Error_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "09731d05652c476dbf4581de5acb2730",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import mmcv\n",
    "import numpy as np\n",
    "import cv2\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "CLASSES = ('tennis-court', 'container-crane', 'storage-tank', 'baseball-diamond', 'plane','ground-track-field',\n",
    "               'helicopter', 'airport','harbor', 'ship', 'large-vehicle', 'swimming-pool', 'soccer-ball-field',\n",
    "               'roundabout', 'basketball-court', 'bridge', 'small-vehicle', 'helipad')\n",
    "ann = mmcv.load(\"./result/post_nms.pkl\")\n",
    "ret = {}\n",
    "names = ann.keys()\n",
    "for name in tqdm(names):\n",
    "    ret[name] = {}\n",
    "    result = ann[name]\n",
    "    segs = result['vis']\n",
    "    bboxes = result[\"bbox\"]\n",
    "    objects = []\n",
    "    for cls in range(18):\n",
    "        curr_class = CLASSES[cls]\n",
    "        bbox_cls = np.array(bboxes[cls])\n",
    "        seg_cls = segs[cls]\n",
    "        if( len(bbox_cls) > 0):\n",
    "            for box, rle in zip(bbox_cls, seg_cls):\n",
    "                obj = {}\n",
    "                rle = np.array(rle)\n",
    "                _,_,_,_, score = box\n",
    "                poly = cv2.minAreaRect(rle)\n",
    "                obj[\"poly\"] = poly\n",
    "                obj[\"category\"] = curr_class\n",
    "                obj[\"score\"] = score\n",
    "                objects.append(obj)\n",
    "    ret[name][\"objects\"] = objects\n",
    "mmcv.dump(ret, \"./result/eval_analy.pkl\") \n",
    "                "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loading annotations into memory...\n",
      "Done (t=0.81s)\n",
      "creating index...\n",
      "index created!\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d083f9c5f226426c9b291719d34315ff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# get all images containing given categories, select one at random\n",
    "import sys\n",
    "sys.path.append('D:\\API\\cocoapi\\PythonAPI')  # 你下载的 cocoapi 所在路径\n",
    "import skimage.io as io\n",
    "from pycocotools.coco import COCO   # 载入 cocoz\n",
    "%matplotlib inline\n",
    "coco=COCO(\"/home/xfr/rssid/data/annotation/annos_rscup_val.json\")\n",
    "imgIds = coco.getImgIds()\n",
    "# imgIds = coco.getImgIds(imgIds=[335328])\n",
    "ret = {}\n",
    "for imgid in tqdm(imgIds):\n",
    "    info = coco.loadImgs([imgid])[0]\n",
    "    name = info['file_name'].split(\".\")[0]\n",
    "    ret[name] = {}\n",
    "    annIds = coco.getAnnIds(imgIds=[imgid], iscrowd=None)\n",
    "    anns = coco.loadAnns(annIds)\n",
    "    objects = []\n",
    "    for ann in anns:\n",
    "        obj = {}\n",
    "        bbox = ann['bbox']\n",
    "        poly = np.array(ann[\"segmentation\"][0], np.int32).reshape(4,2)\n",
    "        poly = cv2.minAreaRect(poly)\n",
    "        obj[\"poly\"] = poly\n",
    "        obj[\"category\"] = CLASSES[ann[\"category_id\"]]\n",
    "        objects.append(obj)\n",
    "    ret[name][\"objects\"] = objects\n",
    "mmcv.dump(ret, \"./result/gt_analy.pkl\")  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/xfr/git_mm/mmdetection'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import json\n",
    "import pylab as pl\n",
    "import random\n",
    "import numpy as np\n",
    "import cv2\n",
    "import copy\n",
    "import os\n",
    "\n",
    "CLASSES = ('tennis-court', 'container-crane', 'storage-tank', 'baseball-diamond', 'plane','ground-track-field',\n",
    "               'helicopter', 'airport','harbor', 'ship', 'large-vehicle', 'swimming-pool', 'soccer-ball-field',\n",
    "               'roundabout', 'basketball-court', 'bridge', 'small-vehicle', 'helipad')\n",
    "\n",
    "def get_original_name(name):\n",
    "    datadir = \"/home/xfr/rssid/data/val/images/\"\n",
    "    files = os.listdir(datadir)\n",
    "    img_name = \"\"\n",
    "    for file in files:\n",
    "        img_name = file.split(\".\")[0]\n",
    "        if(img_name == name):\n",
    "            return file\n",
    "\n",
    "def draw_all(annos, imgid, color=(0,1,0), datadir=None, have_label=True):\n",
    "    img = annos['imgs'][imgid]\n",
    "    filename = get_original_name(imgid)\n",
    "    imgdata = cv2.imread(\"/home/xfr/rssid/data/val/images/\"+filename)\n",
    "    for obj in img['objects']:\n",
    "        poly = obj['poly']\n",
    "        poly = cv2.boxPoints(poly)\n",
    "        loc = poly.flatten()\n",
    "        xmin = min(loc[0::2])\n",
    "        ymin = min(loc[1::2])\n",
    "        cv2.polylines(imgdata, np.array([poly], np.int32), 1, (0, 255, 0), 1)\n",
    "        ss = obj['category']\n",
    "        if 'correct_catelog' in obj:\n",
    "            ss = ss+'->'+obj['correct_catelog']\n",
    "        if have_label:\n",
    "            pass\n",
    "            cv2.putText(imgdata, ss, (int(xmin),int(ymin)), 0, 0.5, color, 1)\n",
    "    return imgdata\n",
    "\n",
    "def rect_cross(rect1, rect2):\n",
    "    rect = [max(rect1[0], rect2[0]),\n",
    "            max(rect1[1], rect2[1]),\n",
    "            min(rect1[2], rect2[2]),\n",
    "            min(rect1[3], rect2[3])]\n",
    "    rect[2] = max(rect[2], rect[0])\n",
    "    rect[3] = max(rect[3], rect[1])\n",
    "    return rect\n",
    "\n",
    "def rect_area(rect):\n",
    "    return float(max(0.0, (rect[2]-rect[0])*(rect[3]-rect[1])))\n",
    "\n",
    "def calc_cover(rect1, rect2):\n",
    "    crect = rect_cross(rect1, rect2)\n",
    "    return rect_area(crect) / rect_area(rect2)\n",
    "\n",
    "def calc_iou(r1, r2):\n",
    "    area_r1 = r1[1][0] * r1[1][1]\n",
    "    area_r2 = r2[1][0] * r2[1][1]\n",
    "    int_pts = cv2.rotatedRectangleIntersection(r1, r2)[1]\n",
    "    inter =0.0\n",
    "    if int_pts is not None:\n",
    "        order_pts = cv2.convexHull(int_pts, returnPoints=True)\n",
    "        int_area = cv2.contourArea(order_pts)\n",
    "        inter = int_area * 1.0 / (area_r1 + area_r2 - int_area + 0.0000001)\n",
    "    return inter\n",
    "\n",
    "def box_long_size(box):\n",
    "    return max(box[1][0],box[1][1])\n",
    "\n",
    "def eval_annos(annos_gd, annos_rt, iou=0.5, imgids=None, check_type=True, types=None, minscore=40, minboxsize=0, maxboxsize=400, match_same=True):\n",
    "    ac_n, ac_c = 0,0\n",
    "    rc_n, rc_c = 0,0\n",
    "    if imgids==None:\n",
    "        imgids = annos_rt.keys()\n",
    "    if types!=None:\n",
    "        types = { t:0 for t in types }\n",
    "    miss = {\"imgs\":{}}\n",
    "    wrong = {\"imgs\":{}}\n",
    "    right = {\"imgs\":{}}\n",
    "    \n",
    "    for imgid in tqdm(imgids):\n",
    "        v = annos_rt[imgid]\n",
    "        vg = annos_gd[imgid]\n",
    "        objs_g = vg[\"objects\"]\n",
    "        objs_r = v[\"objects\"]\n",
    "        bg = [obj['poly'] for obj in objs_g]\n",
    "        br = [obj['poly'] for obj in objs_r]\n",
    "        \n",
    "        match_g = [-1]*len(bg)\n",
    "        match_r = [-1]*len(br)\n",
    "        if types!=None:\n",
    "            for i in range(len(match_g)):\n",
    "                if objs_g[i]['category'] not in types:\n",
    "                    match_g[i] = -2\n",
    "            for i in range(len(match_r)):\n",
    "                if objs_r[i]['category'] not in types:\n",
    "                    match_r[i] = -2\n",
    "        for i in range(len(match_r)):\n",
    "            if 'score' in objs_r[i].keys() and objs_r[i]['score']<minscore:\n",
    "                match_r[i] = -2\n",
    "        matches = []\n",
    "        for i,boxg in enumerate(bg):\n",
    "            for j,boxr in enumerate(br):\n",
    "                if match_g[i] == -2 or match_r[j] == -2:\n",
    "                    continue\n",
    "                if match_same and objs_g[i]['category'] != objs_r[j]['category']: continue\n",
    "                tiou = calc_iou(boxg, boxr)\n",
    "                if tiou>iou:\n",
    "                    matches.append((tiou, i, j))\n",
    "        matches = sorted(matches, key=lambda x:-x[0])\n",
    "        for tiou, i, j in matches:\n",
    "            if match_g[i] == -1 and match_r[j] == -1:\n",
    "                match_g[i] = j\n",
    "                match_r[j] = i\n",
    "                \n",
    "        for i in range(len(match_g)):\n",
    "            boxsize = box_long_size(objs_g[i]['poly'])\n",
    "            erase = False\n",
    "            if not (boxsize>=minboxsize and boxsize<maxboxsize):\n",
    "                erase = True\n",
    "            #if types!=None and not types.has_key(objs_g[i]['category']):\n",
    "            #    erase = True\n",
    "            if erase:\n",
    "                if match_g[i] >= 0:\n",
    "                    match_r[match_g[i]] = -2\n",
    "                match_g[i] = -2\n",
    "        \n",
    "        for i in range(len(match_r)):\n",
    "            boxsize = box_long_size(objs_r[i]['poly'])\n",
    "            if match_r[i] != -1: continue\n",
    "            if not (boxsize>=minboxsize and boxsize<maxboxsize):\n",
    "                match_r[i] = -2\n",
    "                    \n",
    "        miss[\"imgs\"][imgid] = {\"objects\":[]}\n",
    "        wrong[\"imgs\"][imgid] = {\"objects\":[]}\n",
    "        right[\"imgs\"][imgid] = {\"objects\":[]}\n",
    "        miss_objs = miss[\"imgs\"][imgid][\"objects\"]\n",
    "        wrong_objs = wrong[\"imgs\"][imgid][\"objects\"]\n",
    "        right_objs = right[\"imgs\"][imgid][\"objects\"]\n",
    "        \n",
    "        tt = 0\n",
    "        for i in range(len(match_g)):\n",
    "            if match_g[i] == -1:\n",
    "                miss_objs.append(objs_g[i])\n",
    "        for i in range(len(match_r)):\n",
    "            if match_r[i] == -1:\n",
    "                obj = copy.deepcopy(objs_r[i])\n",
    "                obj['correct_catelog'] = 'none'\n",
    "                wrong_objs.append(obj)\n",
    "            elif match_r[i] != -2:\n",
    "                j = match_r[i]\n",
    "                obj = copy.deepcopy(objs_r[i])\n",
    "                if not check_type or objs_g[j]['category'] == objs_r[i]['category']:\n",
    "                    right_objs.append(objs_r[i])\n",
    "                    tt+=1\n",
    "                else:\n",
    "                    obj['correct_catelog'] = objs_g[j]['category']\n",
    "                    wrong_objs.append(obj)\n",
    "                    \n",
    "        \n",
    "        rc_n += len(objs_g) - match_g.count(-2)\n",
    "        ac_n += len(objs_r) - match_r.count(-2)\n",
    "        ac_c += tt\n",
    "        rc_c += tt\n",
    "    print(rc_n,rc_c)\n",
    "    print(ac_n, ac_c)\n",
    "    if types==None:\n",
    "        styps = \"all\"\n",
    "    elif len(types)==1:\n",
    "        styps = list(types.keys())[0]\n",
    "    elif not check_type or len(types)==0:\n",
    "        styps = \"none\"\n",
    "    else:\n",
    "        styps = \"[%s, ...total %s...]\"%(list(types.keys())[0], len(types))\n",
    "    p = 1 if ac_n==0 else ac_c*1.0/ac_n\n",
    "    r = 1 if rc_n==0 else rc_c*1.0/rc_n\n",
    "    f = (2*p*r)/(p+r)\n",
    "    report = \"iou:%s, size:[%s,%s), types:%s, accuracy:%s, recall:%s\"% (\n",
    "        iou, minboxsize, maxboxsize, styps, 1 if ac_n==0 else ac_c*1.0/ac_n, 1 if rc_n==0 else rc_c*1.0/rc_n)\n",
    "   \n",
    "    summury = {\n",
    "        \"iou\":iou,\n",
    "        \"accuracy\":1 if ac_n==0 else ac_c*1.0/ac_n,\n",
    "        \"recall\":1 if rc_n==0 else rc_c*1.0/rc_n,\n",
    "        \"f\": f,\n",
    "        \"miss\":miss,\n",
    "        \"wrong\":wrong,\n",
    "        \"right\":right,\n",
    "        \"report\":report\n",
    "    }\n",
    "    return summury\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ec48c113525746da95659e534e31254c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "81048 52725\n",
      "148945 52725\n"
     ]
    }
   ],
   "source": [
    "annos_gd = mmcv.load(\"./result/gt_analy.pkl\")\n",
    "annos_rt = mmcv.load(\"./result/eval_analy.pkl\")\n",
    "summary = eval_annos(annos_gd, annos_rt, iou=0.5, types=CLASSES, minscore=0.05, minboxsize=0, maxboxsize=40000, match_same=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "54e2d20aa61e4ea5a0bbbc636eb49d90",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "P1314\n",
      "P1275\n",
      "P1878\n",
      "P1542\n",
      "P9705\n",
      "P0199\n",
      "P3575\n",
      "P4151\n",
      "P1456\n",
      "P1854\n",
      "P2481\n",
      "P2027\n",
      "P3107\n",
      "P4359\n",
      "P0242\n",
      "P1754\n",
      "P0146\n",
      "P1373\n",
      "P2256\n",
      "P3185\n",
      "P2181\n",
      "P4952\n",
      "P3989\n",
      "P0706\n",
      "P2802\n",
      "P2455\n",
      "P1269\n",
      "P1672\n",
      "P2733\n",
      "P1007\n",
      "P1982\n",
      "P0622\n",
      "P2570\n",
      "P2771\n",
      "P0604\n",
      "P1541\n",
      "P3182\n",
      "P0342\n",
      "P1786\n",
      "P2593\n",
      "P2721\n",
      "P1030\n",
      "P2385\n",
      "P0466\n",
      "P2695\n",
      "P0799\n",
      "P2239\n",
      "P1806\n",
      "P1751\n",
      "P2678\n",
      "P9031\n",
      "P2726\n",
      "P3032\n",
      "P1021\n",
      "P2709\n",
      "P10810\n",
      "P7431\n",
      "P1512\n",
      "P5171\n",
      "P9594\n",
      "P1809\n",
      "P2351\n",
      "P1960\n",
      "P0161\n",
      "P1246\n",
      "P2135\n",
      "P1610\n",
      "P1398\n",
      "P1143\n",
      "P8666\n",
      "P1570\n",
      "P5070\n",
      "P2504\n",
      "P0764\n",
      "P6519\n",
      "P0128\n",
      "P1022\n",
      "P8081\n",
      "P1390\n",
      "P0929\n",
      "P2050\n",
      "P0213\n",
      "P2745\n",
      "P0179\n",
      "P2365\n",
      "P5789\n",
      "P0613\n",
      "P4772\n",
      "P10105\n",
      "P0411\n",
      "P0432\n",
      "P0007\n",
      "P0053\n",
      "P1268\n",
      "P2150\n",
      "P1149\n",
      "P1452\n",
      "P5155\n",
      "P2722\n",
      "P1829\n",
      "P0550\n",
      "P2625\n",
      "P2993\n",
      "P9997\n",
      "P0309\n",
      "P2718\n",
      "P2582\n",
      "P1386\n",
      "P2214\n",
      "P2285\n",
      "P1266\n",
      "P1478\n",
      "P2167\n",
      "P9170\n",
      "P0998\n",
      "P1240\n",
      "P2758\n",
      "P1230\n",
      "P10795\n",
      "P0945\n",
      "P2610\n",
      "P1463\n",
      "P0949\n",
      "P1738\n",
      "P0130\n",
      "P2231\n",
      "P2230\n",
      "P1513\n",
      "P9444\n",
      "P1700\n",
      "P2488\n",
      "P0763\n",
      "P2198\n",
      "P0665\n",
      "P1122\n",
      "P0486\n",
      "P1278\n",
      "P0547\n",
      "P1643\n",
      "P1090\n",
      "P0086\n",
      "P2791\n",
      "P0729\n",
      "P1234\n",
      "P0715\n",
      "P2608\n",
      "P1184\n",
      "P1518\n",
      "P8392\n",
      "P0385\n",
      "P1601\n",
      "P0936\n",
      "P0336\n",
      "P1749\n",
      "P2429\n",
      "P2413\n",
      "P9015\n",
      "P2168\n",
      "P2789\n",
      "P2082\n",
      "P0904\n",
      "P0143\n",
      "P1179\n",
      "P1742\n",
      "P2342\n",
      "P0673\n",
      "P2462\n",
      "P0990\n",
      "P2714\n",
      "P2271\n",
      "P2689\n",
      "P4183\n",
      "P10956\n",
      "P0414\n",
      "P0770\n",
      "P5334\n",
      "P0953\n",
      "P0454\n",
      "P2375\n",
      "P5728\n",
      "P6769\n",
      "P2194\n",
      "P0131\n",
      "P1992\n",
      "P1257\n",
      "P0704\n",
      "P1370\n",
      "P2766\n",
      "P1474\n",
      "P0526\n",
      "P2794\n",
      "P4947\n",
      "P1377\n",
      "P1904\n",
      "P2011\n",
      "P2242\n",
      "P1095\n",
      "P3466\n",
      "P1219\n",
      "P4492\n",
      "P0559\n",
      "P1432\n",
      "P9397\n",
      "P0331\n",
      "P1029\n",
      "P4721\n",
      "P1075\n",
      "P0123\n",
      "P2587\n",
      "P0887\n",
      "P4417\n",
      "P8601\n",
      "P0168\n",
      "P2294\n",
      "P0841\n",
      "P2215\n",
      "P2337\n",
      "P2012\n",
      "P1561\n",
      "P1573\n",
      "P1973\n",
      "P6089\n",
      "P3810\n",
      "P8877\n",
      "P2310\n",
      "P5078\n",
      "P2236\n",
      "P0839\n",
      "P0838\n",
      "P2374\n",
      "P7427\n",
      "P0969\n",
      "P0577\n",
      "P0402\n",
      "P1147\n",
      "P1283\n",
      "P1005\n",
      "P0374\n",
      "P1225\n",
      "P0882\n",
      "P0246\n",
      "P0961\n",
      "P6687\n",
      "P1903\n",
      "P1210\n",
      "P2956\n",
      "P0897\n",
      "P2363\n",
      "P1732\n",
      "P0749\n",
      "P2645\n",
      "P2592\n",
      "P9467\n",
      "P1434\n",
      "P1584\n",
      "P1189\n",
      "P0004\n",
      "P0110\n",
      "P2701\n",
      "P2241\n",
      "P0262\n",
      "P10090\n",
      "P2547\n",
      "P0683\n",
      "P1950\n",
      "P1720\n",
      "P0259\n",
      "P3241\n",
      "P2097\n",
      "P9595\n",
      "P10137\n",
      "P1957\n",
      "P2781\n",
      "P1586\n",
      "P0249\n",
      "P1323\n",
      "P2394\n",
      "P1380\n",
      "P0019\n",
      "P2286\n",
      "P3044\n",
      "P0352\n",
      "P2630\n",
      "P6357\n",
      "P2059\n",
      "P0150\n",
      "P6347\n",
      "P1598\n",
      "P0962\n",
      "P0261\n",
      "P1429\n",
      "P7392\n",
      "P1397\n",
      "P1384\n",
      "P6661\n",
      "P1679\n",
      "P0056\n",
      "P0989\n",
      "P1880\n",
      "P2126\n",
      "P0300\n",
      "P0761\n",
      "P1315\n",
      "P1877\n",
      "P2088\n",
      "P1752\n",
      "P1376\n",
      "P2598\n",
      "P0796\n",
      "P2768\n",
      "P2754\n",
      "P8656\n",
      "P7484\n",
      "P2393\n",
      "P5949\n",
      "P1410\n",
      "P1371\n",
      "P9016\n",
      "P0347\n",
      "P1065\n",
      "P1529\n",
      "P0353\n",
      "P1332\n",
      "P1557\n",
      "P0237\n",
      "P0801\n",
      "P2450\n",
      "P0858\n",
      "P10210\n",
      "P0789\n",
      "P1763\n",
      "P1583\n",
      "P2111\n",
      "P0060\n",
      "P0977\n",
      "P2580\n",
      "P1342\n",
      "P1750\n",
      "P1905\n",
      "P2541\n",
      "P2330\n",
      "P0822\n",
      "P0910\n",
      "P0047\n",
      "P0833\n",
      "P8287\n",
      "P1770\n",
      "P0864\n",
      "P1860\n",
      "P1178\n",
      "P2197\n",
      "P5337\n",
      "P0117\n",
      "P0104\n",
      "P0974\n",
      "P2331\n",
      "P7800\n",
      "P0368\n",
      "P1978\n",
      "P6162\n",
      "P0795\n",
      "P2595\n",
      "P1356\n",
      "P2129\n",
      "P0696\n",
      "P0623\n",
      "P1838\n",
      "P1156\n",
      "P2220\n",
      "P1009\n",
      "P2724\n",
      "P1014\n",
      "P1566\n",
      "P1713\n",
      "P6669\n",
      "P3957\n",
      "P0257\n",
      "P0003\n",
      "P5925\n",
      "P0837\n",
      "P1273\n",
      "P0027\n",
      "P1471\n",
      "P1476\n",
      "P1983\n",
      "P1508\n",
      "P7485\n",
      "P1619\n",
      "P1995\n",
      "P0348\n",
      "P1623\n",
      "P10733\n",
      "P1242\n",
      "P2124\n",
      "\n"
     ]
    }
   ],
   "source": [
    "key_word = \"miss\"\n",
    "for imgid in tqdm(list(summary[key_word]['imgs'].keys())):\n",
    "    if(len(summary[key_word][\"imgs\"][imgid][\"objects\"]) == 0):\n",
    "        continue\n",
    "    img = draw_all(summary[key_word], imgid, color=(0,255,0))\n",
    "    print(imgid)\n",
    "    cv2.imwrite(\"./result/{}/{}.jpg\".format(key_word, imgid), img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "14bcb1ac9653486999c8c975e7518498",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "annos_gd = mmcv.load(\"./result/gt_analy.pkl\")\n",
    "for name in tqdm(annos_gd.keys()):\n",
    "    filename = get_original_name(name)\n",
    "    img = cv2.imread(\"/home/xfr/rssid/data/val/images/\"+filename)\n",
    "    objects = annos_gd[name][\"objects\"]\n",
    "    for obj in objects:\n",
    "        poly = obj[\"poly\"]\n",
    "        poly = cv2.boxPoints(poly)\n",
    "        loc = poly.flatten()\n",
    "        xmin = min(loc[0::2])\n",
    "        ymin = min(loc[1::2])\n",
    "        cv2.polylines(img, np.array([poly], np.int32), 1, (0, 255, 0), 1)\n",
    "        ss = obj['category']\n",
    "        cv2.putText(img, ss, (int(xmin),int(ymin)), 0, 0.5, (0,255,0), 1)\n",
    "    cv2.imwrite(\"./result/{}/{}.jpg\".format(\"right\", imgid), img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "small-vehicle\n",
      "{'none': 17576, 'large-vehicle': 517, 'ship': 81, 'swimming-pool': 1}\n",
      "storage-tank\n",
      "{'none': 1167, 'bridge': 8, 'roundabout': 3}\n",
      "plane\n",
      "{'none': 767, 'helicopter': 18, 'small-vehicle': 1}\n",
      "harbor\n",
      "{'none': 2795, 'ship': 60, 'bridge': 5, 'plane': 1}\n",
      "ship\n",
      "{'none': 6442, 'harbor': 35, 'small-vehicle': 119, 'large-vehicle': 15, 'swimming-pool': 2, 'tennis-court': 1}\n",
      "bridge\n",
      "{'none': 1120, 'harbor': 2, 'ship': 1}\n",
      "tennis-court\n",
      "{'none': 266, 'soccer-ball-field': 3, 'basketball-court': 3}\n",
      "basketball-court\n",
      "{'none': 276, 'tennis-court': 12, 'helicopter': 1, 'soccer-ball-field': 3, 'ground-track-field': 1}\n",
      "baseball-diamond\n",
      "{'none': 259, 'storage-tank': 1, 'tennis-court': 1}\n",
      "ground-track-field\n",
      "{'none': 212, 'soccer-ball-field': 3, 'baseball-diamond': 1}\n",
      "roundabout\n",
      "{'none': 349, 'storage-tank': 3}\n",
      "large-vehicle\n",
      "{'none': 8056, 'small-vehicle': 1742, 'ship': 39, 'swimming-pool': 1}\n",
      "airport\n",
      "{'none': 88}\n",
      "helicopter\n",
      "{'none': 154, 'plane': 9, 'small-vehicle': 2}\n",
      "swimming-pool\n",
      "{'none': 759, 'small-vehicle': 1, 'roundabout': 1, 'soccer-ball-field': 1, 'tennis-court': 2, 'baseball-diamond': 6, 'ship': 1}\n",
      "soccer-ball-field\n",
      "{'none': 204, 'tennis-court': 1, 'basketball-court': 9, 'ground-track-field': 3}\n",
      "container-crane\n",
      "{'none': 56, 'ship': 1}\n",
      "helipad\n",
      "{'none': 1}\n"
     ]
    }
   ],
   "source": [
    "convert = {}\n",
    "for imgid in summary[\"wrong\"][\"imgs\"].keys():\n",
    "    objects = summary[\"wrong\"][\"imgs\"][imgid][\"objects\"]\n",
    "    for obj in objects:\n",
    "        ss = obj['category']\n",
    "        if ss not in convert:\n",
    "            convert[ss] = {}\n",
    "        if 'correct_catelog' in obj:\n",
    "            mm = obj['correct_catelog']\n",
    "            if mm not in convert[ss]:\n",
    "                convert[ss][mm] = 0\n",
    "            convert[ss][mm] += 1\n",
    "for key in convert.keys():\n",
    "    print(key)\n",
    "    print(convert[key])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ship\n",
      "1473\n",
      "bridge\n",
      "188\n",
      "small-vehicle\n",
      "22068\n",
      "airport\n",
      "23\n",
      "swimming-pool\n",
      "278\n",
      "harbor\n",
      "419\n",
      "ground-track-field\n",
      "29\n",
      "large-vehicle\n",
      "579\n",
      "roundabout\n",
      "59\n",
      "storage-tank\n",
      "1025\n",
      "plane\n",
      "259\n",
      "soccer-ball-field\n",
      "36\n",
      "basketball-court\n",
      "31\n",
      "container-crane\n",
      "14\n",
      "tennis-court\n",
      "36\n",
      "helipad\n",
      "2\n",
      "helicopter\n",
      "15\n",
      "baseball-diamond\n",
      "28\n"
     ]
    }
   ],
   "source": [
    "miss = {}\n",
    "for imgid in summary[\"miss\"][\"imgs\"].keys():\n",
    "    objects = summary[\"miss\"][\"imgs\"][imgid][\"objects\"]\n",
    "    for obj in objects:\n",
    "        ss = obj['category']\n",
    "        if ss not in miss:\n",
    "            miss[ss] = 0\n",
    "        miss[ss] += 1\n",
    "for key in miss.keys():\n",
    "    print(key)\n",
    "    print(miss[key])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6a1b6d4ce34645d38c75894b23a7d75f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=593), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "small-vehicle\n",
      "50062\n",
      "tennis-court\n",
      "783\n",
      "bridge\n",
      "546\n",
      "ship\n",
      "13801\n",
      "harbor\n",
      "2443\n",
      "airport\n",
      "104\n",
      "storage-tank\n",
      "3175\n",
      "baseball-diamond\n",
      "273\n",
      "swimming-pool\n",
      "851\n",
      "large-vehicle\n",
      "5371\n",
      "plane\n",
      "2832\n",
      "soccer-ball-field\n",
      "152\n",
      "ground-track-field\n",
      "186\n",
      "basketball-court\n",
      "147\n",
      "roundabout\n",
      "228\n",
      "container-crane\n",
      "14\n",
      "helicopter\n",
      "78\n",
      "helipad\n",
      "2\n"
     ]
    }
   ],
   "source": [
    "annos_gd = mmcv.load(\"./result/gt_analy.pkl\")\n",
    "right={}\n",
    "for name in tqdm(annos_gd.keys()):\n",
    "    objects = annos_gd[name][\"objects\"]\n",
    "    for obj in objects:\n",
    "        cls = obj[\"category\"]\n",
    "        if cls not in right:\n",
    "            right[cls] = 0\n",
    "        right[cls] += 1\n",
    "for key in right.keys():\n",
    "    print(key)\n",
    "    print(right[key])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python xpy3",
   "language": "python",
   "name": "xpy3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}