notebooks/eval_examples.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:17:34|data_loader|INFO]: Init data loader: split=test training=False batch=16...\n",
      "[2022-07-10 14:17:34|data_process|INFO]: Loading data file from ../data/MegaDepth_undistort/data_processed/v2/sift_pts/megadepth_2d3d_q500ov0.35tp3-10_imsz.irat1.5.sift1024.npy\n",
      "[2022-07-10 14:17:36|data_process|INFO]: Loading scene 3D points from ../data/MegaDepth_undistort/data_processed/v2/scene_points3d/test.npy ...\n",
      "[2022-07-10 14:17:44|data_process|INFO]: Done with 3D data loading.\n",
      "[2022-07-10 14:17:44|data_process|INFO]: Fetching scene data for: 53 scenes\n",
      "Feature dir:../data/MegaDepth_undistort/data_processed/v2/desc_cache/SIFT1024\n",
      "[2022-07-10 14:19:00|data_process|INFO]: Finished loading scenes: 49, queries: 7344 im.pts3d:369 #ims(pts3d <20)=13\n",
      "[2022-07-10 14:19:00|data_loader|INFO]: \n",
      "Dataset:megadepth split=test scenes:53 queries: 7344\n",
      "Data processed dir: ../data/MegaDepth_undistort/data_processed/v2\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=coords load_desc=False, merge_p3dm=True\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "\n",
    "from gomatch.data.dataset_loaders import init_data_loader\n",
    "\n",
    "config = dict(\n",
    "    data_root='../data',\n",
    "    dataset='megadepth',\n",
    "    dataset_conf='../configs/datasets.yml', \n",
    "    p2d_type='sift', \n",
    "    p3d_type='coords', \n",
    "    topk=1, \n",
    "    npts=[10, 1024], \n",
    "    outlier_rate=[0, 1],\n",
    ")\n",
    "\n",
    "data_loader = init_data_loader(config, split='test')\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Architcture ablations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:20:28|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>None\n",
      "[2022-07-10 14:20:28|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/arch_ablation/covis1inls1/OracleMatcher.npy\n",
      "[2022-07-10 14:20:29|metrics|INFO]: Query total=7344 evaluated=7344 failed=42\n",
      "[2022-07-10 14:20:29|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=101\n",
      "[2022-07-10 14:20:29|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.98]cm\n",
      "[2022-07-10 14:20:29|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.28/99.21/99.31/99.36 %\n",
      "[2022-07-10 14:20:30|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=63.27 [0.5,1)=31.88 [1,5)=4.22 [5,10)=0.03 [10,20)=0.01 [20,100)=0.02 [100,1000)=0.01\n",
      "[2022-07-10 14:20:30|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[54.58 90.37 94.87 96.38 98.5  98.97 99.34] nsamples=7344\n",
      "[2022-07-10 14:20:30|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt\n",
      "[2022-07-10 14:20:30|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/arch_ablation/covis1inls1/BPnPNet.npy\n",
      "[2022-07-10 14:20:31|metrics|INFO]: Query total=7344 evaluated=7344 failed=17\n",
      "[2022-07-10 14:20:31|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=242\n",
      "[2022-07-10 14:20:31|metrics|INFO]: R_qs=[15.17 31.05 59.78] t_qs=[1.67 3.14 5.31]m/[166.8  313.5  530.73]cm\n",
      "[2022-07-10 14:20:31|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 0.74/1.36/3.13/8.54 %\n",
      "[2022-07-10 14:20:31|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=0.56 [0.5,1)=0.37 [1,5)=1.28 [5,10)=1.58 [10,20)=3.59 [20,100)=28.98 [100,1000)=60.91\n",
      "[2022-07-10 14:20:31|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.34  0.83  1.21  1.59  4.66 11.41 56.94] nsamples=7344\n",
      "[2022-07-10 14:20:31|metrics|INFO]: Matching time (per pair):5.60 ms\n",
      "[2022-07-10 14:20:31|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>../outputs/shared_outputs/release_models/exported_models/GoMatchBVs_best.ckpt\n",
      "[2022-07-10 14:20:31|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/arch_ablation/covis1inls1/GoMatchBVs_best.npy\n",
      "[2022-07-10 14:20:32|metrics|INFO]: Query total=7344 evaluated=7344 failed=14\n",
      "[2022-07-10 14:20:32|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=95\n",
      "[2022-07-10 14:20:32|metrics|INFO]: R_qs=[ 0.36  6.97 29.85] t_qs=[0.03 0.69 3.38]m/[  3.38  69.21 338.29]cm\n",
      "[2022-07-10 14:20:32|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 28.47/38.17/44.59/50.23 %\n",
      "[2022-07-10 14:20:33|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=15.95 [0.5,1)=11.43 [1,5)=13.03 [5,10)=4.86 [10,20)=4.93 [20,100)=14.56 [100,1000)=34.04\n",
      "[2022-07-10 14:20:33|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[10.3  29.08 34.79 37.87 46.46 52.28 75.49] nsamples=7344\n",
      "[2022-07-10 14:20:33|metrics|INFO]: Matching time (per pair):39.13 ms\n",
      "[2022-07-10 14:20:33|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt\n",
      "[2022-07-10 14:20:34|evaluator|INFO]: Load model from ../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt\n",
      "[2022-07-10 14:20:34|evaluator|INFO]: matcher=OTMatcherCls ep=36 step=84960\n",
      "[2022-07-10 14:20:34|evaluator|INFO]: Load state dict: <All keys matched successfully>\n",
      "[2022-07-10 14:20:34|evaluator|INFO]: Init matcher(p3d_type=coords, share_kp2d_enc=None, att_layers=['self', 'cross', 'self'])\n",
      "[2022-07-10 14:20:35|evaluator|INFO]: >>>Start evaluation..\n",
      "[2022-07-10 14:20:35|evaluator|INFO]: \n",
      "Dataset:megadepth split=test scenes:53 queries: 7344\n",
      "Data processed dir: ../data/MegaDepth_undistort/data_processed/v2\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=coords load_desc=False, merge_p3dm=True\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n",
      "[2022-07-10 14:20:35|evaluator|INFO]: Model: OTMatcherCls sc_thres=0.5 ransac_thres=0.001 merge_before_match=True\n",
      "100%|███████████████████████████████████████████████████████████████████████████████████████| 459/459 [09:42<00:00,  1.27s/it]\n",
      "[2022-07-10 14:30:18|metrics|INFO]: Query total=7344 evaluated=7344 failed=26\n",
      "[2022-07-10 14:30:18|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=78\n",
      "[2022-07-10 14:30:18|metrics|INFO]: R_qs=[ 4.09 23.56 63.21] t_qs=[0.37 2.53 5.93]m/[ 36.79 253.36 593.05]cm\n",
      "[2022-07-10 14:30:18|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 13.17/19.42/25.10/30.95 %\n",
      "[2022-07-10 14:30:18|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=7.20 [0.5,1)=5.66 [1,5)=9.29 [5,10)=4.90 [10,20)=5.52 [20,100)=19.61 [100,1000)=45.57\n",
      "[2022-07-10 14:30:18|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 3.81 13.54 17.46 19.92 28.01 34.36 65.04] nsamples=7344\n",
      "[2022-07-10 14:30:18|metrics|INFO]: Matching time (per pair):48.51 ms\n",
      "[2022-07-10 14:30:18|evaluator|INFO]: Evaluation finished, total runtime: 583.62s.\n",
      "[2022-07-10 14:30:19|gomatch_eval|INFO]: Cache results to : ../outputs/benchmark_cache_release/arch_ablation/covis1inls1/GoMatchCoords_best.npy\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from gomatch_eval.ablate_models import ablate_archs\n",
    "\n",
    "# Define models\n",
    "gomatchbvs = '../outputs/shared_outputs/release_models/exported_models/GoMatchBVs_best.ckpt'\n",
    "gomatchcoords = '../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt'\n",
    "bpnpnet = '../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt'\n",
    "\n",
    "# Define output dir\n",
    "dataset = data_loader.dataset\n",
    "tag = f'covis{dataset.topk}inls{dataset.inls2d_thres}'\n",
    "if dataset.normalized_thres:\n",
    "    tag += 'normth'\n",
    "cache_dir = os.path.join('../outputs/benchmark_cache_release/arch_ablation', tag)\n",
    "\n",
    "# Establish oracle upperbound\n",
    "ablate_archs(\n",
    "    None, data_loader, cache_dir, \n",
    "    oracle=True, debug=False, overwrite=False\n",
    ")\n",
    "\n",
    "# Eval ablation models\n",
    "model_ckpts = [bpnpnet, gomatchbvs, gomatchcoords]\n",
    "ablate_archs(\n",
    "    model_ckpts, data_loader, cache_dir, \n",
    "    oracle=False, debug=False, overwrite=False\n",
    ")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Max outlier rate ablations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:30:47|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      "\n",
      "\n",
      ">>>>>None\n",
      "[2022-07-10 14:30:47|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/orate_ablation/covis1inls1/OracleMatcher.npy\n",
      "[2022-07-10 14:30:51|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 1.0] \n",
      "[2022-07-10 14:30:51|metrics|INFO]: Query total=7344 evaluated=7344 failed=42\n",
      "[2022-07-10 14:30:51|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=101\n",
      "[2022-07-10 14:30:51|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.98]cm\n",
      "[2022-07-10 14:30:51|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.28/99.21/99.31/99.36 %\n",
      "[2022-07-10 14:30:52|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=63.27 [0.5,1)=31.88 [1,5)=4.22 [5,10)=0.03 [10,20)=0.01 [20,100)=0.02 [100,1000)=0.01\n",
      "[2022-07-10 14:30:52|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[54.58 90.37 94.87 96.38 98.5  98.97 99.34] nsamples=7344\n",
      "[2022-07-10 14:30:52|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.75] \n",
      "[2022-07-10 14:30:52|metrics|INFO]: Query total=7344 evaluated=7310 failed=8\n",
      "[2022-07-10 14:30:52|metrics|INFO]: Mean n2d=384 n3d=338 n_matches=102\n",
      "[2022-07-10 14:30:52|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.98]cm\n",
      "[2022-07-10 14:30:52|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.47/99.24/99.37/99.39 %\n",
      "[2022-07-10 14:30:52|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7310/102/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=63.67 [0.5,1)=31.98 [1,5)=4.21 [5,10)=0.02 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:30:52|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[54.66 90.45 94.94 96.43 98.53 98.98 99.34] nsamples=7310\n",
      "[2022-07-10 14:30:52|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.5] \n",
      "[2022-07-10 14:30:53|metrics|INFO]: Query total=7344 evaluated=7292 failed=0\n",
      "[2022-07-10 14:30:53|metrics|INFO]: Mean n2d=202 n3d=200 n_matches=102\n",
      "[2022-07-10 14:30:53|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.49 0.96]cm\n",
      "[2022-07-10 14:30:53|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.39/99.10/99.22/99.25 %\n",
      "[2022-07-10 14:30:53|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7292/102/525/5\n",
      "Ratios(%): mean=0.45 [0,0.5)=63.60 [0.5,1)=32.17 [1,5)=4.19 [5,10)=0.03 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:30:53|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[54.55 90.3  94.8  96.3  98.39 98.84 99.2 ] nsamples=7292\n",
      "[2022-07-10 14:30:53|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.25] \n",
      "[2022-07-10 14:30:53|metrics|INFO]: Query total=7344 evaluated=7239 failed=0\n",
      "[2022-07-10 14:30:53|metrics|INFO]: Mean n2d=136 n3d=136 n_matches=102\n",
      "[2022-07-10 14:30:53|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.96]cm\n",
      "[2022-07-10 14:30:53|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 96.83/98.45/98.52/98.52 %\n",
      "[2022-07-10 14:30:54|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7239/102/525/8\n",
      "Ratios(%): mean=0.46 [0,0.5)=63.48 [0.5,1)=32.23 [1,5)=4.25 [5,10)=0.03 [10,20)=0.00 [20,100)=0.01 [100,1000)=0.00\n",
      "[2022-07-10 14:30:54|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[54.05 89.61 94.07 95.57 97.67 98.12 98.48] nsamples=7239\n",
      "[2022-07-10 14:30:54|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0] \n",
      "[2022-07-10 14:30:54|metrics|INFO]: Query total=7344 evaluated=7201 failed=0\n",
      "[2022-07-10 14:30:54|metrics|INFO]: Mean n2d=103 n3d=103 n_matches=103\n",
      "[2022-07-10 14:30:54|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.24 0.48 0.96]cm\n",
      "[2022-07-10 14:30:54|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 96.47/97.92/98.00/98.04 %\n",
      "[2022-07-10 14:30:54|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7201/103/525/10\n",
      "Ratios(%): mean=0.45 [0,0.5)=63.49 [0.5,1)=32.24 [1,5)=4.25 [5,10)=0.02 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:30:54|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[53.83 89.19 93.61 95.1  97.17 97.61 97.96] nsamples=7201\n",
      "[2022-07-10 14:30:54|gomatch_eval|INFO]: Cache results to : ../outputs/benchmark_cache_release/orate_ablation/covis1inls1/OracleMatcher.npy\n"
     ]
    }
   ],
   "source": [
    "from gomatch_eval.ablate_models import ablate_max_orate\n",
    "\n",
    "# Define models\n",
    "gomatchbvs = '../outputs/shared_outputs/release_models/exported_models/GoMatchBVs_best.ckpt'\n",
    "gomatchcoords = '../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt'\n",
    "bpnpnet = '../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt'\n",
    "\n",
    "# Define output dir\n",
    "dataset = data_loader.dataset\n",
    "tag = f'covis{dataset.topk}inls{dataset.inls2d_thres}'\n",
    "if dataset.normalized_thres:\n",
    "    tag += 'normth'\n",
    "cache_dir = os.path.join('../outputs/benchmark_cache_release/orate_ablation', tag)\n",
    "\n",
    "# Establish oracle upperbound\n",
    "ablate_max_orate(\n",
    "    None, data_loader, cache_dir, \n",
    "    oracle=True, debug=False, overwrite=False\n",
    ")\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:30:54|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      "\n",
      "\n",
      ">>>>>../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt\n",
      "[2022-07-10 14:30:54|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/orate_ablation/covis1inls1/BPnPNet.npy\n",
      "[2022-07-10 14:30:59|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 1.0] \n",
      "[2022-07-10 14:30:59|metrics|INFO]: Query total=7344 evaluated=7344 failed=17\n",
      "[2022-07-10 14:30:59|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=242\n",
      "[2022-07-10 14:30:59|metrics|INFO]: R_qs=[15.17 31.05 59.78] t_qs=[1.67 3.14 5.31]m/[166.8  313.5  530.73]cm\n",
      "[2022-07-10 14:30:59|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 0.74/1.36/3.13/8.54 %\n",
      "[2022-07-10 14:30:59|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=0.56 [0.5,1)=0.37 [1,5)=1.28 [5,10)=1.58 [10,20)=3.59 [20,100)=28.98 [100,1000)=60.91\n",
      "[2022-07-10 14:30:59|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.34  0.83  1.21  1.59  4.66 11.41 56.94] nsamples=7344\n",
      "[2022-07-10 14:30:59|metrics|INFO]: Matching time (per pair):5.22 ms\n",
      "[2022-07-10 14:30:59|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.75] \n",
      "[2022-07-10 14:30:59|metrics|INFO]: Query total=7344 evaluated=7310 failed=158\n",
      "[2022-07-10 14:30:59|metrics|INFO]: Mean n2d=384 n3d=338 n_matches=179\n",
      "[2022-07-10 14:30:59|metrics|INFO]: R_qs=[14.   29.78 59.25] t_qs=[1.48 2.98 5.28]m/[147.59 297.57 528.18]cm\n",
      "[2022-07-10 14:30:59|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 1.36/2.29/4.74/10.78 %\n",
      "[2022-07-10 14:31:00|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7310/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=1.02 [0.5,1)=0.70 [1,5)=2.08 [5,10)=2.10 [10,20)=4.52 [20,100)=33.00 [100,1000)=52.61\n",
      "[2022-07-10 14:31:00|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.51  1.51  2.15  2.74  6.82 15.11 62.51] nsamples=7310\n",
      "[2022-07-10 14:31:00|metrics|INFO]: Matching time (per pair):4.51 ms\n",
      "[2022-07-10 14:31:00|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.5] \n",
      "[2022-07-10 14:31:00|metrics|INFO]: Query total=7344 evaluated=7292 failed=240\n",
      "[2022-07-10 14:31:00|metrics|INFO]: Mean n2d=202 n3d=200 n_matches=117\n",
      "[2022-07-10 14:31:00|metrics|INFO]: R_qs=[ 4.08 18.45 48.09] t_qs=[0.4  1.83 4.35]m/[ 40.18 182.83 434.69]cm\n",
      "[2022-07-10 14:31:00|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 12.30/17.77/23.27/30.64 %\n",
      "[2022-07-10 14:31:00|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7292/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=7.43 [0.5,1)=5.54 [1,5)=7.51 [5,10)=4.33 [10,20)=7.05 [20,100)=32.41 [100,1000)=31.39\n",
      "[2022-07-10 14:31:00|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 4.16 12.74 15.84 17.92 26.51 37.09 74.73] nsamples=7292\n",
      "[2022-07-10 14:31:00|metrics|INFO]: Matching time (per pair):3.09 ms\n",
      "[2022-07-10 14:31:00|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0.25] \n",
      "[2022-07-10 14:31:01|metrics|INFO]: Query total=7344 evaluated=7239 failed=166\n",
      "[2022-07-10 14:31:01|metrics|INFO]: Mean n2d=136 n3d=136 n_matches=90\n",
      "[2022-07-10 14:31:01|metrics|INFO]: R_qs=[0.09 0.25 6.34] t_qs=[0.01 0.02 0.64]m/[ 0.8   2.39 64.41]cm\n",
      "[2022-07-10 14:31:01|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 57.86/65.39/69.27/73.11 %\n",
      "[2022-07-10 14:31:01|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7239/102/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=33.22 [0.5,1)=22.20 [1,5)=12.26 [5,10)=2.95 [10,20)=4.18 [20,100)=13.29 [100,1000)=9.22\n",
      "[2022-07-10 14:31:01|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[24.97 54.8  60.51 63.09 70.59 75.94 89.09] nsamples=7239\n",
      "[2022-07-10 14:31:01|metrics|INFO]: Matching time (per pair):3.48 ms\n",
      "[2022-07-10 14:31:01|gomatch_eval|INFO]: \n",
      "\n",
      ">>>>> orate=[0, 0] \n",
      "[2022-07-10 14:31:01|metrics|INFO]: Query total=7344 evaluated=7201 failed=57\n",
      "[2022-07-10 14:31:01|metrics|INFO]: Mean n2d=103 n3d=103 n_matches=84\n",
      "[2022-07-10 14:31:01|metrics|INFO]: R_qs=[0.04 0.07 0.13] t_qs=[0.   0.01 0.01]m/[0.31 0.63 1.27]cm\n",
      "[2022-07-10 14:31:01|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 90.80/92.99/93.33/93.76 %\n",
      "[2022-07-10 14:31:02|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7201/103/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=57.79 [0.5,1)=31.33 [1,5)=5.72 [5,10)=0.44 [10,20)=0.44 [20,100)=1.37 [100,1000)=2.04\n",
      "[2022-07-10 14:31:02|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[48.26 83.44 88.21 89.91 92.59 93.45 95.62] nsamples=7201\n",
      "[2022-07-10 14:31:02|metrics|INFO]: Matching time (per pair):3.31 ms\n",
      "[2022-07-10 14:31:02|gomatch_eval|INFO]: Cache results to : ../outputs/benchmark_cache_release/orate_ablation/covis1inls1/BPnPNet.npy\n"
     ]
    }
   ],
   "source": [
    "# Eval ablation models\n",
    "model_ckpts = [bpnpnet]\n",
    "ablate_max_orate(\n",
    "    model_ckpts, data_loader, cache_dir, \n",
    "    oracle=False, debug=False, overwrite=False\n",
    "\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plot curves for max orate ablation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:56:56|plot|INFO]: >>Max orate:1.0\n",
      "[2022-07-10 14:56:56|metrics|INFO]: Query total=7344 evaluated=7344 failed=17\n",
      "[2022-07-10 14:56:56|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=242\n",
      "[2022-07-10 14:56:56|metrics|INFO]: R_qs=[15.17 31.05 59.78] t_qs=[1.67 3.14 5.31]m/[166.8  313.5  530.73]cm\n",
      "[2022-07-10 14:56:56|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 0.74/1.36/3.13/8.54 %\n",
      "[2022-07-10 14:56:56|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=0.56 [0.5,1)=0.37 [1,5)=1.28 [5,10)=1.58 [10,20)=3.59 [20,100)=28.98 [100,1000)=60.91\n",
      "[2022-07-10 14:56:56|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[0.34 0.83 1.21] nsamples=7344\n",
      "[2022-07-10 14:56:56|metrics|INFO]: Matching time (per pair):5.22 ms\n",
      "[2022-07-10 14:56:56|plot|INFO]: >>Max orate:0.75\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Query total=7344 evaluated=7310 failed=158\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Mean n2d=384 n3d=338 n_matches=179\n",
      "[2022-07-10 14:56:57|metrics|INFO]: R_qs=[14.   29.78 59.25] t_qs=[1.48 2.98 5.28]m/[147.59 297.57 528.18]cm\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 1.36/2.29/4.74/10.78 %\n",
      "[2022-07-10 14:56:57|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7310/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=1.02 [0.5,1)=0.70 [1,5)=2.08 [5,10)=2.10 [10,20)=4.52 [20,100)=33.00 [100,1000)=52.61\n",
      "[2022-07-10 14:56:57|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[0.51 1.51 2.15] nsamples=7310\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Matching time (per pair):4.51 ms\n",
      "[2022-07-10 14:56:57|plot|INFO]: >>Max orate:0.5\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Query total=7344 evaluated=7292 failed=240\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Mean n2d=202 n3d=200 n_matches=117\n",
      "[2022-07-10 14:56:57|metrics|INFO]: R_qs=[ 4.08 18.45 48.09] t_qs=[0.4  1.83 4.35]m/[ 40.18 182.83 434.69]cm\n",
      "[2022-07-10 14:56:57|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 12.30/17.77/23.27/30.64 %\n",
      "[2022-07-10 14:56:58|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7292/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=7.43 [0.5,1)=5.54 [1,5)=7.51 [5,10)=4.33 [10,20)=7.05 [20,100)=32.41 [100,1000)=31.39\n",
      "[2022-07-10 14:56:58|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[ 4.16 12.74 15.84] nsamples=7292\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Matching time (per pair):3.09 ms\n",
      "[2022-07-10 14:56:58|plot|INFO]: >>Max orate:0.25\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Query total=7344 evaluated=7239 failed=166\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Mean n2d=136 n3d=136 n_matches=90\n",
      "[2022-07-10 14:56:58|metrics|INFO]: R_qs=[0.09 0.25 6.34] t_qs=[0.01 0.02 0.64]m/[ 0.8   2.39 64.41]cm\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 57.86/65.39/69.27/73.11 %\n",
      "[2022-07-10 14:56:58|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7239/102/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=33.22 [0.5,1)=22.20 [1,5)=12.26 [5,10)=2.95 [10,20)=4.18 [20,100)=13.29 [100,1000)=9.22\n",
      "[2022-07-10 14:56:58|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[24.97 54.8  60.51] nsamples=7239\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Matching time (per pair):3.48 ms\n",
      "[2022-07-10 14:56:58|plot|INFO]: >>Max orate:0\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Query total=7344 evaluated=7201 failed=57\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Mean n2d=103 n3d=103 n_matches=84\n",
      "[2022-07-10 14:56:58|metrics|INFO]: R_qs=[0.04 0.07 0.13] t_qs=[0.   0.01 0.01]m/[0.31 0.63 1.27]cm\n",
      "[2022-07-10 14:56:58|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 90.80/92.99/93.33/93.76 %\n",
      "[2022-07-10 14:56:59|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7201/103/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=57.79 [0.5,1)=31.33 [1,5)=5.72 [5,10)=0.44 [10,20)=0.44 [20,100)=1.37 [100,1000)=2.04\n",
      "[2022-07-10 14:56:59|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[48.26 83.44 88.21] nsamples=7201\n",
      "[2022-07-10 14:56:59|metrics|INFO]: Matching time (per pair):3.31 ms\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BPnPNet max orates=[1.0, 0.75, 0.5, 0.25, 0] AUC@[1, 5, 10]px\n",
      "[[ 0.34  0.83  1.21]\n",
      " [ 0.51  1.51  2.15]\n",
      " [ 4.16 12.74 15.84]\n",
      " [24.97 54.8  60.51]\n",
      " [48.26 83.44 88.21]]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:57:03|plot|INFO]: >>Max orate:1.0\n",
      "[2022-07-10 14:57:03|metrics|INFO]: Query total=7344 evaluated=7344 failed=14\n",
      "[2022-07-10 14:57:03|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=95\n",
      "[2022-07-10 14:57:03|metrics|INFO]: R_qs=[ 0.36  6.97 29.85] t_qs=[0.03 0.69 3.38]m/[  3.38  69.21 338.29]cm\n",
      "[2022-07-10 14:57:03|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 28.47/38.17/44.59/50.23 %\n",
      "[2022-07-10 14:57:03|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=15.95 [0.5,1)=11.43 [1,5)=13.03 [5,10)=4.86 [10,20)=4.93 [20,100)=14.56 [100,1000)=34.04\n",
      "[2022-07-10 14:57:04|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[10.3  29.08 34.79] nsamples=7344\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Matching time (per pair):49.05 ms\n",
      "[2022-07-10 14:57:04|plot|INFO]: >>Max orate:0.75\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Query total=7344 evaluated=7310 failed=233\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Mean n2d=384 n3d=338 n_matches=80\n",
      "[2022-07-10 14:57:04|metrics|INFO]: R_qs=[0.12 0.35 7.44] t_qs=[0.01 0.03 0.82]m/[ 1.03  3.31 81.72]cm\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 53.44/63.88/68.14/71.65 %\n",
      "[2022-07-10 14:57:04|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7310/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=30.99 [0.5,1)=20.70 [1,5)=14.23 [5,10)=3.25 [10,20)=3.37 [20,100)=8.98 [100,1000)=14.80\n",
      "[2022-07-10 14:57:04|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[21.44 52.39 59.09] nsamples=7310\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Matching time (per pair):33.69 ms\n",
      "[2022-07-10 14:57:04|plot|INFO]: >>Max orate:0.5\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Query total=7344 evaluated=7292 failed=206\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Mean n2d=202 n3d=200 n_matches=66\n",
      "[2022-07-10 14:57:04|metrics|INFO]: R_qs=[0.06 0.12 0.24] t_qs=[0.01 0.01 0.02]m/[0.52 1.09 2.49]cm\n",
      "[2022-07-10 14:57:04|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 84.22/91.27/92.59/93.25 %\n",
      "[2022-07-10 14:57:05|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7292/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=51.96 [0.5,1)=29.53 [1,5)=11.02 [5,10)=1.06 [10,20)=0.84 [20,100)=1.47 [100,1000)=1.24\n",
      "[2022-07-10 14:57:05|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[39.94 79.76 85.88] nsamples=7292\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Matching time (per pair):27.19 ms\n",
      "[2022-07-10 14:57:05|plot|INFO]: >>Max orate:0.25\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Query total=7344 evaluated=7239 failed=39\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Mean n2d=136 n3d=136 n_matches=62\n",
      "[2022-07-10 14:57:05|metrics|INFO]: R_qs=[0.06 0.1  0.21] t_qs=[0.   0.01 0.02]m/[0.49 0.98 2.06]cm\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 89.42/95.67/96.70/96.99 %\n",
      "[2022-07-10 14:57:05|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7239/102/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=56.74 [0.5,1)=30.60 [1,5)=10.11 [5,10)=0.76 [10,20)=0.41 [20,100)=0.53 [100,1000)=0.31\n",
      "[2022-07-10 14:57:05|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[43.34 84.38 90.39] nsamples=7239\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Matching time (per pair):24.79 ms\n",
      "[2022-07-10 14:57:05|plot|INFO]: >>Max orate:0\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Query total=7344 evaluated=7201 failed=10\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Mean n2d=103 n3d=103 n_matches=56\n",
      "[2022-07-10 14:57:05|metrics|INFO]: R_qs=[0.06 0.12 0.23] t_qs=[0.01 0.01 0.02]m/[0.53 1.08 2.23]cm\n",
      "[2022-07-10 14:57:05|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 89.15/96.21/97.25/97.54 %\n",
      "[2022-07-10 14:57:06|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7201/103/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=56.86 [0.5,1)=30.63 [1,5)=11.02 [5,10)=0.65 [10,20)=0.36 [20,100)=0.23 [100,1000)=0.11\n",
      "[2022-07-10 14:57:06|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[42.38 84.67 90.9 ] nsamples=7201\n",
      "[2022-07-10 14:57:06|metrics|INFO]: Matching time (per pair):24.64 ms\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GoMatch max orates=[1.0, 0.75, 0.5, 0.25, 0] AUC@[1, 5, 10]px\n",
      "[[10.3  29.08 34.79]\n",
      " [21.44 52.39 59.09]\n",
      " [39.94 79.76 85.88]\n",
      " [43.34 84.38 90.39]\n",
      " [42.38 84.67 90.9 ]]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 14:57:10|plot|INFO]: >>Max orate:1.0\n",
      "[2022-07-10 14:57:10|metrics|INFO]: Query total=7344 evaluated=7344 failed=42\n",
      "[2022-07-10 14:57:10|metrics|INFO]: Mean n2d=780 n3d=442 n_matches=101\n",
      "[2022-07-10 14:57:10|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.98]cm\n",
      "[2022-07-10 14:57:10|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.28/99.21/99.31/99.36 %\n",
      "[2022-07-10 14:57:11|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7344/101/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=63.27 [0.5,1)=31.88 [1,5)=4.22 [5,10)=0.03 [10,20)=0.01 [20,100)=0.02 [100,1000)=0.01\n",
      "[2022-07-10 14:57:11|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[54.58 90.37 94.87] nsamples=7344\n",
      "[2022-07-10 14:57:11|plot|INFO]: >>Max orate:0.75\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Query total=7344 evaluated=7310 failed=8\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Mean n2d=384 n3d=338 n_matches=102\n",
      "[2022-07-10 14:57:11|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.98]cm\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.47/99.24/99.37/99.39 %\n",
      "[2022-07-10 14:57:11|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7310/102/525/1\n",
      "Ratios(%): mean=inf [0,0.5)=63.67 [0.5,1)=31.98 [1,5)=4.21 [5,10)=0.02 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:57:11|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[54.66 90.45 94.94] nsamples=7310\n",
      "[2022-07-10 14:57:11|plot|INFO]: >>Max orate:0.5\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Query total=7344 evaluated=7292 failed=0\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Mean n2d=202 n3d=200 n_matches=102\n",
      "[2022-07-10 14:57:11|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.49 0.96]cm\n",
      "[2022-07-10 14:57:11|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 97.39/99.10/99.22/99.25 %\n",
      "[2022-07-10 14:57:12|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7292/102/525/5\n",
      "Ratios(%): mean=0.45 [0,0.5)=63.60 [0.5,1)=32.17 [1,5)=4.19 [5,10)=0.03 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:57:12|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[54.55 90.3  94.8 ] nsamples=7292\n",
      "[2022-07-10 14:57:12|plot|INFO]: >>Max orate:0.25\n",
      "[2022-07-10 14:57:12|metrics|INFO]: Query total=7344 evaluated=7239 failed=0\n",
      "[2022-07-10 14:57:12|metrics|INFO]: Mean n2d=136 n3d=136 n_matches=102\n",
      "[2022-07-10 14:57:12|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.25 0.48 0.96]cm\n",
      "[2022-07-10 14:57:12|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 96.83/98.45/98.52/98.52 %\n",
      "[2022-07-10 14:57:12|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7239/102/525/8\n",
      "Ratios(%): mean=0.46 [0,0.5)=63.48 [0.5,1)=32.23 [1,5)=4.25 [5,10)=0.03 [10,20)=0.00 [20,100)=0.01 [100,1000)=0.00\n",
      "[2022-07-10 14:57:13|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[54.05 89.61 94.07] nsamples=7239\n",
      "[2022-07-10 14:57:13|plot|INFO]: >>Max orate:0\n",
      "[2022-07-10 14:57:13|metrics|INFO]: Query total=7344 evaluated=7201 failed=0\n",
      "[2022-07-10 14:57:13|metrics|INFO]: Mean n2d=103 n3d=103 n_matches=103\n",
      "[2022-07-10 14:57:13|metrics|INFO]: R_qs=[0.03 0.06 0.1 ] t_qs=[0.   0.   0.01]m/[0.24 0.48 0.96]cm\n",
      "[2022-07-10 14:57:13|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 96.47/97.92/98.00/98.04 %\n",
      "[2022-07-10 14:57:13|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=7201/103/525/10\n",
      "Ratios(%): mean=0.45 [0,0.5)=63.49 [0.5,1)=32.24 [1,5)=4.25 [5,10)=0.02 [10,20)=0.01 [20,100)=0.00 [100,1000)=0.00\n",
      "[2022-07-10 14:57:13|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10]px AUC=[53.83 89.19 93.61] nsamples=7201\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Oracle max orates=[1.0, 0.75, 0.5, 0.25, 0] AUC@[1, 5, 10]px\n",
      "[[54.58 90.37 94.87]\n",
      " [54.66 90.45 94.94]\n",
      " [54.55 90.3  94.8 ]\n",
      " [54.05 89.61 94.07]\n",
      " [53.83 89.19 93.61]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x360 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "from gomatch_eval.plot_curves import *\n",
    "\n",
    "cache_root = Path(\"../outputs/benchmark_cache_release/orate_ablation/covis1inls1\")\n",
    "model_caches = {\n",
    "    'BPnPNet': 'BPnPNet.npy',\n",
    "    'GoMatch' : 'GoMatchBVs_best.npy',  \n",
    "    'Oracle' : 'OracleMatcher.npy',  \n",
    "}\n",
    "\n",
    "model_colors = {\n",
    "    'BPnPNet': Color.GREY,\n",
    "    'GoMatch' : Color.BRIGHTGREEN,\n",
    "    'Oracle' : Color.YELLOW,\n",
    "}\n",
    "\n",
    "auc_thresholds = [1, 5, 10]\n",
    "model_results = {}\n",
    "for model_name, cache_name in model_caches.items():    \n",
    "    max_orates, aucs = load_ablation_cache(\n",
    "        cache_root / cache_name, auc_thresholds, key_name='Max orate'\n",
    "    )\n",
    "    model_results[model_name] = aucs\n",
    "    print(f\"{model_name} max orates={max_orates} AUC@{auc_thresholds}px\\n{aucs}\")\n",
    "\n",
    "\n",
    "# Plotting\n",
    "plot_orate_ablation_curves(model_results, model_colors, max_orates, auc_thresholds)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 17:52:03|data_loader|INFO]: Init data loader: split=kings training=False batch=16...\n",
      "[2022-07-10 17:52:03|data_process|INFO]: Loading data file from ../data/cambridge/data_processed/query-netvlad10-sift/cambridge_2d3d.npy\n",
      "[2022-07-10 17:52:04|data_process|INFO]: Loading scene 3D points from ../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy ...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n",
      "../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 17:52:05|data_process|INFO]: Done with 3D data loading.\n",
      "[2022-07-10 17:52:05|data_process|INFO]: Fetching scene data for: 1 scenes\n",
      "Feature dir:../data/cambridge/data_processed/query-netvlad10-sift/desc_cache/SIFT1024\n",
      "[2022-07-10 17:52:12|data_process|INFO]: Finished loading scenes: 5, queries: 343 im.pts3d:258 #ims(pts3d <20)=0\n",
      "[2022-07-10 17:52:12|data_loader|INFO]: \n",
      "Dataset:cambridge_sift split=kings scenes:1 queries: 343\n",
      "Data processed dir: ../data/cambridge/data_processed/query-netvlad10-sift\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=visdesc load_desc=True, merge_p3dm=False\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n",
      "[2022-07-10 17:52:12|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>Visual Desc Matching\n",
      "[2022-07-10 17:52:12|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/cambridge_sift/kings/sift_inls1.merge_after_match/VisMatcher.npy\n",
      "[2022-07-10 17:52:13|gomatch_eval|INFO]: >>>>> covis_k=10 \n",
      "[2022-07-10 17:52:13|metrics|INFO]: Query total=343 evaluated=343 failed=0\n",
      "[2022-07-10 17:52:13|metrics|INFO]: Mean n2d=721 n3d=1426 n_matches=1107\n",
      "[2022-07-10 17:52:13|metrics|INFO]: R_qs=[0.56 1.08 2.64] t_qs=[0.23 0.42 0.89]m/[22.57 42.02 88.76]cm\n",
      "[2022-07-10 17:52:13|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 3.21/26.82/57.14/79.01 %\n",
      "[2022-07-10 17:52:13|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=343/31/115/2\n",
      "Ratios(%): mean=46.92 [0,0.5)=6.96 [0.5,1)=10.80 [1,5)=37.63 [5,10)=19.40 [10,20)=11.11 [20,100)=8.01 [100,1000)=5.66\n",
      "[2022-07-10 17:52:13|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 1.28 24.88 44.24 55.76 79.24 85.92 94.23] nsamples=343\n",
      "[2022-07-10 17:52:13|metrics|INFO]: Matching time (per pair):6.31 ms\n",
      "[2022-07-10 17:52:13|data_loader|INFO]: Init data loader: split=old training=False batch=16...\n",
      "[2022-07-10 17:52:13|data_process|INFO]: Loading data file from ../data/cambridge/data_processed/query-netvlad10-sift/cambridge_2d3d.npy\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 17:52:16|data_process|INFO]: Loading scene 3D points from ../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy ...\n",
      "[2022-07-10 17:52:17|data_process|INFO]: Done with 3D data loading.\n",
      "[2022-07-10 17:52:17|data_process|INFO]: Fetching scene data for: 1 scenes\n",
      "Feature dir:../data/cambridge/data_processed/query-netvlad10-sift/desc_cache/SIFT1024\n",
      "[2022-07-10 17:52:20|data_process|INFO]: Finished loading scenes: 5, queries: 182 im.pts3d:177 #ims(pts3d <20)=6\n",
      "[2022-07-10 17:52:20|data_loader|INFO]: \n",
      "Dataset:cambridge_sift split=old scenes:1 queries: 182\n",
      "Data processed dir: ../data/cambridge/data_processed/query-netvlad10-sift\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=visdesc load_desc=True, merge_p3dm=False\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n",
      "[2022-07-10 17:52:20|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>Visual Desc Matching\n",
      "[2022-07-10 17:52:20|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/cambridge_sift/old/sift_inls1.merge_after_match/VisMatcher.npy\n",
      "[2022-07-10 17:52:20|gomatch_eval|INFO]: >>>>> covis_k=10 \n",
      "[2022-07-10 17:52:20|metrics|INFO]: Query total=182 evaluated=182 failed=0\n",
      "[2022-07-10 17:52:20|metrics|INFO]: Mean n2d=763 n3d=1234 n_matches=922\n",
      "[2022-07-10 17:52:20|metrics|INFO]: R_qs=[ 0.96  2.54 76.09] t_qs=[ 0.33  0.84 19.43]m/[  33.27   84.12 1942.87]cm\n",
      "[2022-07-10 17:52:20|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 0.00/17.58/36.81/52.75 %\n",
      "[2022-07-10 17:52:20|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=182/25/86/1\n",
      "Ratios(%): mean=inf [0,0.5)=4.41 [0.5,1)=7.02 [1,5)=30.16 [5,10)=13.90 [10,20)=8.78 [20,100)=6.63 [100,1000)=22.61\n",
      "[2022-07-10 17:52:20|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.49 16.08 32.89 41.23 58.49 63.89 75.16] nsamples=182\n",
      "[2022-07-10 17:52:20|metrics|INFO]: Matching time (per pair):5.97 ms\n",
      "[2022-07-10 17:52:20|data_loader|INFO]: Init data loader: split=shop training=False batch=16...\n",
      "[2022-07-10 17:52:20|data_process|INFO]: Loading data file from ../data/cambridge/data_processed/query-netvlad10-sift/cambridge_2d3d.npy\n",
      "[2022-07-10 17:52:20|data_process|INFO]: Loading scene 3D points from ../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy ...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 17:52:21|data_process|INFO]: Done with 3D data loading.\n",
      "[2022-07-10 17:52:21|data_process|INFO]: Fetching scene data for: 1 scenes\n",
      "Feature dir:../data/cambridge/data_processed/query-netvlad10-sift/desc_cache/SIFT1024\n",
      "[2022-07-10 17:52:23|data_process|INFO]: Finished loading scenes: 5, queries: 103 im.pts3d:209 #ims(pts3d <20)=0\n",
      "[2022-07-10 17:52:23|data_loader|INFO]: \n",
      "Dataset:cambridge_sift split=shop scenes:1 queries: 103\n",
      "Data processed dir: ../data/cambridge/data_processed/query-netvlad10-sift\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=visdesc load_desc=True, merge_p3dm=False\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n",
      "[2022-07-10 17:52:23|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>Visual Desc Matching\n",
      "[2022-07-10 17:52:23|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/cambridge_sift/shop/sift_inls1.merge_after_match/VisMatcher.npy\n",
      "[2022-07-10 17:52:23|gomatch_eval|INFO]: >>>>> covis_k=10 \n",
      "[2022-07-10 17:52:23|metrics|INFO]: Query total=103 evaluated=103 failed=0\n",
      "[2022-07-10 17:52:23|metrics|INFO]: Mean n2d=782 n3d=1429 n_matches=1091\n",
      "[2022-07-10 17:52:23|metrics|INFO]: R_qs=[0.71 1.86 6.07] t_qs=[0.11 0.22 0.57]m/[11.02 21.57 57.49]cm\n",
      "[2022-07-10 17:52:23|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 8.74/41.75/68.93/76.70 %\n",
      "[2022-07-10 17:52:23|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=103/14/40/1\n",
      "Ratios(%): mean=120.88 [0,0.5)=2.37 [0.5,1)=3.26 [1,5)=24.58 [5,10)=18.99 [10,20)=18.30 [20,100)=14.54 [100,1000)=15.31\n",
      "[2022-07-10 17:52:23|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.    7.5  20.22 31.93 61.2  70.   83.71] nsamples=103\n",
      "[2022-07-10 17:52:23|metrics|INFO]: Matching time (per pair):5.86 ms\n",
      "[2022-07-10 17:52:23|data_loader|INFO]: Init data loader: split=stmarys training=False batch=16...\n",
      "[2022-07-10 17:52:23|data_process|INFO]: Loading data file from ../data/cambridge/data_processed/query-netvlad10-sift/cambridge_2d3d.npy\n",
      "[2022-07-10 17:52:23|data_process|INFO]: Loading scene 3D points from ../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy ...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "../data/cambridge/data_processed/query-netvlad10-sift/scene_points3d/all.npy\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2022-07-10 17:52:24|data_process|INFO]: Done with 3D data loading.\n",
      "[2022-07-10 17:52:24|data_process|INFO]: Fetching scene data for: 1 scenes\n",
      "Feature dir:../data/cambridge/data_processed/query-netvlad10-sift/desc_cache/SIFT1024\n",
      "[2022-07-10 17:52:36|data_process|INFO]: Finished loading scenes: 5, queries: 530 im.pts3d:281 #ims(pts3d <20)=5\n",
      "[2022-07-10 17:52:36|data_loader|INFO]: \n",
      "Dataset:cambridge_sift split=stmarys scenes:1 queries: 530\n",
      "Data processed dir: ../data/cambridge/data_processed/query-netvlad10-sift\n",
      "Settings=(\n",
      "  topk=1, random_topk=False, orate=[0, 1], npt=[10, 1024],\n",
      "  p2d_type=sift, p3d_type=visdesc load_desc=True, merge_p3dm=False\n",
      "  inls_thres=1 normalized_thres=False quant_desc=None\n",
      ")\n",
      "\n",
      "[2022-07-10 17:52:36|gomatch_eval|INFO]: \n",
      "\n",
      "\n",
      ">>>>>Visual Desc Matching\n",
      "[2022-07-10 17:52:36|gomatch_eval|INFO]: Load cache: ../outputs/benchmark_cache_release/cambridge_sift/stmarys/sift_inls1.merge_after_match/VisMatcher.npy\n",
      "[2022-07-10 17:52:36|gomatch_eval|INFO]: >>>>> covis_k=10 \n",
      "[2022-07-10 17:52:36|metrics|INFO]: Query total=530 evaluated=530 failed=0\n",
      "[2022-07-10 17:52:36|metrics|INFO]: Mean n2d=768 n3d=1889 n_matches=1175\n",
      "[2022-07-10 17:52:36|metrics|INFO]: R_qs=[0.57 1.64 9.73] t_qs=[0.15 0.36 2.81]m/[ 15.44  36.31 281.44]cm\n",
      "[2022-07-10 17:52:36|metrics|INFO]: Localize recall@(<5cm5deg/0.25m2deg/0.5m5deg/1m10deg): 5.66/35.09/53.58/65.66 %\n",
      "[2022-07-10 17:52:36|metrics|INFO]: \n",
      "Reproj Sample/N(mean/max/min)=530/27/155/1\n",
      "Ratios(%): mean=inf [0,0.5)=5.48 [0.5,1)=8.26 [1,5)=30.98 [5,10)=15.20 [10,20)=9.80 [20,100)=9.12 [100,1000)=15.80\n",
      "[2022-07-10 17:52:36|metrics|INFO]: \n",
      "ReprojThres=[1, 5, 10, 15, 50, 100, 500]px AUC=[ 0.81 17.3  32.8  42.53 62.81 69.52 79.13] nsamples=530\n",
      "[2022-07-10 17:52:36|metrics|INFO]: Matching time (per pair):6.05 ms\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = '0'\n",
    "import torch\n",
    "\n",
    "from gomatch_eval.benchmark import  run_benchmark\n",
    "\n",
    "# Define models\n",
    "gomatchbvs = '../outputs/shared_outputs/release_models/exported_models/GoMatchBVs_best.ckpt'\n",
    "gomatchcoords = '../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt'\n",
    "bpnpnet = '../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt'\n",
    "\n",
    "# Run\n",
    "splits = ('kings', 'old', 'shop', 'stmarys')\n",
    "\n",
    "for split in splits:\n",
    "    run_benchmark(\n",
    "        root_dir='..', \n",
    "        dataset_name='cambridge_sift',\n",
    "        p2d_type='sift',\n",
    "        split=split,\n",
    "        ckpt=gomatchbvs,\n",
    "        vismatch=True,\n",
    "        merge_before_match=False,\n",
    "        overwrite=False,\n",
    "        debug=False, \n",
    "        oracle=False,\n",
    "        odir='outputs/benchmark_cache_release',\n",
    "        covis_k_nums=[10],\n",
    "    )\n",
    "    \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Eval on 7Scenes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = '0'\n",
    "import torch\n",
    "\n",
    "from gomatch_eval.benchmark import  run_benchmark\n",
    "\n",
    "# Define models\n",
    "gomatchbvs = '../outputs/shared_outputs/release_models/exported_models/GoMatchBVs_best.ckpt'\n",
    "gomatchcoords = '../outputs/shared_outputs/release_models/exported_models/GoMatchCoords_best.ckpt'\n",
    "bpnpnet = '../outputs/shared_outputs/release_models/exported_models/BPnPNet.ckpt'\n",
    "\n",
    "# Run\n",
    "splits = ('chess', 'fire', 'heads', 'office', 'pumpkin', 'redkitchen', 'stairs')\n",
    "\n",
    "for split in splits:\n",
    "    run_benchmark(\n",
    "        root_dir='..', \n",
    "        dataset_name='7scenes_sift_v2',\n",
    "        p2d_type='sift',\n",
    "        split=split,\n",
    "        ckpt=gomatchbvs,\n",
    "        vismatch=True,\n",
    "        merge_before_match=False,\n",
    "        overwrite=False,\n",
    "        debug=False, \n",
    "        oracle=False,\n",
    "        odir='../outputs/benchmark_cache_release',\n",
    "        covis_k_nums=[10],\n",
    "    )\n",
    "    \n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:bpnp]",
   "language": "python",
   "name": "conda-env-bpnp-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}