compute_map.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "f8225fa7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from tqdm.notebook import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0670ddb0",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a =Image(\"D:\\practice\\博客\\目标检测\\img\\lip_image005.png\")\n",
    "a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e3d75223-2ab1-4eaa-869b-8c3bc15b4855",
   "metadata": {},
   "outputs": [],
   "source": [
    "# from PIL import Image\n",
    "# b = Image.open(r\"D:\\practice\\博客\\目标检测\\img\\lip_image005.png\")\n",
    "# b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5d637646-12c8-436b-8108-345e243d2354",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "query box : [ 0.  0. 10. 10.]\n",
      "box : [[ 5.  5. 15. 15.]\n",
      " [ 0.  0.  9.  9.]\n",
      " [ 1.  1.  8.  8.]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "10.0"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "query_box = [0,0,10,10]\n",
    "boxes = [(5,5,15,15),(0,0,9,9),(1,1,8,8)]\n",
    "query_box = np.array(query_box,dtype=np.float)\n",
    "boxes = np.array(boxes,dtype=np.float)\n",
    "print(\"query box : {}\".format(query_box))\n",
    "print(\"box : {}\".format(boxes))\n",
    "query_box[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "a02f650b-6176-4551-a876-ab1cd97efb92",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9.0"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boxes[1,2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "307e18d6-cdd2-4f86-a041-bfefebf2758e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9.0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boxes[1][2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "71e99655",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# compute all the IOUs between the N annostation boxes and the query_boxes\n",
    "def compute_iou(boxes, query_boxes):\n",
    "    \"\"\"\n",
    "    Args:\n",
    "        boxes:       (N,4) ndarray of float\n",
    "        query_boxes: (4)   ndarray of float  xmin,ymin,xmax,ymax\n",
    "     Returns\n",
    "        overlaps:    (N)   ndarray of overlap between boxes and query_boxes\n",
    "    \"\"\"\n",
    "    N = boxes.shape[0]\n",
    "    ious = np.zeros((N),dtype=np.float64)\n",
    "    query_box_area=(\n",
    "        (query_boxes[2]-query_boxes[0])*(query_boxes[3]-query_boxes[1])\n",
    "    )\n",
    "    for i in range(N):\n",
    "        iw=(\n",
    "#             min(boxes[i][2],query_boxes[2])-max(boxes[i][0],query_boxes[0])\n",
    "             min(boxes[i,2],query_boxes[2])-max(boxes[i,0],query_boxes[0])\n",
    "        )\n",
    "        if iw>0:  \n",
    "            ih=(\n",
    "#                 min(boxes[i][3],query_boxes[3])-max(boxes[i][1],query_boxes[1])\n",
    "                 min(boxes[i,3],query_boxes[3])-max(boxes[i,1],query_boxes[1])\n",
    "            )\n",
    "            if ih>0:\n",
    "                oa=iw*ih\n",
    "                box_area = np.float64(((boxes[i,2]-boxes[i,0])*(boxes[i,3]-boxes[i,1])))\n",
    "                ua = query_box_area+box_area-oa\n",
    "                ious[i]=oa/ua\n",
    "    return ious"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "ba6d9d71-2c34-4c4c-b57e-769ea106f590",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.14285714, 0.81      , 0.49      ])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ious = compute_iou(boxes,query_box)\n",
    "ious"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2d739bcb-5c7b-44ac-b11d-8bad300a3953",
   "metadata": {},
   "outputs": [],
   "source": [
    "# check if true positive or false positive\n",
    "def check_if_true_or_false_positivate(annotations,detections,iou_threshold):\n",
    "    \"\"\"\n",
    "    All detected bounding boxes will be assigned to the GT box that has the largest IOU. \n",
    "    Then check whether they are TP or FP by iou_threshold.\n",
    "    Here, if multiple detected bounding boxes are mapped to the same GT box, then only the highest scored bounding box needs to be assigned to that GT. \n",
    "    So note that the detections passed to this function need to be pre-sorted.\n",
    "    \n",
    "    Args:\n",
    "        annotations:      (N,4) ndarray of float\n",
    "        detections:       (M,5) ndarray of float  xmin,ymin,xmax,ymax,score\n",
    "        iou_threshuold    default:0.5  float\n",
    "        \n",
    "    Returns\n",
    "        scores：          (M)   list of detection boxs's score\n",
    "        false_positivate  (M)   list if detection box is FP,value=1.[1,0,0]\n",
    "        true_positivate:  (M)   list if detection box id TP,value=1.[0,1,1]\n",
    "    \"\"\"\n",
    "    annotations = np.array(annotations,dtype=np.float64)\n",
    "    scores,false_positivate,true_positivate = [],[],[]\n",
    "    detected_annotations = [] # a GT box should be mapped only one predicted box at most\n",
    "    for query_box in detections:\n",
    "        scores.append(query_box[4])\n",
    "        if len(annotations)==0:\n",
    "            false_positivate.append(1)\n",
    "            true_positivate.append(0)\n",
    "            continue\n",
    "        ious = compute_iou(boxes=annotations,query_boxes=query_box[:4])\n",
    "        assigned_annotation = np.argmax(ious)   # assigned anno index \n",
    "        max_iou = ious[assigned_annotation]\n",
    "        if max_iou>=iou_threshold and assigned_annotation not in detected_annotations:\n",
    "            true_positivate.append(1)\n",
    "            false_positivate.append(0)\n",
    "            detected_annotations.append(assigned_annotation)\n",
    "        else:\n",
    "            false_positivate.append(1)\n",
    "            true_positivate.append(0)\n",
    "    return scores,false_positivate,true_positivate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d0529e5d-6db4-4671-8600-d90273649a4d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "a= [1. 2. 3. 4. 5.]\n",
      "b= 4\n",
      "c= 5.0\n"
     ]
    }
   ],
   "source": [
    "a = np.array([1,2,3,4,5],dtype=np.float64)\n",
    "b = np.argmax(a)\n",
    "c = np.max(a)\n",
    "print(\"a=\",a)\n",
    "print(\"b=\",b)\n",
    "print(\"c=\",c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "1bfd5342-17b1-4ec8-bb8d-96b0bb5f935d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1. 1. 6.]\n",
      "[2. 2. 4.]\n",
      "[4. 9. 5.]\n",
      "anno: [[ 1.  1.  6.  6.]\n",
      " [ 2.  2.  4.  5.]\n",
      " [ 4.  9.  5. 19.]]\n",
      "len of anno: 3\n",
      "detec: [[1.  1.  5.  6.  0.5]\n",
      " [2.  2.  4.  5.  0.8]]\n",
      "size of detec: 10\n"
     ]
    }
   ],
   "source": [
    "anno = np.array([(1,1,6,6),(2,2,4,5),(4,9,5,19)],dtype=np.float64)\n",
    "detect = np.array([[1,1,5,6,0.5],[2,2,4,5,0.8]],dtype=np.float64)\n",
    "for i in anno:\n",
    "    print(i[:-1])\n",
    "print(\"anno:\",anno)\n",
    "print(\"len of anno:\",len(anno))\n",
    "print(\"detec:\",detect)\n",
    "print(\"size of detec:\",detect.size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e71a1dd2-bc23-41a4-ac6b-c4db0b818068",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.5, 0.8]\n",
      "[0, 0]\n",
      "[1, 1]\n"
     ]
    }
   ],
   "source": [
    "scores,false_positivate,true_positivate=check_if_true_or_false_positivate(anno,detect,0.1)\n",
    "print(scores)\n",
    "print(false_positivate)\n",
    "print(true_positivate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "4f10a183-353a-4efb-8bdc-5d2c44b0a3bb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAnEAAAIXCAYAAADzIAh/AAAgAElEQVR4nOzdCVzUZf4H8A/3IXILcnjhhQeimKWWR5q1LqnlAbZbtuW21l/L2i673U6zWqvt0I6t3a1doTRT0zLTTNeyvPC+UFMOARUB5Yb5z/eBYTkGGBD4zW/m897XLD9+85uZ7xw2H57n+T2Pg8EIRERERKQrjloXQERERERNxxBHREREpEMMcUREREQ6xBBHREREpEMMcUREREQ6xBBHREREpEMMcUREREQ6xBBHREREpEMMcUREREQ6xBBHREREpEMMcUREREQ6xBBXadasWdixY0e91z/66KPo3r07HBwc1LHyu/wcN26cujR0WyIiIqKW5mAwavVHyfwKjyzrhIX3DDBzZQa+evgubBr1PhbeGNzqpZgjAeyKK65AYmIipk2bVu9xEtZ8fX3x2Wef1dgvv8fFxeHYsWMq6LWlb7/9FvPmzcN7772HwYMHt9pj7Ny5Uz3348eP47rrrlOvRWOWLFmifl64cAHbt29Xr7GE36Ye01qa+7yE1C01m0ydOrXqvZf6/f39ERERoe67Otnfku+Trb43Js15fsnJyerfoxwr74HUv379eixevLjqPbLkGCIiq2dopqR3JhgeXnXG0qMNiyc8bFidUc/VuxYbJjy02mDpvbW0BQsWSJBVPxsixxj/I19n//nz5+u9rrUYA6Oq1/jlqh5bfrYGud+YmJga++R3efyGPPLII3Vej9qvsSXHWEJe/6Zq7vMSf/rTnwzGwF/jd2OIq/pdtuV5mLu05GdED+/N5Wju85PrjcFM1WsMf+r9qH0bS44hIrJ2zetOzfwKn6yNxKgrLW05G4Cb/mT8y37ZHvNXD7wJs7AJ2zPrv4c9707ExIlNuLxbz2PVIq0Nxi9h9Ze+MQzUe5ypu1RaI2qTFgIhrSxtRVoLpFWktVrfTKSVTy7VSQvGwoULG73t559/XuN3eY2lRaepxzRGWlClxaYpmvu85HHk/a7eYiufn+r3JS07xn9bdS7GAKS64FuKHt6by3E5z09ea3nNs7OzVUu5udY1S44hIrJmzpYc9NXDE7HkcN39h/5o3N/gLSMx64OFiA0Cgm+8FbETn8SSYe8j/NO7zN4fat9f71l4/5VYmKJi5J9qdblKN+0fU3Drylmo3lGbsfoR3HW68ecl//GW4OXn56e+eE1hzBzTl5e50CRdmfIFJ10zeiGh1JIAKF1M8mVXXUxMjNpn6m4z5+WXX67xu7zW0mVV/TWy5JjW0tzndffdd+ORRx6psa/262iufgnctZ9vffT+3shtTXVJ6JJt+TdWmwQnuT9z14nmPj8iInthUYiLfWUlYk2/qOC0BN2fXYlZA5vyUNIaF4nXUoBZ1e9PQxK+TGN8Ggtx8oVi7gtMWmakJUiur+/LyBrJc5Ev5oaYWh/NjeuSL3S5vaXPWV5r+QJuqCXKkmNawuU8L/mMyHGmECG/S+ts9Za52mO2TK29ltLzeyMBTv7gMf3RI7eTf1vys/Z404bGoF7u85P3pfp7ZG4snSXHEBFZM4tCXHUZP2/Cod6z8GD1AFdPi1htwTcuRKMdIeq+NmFUZQtedYfeuwsT36t7kycnflV35/iGH0a+JKqHMvlykAHU9ZGQJl/W1Qd7C+mClS8AS74w5QuwoS7b6uSLz9KWG0tJ7fIcpJXH1C0sA7zluZn7cm6sVrm+sectj2l6Xes7g9eSY1pSc5+XvFZCXq/qLUGmL35zgUTChnw+Guuqs5X3RsJb9RN/5DMs/86kRU6ei4RZ+WxLa1pDww8u5/mZ7rf66yavv4RGUwunJccQEVm9pg2hkxMUJhgW76q1O2O14eEJi43XmqGum2CYIJeqkxfOGFY/ZP7EiDOrHjZ7koPZEynqeVx1H++YraZK7UHbppMbzJEBz3LdunXrGrxPrcDCExvkGOMXqnqupoHrMlC8vudlOmmi9oBvOd7c/obIY8ljN3QSgiXH1He7prw3zX1eptvV/uzI7zJI3hw56cGe3puGTtyQ5yWvk5xIIK9LQ1ry+Ql5PKn/co8hIrImTTqxYc+7T+Kr3rNwU1O6UYNisXDlSqx8tnoHajBiX3ke3aVl7eGvkGHanfkVXnvvEGJ//79xcK1BWhekdUS6Uk0XaSWoj+k6vXe1SAuDdP9KF5UMDpfWRGldqe95mbqyqk+lUf33ppzIYWrxaGhMlaXHSItX9Yt01ckA+Nr765sOo7nPy3Q7aUmqTo6Xz5O5LlBLp36xhfem+nHmSKuctC7K69TYmLaWfH5C3rPGTsqw5BgiImtieXfq7iV4cq2cXHAFgivHxR2qdUjtbs3YBsfNDcCsle8j/OG7cNfEFDy/8iacfsV4n+Ofx8J6btMS3ammkxlqf9nIF6iENfmSqd31JftlPJDeSSiQsCNfxKbB+RISpGvLXNAwvQ7murbkS7ah7joJUfJY1e9XviTltTSNZ7LkmNrMffnLPrmdpSG7uc/LdLvawaI+0q1Ye0xXfWzhvTGRf1vSLSnk+chzMNed3NCJDZfz/OS28pi1u7erv2+WHENEZO0sDHF7sOTpakFJta5Va1mzcExcXcHqpIlO706sDGKxeP6V+u+hJc5OrX4yQ3Wm1hVz/xGXMUEymevl0npMnLQymEKBBAQJrvK8ZCqJ+lqL5Eu29gkfpkHg9ZEgLPcp46Cq36/pucsXsCXHtKbmPK+GbifvV+2aJfDUbrWrj628N/IZN51Banpe0rIotzVNrivktTF3ssPlPj8hIa/2H121b2fJMUREVs+SPlcZjzbhocWGxfWMY2twTJxJvRP6VoyPm1A5bs7yCYQtfNxqZILW+sbsmCbsrT6JqzCNh6u935qgGeP1ZFySJeR+a48Tqj3hqmzLZKnVx0vJ+Kfqv8u2PGbtCWUbO8YSTR0TJ5r7vGSsVu3xb/WNXZP7b84YKz2/N/VdL7eXMWemMXG1a2rJ5yefB3O1V7+dJccQEVk7i1riPjku87Vdge0Pf4XklkyQVd2ysXheWtN2L8HEp+/CxNPPY2X1JbrUfjNdppXMdqfif9250rIgXUPSuiB/fUuLQPXWBGn1kFYBIS0I0mpgWh/VdGaeqXuooZaDtiRdWlKrqaVCnp+0btRuPamP6fk0RlqFpHVSWgVNY4Zqd4+ZliyqfsagHC8X2Sfd13K72i0vlhzTWpr7vOS1lddOPh+maWnkdua6cs0tu2UJPb839Y1DlMdu6txuzX1+8t6YlkWT/aZlw6rfzpJjiIisXRPWTm1gjVNLulMliH0aXjl57x4smfgkVPQaX09gqzXRr1nN7sYlW9LUMXFERES2oMnzxDWLhC3VkhaJ7ZlXAK9UnOVqNqQNnIWVH4SrueK2Z8bWmSuOqDZpfSQiIrI3zWiJm4Xu7y1B/Z2b1UVi1sujsOnRirNOVw77UbWyNXzWai31nAlrqSY9FhEREZFOtEx3aoP2YMm7wCxTl2mTQtn/1l4lIiIiov9pQogjIiIiImvRpBUbiIiIiMg6MMQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHJHObdq0CQ4ODuqnHvzhD39Q9da+dO/evcZxv/76q9nj5JKQkNDo4/zjH/+o9/YZGRl1jj937hzmzp2L0NBQODo6wsfHBzfccAOys7Orjtm/fz+GDRsGFxcXREdHY+vWrXXu59VXX0W3bt1QWFjYjFfn8iUmJqJXr15wcnJSzzUpKUmTOixx8803qxrnzJlj9nrTZ7v6Rd6XqVOn4tChQ21cLZH1cda6ACKyPxKSagcgDw8Ps8fee++9+P3vf19jX8+ePS1+rI8//hiRkZE19vn7+9f4PT09HVdeeaUKZ88//7y6/7PGULfhu+9QXFysjiktLcWNN96IAQMGYO3atfjPf/6D66+/HikpKfD19VXHSPCcN28evv32W7i7u1tcY0s5e/Yspk+fjtjYWHzwwQdwc3Nr0mvVlrKysvDll1+qbalVwm99r9lLL72Ea6+9Vr0X27dvV6/xV199heTkZBW6iewVQxwRtTlpUbnqqqssOrZLly4WH2tO//79MXjw4AaPufvuu1XL2d69e6sCmbj5ppuqto8dO4aTJ0/ixx9/RMeOHTFq1CjV2rdt2zbVYif+9Kc/4dZbb1WBQwtHjhyBwWDAjBm3Y+TIkZrUYKl//vOfqtbf/va3WLNmDb744gvccsstZo/t2atX1WdgxIgR6j268847VUB//PHH27JsIqvC7lQiK3f48GHVuiLdSNKCJQFixowZKCoqqvc2O3bsQHx8vGqlkG41+Sn3cerUqRrH5efn4+GHH0ZYWJi6b2kNGzRoEJYuXVp1zIkTJ9R9BQQEVHUzSkix5m66ppDWs5UrV6oWv+oBrjZT96inp6f6Ka12cjHtl9dMWhf/+te/tn7RZkg39dVXX6224+KmqaBsCnKWfh5EWlqaCqNBQUHq/Zb3fcqUKcjMzKw6Jjc3t8bnJjAwEPc/8AAuXbpkcb2LFy9Wr/e//vUvdR/vv/++xbcdOnSo+imhmsieMcQRWbE9e/aolqTNmzfjlVdewXfffadCggQHUzefOSeNwaRPnz5499138f333+P1119XX85yXzL2y+TPf/4zFi1apL6QN27ciGXLlqnWkOrHXHfddaq16c0331RjlD7+xz9wxRVX4MKFC1XHmMafyU9LlJWVwc/PT92mQ4cOakxU9bFn1T3zzDPqS17Ch4xHW7VqlUWPUb1+eRwJqDIGS8a1VSevrTAFG2dnZ/V4EoB++umnquN69+6tAtzChQvVc3/nnXdUkDa9Fvfcc496vWt31TZGWqOkq9aSS0OefvppVZOQ7kepfcmSJep3Sz8Psi8qKkp1FUsL1/r1641ha4kKW6b3R4K/hEUJYfK5kWPkPXrX+NjS3SzPpzHSmiktmxIW5fX63e9+pz5/loYyua2QoElk1wxEZLVGjx5tcHNzM2RlZdV7jPFLWb411c/6GAOA4eLFiwZjEDIYw1jV/sjISMOkSZPqvZ3xC17d9xtvvNFgnf/85z/VcfKzMcbQqC7ffvutujz55JOqru7duxvy8vKqjktPTzfcddddhs8+/9ywZcsWw7///W/DkCFD1ON88MEHjT7O119/re579erVhh9++MHw9ttvG4yBUT1WUlJS1XELFixQ9+nq6mqYOHGi4ZtvvjEs/+ILQ9++fQ3GMGcwBumqY1esWKGOk+ONwdDw3nvvqf0zZ840jBkzptGazPn444/V/VlyaYzps/D5smUNHlff5+HOO+9Uz+vgwYP13tb0em3fvr3G/mXLl6v9a9eubbROeRw59tChQzXqNgZRs88nMfEzQ0lJicEYIA3G0G3o0qWL2l/9vSGyRwxxRFbq0qVL6ovqT3/6U4PHmQtxEoYeffRRQ+fOnesEgbvvvrvqONOX9rx589Tt5UuyuvLycnUfgYGBhr/+9a+GXbt2GcrKylr2iRr+FwAk3DWkuLjYEBUVZWjXrp36Um+qkydPGpydnQ0TJkyo2vfiiy+qx+7Zs6cKNyYSIiXk/P73v69xH4WFhSp8SAgSEhDluOTkZPX6zZ492+Dn52cIDg42zJ8/X72GDZGgLIHIkktj6gtxln4e/P39Ddddd12DjzF06FAVcOX1r36Rx5D7k8dpiBwn78FVV11Vtc/0OZPXrPrny/R8al/CwsJUoCaydzyxgchKmborO3Xq1OTbSpeonEH5wgsvYMiQIfD29lZdijKWraCgoOo46SKV+5dxSQsWLFDdiNIl9tprr6FHjx7qNtLd+Je//AXPPvus6n6VLsWZM2eq+27fvn2LPNebJk1S3aUypuz++++v9zgZgyYnDhiDgupSq33WaWPkJIkxY8bUmI5FxnyJ8ePHqxpMZOyhnBAhXcnVyRmf0rUqpEv7jjvuUGe0RkREqC5NY6jDwYMHYQwrqvtXHlPGq9VHupXl/WlNln4ezp8/D2OYavC+5ExeGUco74U5ctZpQ2QKFOkali7U6l3y8r4aA7XqnpWzfquTLmx53+T9kfF34eHhjT1lIrvAEEdkpeTLXZw+fbpJt8vJycHq1atV8JKwYyLjtyRYVNeuXTvMnz9fXWTgunzRS4iSsy1l+gYhX+offvihuhw9elR9CT/55JMqwMi4qJYkIbIxhsoxV5YcW9/tq4c1mTLE0mNrk7FnEmYefPBB9buM15NxXsHBweoiJ6DIe9FQiJOzNBu6vnY9TdWUz4OMTzN3skN1Mg5NgrwEf3MkZDXENE5P5uSTS21ygkPtEBfRvXujZxgT2SOGOCIrJQPxR48erb7kpYXC1GLUGGlhEdJiVJ2EsIbIl/Ptt9+O3bt3q4HvMoDddCamicw59sQTT6gzMX/55ZcmPJuGLVu2XJ3sMHz48AaPKykpwSeffAIvL686kwNbQgbOywB6mUfNROaHk5MrZN4xOWnEFNqkxUnO6pSWNnNkOg9pnZTWQ1OrVHl5eY0zNCUkNRa8JkyYoOY+ay1N+TzcdNNN+Oijj9RzkwmDzZFjpMVRPo9du3ZtUi0yQe/PP/+MyZMn47777qtzvQRNOblGWgSbeoIIkT1iiCOyYm+88YZqgRg4cKA6A1C6OGW1gRUrVuC9994z250pXWXXXHONChjSKiJftNJ9+Pbbb9f5IpczK+VLWVqjpOVPugGlJUSCjQQ4mTdN5lCT7jgJcK6urtiwYQP27dunwpyJtMpIq5MEzttuu63e5yOtPHFxcarrzNRdK7W9/PLLqotSumlNpHVLWvvkuUjXprRIypm08tgyP1j1FjK53d///nfVzWfqDpTuN+kulOcmr4k8F3lN5DGlW9FEWvTeXbwYU6dMwcSJEzF79mxjEMs3BpWn1Jmq5uYhk2AmjymvTfU57KQrWrr+ZF6zi8YAJzWZWp7qI2GlNQNLUz4Pzz33nPpsSZerbMuZqhdycrB2zRr1fsh7JK1nEuLlM/LYY4+p11fCq7y3X3/9tTpjVa4zxxQcZbJeeYzapLteQrYEdXMhj4hq0XJAHhE1Ts4UnDp1qhrMLychGAON4Q9/+IMaYC/MndiQmppqmDx5ssHd3d3g4uJiuP766w379+83hISEGG6//faq44xfwoZBgwap4+S+w8PDDcYvUjXYXmRmZqrHkjNHZfC+DEiXQe2vv/56jZMATGdYys+GZGdnG4yh0RAaGqrO/JTHNIYKNRj+woULNY41BiCDMcAajGFS3bfUOG7cOMO6devq3K88JzlGTlwwuf/++9XJCvL85bqAgACDMTwajhw5Yra2L7/80hATE6PqkjNQ5eSHAwcOmD3WGEbU/eXk5NTYLyc7/PGPf1Tvla+vr3p9W+NEkPrUd2KDpZ8HkZKSok54kZMz5P2R5zlt2jT1WTCR5/nUU08ZIiIi1DHyevXp00d9dox/ZJitTU5KMQZKQ//+/eutXz5Tcgax3FdDz4eIKjjI/2mYIYmIiIioGTjZLxEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOMcQRERER6RBDHBEREZEOccWGViYzmcvyPbJMkGn5GyIiIrJuMo3uxYsXERIS0uy1mlsbQ1wrkwAXHh6udRlERETUDCkpKQgLC9O6DLMY4lqZtMAJ+RCYW+eSiIiIrE9eXp5qhDF9j1sjhrhWZupClQAnC1ETERGRfljzUCjr7OQlIiIiogYxxBERERHpEEMcERERkQ4xxBERERHpEEMcERERkQ4xxBERERHpEEMcERERkQ4xxBERERHpEEMcERERkQ5xxQYiO1JWbsAvJ7ORmVeIoPbuGNLVD06O1jsbORER1c+uWuI2b96MG2+8EQEBAWoZjS+//LLR2/zwww+Ijo6Go6OjWkNtyZIlbVApUcv7Zv8ZDFu4EZPf/wl3L92tfsrvsp+IiPTHrkLcpUuXMGjQIHzw4YcWHX/y5EmMGTMG1xovBw4cwF/+8hfcc889WP7FF61cKVHLkqB25yc7kZJTWGN/qvF32c8gR0SkP3bVnfqb3/xGXSy1ePFidO3aFa8vWqR+j4yMxC+//IIFL72EyTff3FplErUo6UJ9atUBGMxcJ/ukM1Wuv65PMLtWiYh0xK5CXFNJ92tsbGyNfRICpUu1pKQELi4udW5TVFSkLiZ5eXkVGwUFxlebLze1vV9+vVCnBa46CXJyvYyVGxrh33aFERHRZWGqaEBaWho6duxYY19wcLD6ee7cuTrXiQULFmD+/Pl19mfv3IvSdu1apU6ihhxPK7boODnZgYiI9IMhrhFyAkR1BoPB7H6TefPm4YEHHqj6XVri5ISITL+OuBgY2HqFEtXDwSEPSDra6HFytioREekHQ1wDQkNDkZ6eXmNfZmam+unvb77byc3NTV1qK3d3g8HDo+WLJGpE367u6OD1K7Iumm+Rkz9HwnwqphshIiL9sKuzU5tqxIgRWLNmTY1933zzDYYMGWJ2PByRNZKTFWaPiDB7nak9+bkJfXlSAxGRzthViLt48SKSkpLURZw4cUJtnz59Wv3++OOPY8aMGVXH33333eqYBx98EIcOHcJHH32kTmqY99hjmtRP1FxywkKHdq519ksL3N9vjcEN/eqO7yQiIutmV92pO3bswOjRo6t+N41du/322/Hxxx+rExlkbjgTmV5kw4YNmDNnDhYtWoSQkBC8++67nF6EdOfnE9nIulQMTxdHhHi5Ijm7EHOHd8JDsf3ZAkdEpFN2FeJGjRpVdWKCORLkahs5ciT27NnTilURtS75zCfsSFHbk6NDkXwmV4W47oHtGOCIiHTMrrpTiezRvtRc7M/Ig6uTAyYMCNW6HCIiaiEMcUQ2LnFnRSvc+D7B8PPkCTlERLaCIY7IhiVnXcK2UxcgvaZTBoVrXQ4REbUghjgiG/Z5ZSvc2J4d0NGn7vyFRESkXwxxRDYq7UIhNh47q7anxrAVjojI1jDEEdmoZbtSUW4AhnXxQ7dAT63LISKiFsYQR2SDzl8qwdqDGWo7bjBb4YiIbBFDHJENWpmUitJyA6JCvNEv1FvrcoiIqBUwxBHZmItFZVix94zajudYOCIim8UQR2Rj1uxNR35JGbr7e+KKrr5al0NERK2EIY7IhhSVlmNZUpralrFwjg5cVouIyFYxxBHZkPUHM5FdUIKQ9m4Y2TNQ63KIiKgVMcQR2YiyckPVEltxg8K4uD0RkY1jiCOyEZuPnkV6XhH8PFwwtk+w1uUQEVErY4gjsgEGgwEJOypa4SZHh8Ldhf+0iYhsHf9LT2QDtv96Acnn8+Hp4oTYqBCtyyEiojbAEEdkA0ytcBP7d4SXm5PG1RARUVtgiCPSuf1pudiTngtnRwdMig7VuhwiImojDHFEOpdY2Qo3PjIIAV6uGldDRERthSGOSMdOnM3Hj79mQ2YTmcIltoiI7ApDHJGOLaucF25090CE+rprXA0REbUlhjginTqTW4T1R7PU9lS2whER2R2GOCKdWr4rFeUG4MpOvugR1E7rcoiIqI0xxBHpUHZ+CdYeyFDb8YPZCkdEZI8Y4oh0aNWeNBSVlaNfcHv0D/PWuhwiItIAQxyRzuQXl2HFnnS1HTc4HA4OXOieiMgeMcQR6czafWdw0Rjkuvh54KpuflqXQ0REGmGII9KR4lIDPt+dqrbjY8LhyFY4IiK7xRBHpCMbDmXiXH4JgrzcMKpXB63LISIiDTHEEelEWbkBibsqJveNGxQKFye2whER2TOGOCKd2Jp8Dik5hfBxc8b1fTtqXQ4REWmMIY5IBwwGAxIqF7qfPDAU7i78p0tEZO/4TUCkA7tO5eDI2UvwMIa32P4hWpdDRERWgCGOSAcSdpxWP2/s2xHeHs4aV0NERNaAIY7Iyh1Kz8OutFw4Ozrg5kFhWpdDRERWgiGOyMol7qwYC3d97w4I9HLVuBoiIrIWDHFEVuzU+QJsOXEeMpnIlBgudE9ERP/DEEdkxT6vPCN1REQAOvl5aFwNERFZE4Y4IiuVmVuEb49mqW1Z6J6IiKg6hjgiK/XF7lS1SsPgcB/0CvbSuhwiIrIyDHFEViinoBSrD2So7fjBnTSuhoiIrBFDHJEVWrU3DUWl5Yjs4IXocG+tyyEiIivEEEdkZfKLy7AiKV1txw8Oh4MDF7onIqK6GOKIrMy6/RnILSpFZ18PDI3w17ocIiKyUgxxRFakpMyAxN2pajsuJgxOjmyFIyIi8xjiiKzI94czcfZSMQLbueLa3kFal0NERFaMIY7ISsh0Igk7K1rhpg0MhYsTW+GIiKh+DHFEVuKn4+dx6kIBvN2ccUO/jlqXQ0REVo4hjsgKGAzSClexxNZN0SHwdHXSuCIiIrJ2DHFEVmBPSi4OZV6Em7MjJkSFal0OERHpAEMckRVI2HFa/YztGwwfD2eNqyEiIj1giCPS2JGMi9iekqOmE5k8MEzrcoiISCcY4og09lnlWA9/LawAACAASURBVLhxPTsgyNtN42qIiEgvGOKINJSSXYAfks+p7SkxbIUjIiLLMcQRaWjZzlQYjD+v6eaPLgGeWpdDREQ6whBHpJGzF4vxzeFMtR0XE65xNUREpDcMcUQaWbE7FaXlBgwK9UZkSHutyyEiIp1hiCPSQF5hKVbtP6O24wZ30rgaIiLSI4Y4Ig2s3puOgpJy9Axsh5jOPlqXQ0REOmR3Ie7dd99FWFgYHB0dER0djS1btjR4/BtvvIGIiAg4OTkhKCgIDzzwAAoLC9uoWrJFhcbwtjwpTW3Hx4TDwYEL3RMRUdPZ1dTwiYmJmD17Nt5//31cffXVWLxkCUaPHo0TJ06gU6e6XVr//ve/VWj79NNPMXz4cBw5cgTTp09X1y1atKityycbse5ABnIKSxHm7Y6rewRoXQ4REemUXbXELVy4ELNmzcLMmTMRGRmJ141BrGvXrqp1zpytW7di1KhRuOWWW9ClSxeMGzcOd9xxB7Zt29bGlZOtKC0zIHFXqtqOiwlTqzQQERE1h92EuOLiYuzYsQM33HBDjf2xsbHYvHmz2duMGDFCBblffvlF/S4tditWrMCkSZNavV6yTZuOZCHzYhH8PV0wNjJY63KIiEjH7KY79dy5ilnxg4NrfnF27NgRqampZm8THx+PrKwsXHXVVTAYDGrfPffcg0cffbTexykqKlIXk7y8vMstnWxEufEzlFC5xNbU6DC4OrMVjoiIms9uWuJMag8il3BW38DyTZs24amnnlJj6Pbu3YsvvvgCy5cvx/PPP1/v/S9YsAA+Pj5Vl/BwTuJKFX4+kY2T2QVo5+qE8f07al0OERHpnN20xAUEVAwgP3PmTI39GRkZCA0NNXubxx57DHfddZcaQyf69++PS5cuYcaMGXj88cfVGa61zZs3T50MYSItcQxyJH8sJOyoaIW7eUAI2rk5aVwRERHpnd20xLm6umLw4MFYt25djf1r1qxRY9/MkcBWO6jJVCPC1L1am5ubG7y9vasu7dtzJn4C9qXmYn9GHlydHDBhgPk/GoiIiJrCblrixCOPPKKmCBkyZAiGDRuGJUuWqJMV7r77bnW9tLBJq9mLL76ofp88eTJeeOEFDBo0SI2LO3bsmBoPN2XKlKowR2SJxMqxcOP7BMPP00XjaoiIyBbYVYiLi4tTJzg88cQTqls1KioKGzduROfOndX1J0+erNHyJsfJeDkJfykpKejQoYMKcBLsiCyVnHUJ205dgMwmMmUQu9aJiKhl2FWIE3J2qVzM+eGHH2r87uzsjKefflpdiJrr88pWuLE9O6Cjj5vG1RARka2wmzFxRFpIu1CIjcfOqu2pMWyFIyKilsMQR9SKlu1KRbkBGNbFD90CPbUuh4iIbAhDHFErOX+pBGsPZqjtuMFshSMiopbFEEfUSlYmpaK03ICoEG/0C/XWuhwiIrIxDHFEreBiURlW7K2YWDqeY+GIiKgVMMQRtYI1e9ORX1KG7v6euKKrr9blEBGRDWKII2phRaXlWJaUprZlLJxjPWvzEhERXQ6GOKIWtv5gJrILShDS3g0jewZqXQ4REdkohjiiFlRWbqhaYituUBicHNkKR0RErYMhjqgFbT56Ful5RfDzcMHYPsFal0NERDaMIY6ohRgMBiTsqGiFmxwdCncX/vMiIqLWw28Zohay/dcLSD6fD08XJ8RGhWhdDhER2TiGOKIWYmqFm9i/I7zcnDSuhoiIbB1DHFEL2J+Wiz3puXB2dMCk6FCtyyEiIjvAEEfUAhIrW+HGRwYhwMtV42qIiMgeMMQRXaYTZ/Px46/ZkNlEpnCJLSIiaiMMcUSXaVnlvHCjuwci1Ndd42qIiMheMMQRXYYzuUVYfzRLbU9lKxwREbUhhjiiy7B8VyrKDcCVnXzRI6id1uUQEZEdYYgjaqbs/BKsPZChtuMHsxWOiIjaFkMcUTOt2pOGorJy9Atuj/5h3lqXQ0REdoYhjqgZ8ovLsGJPutqOGxwOBwcudE9ERG2LIY6oGdbuO4OLxiDXxc8DV3Xz07ocIiKyQwxxRE1UXGrA57tT1XZ8TDgc2QpHREQaYIgjaqINhzJxLr8EQV5uGNWrg9blEBGRnWKII2qCsnIDEndVTO4bNygULk5shSMiIm0wxBE1wdbkc0jJKYSPmzOu79tR63KIiMiOMcQRWchgMCChcqH7yQND4e7Cfz5ERKQdfgsRWWjXqRwcOXsJHsbwFts/ROtyiIjIzjHEEVkoYcdp9fPGvh3h7eGscTVERGTvGOKILHAoPQ+70nLh7OiAmweFaV0OERERQxyRJRJ3VoyFu753BwR6uWpcDREREUMcUaNOnS/AlhPnIZOJTInhQvdERGQdGOKIGvF55RmpIyIC0MnPQ+NqiIiIKjDEETUgM7cI3x7NUtuy0D0REZG1YIgjasAXu1PVKg2Dw33QK9hL63KIiIiqcJ4EonrkFJRi9YEMtR0/uJPG1RDZLvlD6ZeT2cjMK0RQe3cM6eoHJ0cuaUfUGIY4onqs2puGotJyRHbwQnS4t9blENmkb/afwVOrDqjl7EzCfdzx3IS+uKEfl7Yjagi7U4nMyC8uw4qkdLUdPzgcDg5sFSBqaRLg7vxkZ40AJ1KNv8t+uZ6I6scQR2TGuv0ZyC0qRWdfDwyN8Ne6HCKbI12o0gJnMHOdaZ9cL8cRkXnsTiWqpaTMgMTdqWo7LibMdsfmlJQA+flaV0F26psDWXVa4KqT6CbXy1g5/iFFZB5DHFEtGw9n4uylYgS2c8W1vYO0LqflVXYNX0xJx3nj/4jaSl5JOdafKcFXacX46XypRbeRkx2IyDyGOKJqpOsmcWdFK9y0gaFwcbLBVjjHilEUOb4dkB4aoHExZOuKS8vxy+kcbEg+j62nc1HaxO5ROVuViMxjiCOq5qfj53HqQgG83Zxt/sw4g6srDB5cgYJanvwxtC81FxuPZGJT8jlcKi6ruq67vyfG9O6AEd0D8eCKvci6WFzv/chZqjLdCBGZxxBHVMlgMCChcqH7m6JD4OnqpHFFRPoh/36Ss/LxvTG4bTh6Vg1JMAnycsN1vQIxqlcQugV6Vu2fPSIC89ceqvc+/xLb13bHpBK1AIY4okp7UnJxKPMi3JwdMSEqVOtyiHThTE6RCm7fHcnCr9kFVfulNXtUj0Bc26sD+oa2h6OZaXqGdw/A/PGReHvzcbMtcmk5BXX2EdH/MMQRVUrYcVr9jO0bDB8P/tMgqo+sZrL5aBa+O5yF/Rl5VfvdnBwxrJs/xhiD2+AufhaNKZUgd5XxNgdOZsGQmoaI6J44mlOCR7/cj+e/PoxregagdzAn2yYyh99UREZHMi5ie0qO6rqZPDBM63KIrE5BcTm2nTiHDUey8POpbJjOT5DezsHhviq4DTMGsuYMQ5B/dwNC2iPE4Ar/Lr64ysMD3xsD4tpDmbh3aRJW/t9wuLtweANRbQxxREafVY6FG9ezA4K83TSuhsg6lJYZsOv0BWw0BqotxgBXWFpedV1kkBfGGoPbCOO/Gf92Li36uLJCysIpUdjxxmbsO5OHV9cdwZOxfVr0MYhsAUMc2b2U7AL8kHxObU+JYSsc2Tc5QeHQmYvYeCQL3x/NwoXC/83nJmeLXte7A0b1DEKYX+tO/RHg5Ya/ThmAW/+xHe9sOaHmbLy6B6fEIaqOIY7s3rKdqWp2+Gu6+aNLgGejxxPZolPnCypOUDichfS8oqr9/h4uuLZnoApRPYPatek6wmMigzBzaGd8+NMpzP1sN9bPHQFfT9c2e3wia8cQR3bt7MVifHM4U23HxYRrXA1R2zpn/PxvOpKlxrkdOXupar+nixNGRPjj2l5BiO7ko+k0H4+P74PNyedwJOsSHl+xD2/fMqhNgySRNWOII7u2YneqmkE+OtQbkSHttS6HqNVdLCrDf4+dVd2lu1Jzqhabl6A2tLOfmoh3SFd/uLs4alqniYerE96KH4jfvrMVK/aewdjINA57IKrEEEd2K6+wFKv2n1Hb0wd30rgaotZTXGrALyfPq+C21fiz+tJXA0K8MdYY3K7uHghvK51ap3+YD+aN64XnvzmMx77chyu7+qGTP4c+EFnnv1iiNrB6bzoKSsrRM7AdYjr7aF0OUYtSS1+l5aqpOjYln8VFM0tfjdLR2dizRkZgw+FMYwjNxtzEJCTedRWcnayjtZBIKwxxZJcKjeFteVKa2o6PCecYG7IJzVn6Si+ku/f1uGiMe2MLfvo1G4s3HcecMT20LotIUwxxZJfWHchATmEpwrzdOW0B6Z4sfbXJGNzWN2PpKz0J9/PESzf1w/8lJOHl745ihDGURof7al0WkWYY4sjuyASmibtS1XZcTBgX2CZdasmlr/RkUnQo1h/MxPI96bh36W6svfcatHPjVxnZJ37yye7IlAqZF4vg7+mCsZHBWpdDZLHGlr6SFrfhzVz6Si9k6MMLk/rj51+zcexcPl5Yewgv3tRf67KINMEQR3al3GBAQuUSW1Ojw+DqbFutFGR7TEtffW8MbpuPt93SV9bMx/gH2Btx0Zj6wTZ8vO2UCq/j+vIPMrI/dndqz7vvvouwsDA4OjoiOjoaW7ZsafD4CxcuYPbs2fD391e36dGjB9auXdtG1VJL+/lENk5mF6CdqxPG9++odTlEZskJCgfT8/DOpuP43Uc/44nVB/CtMcRJgJOlr26/shP+/vsYvGkMMpMGhtpVgDMZFhGAOSMi1PaDy/ciK69Q44qI2p5dtcQlJiaqQPb+++/j6quvxuIlSzB69GicOHECnTrVnSesuLgYY8aMQXDHjli9ejXCw8Nx+vRptG/PSWH1SL4YE3ZUtMLdPCAE7dxst8uJ9EmWvtqkxrllIi3XOpa+smZ/HtcT3x89i73puXh42V58dPsVfG3IrthViFu4cCFmzZqFmTNnqt9fX7QIq1etUq1zL774Yp3jP/roI2RlZWHbtm1wcan4S7dz585tWjO1nH2puWoAuKuTAyYMCNW6HCJFD0tfWSs3Zye8NT0a1//tv1h3OAufbDuF24Z20bosojajmxAn3Zq//PILMjMzUV5eXuO62267rdHbS6vajh078OSTT9bYHxsbi82bN5u9zRdffIFRo0ap1ruEhASEhITg9ttvxyOPPAInJ/OtOEVFRepikpeXZ/Y4anuJlWPhxvcJhp+n/XU/kfWQpa+2Jp/FRmPw2KmDpa+sWc+g9nhmfCQeX3UAz3x1EMO7+6N7B/aWkH3QRYiTrszJkyejpKQErq6uNZrLZduSEHfu3Dn1Mzi45uDXjh07IjU11extjh49inXr1uGOO+7Axo0b1e933nknSktL8dRTT5m9zYIFCzB//nwLnxm1leSsS9h26oI6i2/KIC50T21P70tfWbMZQ7tggzEQyzx5c5Ym4ct7hsPVmQGYbJ8u/mtx77334v9mz8aLL7wAT8/Lm2m89ngJGSdV3xgKafELCgrCe++9p1reYmJikJaWhueee67eEDdv3jw88MADVb9LS5yMpSNtfV7ZCje2Zwd09NHHMkOkf5YsfTXS+JkM1snSV9bK0fjX2atTojD2jc1IMr7ei9YfxaO/6a11WUStThchTk4mmHvffZcV4AICKmblP3PmTI39GRkZCA01Pz5KwpeMhaveddqnTx9kZ2er7llpFazNzc1NXch6pF0oxMZjZ9X21BgGampdDS995YrrenXQ7dJX1izI2x2vTh6AOz7ZgTd/SMZoY0C+qpu/1mURtSpdhLgJEyZg+/Yd6NatW7PvQwLX4MGDVffoTTfdVLV/zZo1mDZtmtnbjBw5Up3cIC1yMr2IOHLkiJpuxFyAI+u0bFeqmhR1WBc/fnFSq6lv6SsvNydc2z3QGCqC0M8Glr6yZjf0C8aMIZ3wz19O476E3Vg3dwR8PDj+lWyXbkLcfffdiwMH9iNqwAC4ODvXud4SckLC9OnTMWTIEAwbNgxLlixR04vcfffd6voZM2ao1jfTmar33HOPOqN17ty5xse/T42Jk27URx99tGWfILWa85dKsPZghtqOG8xWOGpZpqWv5MzSfWfsZ+kra/ZUbB/89/g5JJ/Lx9Mr9+ON+IFal0TUanQR4kxTgtR3woB0X1giLi5OneDwxBNPqG7VqKgodcKCadqQkydPVrW4CQl0cuaqjMl7++231STBMuZNwiDpw8qkVDWAPCrEG/1CvbUuh2wAl76ybrKO6pvG4DZx8Y/4bHcaxkYGYWI0pxQi26SLEGdpSLOEtK7JxZwffvihzr6hQ4eqqU1If2QahxV7K8ZAxnMsHF0GLn2lL4M6+eKhMT3wspzgsGIfBnfxRZgvh1KQ7dFFiCNqjjV705FfUqbOAryiq6/W5ZDOyB+Ph85cVFOCfH80CxcKS6uuk6WvZEqQUcbgFu7noWGVVJ/Zo7ur9+7nUxfwQOIe/OePV3HCZLI5uglx0kom49P27t2rpgSRrlAZm3bNNddoXRpZoaLScixLSlPbMhaOg8nJUqfPF6jQZm7pq9Gy9FWvIPQK5tJX1s7ZyRFvxkdj7BtbsOXEeXyw5QRmjYzQuiyiFqWLEPfpp5/i1ltvVWeRPvTQQ+ovZFm4Xs4eletuueUWrUskK7P+YCayC0oQ0t4NI41fvEQN4dJXtqmzfzu8OKEf5i7bg5fWHcY1PQLQL9RH67KIWowuQtwzzzyDRYsW4f7776/aJ2eLyj45W5QhjqqTCVZNS2zFDQrjFy+Z1dDSV1d19sUYY3C7shuXvtK7qYPDsP5QBlbtz8CcpbuxZs418OBJJ2QjdBHijh8/bnYakYkTJ+LBBx/UoCKyZpuPnkV6XhH8PFwwtk9w4zcgu8Glr+yPdHsvuDkK209dwOGsS1jw9SH8ZWI/rcsiahG6+C9Vp06d8N1336F79+419ss+0/QgREK62hN2VLTCTY4OZSsKNbj0VTd/T7WCwsheXPrKlvm1c8WiadGY/vef8f6Pv+La3kFqRQcivdNFiHvsscfUhLy7du3C1Vdfrf6ykvnbZE3TxYsXa10eWZHtv15A8vl8NZYpNipE63JII/9b+ioLG45m1bv0VdcAD56gYCdkbOys4V2xZOtJ3P/5HmyYew38vRjcSd90EeIkwHXs2BEvv/wy/vnPf6p9AwYMwIoVK1SXKpGJqRVuYv+Oarkjsi9c+ooa8uhvemNz8lkcyLiIR5bvw/u3xTDEk67pIsQJWe+0+pqnRLXtT8vFnvRcODs6YBJnaLcbsvTVFpkShEtfUSPcXZzwt/iB+O3bW7HmYAYStqdg+pBOWpdF1Gy6CXFEjUmsbIUbHxmEAC9Xjauh1sSlr6i5+oR44/EbeuGZNYfw5KoDGBrhj64B7bQui6hZrDbEeXh4ICUlBQEBAXB3d2+wybugoKDe68g+nDibjx9/zVZf4lO4xJZNkqWvdp++oM4sNbf0lbS4jeTSV2SBmVd3w3eHsvCD8XN039LdWHb3MLg48SQo0h+rDXFLliyBl5dX1TbHLVBDllXOCze6eyBCfd01roZaCpe+otbgaPxrb9G0ARj7xmZsT8nBWxuO4YFxvbQui6jJrDbEzZgxo2r79ttv17ASsnZncouw3vgFL6ayFc4mcOkram0hvh54ZXIU7vr3Lry28RhGGf8giOnsp3VZRE1itSGuOplaxMXFBf3791e/r1y5Eh9++KH6XVZzcHXl+Cd7tnxXqhoTdWUnX/QI4tgWveLSV9TWZBqi6YMysdT435A5Cbvxzb3XoL07u+NJP3QR4mbOnIknn3pKhbYTJ06os1Tj4+PVuqmX8vPx+qJFWpdIGsnOL8HaAxlqWxa6J33h0lektfkT+2LrifM4eb4Az351EK9MGaB1SUQW00WI279/PwZGR6vtzz77DGPHjsV//vMfbN26VQU6hjj7tWpPGorKytEvuD2iwry1LocswKWvyJp4u7uoaUdufu9HfLo9BWN6B2F8/45al0VkEV38V1IGN5eXV5yJ9s0332DSpElqW5bjOnfunJalkYbyi8uwYk+62pZWOI6Psl5c+oqs2ZCufrh/dHf8dWMyHlq+FzGdfBHswxOkyPrpIsQNHz4czz33HMaNG4eNGzeq5baEdK0GB3OBc3u1dt8ZFQa6+Hngqm4ckGxtLFv6qgO6BngygJPm7hvT0/hZPau69R/8fA/+eccQdRYrkTXTRYj729/+hri4OCxduhTz589H9+7d1X7pWh01apTG1ZEWpEvu892pajs+JpzLKFkRLn1FeuTq7Ig3p0dj3JtbsOHYWXy09SRmXtNN67KIGqSLEBcVFYWDBw/W2f/KK6/AyYkzstujDYcycS6/BEFebqo1h7RV39JXrk4OGN7VH9cag9sVXPqKrFxEoBeeje2Lh1fsw/NfH8Y1PQPQO5hjbcl66SLE1UdWciD7I+OrEndVTO4bNyiUwUAjhSXl+Ol4w0tfDYsIQDs3/qFF+vG7KzupPxLXGi/3Lk3Cyv8brtZcJbJGVhviuOwW1Wdr8jmk5BTCx80Z1/flWWRtqcGlrzp4YUxvLn1F+ibfNQunRGHHG5tVq/Kr647gydg+WpdFZJbVhjguu0XmyGD5hMqF7icPDOX8YW2AS1+RvQnwcsNfpwzArf/Yjne2nFDDAa7uEaB1WUR1WG2I47JbZM6uUzlqNn8PY3iL7R+idTm6I13ROYUlajslO1/9Xt8KCFz6iuzZmMggzBzaGR/+dApzP9uN9XNHwNeTqwORdbHaEFfd2rVr1QkM119/fY393377LcrKyvCb3/xGo8qorSXsOK1+3ti3IyeDbSLphn5783FkXayY6uPfO1Px7ZEszB4RgeHdK1oZ6lv6SkLziIgAFdwGcukrshOPj++DzcZ/N0eyLuHxFfvw9i2D+EcLWRVdfAs++OCDeO211+rslwmA5TqGOPtwKD0Pu9Jy4WwMEDcPCtO6HF2RADd/7aE6+yXQyf5J/TsiJbuAS18RVePh6oS34gfit+9sxYq9ZzA2Mg1TYvjfHrIeughxhw8fRt++fevsj4yMVNeRfUjcWTEW7vreHRDoxW4NS0mXqbTANeTLfWeqtqNk6ateHXB1j0D4sLWT7Fz/MB/MG9cLz39zGI99uQ9XdvVDJ39PrcsiUnTxp7W3tzeOH6/7JXTs2DF4evIfkz04db4AW06ch3RkTInhQvdNcSAtt6oLtSE39gnGv2ZcgdemROG3UR0Z4IgqzRoZgeHG8CYrxMxNTEJpWXnjNyJqA7oIcdOmTcPs2bNrBLnk5GTcd999aiUHsn2fV56RKuOyOvEsyCY5n19i0XHRsl4k1y4lqkOGFrweF62mNfrp12ws3tRwyzZRW9FFiJOVGWS6EVluKywsTF169OiBDh064NVXX9W6PGplmblF+PZoltqWhe6pafw9LZuzzdLjiOxRuJ8nXrqpn9p++bujSEq5oHFFRDoZE+fj44Nt27Zh/fr1SEpKUhMBDxgwACNGjNC6NGoDX+xOVeO6Bof7oFewl9bl6E7fUG908HJtsEtVFqSX44iofpOiQ7H+YCaW70nHvUt3Y+2916Cdmy6+RslG6ebTJ6d1jxs3TgU3Nzc3nuZtJ2RNztUHMtR2/OBOGlejT9IVJNOImDs71eT/jNdz2hCihsn3zguT+uPnX7Nx7Fw+XjD+m3rxpv5al0V2TBfdqTKVyPPPP4/AwEDVCnfy5Em1/+mnn8bf//53bYujVrVqbxqKSsvVkk7R4Wwpai6ZB27++EjVIledtMDJftM8cUTUMB9PF7wRFw1pR/h42yl8W/lHJpEWdNES98ILL+C9997Dm2++idtuu61qf1TUALzyykLceeedGlZHrSW/uAwrktLVdvzgcLa+XiYJald181dnq8rJDjIGTrpQ2QJH1DTDIgIwZ0QE/vbDcTy4fC++6+SDDu3dtS6L7JAuWuI++OAD/OMf/8Dvfvc7ODr+r+QBA6KwZ88eDSuj1rRufwZyi0rR2dcDQyP8tS7HJkhgiwr3wahegeonAxxR8/x5XE81p+LZS8V4eNletcYwUVvTRYhLSUlRZ6PWJt2spaWlZm5BeldSZkDi7lS1HRcTxrBBRFbFzdkJb02PNv50xLrDWfhk2ymtSyI7pIsQJ2eibt68uc7+zz77DFdeeaUGFVFr23g4U/2FG9jOFdf2DtK6HCKiOnoGtccz4yPV9jNfHURyVp7GFZG90cWYuOeeew5TpkxBamqqarJevnw5Dh06hA8//BDr1q3TujxqYTKdSOLOila4aQND4eLEVjgisk4zhnbBhsNZWH8kC3OWJuHLe4bD1VkX7SNkA3TxSbvxxhuxatUqdRGPPPII9u3bh2+++QbXXXedxtVRS/vp+HmculAAbzdn3NCvo9blEBHVy9HRAa9OiUKApwuS0nKxaP1RrUsiO2L1LXFlZWX473//q7pNzXWpkm2RltaEyoXub4oOgaerk8YVERE1LMjbHa9OHoA7PtmBN39IxujeHdSZ4EStzepb4pycnDB69GhcuMAlTuzBnpRcHMq8qAYLT4gK1bocIiKL3NAvGDOGdIKcpHpfwm7kFFi2ZjHR5bD6ECcGDhyI48e54LA9SNhxWv2M7RsMHw+rbygmIqryVGwfdA/wxOmcQjy9cr/W5ZAd0EWIe/nllzF37lx89dVXOHPmDHJzc2tcyDYcybiI7Sk5ajqRyQPDtC6HiKhJZB3VN+MHqv+GfbY7DSuT0rQuiWycLpo6rr/+evVTTnAwh5Ms2obPKsfCjevZAUHebhpXQ0TUdIM6+eKhMT3w8vqjeHTFPgzu4oswX0+tyyIbpYsQt2nTJq1LoFaWkl2AH5LPqe0pMWyFIyL9mj26OzYeycLPpy7ggcQ9+M8fr+KE5dQqrDrE5efn4+GHH0ZCQgKKi4sRGxuLt956CwEBXKzb1izbmQppT72mmz+6BPCvViLSL2cnR7wZH42xb2zBlhPn8cGWE5g1MkLrT+Ju8AAAIABJREFUssgGWfWYuPnz52PJkiWYPHky7rjjDqxevRqzZs3SuixqYWcvFuObw5lqOy4mXONqiIguX2f/dnhxQj+1/dK6w9iflqNxRWSLrLol7j//+Q8+/fRTxMfHq99vvfVWNV+czB0nU4+QbVixOxWl5QZEh3ojMqS91uUQEbWIqYPDsP5QBlbtz8CcpbuxZs418ODcl9SCrLolTpbZGjFiRNXvQ4YMgaOjI9LT0zWsilpSXmGp8T9wZ9T29MGdNK6GiKjlODg4YMHNUQhp74bDWZew4OtDWpdENsaqQ5ycderq6lpjn4S40tJSjSqilrZ6bzoKSsrRM7AdYjr7aF0OEVGL8mvnikXTotX2+z/+iu8PZ2lcEdkSq+5OFbfNmAF3t/9NNyEB7q677oKXl1fVvi+++EKL0ugyFRrD2/LKeZTiY8LVX61ERLZmZM9AzBreFUu2nsT9n+/BhrnXwN+L0yjR5bPqEPeHP/zBon2kT+sOZCCnsBRh3u64ugfPOCYi2/Xob3pjc/JZHMi4iEeW78P7t8XwD1e6bFYd4j766COtS6BWUlpmQOKuVLUdFxPGOZSIyKa5uzjhb/ED8du3t2LNwQwkbE/B9CEcB0yXx6rHxJHt2nQkC5kXi+Dv6YKxkcFal0NE1Or6hHjjiRt6q+0nVx3AyXOXNK6I9I4hjtpcucGAhMoltqZGh8HVma1wRGQf7ry6K0ZGBCC/pAz3Ld2NkrJyrUsiHWOIozb384lsnMwuQDtXJ4zv31HrcoiI2oyjowMWTRsAX3dnbE/JwVsbjmldEukYQxy1KZk2JmFHRSvczQNC0M6NE18SkX0J8fXAK5Oj1PZrG49h56lsjSsivWKIoza1LzUX+zPy4OrkgAkDQrUuh4hIE7FRIZg+KAzlBmBOwm7kFZZoXRLpEEMctanEyrFw4/sEw8/TReNqiIi0M39iX3T29cDJ8wV49quDWpdDOmR3Ie7dd99FWFiYWvkhOjoaW7Zsseh2CQkJak6fSZMmtXKFtis56xK2nboAmU1kyiAudE9E9s3b3UVNOyL/Tfx0ewrW7jujdUmkM3YV4hITEzF79mw8++yzOHDgAK4dMwajR4/G6dOnG7zdqVOnMGfOHAwfPryNKrVNn1e2wo3t2QEdfThbORHRkK5+uH90d7X90PK9yMgp1Lgi0hO7CnELFy7ErFmzMHPmTERGRuL1RYvQtWtX1TpXn7KyMsTHx+Pll19Gz54927Ba25J2oRAbj51V21Nj2ApHRGRy35ieiAnzQXZBCR78fA/KZaAckQXsJsQVFxdjx44duOGGG2rsj42NxebNm+u93XPPPYfg4GDceeedFj1OUVERcnNzqy55eXmXVbetWLYrVQ3gHdbFD90CPbUuh4jIarg6O+LN6dHwcHHEBuMfux9tPal1SaQTdhPizp07p35KIKuuY8eOSE1NNXubrVu34u2338aHH35o8eMsWLAAPj4+VZfwcLY6nb9UgrUHM9R23GC+HkREtUUEeuHZ2L5q+/mvD+NwRq7GFZEe2E2IM6m94LDMW2ZuEWJpQZs2bRr+9a9/ISDA8sXZ582bh5ycnKpLSkrKZdesdyuTUlFabkBUiDf6hXprXQ4RkVX63ZWdMD4yCEVl5bh3aRIKS8q0LomsnLPWBbQVUxA7c6bm2T8ZGRkIDa07X9nx48eRlpaG8ePH17lOQl9ycjIiIiLqXOfm5qYuVOFiURlW7K14zeM5Fo6IqF7y3bJwShR2vLEZ+87k4dV1R/BkbB+tyyIrZjctca6urhg8eDDWrVtXY/+aNWswYsSIOsf37t0b+/btq3GZOHGiOptVttlNapk1e9PVGoHd/T1xRVdfrcshIrJqAV5u+OuUAWr7nS0n8N9j5zSuiKyZ3bTEiUceeQTTp0/HkCFDMGzYMCxZsgQnTpzA3Xffra6fMWOGCmcvvvgi3N3d0a9fvxq39/PzUz9r7yfzikrLsSwpTW3LWDhHM93WRERU05jIIMwc2hkf/nQKcz/bjfVzR8DX01XrssgK2VWIi4uLUyc4PPHEE6pbNSoqChs3bkTnzp3V9SdPnlSTAFPLWH8wU50yH9LeDSN7BmpdDhGRbjw+vg82J5/DkaxLeHzFPrx9yyCz47fJvjkYZGQ/tRqZZkTOUv1p1xG06xCkdTltpqzcgDv/tQPpeUWYOzICsQNCtC6JiKyYQ0EBQtJOwP+KaMCT0xCJfak5+O07W9WJYX+bFo0pMWFal2RXTN/fcpKit7d1npTHZidqFZuPnlUBzs/DBWP7BDd+AyIiqqF/mA/mjeulth/7ch9On8/XuCKyNgxx1OKkcTdhR8XUKpOjQ+Huwo8ZEVFzzBoZgeFd/XCxuAxzE5NQWlaudUlkRfjtSi1u+68XkGz8i9HTxQmxUexGJSJqLidHB7weFw0fN2f89Gs2Fm86rnVJZEUY4qjFmVrhJvbvCC83J42rISLSt3A/T7x0U8WsCC9/dxRJKRc0roisBUMctaj9abnYk54LZ+Nfj5Oi606iTERETSf/PZ08IESdNHbv0t24VFSqdUlkBRjiqEUlVrbCydIxAV6c14iIqCXI9CIvTOqPcB93HDuXjxfWHtK6JLICDHHUYk6czcePv2bD0QGYwiW2iIhalI+nC96Ii4ZMF/fxtlP49kCG1iWRxhjiqMUs21nRCje6eyBCfd01roaIyPYMiwjAnBEV63Y/uHwvsvIKNa6ItMQQRy3iTG4R1h/NUttT2QpHRNRq/jyuJ6JCvHH2UjEeXrYXnLPffjHEUYtYvisV5cb/jlzZyRc9gtppXQ4Rkc1yc3bCW9OjjT8dse5wFj7ZdkrrkkgjDHF02bLzS7C2cmyGLHRPREStq2dQezwzPlJtP/PVQSRn5WlcEWmBIY4u26o9aSgqK0e/4PaICrPO9eWIiGzNjKFdcF2vDigsLcecpUkoLuVqDvaGIY4uS35xGVbsSVfb0gonp8ETEVHrc3R0wKtTohDg6YKktFwsWn9U65KojTHE0WVZu++MWtOvi58Hrurmp3U5RER2JcjbHa9OHqC23/whGdtOnNe4ImpLDHHUbMWlBny+O1Vtx8eEw5GtcEREbe6GfsGYMaQT5CTV+xJ2I6egROuSqI0wxFGzbTiUiXP5JQjycsOoXh20LoeIyG49FdsH3QM8cTqnEE+v3K91OdRGGOKoWWT9vsRdFZP7xg0KhYsTW+GIiLTSzs0Zb8YPhJOjAz7bnYaVSWlal0RtgCGOmmVr8jmkGP/i8zH+h+P6vh21LoeIyO4N6uSLh8b0UNuPrtiH1Av5GldErY0hjppMZgdPqFzofvLAULi78GNERGQNZo/ujis7+yKnsBQPJO5RvSZku/jtS02261QOjpy9BA9jeIvtH6J1OUREVMnZyRFvxkejnasTtpw4jw+2nNC6JGpFDHHUZAk7TqufN/btCG8PZ42rISKi6jr7t8OLE/qp7ZfWHcb+tByNK6LWwhBHTXIoPQ+70nLh7OiAmweFaV0OERGZMXVwGCb0C0ZxmQFzlu5GQXGZ1iVRK2CIoyZJ3FkxFu763h0Q6OWqcTVERGSOrJ6z4OYohLR3w+GsS3j560Nal0StgCGOLHbqfIEaYyGTiUyJ4UL3RETWzK+dKxZNi1bb7/34K74/kqVxRdTSGOLIYp9XnpE6IiIAnfw8NK6GiIgaM7JnIGYN76q27/9sD85fLNK2IGpRDHFkkczcInx7tOKvOFnonoiI9OHR3/RG32AvZBoD3CPL96lposg2MMSRRb7YnarmGxoc7oNexv8YEBGRPri7OOFv8QPh5uSINQczkLA9ReuSqIUwxFGjcgpKsfpAhtqOH9xJ42qIiKip+oR444kbeqvtJ1cdwMlzlzSuiFoCQxw1atXeNBSVliOygxeiw721LoeIiJrhzqu7YmREAPJLynDf0t0oKSvXuiS6TAxx1KD84jKsSEpX2/GDw9Vp60REpD+Ojg5YNG0AfN2dsT0lB29tOKZ1SXSZGOKoQev2ZyC3qBSdfT0wNMJf63KIiOgyhBj/W/7K5Ci1/drGY9h5KlvjiuhyMMRRvUrKDEjcnaq242LC4OTIVjgiIr2LjQrB9EFhKDcAcxJ2I6+wROuSqJkY4qheGw9n4uylYgS2c8W1vYO0LoeIiFrI/Il9VQ/LyfMFeParg1qXQ83EEEdmyXQiiTsrWuGmDQyFixNb4YiIbIW3u4uadkQ6WD7dnoK1+85oXRI1A0McmfXT8fM4daEA3m7OuKFfR63LISKiFjakqx/uH91dbT+0fC8ycgo1roiaiiGO6pDZvBMqF7q/KToEnq5OGldERESt4b4xPRET5oPsghI8+PkelJdzNQc9YYijOpJScnEo8yLcnB0xISpU63KIiKiVuBr/O//m9Gh4uDhiw7Gz+GjrSa1LoiZgiKM6EnecVj9j+wbDx8NZ42qIiKg1RQR64dnYvmr7+a8P43BGrsYVkaUY4qiGIxkX1SSQMth18sAwrcshIqI28LsrO2F8ZBCKyspx79IkFJaUaV0SWYAhjmr4rHIs3LheHRDk7aZxNURE1BZkNZ6FU6IQ5OWKfWfy8Oq6I1qXRBZgiKMqKdkF+CH5nNqeGhOucTVERNSWArzc8NcpA9T2O1tO4L/HzmlcETWGIY6qLNuZCjkv6Zpu/ugS4Kl1OURE1MbGRAZh5tDOanvuZ7txIb9Y44qoIQxxpJy9WIxvDmeq7Ti2whER2a3Hx/dBrw7tkJZbhMdX7FPTTpF1YogjZcXuVJSWGxAd6o3IkPZal0NERBrxcHXCW/ED4ezogBV7z2D5rjStS6J6MMQR8gpLsWp/xZIr0wd30rgaIiLSWv8wH8wb10ttP/blPpw+n69xRWQOQxxh9d50FJSUo2dgO8R09tG6HCIisgKzRkZgeFc/XCwuw9zEJJSWlWtdEtXCEGfnCo3hbXlSRVN5fEy4Os2ciIjIydEBr8dFw8fNGT/9mo3Fm45rXRLVwhBn59YdyEBOYSnCvN1xdY8ArcshIiIrEu7niZdu6qe2X/7uKJJSLmhcEVXHEGfHSssMSNyVqrbjYsLUX11ERETVTYoOxeQBISgrN+DepbtxqahU65KoEkOcHdt0JAuZF4vg7+mCsZHBWpdD/9/efcBHVaX9A38y6YV00gkhhE6IEGCFFWRhATEKAlJW/8Cqf1dcsK8dV2QFpCgrFlwLuu6rC6GuKCD6CiKiCEhoIYGEQBoppPd633lOmOwkTMjMJJkz5ff9fIZc7tzJPPeeO3OenHvOuQAAZoi72SyfNpjCvFwopaCSlu9Jkh0SXIMkzkY1KgptvnaLrbtjQsnJAa1wAACgm5f6j/03Z8cQd5v+5Eg6fZOYKzskICRxNuuXtCK6VFRF7k72NGVwkOxwAADAzI2K9KPFYyLF8lPbT1N+WbXkiABJnA3i2bc3H29qhZs+JJjcne0lRwQAAJbgyYl9KDrYk65W1NLT207jbg6SIYmzQWeySulsbhk52dvRnUNCZIcDAAAWwtnBnt6eG6P+qaJ9yfn0P0fSZYdk05DE2aD4a33hpgwIJB83R8nRAACAJekT0I1entJfLL/81TlKzS+THJHtQhJnY1LzK+hIejHxbCIzh+JG9wAAYLj5N/ek3/ftTtX1jbR400mqrcfdHGRAEmdjtl5rhZvQpzsFeTlLjgYAACyRSmVHa2dGk5+bI53MLqV1316QHZJNQhJnQ7KLq2l/ylWxfPcwtMIBAIDxAjxdaO2MIWJ5/cFUOpJWKDki24MkzoZsO5FFjQrRqJ4+1MvfTXY4AABg4SYPCqT5I3oQD1J9dHMClVTVyQ7JpthcErdhwwYKDQ0llUpFMTExdOjQoTa3/fDDD2n06NHk4uIiHuPHj6ejR4+aMNrOU1hRR3vONU3OODsWrXAAANA5XoobQL393CijpJr++sVZ2eHYFJtK4uLj42nRokW0bNkySkxMpN+pk7Jx48ZRRkaGzu33799P8+bNoyNHjtDJkycpIiKCbrnlFsrOzjZx5B33xcksqm9UxPw+g0I8ZYcDAABWwt3ZgdbPuUncf3tLQra6vrG8OtJS2Sk2NFPf8OHDacSIEaI1TiMqKopmz55NK1asaPf1DQ0N5OrqSh999JFI7vRRWlpKXl5e9POJ8+TePcDo2DuivKaB/t8nR6myroFejRtII3v5SIkDAKAtdlVVFJydRr7DY4jc0N3DEr35vxdo1bcXyMvFgb597BYK9bbsctTU3yUlJeTpaZ6NHzbTEldbW0vHjx+nyZMnt1gfFxdHP/zwg16/o7KyUiRyvr6+XRFil9l9+opI4Hr7utHwCG/Z4QAAgBVaNK43jQz3ppLqenoi/hQ1NNpMG5E0NpPEFRQUiJ+BgYEt1gcFBVFWVpZev+O5556jsLAwmjBhQpvb1NTUiOxd8ygrkzsJYk19I2271rTNfeFUdrjRPQAAdD4HexWtnxMj7sl9KK2QPjyUJjskq2czSZyGXaskhq8mt16ny5o1a2jjxo305ZdfikEObXnttddE86vmwUmfTN+ey6OiqjoK7uZMY/v4S40FAACsW7ivO624c5BYXrkvmc5ml0iOyLrZTBLn5+cnfubk5LRYn5ubSyEhN75/6Ouvv05LliwRI1mjo6NvuC231vH1c80jMzOzY4F3ADdla26xNXtoqOh0CgAA0JXujg2lOwcFUm2DQos3JVBVbYPskKyWzSRxTk5OFBsbS/v27Wuxfvfu3TRmzJg2X7d27Vp64YUX6ODBg+L17XF2dhYdIDWPbt26dTh2Y/1w4SpdKashH1dHmjAgsP0XAAAAdBBf3XpterS4ApScX0Gr9ibJDslq2UwSx5555hl677336OOPP6akpCR64oknKC0tjRYuXCienz9/vkjYNPgSKr/m35s2ielFuNWOH+Xl5bJ2QW98mXjz8aZWuBkxIeTiaFNFDQAAEvm4O9G6WTFi+f2fLtOB8/mSI7JONlWz81Qi77zzDr344os0cOBA+u6778RccOHh4eL5S5cutZgDbt26dSIZmjljhhgAoXnw5VVzd+xyMaUWVpKboz3FRQfLDgcAAGwM98N+aHSEWH58yykqLK+RG5AVsql54mSQNU/cU9tO0+krpTR3aCjd/9sIk70vAIAxME+cdaqua6A73vmREnPL6fYBgfTBvGF6DSY0B5gnDqQ4m10qEjgHlR1Ni7nxoA0AAICu4uJoT2/NuYmc7VW0+1wubT4mb7CfNUISZ4Xir/WFm9I/gPw8nCRHAwAAtmxAsCe9OLmfWF6yK5EuFVRIjsh6IImzMmlXK+mny0XEs4nMHIYb3QMAgHzcrWdspJ+4e9CjmxKorqFRdkhWAUmcldl2bV64cb39KcS77UmJAQAATEWlsqN1s4aQt4sDHcssobe/S5EdklVAEmdFckpr6NsLTcO470YrHAAAmJFgb1daM6NpwvzX96fQr+lFkiOyfEjirMj2E1nE9xse2cObogLcZYcDAADQAk95xbMmcF21eHMClVXXyQ7JoiGJsxJFlXW0JzFXLPON7gEAAMzR0qkDKdzblS4VVtGyr87JDseiIYmzErtOZVNNQyMNCuxG0aHmOZ8NAACAp4ujmHaEB+B9diyT9pzJaf9FoBOSOCtQWdtAO09dEcvcCmcpEykCAIBtGhHhQ4+P6y2W/7L9NOWWVEuOyDIhibMC/FdMuTqR6+njSr/p5SM7HAAAgHY9Or4PDQv1oqKqOnpq6ylqbMQNpAyFJM7C1dYrtDUhSyzPGRZGKrTCAQCABXByUNH6uTHk6qii71Ku0seHL8kOyeIgibNw3yXlUUFlHQV4ONOtfbvLDgcAAEBvkf4etCxuoFh+dW8yJeeWSo7IsiCJs2ANjQrFn2ia3Hf20BBytEcrHAAAWJZ7RvYQt4nkwXmPbDpJ1XUNskOyGEjiLNjh1ALKLKkmL2cHmjQwSHY4AAAABuPBeKtnRlOAhxOdySmjtfvOyw7JYiCJs1CKotDmaze6n3FTCLk4oigBAMAy+Xk40xszh4jldw+l0Y8pBZIjsgyo+S3UifQSOn+1QnQIjRscLDscAACADhnfP4AeuDlcLD+2JYGKK2slR2T+kMRZqM3HM8TPOwYGkaerg+RoAAAAOu6FKQOob3d3yi6toRd2nhFXnaBtSOIsUNKVMjqRXUoOKjuaPjRUdjgAAACdwtXJnt6ec5Oo33aezqHtJ7Jlh2TWkMRZoPhfm/rCTerXnfw9nCRHAwAA0HkGh3rRcxP7iuXn/3OGMgorJUdkvpDEWZj0wio6lFZIPJnIzGG40T0AAFifh8ZG0ugIH3E3osfiT1J9Q6PskMwSkjgLs/XaiNQxkX7Uw8dVcjQAAACdz15lR3+fHSOm0Pr5chG99/1F2SGZJSRxFiSvtIa+uZAvlvlG9wAAANYqzMeNVt41SCyv+t8LdDKzWHJE5gdJnAXZkZAl7tIQG+ZFfQM9ZIcDAADQpabFhNCMIcGi7ntkUwJV1NTLDsmsIImzECVV9fRlYq5YnhPbQ3I0AAAAXY/v5rB82mAK83KhlIJKWr4nSXZIZgVJnIXYdTqbauobqX93D4oJ85QdDgAAgEl4uTnSm7Nj1Akd0SdH0umbaw0agCTOIlTWNtDOk1fE8pzYMPGXCQAAgK0YFelHi8dEiuWntp+m/LJqyRGZByRxFmDf2VwqramncG9XujnSV3Y4AAAAJvfkxD4UHexJVytq6eltp3E3B0ISZ/bqGhSKT8gSy7OHhYph1wAAALbG2cGe3p4bo/6pon3J+fQ/R9JlhyQdkjgztz85T/zV4e/uRL/rFyA7HAAAAGn6BHSjl6f0F8svf3WOUvPLJEckF5I4M8ZDquN/bWqFm3VTCDnaoxUOAABs2/ybe9Lv+3an6vpGWrzpJNXW2+7dHJDEmbGfLxZSenEVeTo70ORBQbLDAQAAkE6lsqO1M6PJz82RTmaX0rpvL8gOSRokcWaKO2xuvnaj+7tigsnNyV5yRAAAAOYhwNOF1s4YIpbXH0ylI2mFkiOSA0mcmTqZWUpJeeWiA+ed0SGywwEAADArkwcF0vwRPYgHqT66OYFKqupkh2RySOLMVPzxDPEzbmAgebk6SI4GAADA/LwUN4B6+7lRRkk1/fWLs7LDMTkkcWbofG45HcssIZ5NZMZNobLDAQAAMEvuzg60fs5NYvqtLQnZ9MXJbNkhmRSSODO05VpfuIl9u1OAp7PkaAAAAMzX0B7e9JfxUWL52Z1nKKu4UnJEpoMkzsxkFlXRwdQCsXz3sDDJ0QAAAJi/ReN608hwbyqprqcn4k+JKbpsAZI4M7Pt1yziU++WXr7U089NdjgAAABmz8FeRevnxJC7kz0dSiukDw+lyQ7JJJDEmZGr5bX0dXKeWJ6NVjgAAAC9hfu604o7B4nllfuS6Wx2ieSIuh6SODOyMyGL6hsVignxpP7B3WSHAwAAYFHujg2lOwcFUm2DQos3JVBVbYPskLoUkjgzUVZdT7vO5ojlubE9JEcDAABgeezs7Oi16dEU3M2ZkvMraNXeJNkhdSkkcWbiy9NXqKqukfr4u9OwcC/Z4QAAAFgkH3cnWjcrRiy//9NlOnA+X3JEXQdJnBmoVidv26/NbTNnWJj4SwIAAACMM7aPPz00OkIsP77lFBWW18gNqIsgiTMD+xJzxbDoUE8X+m2Un+xwAAAALN6zt/WjgYEelKdO4J7Zfkbck9zaIImTrL5BofgTWWJ59rBQMes0AAAAdIyLoz29NecmcrZX0e5zubT5WKbskDodkjjJvj+fL/5K8HVzpAn9A2WHAwAAYDUGBHvSi5P7ieUluxLpUkGF5Ig6F5I4iRoVhTZfu8XW3TGh5OSAVjgAAIDOdP9vI2hspB9V1jXQo5sSqK6hUXZInQZJnES/pBXRpaIqMcP0lMFBssMBAACwOiqVHa2bNYS8XRzoWGYJvf1diuyQOg2SOEm4g+Xm402tcNOHBJO7s73kiAAAAKxTsLcrrZkRLZZf359Cv6YXSY6ocyCJk+RMVimdzS0jJ3s7unNIiOxwAAAArFpcdDDNHRpKjQrR4s0JVFZdJzukDkMSJ0n8tb5wUwYEko+bo+RoAAAArN/SqQMp3NuVLhVW0bKvzskOp8OQxEmQml9BR9KLiWcTmTkUN7oHAAAwBU8XRzHtCNe/nx3LpD1ncmSH1CFI4iTYeq0VbkKf7hTk5Sw5GgAAANsxIsKHHh/XWyz/Zftpyi2plhyR8ZDEmVh2cTXtT7kqlu8ehlY4AAAAU3t0fB8aFupFRVV19NTWU9TYaJl3c0ASZ2LbTmSJTpWjevpQL3832eEAAADYHCcHFa2fG0Oujir6LuUqfXz4kuyQjIIkzoQKK+poz7lcsTw7Fq1wAAAAskT6e9CyuIFi+dW9yZScWyo5IsMhiTOhL05mUX2jQtHBnjQoxFN2OAAAADbtnpE9aEr/AKppaKRHNp2k6roG2SEZBEmciVTU1NPO002jYOagLxwAAIB0dnZ2tHpmNAV4ONGZnDJau++87JAMgiTORPYlF4j7tvX2daPhEd6ywwEAAAA1Pw9nemPmELH87qE0+jGlQHJE+kMSZyJbzuaLn9wXTmWHG90DAACYi/H9A+iBm8PF8mNbEqiwvIaOppn/rblsLonbsGEDhYaGkkqlopiYGDp06NANt9++Ywf16dNHbM8/d+7cadT7VtQ2iskFHVU2d8gBAADM3gtTBlDf7u6UXVpDI1bvp3s/OSY7pHbZVEYRHx9PixYtomXLllFiYiL9bvx4GjduHGVkZOjc/ueff6aZM2bQAw88QOfOnRM/p0+fTr/88otR789Tiyz7OokOp1pOUy0AAIAtcHWyp3uG9xDLVXWNkqPRj52iJjsIUxk+fDiNGDFCtMZpREVF0ezZs2nFihXXbc/rS0pK6Ouvv25eN2nSJPL19aVNmzbp9Z6lpaWkewmwAAAVzUlEQVTk5eVFgY9vJZWLq1jHHSj/OX842atwWRUAgNlVVVFwdhr5Do8hcsMcmmB6DY0KjVq9nzKv3cGhsbqKcv9+t8gDPD3Nc0YJB9kBmEptbS0dP36clixZ0mJ9XFwc/fDDDzpfc/DgQXruuedarLv99ttp1apVbb5PTU2NeGhwEscaaiqa1+WoT4xjZ9NocJCHwfsBAGCNVOokrqyighzS04lcXGSHAzboaEYppef+tx+cpt4257Yum0niCgqaLmEGBga2WB8UFERZWVk6X5OXl3fd9vz/3NzcNt/ntddeo6VLl163/uqGBS3+////rk/UAAAAIFNhYaG4omaObCaJ07BrNTKUM+zW69rb/ka45e6JJ55o/j83w4aHh4t+d+baHGsrysrKKCwsjDIzM6lbt26yw7FpKAvzgvIwHygL88FX0nr06EE+Pj6yQ2mTzSRxfn5+4mdOTk6L9dyqFhISovM1AQEB122vq3VOm7Ozs3i0xgkckjjzwF+MKAvzgLIwLygP84GyMB8qM55Vwnwj62ROTk4UGxtL+/bta7F+9+7dNGbMGJ2vGTt2LO3du7fFuj179tCtt97aZXECAAAA6MNmWuLYM888Q3PnzhUjVEeNGkX/+Mc/KC0tjRYuXCienz9/vmjG1oxU5cuio0ePptWrV9O0adPoP//5j0gCjxw5InM3AAAAAMh+qa5e+FZq0KBB4hLpyy+/LEaYVlZWirnjhgxput3GunXryMHBge666y7xf74WHq1+7rWVK+nVV1+lixcv0kcbN9LvJ0ww6H35d/J8dPwT5EJZmA+UhXlBeZgPlIX5MPeysKl54gAAAACshc30iQMAAACwJkjiAAAAACwQkjgAAAAAC4QkDgAAAMACIYnrBBs2bKDQ0FAxIWBMTAwdOnTohttv37GD+vTpI7bnnzt37jRRpNbPkLL48MMPxRQyLi4u4jF+/Hg6evSoCaO1boZ+LjQ2b94s7pTC0/pA5zC0LIqLi2nRokXk6+srXhMVFSXmyITOYWh5vPnmmxQZGUn29vZihgWe/qq6utpE0Vovvm/6HXfcIW4GwN85PI1Ye/ie6lxmXHY8JRlPVSaVAh2irnAUdeEr6oRAOXfunPLY448r6g+akp6ernP7n376iUcDKytXrlSSkpLET/7/kSNHTBy59TG0LO655x7l3XffVRISEkRZ3HfffYqTk5OSlZVl4sitj6FloXH58mXF399fUSfXytSpU00UrXUztCxqamqUoUOHKrdNmaL8+OOPokzUSYZy8uRJE0dunQwtj88++0xs//nnnyuXLl1S9u3bp6iTa+Vx9eugY9R/mChLlixRtu/YIerhnTt33nD7tLQ0UVZcZlx2XIZcNtu2bzdRxNdDEtdBsbGxysKFC1us6927t/L888/r3H7WrFnKpEmTWqybOHGiMmfOnC6L0VYYWhat1dfXK46Ojsqnn37aFeHZFGPKgo//zTffrHz00UfKggULkMR1EkPL4r333lPCwsKU2tpaU4Rncwwtj0WLFinjxo1rse6pp55SRo0a1WUx2iJ9krhnn31WlJW2hx56SBkxYkRXhnZDuJzaAeovOTp+/DhNnjy5xfq4uDjRTKsLN8VOmTKlxbrbb7+dvv/++y6L0xYYUxat8eTPDQ0N4hISGM/Ysvjb3/4m7kt8//33d3WINsOYstixY4e4tSBfTvXy8qL+/fvTypUrxWcDOsaY8uDbQh4+fLi5qwffZYi74KC7gelxGXFZabvttttE2dTV1UmJyTynILYQBQUF4idXPNqCgoIoKytL52vy8vKu257/n5ub2zVB2ghjyqK15557TvRxmGDgHTmgJWPKgiupd955h5KSkro8PltiTFlcuHBB3F7wvvvuo/3794v/c2JdX19PL730UpfHbM2MKY85c+ZQfn4+/eY3v+ErZ2Ldww8/TM8++2zXBgvXyc7OFmWlTVOWXLatnzMFJHGdgDtEauMPWut17W0PncPQstBYs2YNbdy4kX755RcxyAE6Tt+yKCsro1mzZtG//vUv0cEYOp8hn4vGxkbRef79998XHemHDRsmKi9uKUUS1zkMKQ++SsPH/YMPPhCJXEpKirjfd0hICC1ZssQU4YKWtupvfeqZroAkrgM0FU5OTk6L9dyqxh8wXfjLsfX2ulrnwDDGlIXG66+/Lr4MuTUoOjq6y2K0FYaWBd+TmJOE1t0MGH8xpqamipF5YDhjPhfcGu3o6CgSOI0BAwZQUVGRuBzo5OTUdQFbOWPK4/nnn6cHH3yQHnjgAfH/wYMHU0VFBc2fP59eeOEFMUoSTIPL6MqVKy3Wcf3NZHXDQel3AH+ZxcbGiksP2nbv3i36MegyduxY2rt3b4t1PHSf+6CA8YwpC7Z27VrxRch9Ffn10HGGlkW/fv3ozJkzLR5Tp04VN53mZU4qwDjGfkfxZW1ukdM4f/68qKSQwHWMMeXBCVvrRE2TYOMqjmlxGXFZafv6669pxIgR4g8fKaQNqbASmuHiGzduFEOOH782XJyH5bN58+a1GHV0+PBhMQpm1apVYloL/kmYYqRTGFoWq1evbh4erv7LuPlRVlYmaxeshqFl0RpGp3YeQ8siIyNDcXBwUBYvXqyokzflq6++Ujw9PZXly5fL2gWrYmh5LF26VIya37Rpk5ji4ptvvlHCw8PFTAfQMfxdz1NM8YPr4XXr1ollzXQvXA5cHhqaKUaefPJJUXZchphixArwXGPBwcGiMIcMGaIcPHiw+Tl15i4qJG1bt21TIiMjxfY8XJnnqIHOYUhZ8Hb8wW394C9N6DhDPxfakMR1LkPLguezHD58uNiepxtZsWKFmAIGOoch5VFXV6e88sorInFTqVRKYGCg8uc//1kpKiqSELl1OXDggM46QHP8+SeXh7bvv/9eiY6OFmUXEhIipuSRyY7/kdIECAAAAABGQ584AAAAAAuEJA4AAKyC9mCMztgOgJnz+YIkDgAArEJW1mW97iyh73YAzJzPFyRxAAAWrLGxgbKz02WHIV1NTTU5OLSc306f7Roa6qm0tFjv9+Htr17Npdxc/e4EY20qKyvEMbQVrc8Xc/u8YbJfAAALxRUqJxN1dbWyQ5EuPz+HgoJCDdquvLxMffwy1Ut25Onp3e5rOXlTlEYqKMhXb+/T0ZAtCrdEXb2aoz5e2RQR0YecnW3jzjba54s5ft7QEgcAYKHc3NzJzy/A4NeVl5d26naycWsJT4jLLSaGbOfh0c2g4+fvH0jduwfbXALHuCUqMDDUopM3Q8/n1ueLsZ+3roQkDgDAhnAn7cLC/E7bzhxwC1n37u3ffFzf7aBtsu4R2lHGnM+WcL4giQMAMDG+qTlXhsXFTX2x/vnPf5Kzs3OXv29dXR1lZqa1O9pO3+3MQW1tjfjp6HjjW4Lpux1YH2POZ0s5X9AnDgBsyh//+EeRNGkEBgbS9OnTaeXKleTt3X6/KHNVXV0l+itxny2+UT1f+vP19W9+vri4UPTlaWxUxGUi7uvD+FKRj4+fQduVlZWIzt2ciIaGRlBBQa5Y5sqS39Pb24+0cYXIrSD29g5iYEBR0VURY3T08OvuC1pfX0fnz58lNzcPioiIane/O6sVjve1oCBP9P3iZd4PX9/u7f5eXXh/8/KuiASAG65qamooICD4ukuRmuPL63m/udM8H+erV/OoZ8/e5OHhadD78uv5HOBjyg8+J3gf3N27XRefPuXB50F2dobYB5XK/rr4q6oqxchNft+ePaPEPmt+P1929PY27Kbwhv4+3i4nJ0skZ3z+1dfXqz/PIeTi4tq8jb7nfWvtnS/tfd40cetznDmunJxMsR0/+DjzvvFngC/h8jl55Uq66Ivp5eVDPXpEisvb/P5I4gDA5kyaNIk+/fRT8aWfmJhI99xzj/oLtog2bdokOzSjcCXFlVNYWIT4cuf/JyWdIk9Pr+b+PJoKUKXKI767UFsVlD7bdevmRcHBPSgj46JI6MLDe4v1XNlcuHCWeKCAdoWbnp5KvXr1ax7hxwlNSkqSiLN1YsAVMpeLPp3HmyrnRnJyunErZnvbcQLFCRzvU1MyUEfJyWdEEsb7agh+7cWLyer97du8b1yZ87revQc03yid40lLO69e17+5tefKlQxxU3s+ntqJiD74dampyeryCmxOoquqKsRxHjjwphajdvUpDy7LlJRz4pzSHANOsjhp0nB1dROv5fOABz2EhvYUyQkfAz7/unXzFEmJvgz5fby/Fy+eF8mT5lzj+FJTz1HfvoObj6m+57229s4XfT5v+h5nzbnB54umzHm0dFraRerTZ5D4P78+LKyXev+qRBKn+X28L7icCgA2x8XFRbTAhYaG0sSJE2nBggW0a9eu5udLSkroT3/6k/pL2VP9Re5E48aNo1OnTrX4Hbz90KFDRSXj4eEhWvM0PvvsM/EcV9g+Pj70hz/8Qf2l33X9y7iVRFOhMK7A+FFd3bVTQfD7ao8I5ffnSlXTgsK4suXKVbsvFb+OK2hduOIcNGgoRUUNaPf9uSLVp1JubzuuSLnTviZGroi58jemTyAnOTzSVTs55X3iwRDc2qJRUVEmjpf25TpueSspKRLJSnuDNFrjWLlVSLsV1M5OJc5BbtXR0Lc8OKnlGLSTWE6ydF3259eHhPRsbl3i1zWdf1UG7YMhv6+kpFAkWtp/LHB83HrFsXdEe+eLPp83fY8zj3blxEw7aefzR5Psa+OWPj4/NEpLS5DEAYBtS0tLoy+++EIka4y/fG+77Ta6cuUK7d+/X7TUjRgxgkaPHi1a69ju3btp6tSpNG3aNPH84cOHaeTIkc2/ky+Z8OXZ8+fP0969eyk1NZXmzZvXZfvAlUnrDudcAWpX3l31vq1x8sItQBocl5eXr2iha7pcWS/Wu7t7tDnSUdf+tKa5/NjeaEl9tuOKv/W+cKXKlbChiosL1Pt2/WVQ3l9OPLTxJb7WuKI3BlfuHh4tWw15H/r3H9IiUdS3PLiFlUfv6kNXefH/jZ0gV5/fx61VfKmxNd6Hiopyo96X6XO+6PN50/84l4rz73rXn/+csPL2mvcpKyvG5VQAsD2ctDk4OIiETdPZed26deLngQMH6Ndff1VXEqXNrQ5r1qyh+Ph42rp1Kz344IP0yiuv0L333ktLly5t/p1DhgxpXr7vvvual3v16kXvvvuuSATLy8tFq50101zu4uOqaUkJD48USQb3M+I+T9zyEBQU1qHpKri1xN9fv1Y4fbZrjZMpQwd28PacaOhqReF1/JzmuHALl6NjjkiWNK1d3GeKpzAxBicfut5XF33Kg5MOQ1sDTYn3V/2v2Adt/Jk29FK0NmPPF130Oc68H+1NUK3Bny1upeUElhNEvjyMJA4AbM748ePp/fffp8rKSvrggw8oKSmJFi9eLJ47fvy4aElzd2/5Vz5XwNyixo4dO0YPP/xwm78/ISGB/vrXv9LRo0fVX+BXRcXCMjIyaMCA9i8TWjKu/PkyXusBC1yB8aNpAEE+XbiQKC6ZGlPhcp85frT3Wn2304X3Q9/KVaNpQIG9zhZQjkMz4IDxOaFptdS0+Pn4+BvcB09D03leX+2VB+97V7fkdgTvL7cw8uX7ztKR86Ut7R9nB50tsm3x8elOhYV54nXc8orLqQBgc7g1rHfv3hQdHU3r168XoweXLVsmnuOWEn9/f0pOTm7xSElJoaefflpsw33q2lJRUUFjxoxRV8bdaMuWLXTmzBnas2ePeI6TQ9m4ctYklRq6JkHVZztFR93DSYn2ZThuNeDWJg1OXEJCepCfX/c2b3fFCXPr99bGnd71qbz13U5Xixv3v2o9qlMfXl7eOi/D8nHhylyjsrJcHE9O2jhG7vSuK4Hj48D95250PBgfc74DhS7afcT0LQ83t246+7RxkmMMffdDX9x/UPuyvTa+s0Jr+pzP+p4v+tD3OPPl1dra649zW63APHiCzy++bM9955DEAYDN4wRu+fLloh/csGHDROsZX27lRE/74efX1Gl8+PDh9O233+r8XZzw8WXTVatW0S233EL9+vWjvLyOdbTuTM7OrqJy1lRoxcVFOkcQ6rNddXWl6KOjwa0NPD0GDxLQ4JanpqkYWlagXEm5ul7fp4kHGZw9e0KMjNSFEzxOhrkT+43oux3jZED7fqA8epD7Gxkz0StfLissvNpidC0nPjzVBT+nwf2gnJxc6MKFc3T69DFKTEygixeTxKAQ7QqcR6zysdAeFKGLn1+gen+rqaTkvwkCH3N+nXYyqm95+PsHiCSHy1iDk0G+/GdMHqbvfuiLWy2bjmtBi/V8ObR1KzBr73w25HzRh77HmS+fcwtd0+XhJjxIpa1W0Ka+dj5i3zkxxeVUALB5t956q+jTxoncW2+9RaNGjaK4uDh64403RBKWnZ0tBjPcddddFBsbK5I+HrEaFRVFc+fOFV+o3NrGLXXh4eHii5Z/z8KFC0VL3EsvvdQlcXOlxBUr/9QeUcfLPFKO++LwyEjty0NcSfEoN65Quc8ft/7oqrj02Y7/z4nDlSsl1yrBGtEPiFsXNLhfFbfA8BQMXPnwJTBNRcb9e1rjS7EODvZt9u/q7FY4xlN68KCDhoZGdWx8Sa3huvuDcmXc1P9KodLSpgEujo5chdqJ99F0dOfjzfPb8ShV3lfNFBn8+7SnrOD34+MaERErKmwuL04kuQLnY8pTSjCOgX9He/0H+Y+OqKiB6rLIpKKi/Gv92RRx+U27/PUtD14fGdlPzBPHyQInPFyunHyWlBSo98VRlJW+558++2HI+cy/i6dsycnJUO9vgdgvLgOe+03X5dD2zmd9zhdD4tP3OPMxDQkJp8uXU5vPF75MeqP+iNwK2ZzkKQAANmTBggXK1KlTr1v/+eefK+pKQMnIyFBKS0uVRx55RPHz8xPrAgMDlXvvvVc8p7F9xw5l8ODB4nkPDw9l+vTpzc/9+9//VkJCQsRzI0eOVHbt2sV/jisJCQni+QMHDoj/FxUVif9/8sknipOTUxfveecqKSlSkpNPm/Q9GxrqlcuXUzptO5nOnTupqJOB69aXlpYoFy6clRCR7TLH8+Xs2RNKRUW5zueys9MV9R+OYtmO/7lh6gkAANAK9+vhS2T9+kWb7D25tY9btdprldJ3O5m45YVbYVq3/mjuKGDu9+y0JuZ4viQmnqCIiL7N06hwKzdP9sv9efmSO09MzZDEAQCAQbjjOE9Syokc93/jhMPQkZy2ji+tccLGl8X+Oy1L0zISONvFt67jS7aaSYC5HyVfnuV1fGmbL8VqbrvFkMQBAAAAWCCMTgUAAACwQEjiAAAAACwQkjgAAAAAC4QkDgAAAMACIYkDAAAAsEBI4gAAAAAsEJI4AAAAAAuEJA4AAADAAiGJAwAAALBASOIAAAAALBCSOAAAAAALhCQOAAAAwAIhiQMAAACwQEjiAAAAACzQ/wFe/ePqufG8pwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ap = Image(r\"D:\\practice\\博客\\目标检测\\img\\RP.png\")\n",
    "ap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2fb51176-3a12-4b6f-a48b-87b534575073",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Average Precision Calculation\n",
    "def compute_ap(recall,precision):\n",
    "    \"\"\" \n",
    "    Compute the average precision, given the recall and precision curves.\n",
    "    # Arguments\n",
    "        recall:    The recall curve (list).\n",
    "        precision: The precision curve (list).\n",
    "    # Returns\n",
    "        The average precision as computed in py-faster-rcnn.\n",
    "    \"\"\"\n",
    "    mrec = np.concatenate(([0.],recall,[1.]))\n",
    "    mpre = np.concatenate(([0.],precision,[0.]))   # 参考上图的RP曲线，看两端\n",
    "    \n",
    "    # compute the precision envelope\n",
    "    for i in range(mpre.size-1,0,-1):\n",
    "        mpre[i-1]=np.maximum(mpre[i-1],mpre[i])\n",
    "    \n",
    "    i=np.where(mrec[1:]!=mrec[:-1])[0]      # where X axis (recall) changes value\n",
    "\n",
    "    ap = np.sum((mrec[i+1]-mrec[i])*mpre[i+1])\n",
    "    return ap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "a1c84eae-79df-46bf-abee-a5689755a075",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.        , 1.        , 0.5       , 0.66666667, 0.        ])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p = np.concatenate(([0.],[1.0,0.5,2/3],[0.]))\n",
    "p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "664929e2-767a-4abe-b28e-7ef512802c6f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.        , 0.33333333, 0.33333333, 0.66666667, 1.        ])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r  = np.concatenate(([0.],[1/3,1/3,2/3],[1.]))\n",
    "r"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "17a9fa70-cd8c-4355-9258-f16dec7d39ff",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "i:4\n",
      "p:0.6666666666666666\n",
      "i:3\n",
      "p:0.6666666666666666\n",
      "i:2\n",
      "p:1.0\n",
      "i:1\n",
      "p:1.0\n",
      "p:[1.         1.         0.66666667 0.66666667 0.        ]\n",
      "i:[0 2 3]\n"
     ]
    }
   ],
   "source": [
    "for i in range(p.size-1,0,-1):\n",
    "    print(\"i:{}\".format(i))\n",
    "    p[i-1]=np.maximum(p[i-1],p[i])\n",
    "    print(\"p:{}\" .format(p[i-1]))\n",
    "i = np.where(r[1:]!=r[:-1])[0]\n",
    "print(\"p:{}\".format(p))\n",
    "print(\"i:{}\".format(i))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "37c55237-e77e-4a12-ab04-a811a88c2a30",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.554779"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pre =[1.0,0.5,0.667]\n",
    "rec = [0.33,0.33,0.667]\n",
    "ap = compute_ap(rec,pre)\n",
    "ap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "12d07b48-0413-4b5f-a224-df32c121c77e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# get ground truth box with image id and label\n",
    "def get_real_annotations(table):\n",
    "    res = dict()\n",
    "    ids = table['ImageID'].value.astype(np.str)\n",
    "    labels = table['LabelName'].values.astype(np.str)\n",
    "    xmin = table['XMin'].values.astype(np.float32)\n",
    "    xmax = table['XMax'].values.astype(np.float32)\n",
    "    ymin = table['YMin'].values.astype(np.float32)\n",
    "    ymax = table['YMax'].values.astype(np.float32)\n",
    "    \n",
    "    for i in range(len(ids)):\n",
    "        id = ids[i]\n",
    "        label = labels[i]\n",
    "        if id not in res:\n",
    "            res[id] = dict()\n",
    "        if label not in res[id]:\n",
    "            res[id][label] = []\n",
    "        box = [xmin[i], ymin[i], xmax[i], ymax[i]]\n",
    "        res[id][label].append(box)\n",
    "    \n",
    "    return res\n",
    "\n",
    "def get_detections(table):\n",
    "    res = dict()\n",
    "    ids = table['ImageID'].values.astype(np.str)\n",
    "    labels = table['LabelName'].values.astype(np.str)\n",
    "    scores = table['Conf'].values.astype(np.float32)\n",
    "    xmin = table['XMin'].values.astype(np.float32)\n",
    "    xmax = table['XMax'].values.astype(np.float32)\n",
    "    ymin = table['YMin'].values.astype(np.float32)\n",
    "    ymax = table['YMax'].values.astype(np.float32)\n",
    "\n",
    "    for i in range(len(ids)):\n",
    "        ida = ids[i]\n",
    "        label = labels[i]\n",
    "        if id not in res:\n",
    "            res[id] = dict()\n",
    "        if label not in res[id]:\n",
    "            res[id][label] = []\n",
    "        box = [xmin[i], ymin[i], xmax[i], ymax[i], scores[i]]\n",
    "        res[ida][label].append(box)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "dff6c413-26a6-4dfe-835b-938f3208c086",
   "metadata": {},
   "outputs": [],
   "source": [
    "# mean Average Precision Calculation\n",
    "def mean_average_precision_for_boxes(ann,pred,iou_threshould=0.5, exclude_not_in_annotations=False, verbose=True):\n",
    "    \"\"\"\n",
    "    :param ann: path to CSV-file with annotations or numpy array of shape (N, 6)\n",
    "    :param pred: path to CSV-file with predictions (detections) or numpy array of shape (N, 7)\n",
    "    :param iou_threshold: IoU between boxes which count as 'match'. Default: 0.5\n",
    "    :param exclude_not_in_annotations: exclude image IDs which are not exist in annotations. Default: False\n",
    "    :param verbose: print detailed run info. Default: True\n",
    "    :return: tuple, where first value is mAP and second values is dict with AP for each class.\n",
    "    \"\"\"\n",
    "    valid = pd.DataFrame(ann, columns=['ImageID', 'LabelName', 'XMin', 'XMax', 'YMin', 'YMax'])\n",
    "    preds = pd.DataFrame(pred, columns=['ImageID', 'LabelName', 'Conf', 'XMin', 'XMax', 'YMin', 'YMax'])\n",
    "    ann_unique = valid['ImageID'].unique()\n",
    "    preds_unique = preds['ImageID'].unique()\n",
    "    \n",
    "    if verbose:\n",
    "        print('Number of files in annotations: {}'.format(len(ann_unique)))\n",
    "        print('Number of files in predictions: {}'.format(len(preds_unique)))\n",
    "    \n",
    "    # Exclude files not in annotations!\n",
    "    if exclude_not_in_annotations:\n",
    "        preds = preds[preds['ImageID'].isin(ann_unique)]\n",
    "        preds_unique = preds['ImageID'].unique()\n",
    "        if verbose:\n",
    "            print('Number of files in detection after reduction: {}'.format(len(preds_unique)))\n",
    "\n",
    "    unique_classes = valid['LabelName'].unique().astype(np.str)\n",
    "    if verbose:\n",
    "        print('Unique classes: {}'.format(len(unique_classes)))\n",
    "        \n",
    "    all_detections = get_detections(preds)\n",
    "    all_annotations = get_real_annotations(valid)\n",
    "    if verbose:\n",
    "        print('Detections length: {}'.format(len(all_detections)))\n",
    "        print('Annotations length: {}'.format(len(all_annotations)))\n",
    "    \n",
    "    average_precisions = {}\n",
    "    for zz, label in enumerate(sorted(unique_classes)):\n",
    "        # Negative class\n",
    "        if str(label) == 'nan':\n",
    "            continue\n",
    "\n",
    "        false_positives = []\n",
    "        true_positives = []\n",
    "        scores = []\n",
    "        num_annotations = 0.0\n",
    "\n",
    "        for i in range(len(ann_unique)):\n",
    "            detections = []\n",
    "            annotations = []\n",
    "            id = ann_unique[i]\n",
    "            if id in all_detections:\n",
    "                if label in all_detections[id]:\n",
    "                    detections = all_detections[id][label]\n",
    "            if id in all_annotations:\n",
    "                if label in all_annotations[id]:\n",
    "                    annotations = all_annotations[id][label]\n",
    "\n",
    "            if len(detections) == 0 and len(annotations) == 0:\n",
    "                continue\n",
    "\n",
    "            num_annotations += len(annotations)\n",
    "            scr, fp, tp = cehck_if_true_or_false_positive(annotations, detections, iou_threshold)\n",
    "\n",
    "            scores += scr\n",
    "            false_positives += fp\n",
    "            true_positives += tp\n",
    "\n",
    "        if num_annotations == 0:\n",
    "            average_precisions[label] = 0, 0\n",
    "            continue\n",
    "\n",
    "        false_positives = np.array(false_positives)\n",
    "        true_positives = np.array(true_positives)\n",
    "        scores = np.array(scores)\n",
    "\n",
    "        # sort by score\n",
    "        indices = np.argsort(-scores)\n",
    "        false_positives = false_positives[indices]\n",
    "        true_positives = true_positives[indices]\n",
    "\n",
    "        # compute false positives and true positives\n",
    "        false_positives = np.cumsum(false_positives)\n",
    "        true_positives = np.cumsum(true_positives)\n",
    "\n",
    "        # compute recall and precision\n",
    "        recall = true_positives / num_annotations\n",
    "        precision = true_positives / np.maximum(true_positives + false_positives, np.finfo(np.float64).eps)\n",
    "\n",
    "        # compute average precision\n",
    "        average_precision = compute_ap(recall, precision)\n",
    "        average_precisions[label] = average_precision, num_annotations, precision, recall\n",
    "        if verbose:\n",
    "            s1 = \"{:30s} | {:.6f} | {:7d}\".format(label, average_precision, int(num_annotations))\n",
    "            print(s1)\n",
    "        present_classes = 0\n",
    "        precision = 0\n",
    "        for label, (average_precision, num_annotations, _, _) in average_precisions.items():\n",
    "            if num_annotations > 0:\n",
    "                present_classes += 1\n",
    "                precision += average_precision\n",
    "        mean_ap = precision / present_classes\n",
    "        if verbose:\n",
    "            print('mAP: {:.6f}'.format(mean_ap))\n",
    "        return mean_ap, average_precisions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8c92972c-cff5-4c6b-b7b3-61340f96cde9",
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "置信度的作用\n",
    "1.在check_if_true_or_false_positive()之前，检测框需要靠置信度进行排序\n",
    "2.在计算Average Precision之前，列表true_positives和列表false_positives需要靠置信度进行排序\n",
    "\n",
    "更低的置信度阈值，map会更高\n",
    "\"\"\""
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8 (XPython)",
   "language": "python",
   "name": "xpython"
  },
  "language_info": {
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "version": "3.8.10"
  },
  "toc-autonumbering": false,
  "toc-showcode": true,
  "toc-showmarkdowntxt": true,
  "toc-showtags": true
 },
 "nbformat": 4,
 "nbformat_minor": 5
}