20_Final_Project/thesis_model_test.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix, precision_recall_curve\n",
    "\n",
    "base_dirs = ['./2nd_Battery/unet',\n",
    "            './2nd_Battery/unet-mini',\n",
    "            './2nd_Battery/unet-sgd',\n",
    "            './2nd_Battery/unet-rmsprop',\n",
    "            './2nd_Battery/unet-dice-loss',\n",
    "            './2nd_Battery/unet-focal-loss',\n",
    "            './2nd_Battery/unet-l1',\n",
    "            './2nd_Battery/unet-l2',\n",
    "            './2nd_Battery/unet-l1+l2']\n",
    "colors = ['red', 'orange', 'yellow', 'pink', 'green', 'gold', 'magenta', 'cyan', 'violet']\n",
    "losses = [0.2072, 0.1324, 0.1787, 0.1666, 0.3879, 0.0112, 0.0357, 0.0241, 0.0550]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unet - precision: 0.582, recall: 0.895, accuracy: 0.958, f1: 0.685, iou: 0.550\n",
      "unet-mini - precision: 0.708, recall: 0.929, accuracy: 0.973, f1: 0.794, iou: 0.676\n",
      "unet-sgd - precision: 0.541, recall: 0.960, accuracy: 0.956, f1: 0.681, iou: 0.530\n",
      "unet-rmsprop - precision: 0.855, recall: 0.789, accuracy: 0.984, f1: 0.794, iou: 0.696\n",
      "unet-dice-loss - precision: 0.509, recall: 0.934, accuracy: 0.946, f1: 0.642, iou: 0.492\n",
      "unet-focal-loss - precision: 0.894, recall: 0.762, accuracy: 0.984, f1: 0.818, iou: 0.711\n",
      "unet-l1 - precision: 0.670, recall: 0.885, accuracy: 0.971, f1: 0.752, iou: 0.618\n",
      "unet-l2 - precision: 0.778, recall: 0.788, accuracy: 0.977, f1: 0.777, iou: 0.649\n",
      "unet-l1+l2 - precision: 0.714, recall: 0.908, accuracy: 0.976, f1: 0.790, iou: 0.667\n"
     ]
    }
   ],
   "source": [
    "lst_rst, lst_cdata = [], []\n",
    "idx = 0\n",
    "\n",
    "for base_dir in base_dirs:\n",
    "    result_dir = os.path.join(base_dir, 'result')\n",
    "\n",
    "    ##\n",
    "    lst_data = os.listdir(os.path.join(result_dir, 'numpy'))\n",
    "\n",
    "    lst_gt = [f for f in lst_data if f.startswith('gt')]\n",
    "    lst_pr = [f for f in lst_data if f.startswith('pr')]\n",
    "\n",
    "    lst_gt.sort()\n",
    "    lst_pr.sort()\n",
    "\n",
    "    ##\n",
    "    # id = 0\n",
    "    length = 1000\n",
    "    tp_list, tn_list, fp_list, fn_list, rst_list = [], [], [], [], []\n",
    "\n",
    "    for id in range(0, length):\n",
    "        # img = np.load(os.path.join(result_dir,\"numpy\", lst_img[id]))\n",
    "        gt = np.load(os.path.join(result_dir,\"numpy\", lst_gt[id]))\n",
    "        pr = np.load(os.path.join(result_dir,\"numpy\", lst_pr[id]))\n",
    "\n",
    "        tp = np.sum(np.logical_and(gt == 1, pr == 1))\n",
    "        tn = np.sum(np.logical_and(gt == 0, pr == 0))\n",
    "        fp = np.sum(np.logical_and(gt == 0, pr == 1))\n",
    "        fn = np.sum(np.logical_and(gt == 1, pr == 0))\n",
    "\n",
    "        precision = tp / (tp + fp)  # precision = TP / (TP + FP)\n",
    "        recall = tp / (tp + fn) # recall = TP / (TP + FN), SE(Sensitivity), hit rate\n",
    "        accuracy = (tp + tn) / (tp + tn + fp + fn)\n",
    "        f1 = 2 * precision * recall / (precision + recall)\n",
    "        iou = tp / (tp + fn + fp) \n",
    "\n",
    "        tp_list.append(tp)\n",
    "        tn_list.append(tn)\n",
    "        fp_list.append(fp)\n",
    "        fn_list.append(fn)\n",
    "\n",
    "\n",
    "        # output_binary = (pr > 0).astype(np.int32)\n",
    "        # label_flat = gt.flatten().astype(np.int32)\n",
    "        # output_flat = output_binary.flatten().astype(np.int32)\n",
    "\n",
    "        # accuracy = accuracy_score(label_flat, output_flat)\n",
    "        # precision = precision_score(label_flat, output_flat)\n",
    "        # recall = recall_score(label_flat, output_flat)\n",
    "        # f1 = f1_score(label_flat, output_flat)\n",
    "\n",
    "        rst_list.append((precision, recall, accuracy, f1, iou))\n",
    "    \n",
    "    # avg_precision = sum(item[0] for item in rst_list) / len(rst_list)\n",
    "    # avg_recall = sum(item[1] for item in rst_list) / len(rst_list)\n",
    "    # avg_accuracy = sum(item[2] for item in rst_list) / len(rst_list)\n",
    "    # avg_f1 = sum(item[3] for item in rst_list) / len(rst_list)\n",
    "    # avg_iou = sum(item[4] for item in rst_list) / len(rst_list)\n",
    "    avg_precision = np.mean([item[0] for item in rst_list])\n",
    "    avg_recall = np.mean([item[1] for item in rst_list])\n",
    "    avg_accuracy = np.mean([item[2] for item in rst_list])\n",
    "    avg_f1 = np.mean([item[3] for item in rst_list])\n",
    "    avg_iou = np.mean([item[4] for item in rst_list])\n",
    "    print(f'{os.path.basename(base_dir)} - precision: {avg_precision:.3f}, recall: {avg_recall:.3f}, accuracy: {avg_accuracy:.3f}, f1: {avg_f1:.3f}, iou: {avg_iou:.3f}')\n",
    "\n",
    "    total_tp = np.sum(tp_list)\n",
    "    total_tn = np.sum(tn_list)\n",
    "    total_fp = np.sum(fp_list)\n",
    "    total_fn = np.sum(fn_list)\n",
    "    y_true = np.concatenate([np.ones(total_tp + total_fn), np.zeros(total_tn + total_fp)])\n",
    "    y_score = np.concatenate([np.ones(total_tp), np.zeros(total_fn), np.ones(total_fp), np.zeros(total_tn)])\n",
    "\n",
    "    lst_cdata.append((y_true, y_score))\n",
    "    lst_rst.append((losses[idx], avg_precision, avg_recall, avg_accuracy, avg_f1, avg_iou))\n",
    "    idx += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "avg_loss_list = [item[0] for item in lst_rst]\n",
    "avg_precision_list = [item[1] for item in lst_rst]\n",
    "avg_recall_list = [item[2] for item in lst_rst]\n",
    "avg_accuracy_list = [item[3] for item in lst_rst]\n",
    "avg_f1_list = [item[4] for item in lst_rst]\n",
    "avg_iou_list = [item[5] for item in lst_rst]\n",
    "model_list = ['U-Net', 'Mini', 'SGD', 'RMSProp', 'Dice', 'Focal', 'L1', 'L2', 'L1+L2']\n",
    "\n",
    "plt.plot(model_list, avg_loss_list, '-', color='blue', label=f'Avg Loss')\n",
    "plt.plot(model_list, avg_precision_list, '-', color='green', label=f'Avg Precision')\n",
    "plt.plot(model_list, avg_recall_list, '-', color='red', label=f'Avg Recall')\n",
    "plt.plot(model_list, avg_accuracy_list, '-', color='gold', label=f'Avg Accuracy')\n",
    "plt.plot(model_list, avg_f1_list, '-', color='orange', label=f'Avg F1-Score')\n",
    "plt.plot(model_list, avg_iou_list, '-', color='pink', label=f'mIoU')\n",
    "plt.title('Performance Evaluation of Different Models')\n",
    "plt.legend(bbox_to_anchor=(0.69, 0.4))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unet PR AUC: 0.704, minDist: 0.497, (0.885424512257604, 0.51682025846915)\n",
      "unet-mini PR AUC: 0.780, minDist: 0.375, (0.9224052152557138, 0.6334823172698756)\n",
      "unet-sgd PR AUC: 0.731, minDist: 0.497, (0.9538732855195041, 0.5053105923119683)\n",
      "unet-rmsprop PR AUC: 0.813, minDist: 0.283, (0.7531937969524044, 0.8625253033281114)\n",
      "unet-dice-loss PR AUC: 0.692, minDist: 0.555, (0.931617865125619, 0.44876511258753)\n",
      "unet-focal-loss PR AUC: 0.822, minDist: 0.277, (0.7487617009547043, 0.8840717896362793)\n",
      "unet-l1 PR AUC: 0.754, minDist: 0.393, (0.8766366499467493, 0.626387811957221)\n",
      "unet-l2 PR AUC: 0.760, minDist: 0.349, (0.7813983872652583, 0.7285129156027101)\n",
      "unet-l1+l2 PR AUC: 0.789, minDist: 0.341, (0.9011364203573499, 0.6734219064478764)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkIAAAHHCAYAAABTMjf2AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAAEAAElEQVR4nOydd3hUVfrHP9PSGzWEEEjoSC+hKAgCShFQEEWkSlEkuLbV3+oWy666u5a1JIIKSFdAQUCkSUdaAOm9t4Se3qac3x8nkzsz6cmEBDif55mH5Nx7zz0zJJnvvOd9v69OCCFQKBQKhUKhuAfRl/cCFAqFQqFQKMoLJYQUCoVCoVDcsyghpFAoFAqF4p5FCSGFQqFQKBT3LEoIKRQKhUKhuGdRQkihUCgUCsU9ixJCCoVCoVAo7lmUEFIoFAqFQnHPooSQQqFQKBSKexYlhBQKhUKhUNyzKCGkUChKzIwZM9DpdDkPLy8vGjZsyKRJk7hy5UrOeRs2bHA6z2AwUL16dQYPHsyRI0eKdc9Tp07x/PPPU7duXby8vAgICOCBBx7g888/Jz093d1PUaFQ3OUYy3sBCoXizue9994jIiKCjIwMtmzZwuTJk/n11185ePAgPj4+Oef96U9/IjIyErPZzP79+5kyZQobNmzg4MGD1KhRo9D7LF++nCeffBJPT09GjhxJs2bNyMrKYsuWLbz++uscOnSIb775piyfqkKhuMtQQkihUJSaPn360K5dOwDGjRtHlSpV+PTTT1myZAlDhw7NOa9Lly4MHjw45/tGjRrxwgsvMGvWLN54440C73HmzBmefvpp6tSpw7p16wgJCck5FhUVxcmTJ1m+fLlbnk9qaiq+vr5umUuhUFRs1NaYQqFwO927dwekeCmILl26AHK7qzD++9//kpKSwrRp05xEkJ369evz0ksvAXD27Fl0Oh0zZszIdZ5Op+Odd97J+f6dd95Bp9Nx+PBhnnnmGSpVqkTnzp35+OOP0el0nDt3Ltccb775Jh4eHty6dStnbMeOHfTu3ZvAwEB8fHzo2rUrv//+e6HPS6FQlC9KCCkUCrdjFzZVqlQp8LyzZ88CUKlSpULnXLZsGXXr1uX+++8v9fry4sknnyQtLY0PPviA8ePH89RTT6HT6ViwYEGucxcsWMAjjzySs+5169bx4IMPkpSUxNtvv80HH3xAQkIC3bt3Z+fOnWWyXoVC4R7U1phCoSg1iYmJXL9+nYyMDH7//Xfee+89vL296devn9N5ycnJXL9+PSdH6OWXX0an0/HEE08UOH9SUhKXLl3iscceK7Pn0LJlS+bNm+c01rFjR+bPn8/rr7+eMxYbG8vp06dzokpCCCZMmMBDDz3EihUr0Ol0ADz//PM0bdqUv/3tb6xevbrM1q1QKEqHEkIKhaLU9OzZ0+n7OnXqMHfuXEJDQ53Gx4wZ4/R9tWrVmD17NpGRkQXOn5SUBIC/v78bVps3EyZMyDU2ZMgQXn75ZU6dOkW9evUAmD9/Pp6enjmibO/evZw4cYK//e1v3Lhxw+n6Hj16MHv2bGw2G3q9CsArFBURJYQUCkWpiYmJoWHDhhiNRoKDg2nUqFGeb/z/+Mc/6NKlCykpKSxevJgffvihSAIhICAAkBGlsiIiIiLX2JNPPsmrr77K/PnzeeuttxBCsHDhQvr06ZOzphMnTgAwatSofOdOTEws0vafQqG4/SghpFAoSk379u1zqsYKonnz5jnRo8cff5y0tDTGjx9P586dCQsLy/e6gIAAatasycGDB4u0Hvv2lCtWqzXfa7y9vXON1axZky5durBgwQLeeusttm/fzvnz5/nPf/6Tc47NZgPgo48+olWrVnnO7efnV6R1KxSK24+K1SoUinLj3//+NxkZGbz//vuFntuvXz9OnTrFtm3bCj3XHn1JSEhwGs+rAqwwhgwZwr59+zh27Bjz58/Hx8eH/v375xy3b5kFBATQs2fPPB8mk6nY91UoFLcHJYQUCkW5Ua9ePZ544glmzJhBfHx8gee+8cYb+Pr6Mm7cOCfXajunTp3i888/B6QoqVq1Kps2bXI656uvvir2Gp944gkMBgPff/89CxcupF+/fk4eQ23btqVevXp8/PHHpKSk5Lr+2rVrxb6nQqG4faitMYVCUa68/vrrLFiwgM8++4x///vf+Z5Xr1495s2bx5AhQ2jSpImTs/TWrVtZuHAho0ePzjl/3Lhx/Pvf/2bcuHG0a9eOTZs2cfz48WKvr3r16jz00EN8+umnJCcnM2TIEKfjer2eqVOn0qdPH5o2bcqzzz5LaGgoly5dYv369QQEBLBs2bJi31ehUNweVERIoVCUK+3ataNbt25MnjyZxMTEAs8dMGAA+/fvZ/DgwSxZsoSoqCj+8pe/cPbsWT755BO++OKLnHP/8Y9/MHbsWH788UfeeOMNrFYrK1asKNEahwwZQnJyMv7+/vTt2zfX8W7durFt2zbatWtHdHQ0L774IjNmzKBGjRq88sorJbqnQqG4PeiEEKK8F6FQKBQKhUJRHqiIkEKhUCgUinsWJYQUCoVCoVDcsyghpFAoFAqF4p5FCSGFQqFQKBT3LEoIKRQKhUKhuGdRQkihUCgUCsU9yz1nqGiz2bh8+TL+/v759iNSKBQKhUJRsRBCkJycTM2aNYvUrLmo3HNC6PLlywU2d1QoFAqFQlFxuXDhArVq1XLbfPecEPL39wfkCxkQEFDOq1EoFAqFQlEUkpKSCAsLy3kfdxf3nBCyb4cFBAQoIaRQKBQKxR2Gu9NaVLK0QqFQKBSKexYlhBQKhUKhUNyzKCGkUCgUCoXinkUJIYVCoVAoFPcsSggpFAqFQqG4Z1FCSKFQKBQKxT2LEkIKhUKhUCjuWZQQUigUCoVCcc+ihJBCoVAoFIp7FiWEFAqFQqFQ3LOUqxDatGkT/fv3p2bNmuh0On7++edCr9mwYQNt2rTB09OT+vXrM2PGjDJfp0KhUCgUiruTchVCqamptGzZkpiYmCKdf+bMGR599FEeeugh9u7dy8svv8y4ceNYtWpVGa9UoVAoFArF3Ui5Nl3t06cPffr0KfL5U6ZMISIigk8++QSAJk2asGXLFv73v//Rq1evslqmQqFQKBSKu5Q7Kkdo27Zt9OzZ02msV69ebNu2rdhz3TixATLjQQg3rU6hUCgUCkVZceVKSpnMW64RoeISHx9PcHCw01hwcDBJSUmkp6fj7e2d65rMzEwyMzNzvk9KSpJfLG7J1u2e3Ag4SkDobzQPXkFlkycYa4FnHfCqDd7h4FEbDNVBd0dpRoVCoVAo7hpsNsGAAd+Xydx3lBAqCR9++CHvvvturnG9gCY3gZs14OxwsvTD2VDNxs1qpwgMW0QLv0+pRpw8WRhBFwz6mmAIleLIszZ41QFTmBRPxhqgM9zeJ6dQKBQKxT2AXq/jnXe68fTT7p/7jhJCNWrU4MqVK05jV65cISAgIM9oEMCbb77Jq6++mvN9UlISYWFhrK00hcysvtT2akI9m5HgdEHLK3q40gAO/h9pxv9jVYiFWzUPE1j9V1p7z6GGLRZssWAGUl3vZJBiyVATTLXBI0yKJWMt+TDVAmMI6EzufVEUCoVCobgL2bMnjqtXU+nduz4Affo0KJP73FFCqFOnTvz6669OY2vWrKFTp075XuPp6Ymnp2eu8cHPvUlAQACx+xYyedl/MJ0bQJvKTxFaJZRaOiuVMwXtLxjhQgugBUmef2FZqCAh5BpBQXto47GUUNthMFwF/TXQWUFcBstlsOyC9LxWowN9sBRGHmFaNMlUSxNMxpqgz71ehUKhUCjuBWw2wccfb+Vvf1uHn58H+/e/QK1aAWV2v3IVQikpKZw8eTLn+zNnzrB3714qV65M7dq1efPNN7l06RKzZs0CYMKECURHR/PGG28wZswY1q1bx4IFC1i+fHmJ1xDZ8kkiWz7J9dRrTFv6Eu8e/4UGxx5kYPIY6tbuhmctD2rpLARkCjqf1sHp6kBvbnj3YWEdHUm1DFQKNtNWf5E6aSfBfAEsF4F40F/NFkpX5L86C9jiISsesnblvyh9sBRHplq
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import precision_recall_curve, auc\n",
    "\n",
    "idx = 0\n",
    "for cdata in lst_cdata:\n",
    "    y_true, y_score = cdata\n",
    "\n",
    "    # PR 커브 계산 및 그리기\n",
    "    precision, recall, _ = precision_recall_curve(y_true, y_score)\n",
    "\n",
    "    # AUC 계산\n",
    "    pr_auc = auc(recall, precision)\n",
    "    plt.plot(recall, precision, '-', color=colors[idx], label=f'{os.path.basename(base_dirs[idx])}(AUC = {pr_auc:.3f})')\n",
    "\n",
    "    # (1, 1)에 가장 가까운 점 찾기\n",
    "    min_distance = float('inf')\n",
    "    closest_point = None\n",
    "    for i in range(len(recall)):\n",
    "        distance = ((1 - recall[i])**2 + (1 - precision[i])**2)**0.5\n",
    "        if distance < min_distance:\n",
    "            min_distance = distance\n",
    "            closest_point = i\n",
    "    plt.scatter(recall[closest_point], precision[closest_point], color=colors[idx], marker='o')\n",
    "\n",
    "    print(f'{os.path.basename(base_dirs[idx])} PR AUC: {pr_auc:.3f}, minDist: {min_distance:.3f}, {(recall[closest_point], precision[closest_point])}')\n",
    "    idx += 1\n",
    "\n",
    "plt.plot([0.0, 1.05], [0.0, 1.05], '--', color='navy', label='baseline')\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.title('PR Curve')\n",
    "plt.legend()\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unet ROC AUC: 0.923, minDist: 0.121, (0.03855843283619862, 0.885424512257604)\n",
      "unet-mini ROC AUC: 0.949, minDist: 0.081, (0.02485883834974988, 0.9224052152557138)\n",
      "unet-sgd ROC AUC: 0.955, minDist: 0.063, (0.0434974242368962, 0.9538732855195041)\n",
      "unet-rmsprop ROC AUC: 0.874, minDist: 0.247, (0.0055918602005307965, 0.7531937969524044)\n",
      "unet-dice-loss ROC AUC: 0.939, minDist: 0.087, (0.05330329573045489, 0.931617865125619)\n",
      "unet-focal-loss ROC AUC: 0.872, minDist: 0.251, (0.004573449270014527, 0.7487617009547043)\n",
      "unet-l1 ROC AUC: 0.926, minDist: 0.126, (0.024355441418179306, 0.8766366499467493)\n",
      "unet-l2 ROC AUC: 0.884, minDist: 0.219, (0.013563852969138058, 0.7813983872652583)\n",
      "unet-l1+l2 ROC AUC: 0.940, minDist: 0.101, (0.020355838523738234, 0.9011364203573499)\n"
     ]
    },
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve, roc_auc_score\n",
    "\n",
    "idx = 0\n",
    "for cdata in lst_cdata:\n",
    "    y_true, y_score = cdata\n",
    "\n",
    "    # AUC 계산\n",
    "    roc_auc = roc_auc_score(y_true, y_score)\n",
    "\n",
    "    # ROC 커브 계산 및 그리기\n",
    "    fpr, tpr, _ = roc_curve(y_true, y_score)\n",
    "    plt.plot(fpr, tpr, color=colors[idx], label=f'{os.path.basename(base_dirs[idx])}(AUC = {roc_auc:.3f})')\n",
    "\n",
    "    # (0, 1)에 가장 가까운 점 찾기\n",
    "    min_distance = float('inf')\n",
    "    closest_point = None\n",
    "    for i in range(len(fpr)):\n",
    "        distance = ((0 - fpr[i])**2 + (1 - tpr[i])**2)**0.5\n",
    "        if distance < min_distance:\n",
    "            min_distance = distance\n",
    "            closest_point = i\n",
    "    plt.scatter(fpr[closest_point], tpr[closest_point], color=colors[idx], marker='o')\n",
    "\n",
    "    print(f'{os.path.basename(base_dirs[idx])} ROC AUC: {roc_auc:.3f}, minDist: {min_distance:.3f}, {(fpr[closest_point], tpr[closest_point])}')\n",
    "    idx += 1\n",
    "\n",
    "plt.plot([0.0, 1.05], [0.0, 1.05], '--', color='navy', label='baseline')\n",
    "plt.xlabel('FPR')\n",
    "plt.ylabel('TPR')\n",
    "plt.title('ROC Curve')\n",
    "plt.legend()\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 1000x1000 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "id = 7\n",
    "\n",
    "fig, axs = plt.subplots(3, 3, figsize=(10, 10))\n",
    "idx = 0\n",
    "for base_dir in base_dirs:\n",
    "    result_dir = os.path.join(base_dir, 'result')\n",
    "\n",
    "    ##\n",
    "    lst_data = os.listdir(os.path.join(result_dir, 'numpy'))\n",
    "\n",
    "    lst_gt = [f for f in lst_data if f.startswith('gt')]\n",
    "    lst_pr = [f for f in lst_data if f.startswith('pr')]\n",
    "\n",
    "    lst_gt.sort()\n",
    "    lst_pr.sort()\n",
    "\n",
    "    ##\n",
    "    # img = np.load(os.path.join(result_dir,\"numpy\", lst_img[id]))\n",
    "    # gt = np.load(os.path.join(result_dir,\"numpy\", lst_gt[id]))\n",
    "    pr = np.load(os.path.join(result_dir,\"numpy\", lst_pr[id]))\n",
    "\n",
    "    axs[idx//3,idx%3].imshow(pr, cmap='gray')\n",
    "    axs[idx//3,idx%3].axis('off')\n",
    "    axs[idx//3,idx%3].set_title(f'{os.path.basename(base_dir)}')\n",
    "\n",
    "    idx += 1\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.10.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}