note_tracker.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "%autoreload 2\n",
    "import os, json, glob\n",
    "import matplotlib.pyplot as plt\n",
    "import librosa, mir_eval\n",
    "import numpy as np\n",
    "from f0_analysis_utils import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## F0 (evaluation with pitch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80):\n",
    "    \"\"\"Note tracking based on F0 segmentation\n",
    "\n",
    "    This function takes a file as input, suitable for individually testing this component.\n",
    "\n",
    "    When taking audio segments returned by monophony detector, \n",
    "    the function needs to take audio signal as input\n",
    "\n",
    "    Args:\n",
    "        file (str): filename of the mono audio segment\n",
    "        pitch_diff_thres (float, optional): The F0 curve are cut into note events are points \n",
    "        where the pitch difference betweentwo adjacent frames exceeds this threshold. Defaults to 0.8.\n",
    "        note_dur_thres (int, optional): A valid note event must last longer than this threshold in ms. Defaults to 80.\n",
    "\n",
    "    Returns:\n",
    "        list, list: Estimated intervals: [(onset, offset)], and estimated pitches: [pitch]\n",
    "    \"\"\"\n",
    "    est_intervals = []\n",
    "    est_pitches = []\n",
    "    times, notes = load_audio_file(file)\n",
    "    f0_segments = get_continous_f0_segments(notes, dur_thres=note_dur_thres)\n",
    "    for f0_segment in f0_segments:\n",
    "        note_events = get_note_events_from_f0_segment(notes, f0_segment, f0_change_thres=pitch_diff_thres, dur_thres=note_dur_thres)\n",
    "        for note_event in note_events:\n",
    "            onset = times[note_event[0]]\n",
    "            offset = times[note_event[-1]]\n",
    "            pitch = np.mean(notes[note_event])\n",
    "            pitch_hz = librosa.midi_to_hz(pitch)\n",
    "            est_intervals.append([onset, offset])\n",
    "            est_pitches.append(pitch_hz)\n",
    "    est_intervals = np.array(est_intervals)\n",
    "    est_pitches = np.array(est_pitches)\n",
    "    return est_intervals, est_pitches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eval_f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80):\n",
    "    \"\"\"First call `f0_note_tracker` to estimate intervals and pitches. \n",
    "    Then evaluate the result by comparing estimated note events against \n",
    "    ground truth annotations.\n",
    "    \n",
    "    This evaluates the estimated onset and pitch. \n",
    "    The two parameters to be tuned: `pitch_diff_thres` and `note_dur_thres`\n",
    "\n",
    "    Args:\n",
    "        file (str): The filename of a mono audio segment.\n",
    "        pitch_diff_thres (float): The F0 curve is cut into two notes if the pitch difference between two adjacent frames exceeds this threshold.\n",
    "        note_dur_thres (int): A valid note event must last longer than this threshold (in ms).\n",
    "\n",
    "    Returns:\n",
    "        tuple: scores returned by `mir_eval`\n",
    "    \"\"\"\n",
    "    est_intervals, est_pitches = f0_note_tracker(file, pitch_diff_thres, note_dur_thres)\n",
    "    \n",
    "    anno_file = find_anno(file)\n",
    "    ref_intervals = []\n",
    "    ref_pitches = []\n",
    "    with open(anno_file) as anno:\n",
    "        annotation = json.load(anno)\n",
    "    for note_event in annotation:\n",
    "        onset = note_event[\"time\"][\"start\"]\n",
    "        offset = onset + note_event[\"time\"][\"dur\"]\n",
    "        pitch = note_event[\"pitch\"]\n",
    "        pitch_hz = librosa.midi_to_hz(pitch)\n",
    "        ref_intervals.append([onset, offset])\n",
    "        ref_pitches.append(pitch_hz)\n",
    "    ref_intervals = np.array(ref_intervals)\n",
    "    ref_pitches = np.array(ref_pitches)\n",
    "\n",
    "    score = mir_eval.transcription.evaluate(ref_intervals, ref_pitches, est_intervals, est_pitches)\n",
    "    return score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1440x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# clean tone\n",
    "file = \"ACDC - Back In Black (ver 4 by GuitarManiac09)_Angus Young_45_122s.wav\"\n",
    "times, notes = load_audio_file(file)\n",
    "intervals, pitches = f0_note_tracker(file, pitch_diff_thres=1, note_dur_thres=75)\n",
    "\n",
    "plt.figure(figsize=(20,4))\n",
    "plt.plot(times, notes)\n",
    "for i in range(intervals.shape[0]):\n",
    "    pitch = librosa.hz_to_midi(pitches[i])\n",
    "    plt.hlines(y=pitch, xmin=intervals[i][0], xmax=intervals[i][1], linewidth=2, color='r')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1440x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# origin tone\n",
    "file = \"ACDC - Back In Black (ver 4 by GuitarManiac09)_Angus Young_45_122s.wav\"\n",
    "ORIGIN_AUDIO_DIR = \"/Volumes/MacOnly/UG_proc/all_time_top_by_hits/original_tone/mono_audio_segments_filtered\"\n",
    "y, sr = librosa.load(os.path.join(ORIGIN_AUDIO_DIR, file), sr=None)\n",
    "f0, _, _ = librosa.pyin(\n",
    "    y,\n",
    "    fmin=librosa.note_to_hz(\"C2\"),\n",
    "    fmax=librosa.note_to_hz(\"G6\"),\n",
    "    sr=sr,\n",
    "    center=False,\n",
    ")\n",
    "times = librosa.times_like(f0, sr=sr, hop_length=512, n_fft=2048)\n",
    "# convert F0 sequence to MIDI note sequence\n",
    "notes = librosa.hz_to_midi(f0)\n",
    "\n",
    "intervals, pitches = f0_note_tracker(os.path.join(ORIGIN_AUDIO_DIR, file), pitch_diff_thres=1, note_dur_thres=75)\n",
    "\n",
    "plt.figure(figsize=(20,4))\n",
    "plt.plot(times, notes)\n",
    "for i in range(intervals.shape[0]):\n",
    "    pitch = librosa.hz_to_midi(pitches[i])\n",
    "    plt.hlines(y=pitch, xmin=intervals[i][0], xmax=intervals[i][1], linewidth=2, color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate score for one file\n",
    "file = \"ACDC - Back In Black (ver 4 by GuitarManiac09)_Angus Young_45_122s.wav\"\n",
    "\n",
    "anno_file = find_anno(file)\n",
    "ref_intervals = []\n",
    "ref_pitches = []\n",
    "with open(anno_file) as anno:\n",
    "    annotation = json.load(anno)\n",
    "for note_event in annotation:\n",
    "    onset = note_event[\"time\"][\"start\"]\n",
    "    offset = onset + note_event[\"time\"][\"dur\"]\n",
    "    pitch = note_event[\"pitch\"]\n",
    "    pitch_hz = librosa.midi_to_hz(pitch)\n",
    "    ref_intervals.append([onset, offset])\n",
    "    ref_pitches.append(pitch_hz)\n",
    "ref_intervals = np.array(ref_intervals)\n",
    "ref_pitches = np.array(ref_pitches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "est_intervals, est_pitches = f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OrderedDict([('Precision', 0.5666666666666667),\n",
       "             ('Recall', 0.53125),\n",
       "             ('F-measure', 0.5483870967741935),\n",
       "             ('Average_Overlap_Ratio', 0.7663269044100365),\n",
       "             ('Precision_no_offset', 0.6),\n",
       "             ('Recall_no_offset', 0.5625),\n",
       "             ('F-measure_no_offset', 0.5806451612903225),\n",
       "             ('Average_Overlap_Ratio_no_offset', 0.7452633630384912),\n",
       "             ('Onset_Precision', 0.9333333333333333),\n",
       "             ('Onset_Recall', 0.875),\n",
       "             ('Onset_F-measure', 0.9032258064516129),\n",
       "             ('Offset_Precision', 0.9666666666666667),\n",
       "             ('Offset_Recall', 0.90625),\n",
       "             ('Offset_F-measure', 0.9354838709677419)])"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mir_eval.transcription.evaluate(ref_intervals, ref_pitches, est_intervals, est_pitches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# batch processing\n",
    "FILTERED_AUDIO_DIR = \"/Volumes/MacOnly/UG_proc/all_time_top_by_hits/clean_tone/mono_audio_segments_filtered\"\n",
    "audio_files = sorted(glob.glob(os.path.join(FILTERED_AUDIO_DIR, \"*.wav\")), key=os.path.getsize, reverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metallica - The Day That Never Comes_Harmony_0_0s.wav 0\n",
      "Queen - Bohemian Rhapsody_Gtr 1_0_0s.wav 1\n",
      "Metallica - Seek And Destroy (ver 2)_Gtr. III_0_0s.wav 2\n"
     ]
    }
   ],
   "source": [
    "scores = []\n",
    "i = 0\n",
    "\n",
    "for file in audio_files:\n",
    "    print(file.split(\"/\")[-1], i)\n",
    "    score = eval_f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80)\n",
    "    scores.append(score)\n",
    "    i += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "import csv\n",
    "field_names = score.keys()\n",
    "with open('score.csv', 'w') as csvfile:\n",
    "    writer = csv.DictWriter(csvfile, fieldnames=field_names)\n",
    "    writer.writeheader()\n",
    "    writer.writerows(scores)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Onset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "def onset_note_tracker(file):\n",
    "    y, sr = librosa.load(os.path.join(FILTERED_AUDIO_DIR, file), sr=None)\n",
    "    # optional: normalize the signal\n",
    "    y = y / np.max(abs(y))\n",
    "    est_intervals = []\n",
    "    onsets = librosa.onset.onset_detect(y=y, sr=sr, units='time')\n",
    "    for i in range(len(onsets)-1):\n",
    "        onset = onsets[i]\n",
    "        offset = onsets[i+1]\n",
    "        est_intervals.append([onset, offset])\n",
    "    if len(est_intervals) == 0:\n",
    "        est_intervals.append([0, len(y)/sr])\n",
    "    est_intervals = np.array(est_intervals)\n",
    "    return est_intervals\n",
    "\n",
    "def eval_onset_note_tracker(file, onset_tol=0.05, offset_tol=0.05, offset_ratio=0.2):\n",
    "    est_intervals = onset_note_tracker(file)\n",
    "\n",
    "    anno_file = find_anno(file)\n",
    "    ref_intervals = []\n",
    "    with open(anno_file) as anno:\n",
    "        annotation = json.load(anno)\n",
    "    for note_event in annotation:\n",
    "        onset = note_event[\"time\"][\"start\"]\n",
    "        offset = onset + note_event[\"time\"][\"dur\"]\n",
    "        ref_intervals.append([onset, offset])\n",
    "    ref_intervals = np.array(ref_intervals)\n",
    "\n",
    "    onset_match = mir_eval.transcription.match_note_onsets(ref_intervals, est_intervals, onset_tolerance=onset_tol)\n",
    "    offset_match = mir_eval.transcription.match_note_offsets(ref_intervals, est_intervals, offset_min_tolerance=offset_tol, offset_ratio=offset_ratio)\n",
    "    match = [i for i in onset_match if i in offset_match]\n",
    "\n",
    "    return len(est_intervals), len(ref_intervals), len(match)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1440x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "file = \"ACDC - Back In Black (ver 4 by GuitarManiac09)_Angus Young_45_122s.wav\"\n",
    "y, sr = librosa.load(os.path.join(FILTERED_AUDIO_DIR, file), sr=None)\n",
    "est_intervals = []\n",
    "onsets = librosa.onset.onset_detect(y=y, sr=sr, units='time')\n",
    "\n",
    "times, notes = load_audio_file(file)\n",
    "plt.figure(figsize=(20, 4))\n",
    "plt.plot(times, notes)\n",
    "for onset in onsets:\n",
    "    plt.axvline(onset, color='r')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "FILTERED_AUDIO_DIR = \"/Volumes/MacOnly/UG_proc/all_time_top_by_hits/original_tone/mono_audio_segments_filtered\"\n",
    "audio_files = glob.glob(os.path.join(FILTERED_AUDIO_DIR, \"*.wav\"))\n",
    "# audio_files = sorted(glob.glob(os.path.join(FILTERED_AUDIO_DIR, \"*.wav\")), key=os.path.getsize, reverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6504196255648805 0.7782927771340286 0.7086337260418498\n"
     ]
    }
   ],
   "source": [
    "# onset detection-based note tracker (clean):\n",
    "# over all dataset: precision 0.8796, recall 0.8508, f-score 0.8650\n",
    "# onset detection-based note tracker (origin 0.05 tolerance):\n",
    "# over all dataset: precision 0.6643, recall 0.6969, f-score 0.6802\n",
    "# onset detection-based note tracker (origin 0.1 tolerance):\n",
    "# over all dataset: precision 0.6694, recall 0.7022, f-score 0.6854\n",
    "# onset detection-based note tracker (origin 0.2 tolerance):\n",
    "# over all dataset: precision 0.6977, recall 0.7320, f-score 0.7144\n",
    "# onset detection-based note tracker (origin 0.5 tolerance):\n",
    "# over all dataset: precision 0.7217, recall 0.7571, f-score 0.7390\n",
    "est_total = 0\n",
    "ref_total = 0\n",
    "match_total = 0\n",
    "i = 0\n",
    "for file in audio_files:\n",
    "    est_cnt, ref_cnt, match_cnt = eval_onset_note_tracker(file, onset_tol=0.05, offset_tol=0.05, offset_ratio=0.5)\n",
    "    est_total += est_cnt\n",
    "    ref_total += ref_cnt\n",
    "    match_total += match_cnt\n",
    "    # print(est_cnt, ref_cnt, match_cnt)\n",
    "    i += 1\n",
    "    if i == 500:\n",
    "        break\n",
    "precision = match_total / est_total\n",
    "recall = match_total / ref_total\n",
    "f_score = (2*precision*recall) / (precision + recall)\n",
    "print(precision, recall, f_score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## F0 (evaluation without pitch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "def f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80):\n",
    "    est_intervals = []\n",
    "    times, notes = load_audio_file(file)\n",
    "    f0_segments = get_continous_f0_segments(notes, dur_thres=note_dur_thres)\n",
    "    for f0_segment in f0_segments:\n",
    "        note_events = get_note_events_from_f0_segment(notes, f0_segment, f0_change_thres=pitch_diff_thres, dur_thres=note_dur_thres)\n",
    "        for note_event in note_events:\n",
    "            onset = times[note_event[0]]\n",
    "            offset = times[note_event[-1]]\n",
    "            est_intervals.append([onset, offset])\n",
    "    if len(est_intervals) == 0:\n",
    "        est_intervals.append([0, times[-1]])\n",
    "    est_intervals = np.array(est_intervals)\n",
    "    return est_intervals\n",
    "\n",
    "def eval_f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80):\n",
    "    est_intervals = f0_note_tracker(file, pitch_diff_thres, note_dur_thres)\n",
    "    anno_file = find_anno(file)\n",
    "    ref_intervals = []\n",
    "    with open(anno_file) as anno:\n",
    "        annotation = json.load(anno)\n",
    "    for note_event in annotation:\n",
    "        onset = note_event[\"time\"][\"start\"]\n",
    "        offset = onset + note_event[\"time\"][\"dur\"]\n",
    "        ref_intervals.append([onset, offset])\n",
    "    ref_intervals = np.array(ref_intervals)\n",
    "\n",
    "    onset_match = mir_eval.transcription.match_note_onsets(ref_intervals, est_intervals)\n",
    "    offset_match = mir_eval.transcription.match_note_offsets(ref_intervals, est_intervals)\n",
    "    match = [i for i in onset_match if i in offset_match]\n",
    "\n",
    "    return len(est_intervals), len(ref_intervals), len(match)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "audio_files = glob.glob(os.path.join(FILTERED_AUDIO_DIR, \"*.wav\"))\n",
    "# audio_files = sorted(glob.glob(os.path.join(FILTERED_AUDIO_DIR, \"*.wav\")), key=os.path.getsize, reverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7149220489977728 0.626953125 0.668054110301769\n"
     ]
    }
   ],
   "source": [
    "est_total = 0\n",
    "ref_total = 0\n",
    "match_total = 0\n",
    "i = 0\n",
    "for file in audio_files:\n",
    "    est_cnt, ref_cnt, match_cnt = eval_f0_note_tracker(file, pitch_diff_thres=0.8, note_dur_thres=80)\n",
    "    est_total += est_cnt\n",
    "    ref_total += ref_cnt\n",
    "    match_total += match_cnt\n",
    "    # print(est_cnt, ref_cnt, match_cnt)\n",
    "    i += 1\n",
    "    if i == 50:\n",
    "        break\n",
    "precision = match_total / est_total\n",
    "recall = match_total / ref_total\n",
    "f_score = (2*precision*recall) / (precision + recall)\n",
    "print(precision, recall, f_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.3091715976331361 0.408203125 0.3518518518518518\n"
     ]
    }
   ],
   "source": [
    "est_total = 0\n",
    "ref_total = 0\n",
    "match_total = 0\n",
    "i = 0\n",
    "for file in audio_files:\n",
    "    est_cnt, ref_cnt, match_cnt = eval_f0_note_tracker(file, pitch_diff_thres=0.5, note_dur_thres=0)\n",
    "    est_total += est_cnt\n",
    "    ref_total += ref_cnt\n",
    "    match_total += match_cnt\n",
    "    # print(est_cnt, ref_cnt, match_cnt)\n",
    "    i += 1\n",
    "    if i == 50:\n",
    "        break\n",
    "precision = match_total / est_total\n",
    "recall = match_total / ref_total\n",
    "f_score = (2*precision*recall) / (precision + recall)\n",
    "print(precision, recall, f_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7203389830508474 0.6640625 0.6910569105691058\n"
     ]
    }
   ],
   "source": [
    "est_total = 0\n",
    "ref_total = 0\n",
    "match_total = 0\n",
    "i = 0\n",
    "for file in audio_files:\n",
    "    est_cnt, ref_cnt, match_cnt = eval_f0_note_tracker(file, pitch_diff_thres=0.5, note_dur_thres=70)\n",
    "    est_total += est_cnt\n",
    "    ref_total += ref_cnt\n",
    "    match_total += match_cnt\n",
    "    # print(est_cnt, ref_cnt, match_cnt)\n",
    "    i += 1\n",
    "    if i == 50:\n",
    "        break\n",
    "precision = match_total / est_total\n",
    "recall = match_total / ref_total\n",
    "f_score = (2*precision*recall) / (precision + recall)\n",
    "print(precision, recall, f_score)"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "59d85d0919029fe908500b29ceb55c54859f0fca8114ead43d175dc230632326"
  },
  "kernelspec": {
   "display_name": "Python 3.7.11 64-bit ('GTP': conda)",
   "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": "3.7.11"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}