Loftr/notebooks/visualize_dump_results.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "import os\n",
    "os.chdir(\"..\")\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2\n",
    "from pathlib import Path\n",
    "from src.utils.plotting import make_matching_figure, error_colormap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_prediction_and_evaluation_plot(root_dir, pe, path=None, source='ScanNet'):\n",
    "    img0 = cv2.imread(str(root_dir / pe['pair_names'][0]), cv2.IMREAD_GRAYSCALE)\n",
    "    img1 = cv2.imread(str(root_dir / pe['pair_names'][1]), cv2.IMREAD_GRAYSCALE)\n",
    "    if source == 'ScanNet':\n",
    "        img0 = cv2.resize(img0, (640, 480))\n",
    "        img1 = cv2.resize(img1, (640, 480))\n",
    "\n",
    "    thr = 5e-4\n",
    "    mkpts0 = pe['mkpts0_f']\n",
    "    mkpts1 = pe['mkpts1_f']\n",
    "    color = error_colormap(pe['epi_errs'], thr, alpha=0.3)\n",
    "\n",
    "    text = [\n",
    "        f\"LoFTR\",\n",
    "        f\"#Matches: {len(mkpts0)}\",\n",
    "        f\"$\\\\Delta$R:{pe['R_errs']:.2f}°,  $\\\\Delta$t:{pe['t_errs']:.2f}°\",\n",
    "    ]\n",
    "\n",
    "    if path:\n",
    "        make_matching_figure(img0, img1, mkpts0, mkpts1, color, text=text, path=path)\n",
    "    else:\n",
    "        return make_matching_figure(img0, img1, mkpts0, mkpts1, color, text=text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualization on ScanNet\n",
    "- Prediction and Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "root_dir = Path(\"data/scannet/test\")  # Scannet\n",
    "npy_path = \"dump/loftr_ds_indoor/LoFTR_pred_eval.npy\"\n",
    "dumps = np.load(npy_path, allow_pickle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 750x450 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = make_prediction_and_evaluation_plot(root_dir, dumps[2], source='ScanNet')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualization on MegaDepth\n",
    "- Prediction and Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "root_dir = Path(\"data/megadepth/test\")  # MegaDepth\n",
    "npy_path = \"dump/loftr_ds_outdoor/LoFTR_pred_eval.npy\"\n",
    "dumps = np.load(npy_path, allow_pickle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 750x450 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = make_prediction_and_evaluation_plot(root_dir, dumps[51], source='MegaDepth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8.8 64-bit ('loftr': conda)",
   "name": "python388jvsc74a57bd0e2d1507a0fcefcbd70c2e8d5c2edae879585b2c8df0be6cdbe280bf251175c7f"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8-final"
  },
  "orig_nbformat": 2
 },
 "nbformat": 4,
 "nbformat_minor": 2
}