lostanlen_waspaa2019_fig2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gzip\n",
    "import h5py\n",
    "import librosa\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "import pandas as pd\n",
    "import soundfile as sf\n",
    "from collections import OrderedDict, Counter\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "dcase_dir = '/beegfs/vl1019/waspaa2019_data/BirdVox-DCASE-20k_h5'\n",
    "\n",
    "settings = {\n",
    "    \"fmin\": 2000,\n",
    "    \"fmax\": 11025,\n",
    "    \"hop_length\": 32,\n",
    "    \"n_fft\": 1024,\n",
    "    \"n_mels\": 128,\n",
    "    \"sr\": 22050,\n",
    "    \"win_length\": 256,\n",
    "    \"window\": \"hann\",\n",
    "    \"T\": 0.1,\n",
    "    \"alpha\": 1.0,\n",
    "    \"delta\": 0.0,\n",
    "    \"r\": 1.0,\n",
    "    \"eps\": 1e-6,\n",
    "    \"lower_sf_bin\": 15,\n",
    "    \"upper_sf_bin\": 90,\n",
    "    \"padding\": 10}\n",
    "\n",
    "# greater padding for BirdVox-DCASE because there is no signal duplication around boundaries\n",
    "settings[\"padding\"] = 100\n",
    "coni_thresholds_path = '/beegfs/vl1019/waspaa2019_data/coni-knight-thresholds_v-2.csv'\n",
    "coni_thresholds = pd.read_csv(coni_thresholds_path)\n",
    "\n",
    "distance_list = coni_thresholds['nominal_distance'].unique()\n",
    "bird_id_list = coni_thresholds['bird_id'].unique()\n",
    "\n",
    "import tqdm\n",
    "\n",
    "initial_chars =\\\n",
    "    [str(x) for x in range(10)] + [chr(ord('a') + x) for x in range(6)]\n",
    "\n",
    "\n",
    "\n",
    "log_pcen_max_list = []\n",
    "log_pcen_avg_list = []\n",
    "melsf_max_list = []\n",
    "melsf_avg_list = []\n",
    "\n",
    "\n",
    "\n",
    "# Loop over 16 hexadecimal prefixes in BirdVox-DCASE-20k\n",
    "for initial_char in tqdm.tqdm(initial_chars):\n",
    "    h5_name = \"_\".join([\"BirdVox-DCASE-20k_pcen\", initial_char]) + \".h5\"\n",
    "    h5_path = os.path.join(dcase_dir, h5_name)\n",
    "    \n",
    "    # Open HDF5 file\n",
    "    with h5py.File(h5_path, 'r') as f:\n",
    "        log_pcen_max_list.append(np.max(np.log1p(\n",
    "            f[\"pcen\"][:, settings[\"lower_sf_bin\"]:settings[\"upper_sf_bin\"], padding:-padding]), axis=1));\n",
    "        log_pcen_avg_list.append(np.mean(np.log1p(\n",
    "            f[\"pcen\"][:, settings[\"lower_sf_bin\"]:settings[\"upper_sf_bin\"], padding:-padding]), axis=1));\n",
    "        melsf_max_list.append(np.max(np.maximum(\n",
    "            0, np.diff(np.log1p(f[\"mel\"][:, settings[\"lower_sf_bin\"]:settings[\"upper_sf_bin\"], padding:-padding]))), axis=1));\n",
    "        melsf_avg_list.append(np.mean(np.maximum(\n",
    "            0, np.diff(np.log1p(f[\"mel\"][:, settings[\"lower_sf_bin\"]:settings[\"upper_sf_bin\"], padding:-padding]))), axis=1));\n",
    "        \n",
    "feature_matrix = {\n",
    "    \"log_pcen_max\": np.concatenate(log_pcen_max_list, axis=0),\n",
    "    \"log_pcen_avg\": np.concatenate(log_pcen_avg_list, axis=0),\n",
    "    \"melsf_max\": np.concatenate(melsf_max_list, axis=0),\n",
    "    \"melsf_avg\": np.concatenate(melsf_avg_list, axis=0)\n",
    "}\n",
    "\n",
    "feature_strs = [\n",
    "    \"melsf_avg\", \"melsf_max\",\n",
    "    \"log_pcen_avg\", \"log_pcen_max\"]\n",
    "hop_duration = pcen_settings[\"hop_length\"] / pcen_settings[\"sr\"]\n",
    "mtbf_dict = {feature_str: [] for feature_str in feature_strs}\n",
    "\n",
    "for _, row in tqdm.tqdm(coni_thresholds.iterrows()):\n",
    "    for feature_str in feature_strs:\n",
    "        n_false_alarms = np.sum(\n",
    "            row[feature_str] < (feature_matrix[feature_str] - np.min(feature_matrix[feature_str], axis=1)[:, np.newaxis]))\n",
    "        n_samples = np.prod(feature_matrix[feature_str].shape)\n",
    "        mtbf = hop_duration * (n_samples/(n_false_alarms))\n",
    "        mtbf_dict[feature_str].append(mtbf)\n",
    "\n",
    "mtbf_dict[\"bird_id\"] = coni_thresholds[\"bird_id\"]\n",
    "mtbf_dict[\"nominal_distance\"] = coni_thresholds[\"nominal_distance\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.rcParams[\"font.family\"] = \"serif\"\n",
    "\n",
    "\n",
    "x_distances = np.unique(mtbf_df[\"nominal_distance\"])\n",
    "\n",
    "model_colors = [\n",
    "    \"#CB0003\", \n",
    "    \"#E67300\",\n",
    "    \"#990099\",\n",
    "    \"#0000B2\",\n",
    "    \"#009900\",\n",
    "    '#008888',\n",
    "    '#888800',\n",
    "    '#555555',\n",
    "]\n",
    "\n",
    "mtbf_df = pd.DataFrame(mtbf_dict)\n",
    "feature_colors = {\n",
    "    \"melsf_max\": \"#0000B2\",\n",
    "    \"melsf_avg\": \"#E67300\",\n",
    "    \"log_pcen_max\": \"#009900\",\n",
    "}\n",
    "feature_alphas = {\n",
    "    \"#0000B2\": 0.33,\n",
    "    \"#E67300\": 0.5,\n",
    "    \"#009900\": 0.5\n",
    "    \n",
    "}\n",
    "feature_markers = {\n",
    "    \"melsf_avg\": \"-v\",\n",
    "    \"melsf_max\": \"-s\",\n",
    "    \"log_pcen_max\": \"-o\",\n",
    "}\n",
    "feature_labels = {\n",
    "    \"melsf_avg\": \"Averaged spectral flux\",\n",
    "    \"melsf_max\": \"Max-pooled spectral flux\",\n",
    "    \"log_pcen_avg\": \"Averaged PCEN\",\n",
    "    \"log_pcen_max\": \"Max-pooled PCEN\",\n",
    "}\n",
    "\n",
    "plt.figure(figsize=(6, 4))\n",
    "plt.title('Bioacoustic detection of Common Nighthawk', size=12)\n",
    "\n",
    "# Loop over features\n",
    "for feature_str in [\"log_pcen_max\", \"melsf_max\", \"melsf_avg\"]:\n",
    "    \n",
    "    median_mtbf_list = []\n",
    "    lower_mtbf_list = []\n",
    "    upper_mtbf_list = []\n",
    "    \n",
    "    # Loop over distances\n",
    "    for x_distance in x_distances:\n",
    "        \n",
    "        distance_df = mtbf_df[\n",
    "            (mtbf_df[\"nominal_distance\"]==x_distance)]\n",
    "        median_mtbf_list.append(\n",
    "            np.median(distance_df[feature_str]))\n",
    "        lower_mtbf_list.append(\n",
    "            np.quantile(distance_df[feature_str], 0.25))\n",
    "        upper_mtbf_list.append(\n",
    "            np.quantile(distance_df[feature_str], 0.75))\n",
    "        \n",
    "\n",
    "    plt.plot(\n",
    "        x_distances,\n",
    "        np.log10(median_mtbf_list),\n",
    "        feature_markers[feature_str],\n",
    "        markersize=10.0,\n",
    "        linewidth=2.0,\n",
    "        color=feature_colors[feature_str],\n",
    "        label=feature_labels[feature_str])\n",
    "    plt.fill_between(\n",
    "        x_distances, \n",
    "        np.log10(lower_mtbf_list), np.log10(upper_mtbf_list),\n",
    "        color=feature_colors[feature_str],\n",
    "        alpha=feature_alphas[feature_colors[feature_str]])\n",
    "        \n",
    "plt.legend(prop={'size': 11})\n",
    "yticks = np.array([0.1, 0.2, 0.5, 1.0, 2.0, 5.0, 10.0, 20.0, 50.0, 100.0, 200.0])\n",
    "\n",
    "plt.gca().set_xticks(np.arange(0, 501, 100))\n",
    "plt.gca().set_xticklabels(np.arange(0, 501, 100), size=11)\n",
    "plt.gca().set_xticks(np.arange(0, 501, 50), minor=True)\n",
    "plt.gca().set_yticks(np.log10(yticks))\n",
    "plt.gca().set_yticklabels(yticks, size=11);\n",
    "plt.grid(which='minor', linestyle='--', alpha=1.0);\n",
    "plt.grid(which='major', linestyle='--', alpha=1.0);\n",
    "\n",
    "plt.xlim(0, 500)\n",
    "plt.ylim(np.log10(0.5), np.log10(200.0))\n",
    "\n",
    "plt.xlabel(\"CONI-Knight dataset: distance between sensor and source (m)\", size=12)\n",
    "plt.ylabel(\"BirdVox-DCASE-20k dataset: mean time\\nbetween false alarms at half recall (s)\", size=12)\n",
    "\n",
    "\n",
    "ax = plt.gca()\n",
    "ax.spines['top'].set_visible(False)\n",
    "ax.spines['right'].set_visible(False)\n",
    "\n",
    "\n",
    "plt.savefig('lostanlen_waspaa2019_coni_mtbfa-50_semilogy.eps', bbox_inches='tight')\n",
    "plt.savefig('lostanlen_waspaa2019_coni_mtbfa-50_semilogy.png', bbox_inches='tight', dpi=500)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}