3H3lpntD92U2Bi5S74haUdJN9ggHYQ/lyQzG325vOnEAIASlDIoP7yMSk5ZW0NBkV6VvJEkAIDVwmrDAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACEGgAABCECgAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEAQKACAEgQIACFHKM+WB6+Sbb75Rv9/X8+fP9fLly7KrM5N3331Xn3/+ednVKNW6tMHbb7+tu3fv6t13313aexAowAK++eYb/frXv9bt27d17949vfXWWzKzsqs1tefPn+vWrVtlV6NU69AG7q4//vGP+s1vfqObN2/qnXfeWcr70OUFLKDf7+v27du6e/eu3n777UqFCdaHmem73/2u7t69q9/+9rdLex8CBVjA8+fP9f3vf7/sagBTuXXrlr7++uulvT6BAizg5cuXeuutt8quBjCVGzdu6Ntvv13a6xMowILo5kJVLPt3lUABAIQgUAAAIQgUAEAIAgXA3AaDwZVlx8fHun//vh48eKDj4+PXylqtlu7fv6/d3V11u93ccw8PD3V8fKxWq6WzszNJ0tnZ2dj3RLm4sRHAXM7OznRycqLz8/Pc8oODA/3qV79So9HQwcHBa2X7+/uSpEePHqlWq13u7/V6ajabajabr50zGAzUarV0dHSkJ0+ezFzXwWCgTz75ROfn5zo9PZ14fKvVUr1ev6zTaP0XLS/qPYpGoACYS6/XU7vd1mAweC0UplWv19847+HDhzo6OtLOzs5r+2u1mh49eqRmsznz+3S7XfV6PdXrdfV6vYnHZ1/Ue3t7kl6F3MnJSUh5Ue9RBgIFwMwGg4E2Nze1ubmpTz755PKKYxHHx8eq1+tvhEmmVqvN9T5ZPfO61vKcnJy8dhXUaDR0cXERVl7Ue5SBMRQAM2u329rZ2dHjx4/D/io+OTnRw4cPxx6zu7sb8l5XGQwGuVcx9Xr98mpskfKi3qMsXKEAS3Dv7z8tuwq5vvjpX4a+3t7enh4+fKher6dGo7HQa/V6PW1tbU18v8zZ2Zk+/vjjucZUxtUhG5cYVqvVLssWKS/qPcrCFQqAmfR6PW1ubl5u7+/vL3yVMs8X4c7Ojo6OjhZ631H9fv/KssFgsHB5Ue9RFq5QgCWIvhJYJb1e77VxjmazqQ8++GChL/d5rm5qtdqV4y0oB4ECYCbn5+e5A9zZuMq8soHl4aufURFda/OY9Jf/ouVFvceyESgAptbtdvX48eM3pvt++eWXOjk5eSNQ3n//fT19+jT3tUa7bprNpk5PT8fO5Op2u0sNlK2trdwupX6/r83NzYXLi3qPsjCGAmBqvV4v956TDz/88PJu9mF7e3tTzzw6ODhQv9+fenrvMtRqNdXr9Tf+2h8MBtrZ2Vm4vKj3KAuBAmAq4waENzc31Wg03giVRqOhvb293KVXHj169MbrnJ6e6vDwMDeEWq3Wa7O8BoPBTNNkr6r76OscHh6q1Wpdbne73de+qBctL+o9ymDuXnYdlmZra8uzm32yaZzXebB0Gp1OR9vb22VXo3RR7fD555/rRz/60eIVKsm0z1Nvt9s6PDzUYDDQ6enpG10rrVZLh4eHqtfrbyybkpU/ffpUd+7ckZRcuYzruhoOoOyKaLQrbNppw71eT2dnZzo/P1e73dbBwYHu379/+Xo///nP9bOf/ey117nOS69c9Ts7/JkwsyfuPn4Odw4CZc0QKAkCJTFtoFxn69YGywwUurwAACEIFABACAIFABCCQAEAhCBQAAAhCBRgQdd5piSul2X/rhIowAK+853v6M9//nPZ1QCm8u233+rGjeWtuEWgAAu4deuWfve735VdDWAqz58/1zvvvLO01ydQgAXU63V99dVXevbsmf70pz/R/YWV5O76wx/+oGfPnukHP/jB0t6nlNWGzWxfUrawTsPdj8cdP+85wLLdvHlTP/zhD9Xv9/XFF1/o5cuXZVdpJl9//fVS/2KtgnVpg5s3b2pjY2OpP2vhgZIFg7ufpdsNMztx92bkOUBRbt68qffee0/vvfde2VWZWafT0Y9//OOyq1Eq2iBOGVcoTXd/kG24e8/MJq0ZM885AIACFTqGYmY1SXlLjPbNLHft5XnOAQAUr+hB+YZejYMMGyg/NOY9BwBQsKIDpT6mrBZ4DgCgYNfumfLpAP6+JG1sbKjT6bxWPrq9bl68eLH2bSDRDhnagTbIRLTDqgRKLeocd29JaknJA7ayB8Z8sc3DpSTaIEM7JGgH2iAT0Q5Fd3ldKL8Lqy6pG3gOAKBghT8C2MyeSnrg7oPhfe5+P/Kc9JjfSvrXoV13JT2bs+rXBW2QoB0StANtkBluh//g7jPfUl9Gl9eRkjGOY0kys01J7azQzBrpMR8NBcjYc64y2iBmdjHPc5KvE9ogQTskaAfaIBPRDoUHiru3zGzfzPbSXY2RO94bknaUdGkNpjwHAFCyUgbl04Hzq8rakm7Pcg4AoHzrttowoUQbZGiHBO1AG2QWbofCB+UBANfTul2hAACWhEABAIRYlTvlF8ZDuxILtIMk3Vcyy254ynblLPr/1cxO3f1hfM2KNW87mNmBkhmWfUnKnkNUVQt+JqRkVY5WVT8T6YrtjyTtTvt7PfdnyN0r/0/JPSp7Q9sNSSfR56z6v3nbYWR7T9LTsn+WIttg5PzN5GNR/s9SRjtIOpdUG9r+ani7av/m/EwcjP7MVf1uSH+f99J/T5b5u+Pu1yZQ3mioSY03zzmr/m/Wnyn9RTnK2f/V8C9Ulf4t+v81/eB52T9HGe2QfpGO/oHRKPtnKaEdTnP2HVU8WDdnCJS5P0OVH0PhoV2JBX6m/Zx9fY1/bMBKWvT/q5ntecW7d6SF2uGxpE+Gd7h7L7BqhVqgHRo55TWvaJfXLBb9DFU+UMRDuzIz/0zu3nP3N24iTY+/iKtaYeb+/5ou+VPZL88RM7dD+kVSk1Q3sz0z2zGzg3R/Vc37+3Ao6dzMjqTkDw1JJ+G1W00LfTdeh0DhoV2JkJ8pHYxru3sVV3JepA02K/oz55mnHbaUfGnU3P3MkxUrWpJOY6tWqLl+H9Kf/YGkAzNzSYNr9LsxyULfI9chUBAk/Su96e67ZdelSOml/MTFRtdATUNXaWkXTz1djHVtpJ+DD5UsAXWs5Golr2sYI65zoNQKOmfV1WY49kjSB0uqR5lqVxVkXTrr0D+u8b8L2RThQc7+So4rjlGbUH7o7ofuPnD3QyVXK0dVHV8NUpvmoOtwHwoP7Uos9DOl/cWHFf9inacN9qXLRyJcyu7F8GouSjpPO4wbPxosWqGSzNwOaWicD+9z966ZfSRpV9f/Snah75HKB4q7D8ysb2ajszBqaV9oyDmrbpGfKb2cPxme0WNmO1Vrizl/F964YcvMjvL2V8UCn4mumTVGZnZVdYJG9Oe8q+pO2Jnaom12Xbq8sgdwScp/aJeZnY7MWBl7TkXN3A7pX2QXI2FS5T7zeX4XrqN52uFjJTOchs/pVXxAeqZ2SL80P8x5nT1Ve1Xi3MH26O/Ga7Pa8LilAtIvzVMljxHuTXNOVc3SDung49MrXup2Vbu/5vldGCp7qOTD1FJyg1tl/8iY8zOxp1d/id9JxxAqbdZ2SL9cH+vVZ6Mm6ayK9+Skn/E9Jd11O0omGTzNunKjvxuvTaAAAMp1Xbq8AAAlI1AAACEIFABACAIFABCCQAEAhCBQAAAhKn+nPFCWdHmWppKlSX4xVHRfyZz/gbs/GDr2sZK5/dlS6HeU3OPwZHiJlyuOzV73pOI3GuIa4z4UYAFmdqrkbvLDkf01JTdG7k5x7IkkuXtzimOfSPr4OjwIDNcPXV7AEqSrDJznFH2Zs+9Q0n56V/OkYz+W9L8Wqx2wHAQKsDzdadYMG1riZpo11AaSamuwFhkqiEABgg0trjnVKr1D4TDN2MiukidqDmavGbBcDMoDgdJuq7o000O7HmvC4oNp6OwrWUZ8rZ6oieogUIDFbaYzs+4oWdm1OebY99MVfaUkeO4rWf01b1Xf99PVYGtKVgC+o6Hl5YFVQ6AAi+tmy3ubWd5A+rDeDDO0fjW8fH7alfZESQgBK4cxFCDW0h7ClN5/Ul/zZ5tjhXGFAgRKx02W+VCuvtbgUbSoJq5QgGoZiC4vrCgCBVhdd3L2XSi9X8XMNnNuhgRKQ6AAc0pndm1K2jOzg3Ff7qPHTnjdo/Q/GznHHkoaZM9+r+JzznF9sZYXACAEVygAgBAECgAgBIECAAhBoAAAQhAoAIAQBAoAIASBAgAIQaAAAEIQKACAEP8OoP5F3ABvyN0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 6) FFHQ-in-the-wild-20k vs alias_free_t_no_compression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'generated-ours-t-ffhq-u-8325-4-25000'\n",
    "test_real_folder = 'FFHQ-in-the-wild-20k'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]   \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9999986750000001\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 7) FFHQ-in-the-wild-20k vs alias_free_r_compression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'config-r'\n",
    "test_real_folder = 'FFHQ-in-the-wild-20k'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9995815349999999\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 8) FFHQ-in-the-wild-20k vs alias_free_t_compression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_synth_folder = 'config-t'\n",
    "test_real_folder = 'FFHQ-in-the-wild-20k'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cnn_images_test_synth = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_synth_folder)] \n",
    "df_cnn_images_test_real = df_cnn_images_fus[df_cnn_images_fus['image'].str.contains(test_real_folder)]  \n",
    "df_cnn_images_test = pd.concat([df_cnn_images_test_synth, df_cnn_images_test_real]).reset_index().drop('index', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9995660525000001\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_test['label'].values, df_cnn_images_test['fused_scores'].values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC 9) Global"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9995606906259085\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_roc_auc(df_cnn_images_fus['label'].values, df_cnn_images_fus['fused_scores'].values)"
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}