step_1_midi_data_preprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# import library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Current Time:     2019/10/18  17:48:52\n",
      "[info] Python Version:   3.6.8\n",
      "[info] Working Dir:      /host/home/python/musegan_npz/github_repository/ver2/\n"
     ]
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore',category=FutureWarning)\n",
    "\n",
    "import librosa, IPython, datetime, time, os, sys, copy, dill, pickle, mir_eval, glob\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import IPython.display as ipd\n",
    "from scipy.spatial.distance import euclidean, pdist, squareform\n",
    "from time import gmtime, strftime\n",
    "from imageio import imread as imread\n",
    "from imageio import imsave as imsave\n",
    "import librosa.display\n",
    "import tensorflow as tf\n",
    "import IPython.display as ipd\n",
    "import soundfile as sf\n",
    "from pypianoroll import Multitrack, Track\n",
    "from matplotlib import pyplot as plt\n",
    "import pretty_midi\n",
    "from midiutil.MidiFile import MIDIFile\n",
    "#from datetime import datetime\n",
    "#from ops import *\n",
    "#import pydub\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "print (\"[info] Current Time:     \" + datetime.datetime.now().strftime('%Y/%m/%d  %H:%M:%S'))\n",
    "print (\"[info] Python Version:   \" + sys.version.split('\\n')[0].split(' ')[0])\n",
    "print (\"[info] Working Dir:      \" + os.getcwd()+'/')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Synthesis function define done.\n"
     ]
    }
   ],
   "source": [
    "# Ensure File DIR\n",
    "def ensure_dir(file_path):\n",
    "    directory = os.path.dirname(file_path)\n",
    "    if not os.path.exists(directory):\n",
    "        os.makedirs(directory)\n",
    "\n",
    "# Define function for MIDI to wav synthesis\n",
    "def syn_midi(input_midi, output_file, samp_rate):\n",
    "    \n",
    "    tmp_fname = input_midi.split('/')[-1][:-4] + '_tmp_{}.wav'.format(str(datetime.datetime.now().strftime('%H%M%S')))\n",
    "\n",
    "    # run MIDI to audio synthesis\n",
    "    syn_cmd = \"fluidsynth -ni sound_front_lib.sf2 {} -F {} -r {}\".format(input_midi, tmp_fname, samp_rate)\n",
    "    os.system(syn_cmd)\n",
    "\n",
    "    #use only 16-bit mono sound, and do volume normalization\n",
    "    y, sr = librosa.load('{}'.format(tmp_fname), sr=samp_rate, mono=True)\n",
    "    y = y*0.95/np.max([np.abs(np.max(y)), np.abs(np.min(y))])\n",
    "    sf.write(output_file, (y * np.iinfo(np.int16).max).astype(np.int16), sr, 'PCM_16')\n",
    "    try:\n",
    "        os.system('rm {}'.format(tmp_fname))\n",
    "    except:\n",
    "        print ('file is not exist')\n",
    "\n",
    "    #print(\"MIDI to Audio transfer is done, synthesized file name: {}\".format(output_file))\n",
    "    \n",
    "print ('Synthesis function define done.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Read all MIDI files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total files: 24\n",
      "./input_midi/Beatles_20/02_all_tracks/01_A_Hard_Days_Night_all_tracks.mid\n",
      "./input_midi/Beatles_20/02_all_tracks/02_Anna_all_tracks.mid\n",
      "./input_midi/Beatles_20/02_all_tracks/03_Back_In_The_USSR_all_tracks.mid\n",
      "Total files: 24\n",
      "./input_midi/Beatles_20/03_no_drum/01_A_Hard_Days_Night_no_drum.mid\n",
      "./input_midi/Beatles_20/03_no_drum/02_Anna_no_drum.mid\n",
      "./input_midi/Beatles_20/03_no_drum/03_Back_In_The_USSR_no_drum.mid\n",
      "Total files: 24\n",
      "./input_midi/Beatles_20/04_drum_only/01_A_Hard_Days_Night_drum_only.mid\n",
      "./input_midi/Beatles_20/04_drum_only/02_Anna_drum_only.mid\n",
      "./input_midi/Beatles_20/04_drum_only/03_Back_In_The_USSR_drum_only.mid\n"
     ]
    }
   ],
   "source": [
    "all_tracks_mid_flist = np.sort(glob.glob('./input_midi/**/*.mid', recursive=True)).tolist()\n",
    "all_tracks_mid_flist = np.sort([x.replace(' ','') for x in all_tracks_mid_flist if \"all_tracks.mid\" in x]).tolist()\n",
    "print ('Total files: {}'.format(len(all_tracks_mid_flist)))\n",
    "# show file format\n",
    "for x in all_tracks_mid_flist[:3]: print (x)\n",
    "\n",
    "    \n",
    "no_drum_mid_flist = np.sort(glob.glob('./input_midi/**/*.mid', recursive=True)).tolist()\n",
    "no_drum_mid_flist = np.sort([x.replace(' ','') for x in no_drum_mid_flist if \"no_drum.mid\" in x]).tolist()\n",
    "print ('Total files: {}'.format(len(no_drum_mid_flist)))\n",
    "# show file format\n",
    "for x in no_drum_mid_flist[:3]: print (x)\n",
    "    \n",
    "\n",
    "drum_only_mid_flist = np.sort(glob.glob('./input_midi/**/*.mid', recursive=True)).tolist()\n",
    "drum_only_mid_flist = np.sort([x.replace(' ','') for x in drum_only_mid_flist if \"drum_only.mid\" in x]).tolist()\n",
    "print ('Total files: {}'.format(len(drum_only_mid_flist)))\n",
    "# show file format\n",
    "for x in drum_only_mid_flist[:3]: print (x)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test SYN MIDI function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] MIDI: ./input_midi/Michael_Jackson_5/02_all_tracks/22_Michael_Jackson_-_Billie_Jean_all_tracks.mid\n",
      "[info] WAVE: ./pre_processed_data/syn_wav/22_Michael_Jackson_-_Billie_Jean_all_tracks.wav\n"
     ]
    }
   ],
   "source": [
    "file_idx = 20\n",
    "midi_fname = all_tracks_mid_flist[file_idx]\n",
    "output_wav_name = './pre_processed_data/syn_wav/' + midi_fname.split('/')[-1][:-4] + '.wav'\n",
    "ensure_dir(output_wav_name)\n",
    "syn_midi(midi_fname, output_wav_name, 44100)\n",
    "print ('[info] MIDI: {}'.format(midi_fname))\n",
    "print ('[info] WAVE: {}'.format(output_wav_name))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define get_merged_mt function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] MIDI merge function defined.\n"
     ]
    }
   ],
   "source": [
    "def get_merged_mt(multitrack):\n",
    "    \n",
    "    # define track info\n",
    "    TRACK_INFO = (('Drums', 0),\n",
    "                  ('Piano', 0),\n",
    "                  ('Guitar', 24),\n",
    "                  ('Bass', 32),\n",
    "                  ('Strings', 48))\n",
    "    \n",
    "    \"\"\"Merge the multitrack pianorolls into five instrument families and\n",
    "    return the resulting multitrack pianoroll object.\"\"\"\n",
    "    track_lists_to_merge = [[] for _ in range(5)]\n",
    "    for idx, track in enumerate(multitrack.tracks):\n",
    "        if track.is_drum:\n",
    "            track_lists_to_merge[0].append(idx)\n",
    "        elif track.program//8 == 0:\n",
    "            track_lists_to_merge[1].append(idx)\n",
    "        elif track.program//8 == 3:\n",
    "            track_lists_to_merge[2].append(idx)\n",
    "        elif track.program//8 == 4:\n",
    "            track_lists_to_merge[3].append(idx)\n",
    "        elif track.program < 96 or 104 <= track.program < 112:\n",
    "            track_lists_to_merge[4].append(idx)\n",
    "\n",
    "    tracks = []\n",
    "    for idx, track_list_to_merge in enumerate(track_lists_to_merge):\n",
    "        if track_list_to_merge:\n",
    "            merged = multitrack[track_list_to_merge].get_merged_pianoroll('max')\n",
    "            tracks.append(Track(merged, TRACK_INFO[idx][1], (idx == 0),\n",
    "                                TRACK_INFO[idx][0]))\n",
    "        else:\n",
    "            tracks.append(Track(None, TRACK_INFO[idx][1], (idx == 0),\n",
    "                                TRACK_INFO[idx][0]))\n",
    "    return Multitrack(None, tracks, multitrack.tempo, multitrack.downbeat,\n",
    "                      multitrack.beat_resolution, multitrack.name)\n",
    "\n",
    "print ('[info] MIDI merge function defined.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define MIDI object to store MIDI data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MIDI track object is defined.\n"
     ]
    }
   ],
   "source": [
    "class midi_track(object):\n",
    "    def __init__(self):\n",
    "        self.file_name = \"\"\n",
    "        self.pmidi_data = []\n",
    "        self.pmidi_all_tracks_data = []\n",
    "        self.pmidi_no_drum_data = []\n",
    "        self.pmidi_drum_only_data = []\n",
    "        self.tempo = 0        \n",
    "        self.downbeats_list_fixed = []\n",
    "        self.bar_range_list_fixed = []\n",
    "        self.drum_bar_list = []\n",
    "        self.drum_bar_list_bin = []\n",
    "        self.drum_bar_note_num = []\n",
    "        \n",
    "print ('MIDI track object is defined.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# loop over all files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] file  1:  01_A_Hard_Days_Night_all_tracks.mid\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/pretty_midi/pretty_midi.py:100: RuntimeWarning: Tempo, Key or Time signature change events found on non-zero tracks.  This is not a valid type 0 or type 1 MIDI file.  Tempo, Key or Time Signature may be wrong.\n",
      "  RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] file  2:  02_Anna_all_tracks.mid\n",
      "[info] file  3:  03_Back_In_The_USSR_all_tracks.mid\n",
      "[info] file  4:  04_Cant_Buy_Me_Love_all_tracks.mid\n",
      "[info] file  5:  05_Hold_Me_Tight_all_tracks.mid\n",
      "[info] file  6:  06_I_Call_Your_Name_all_tracks.mid\n",
      "[info] file  7:  07_I_Wanna_Be_Your_Man_all_tracks.mid\n",
      "[info] file  8:  08_Money_all_tracks.mid\n",
      "[info] file  9:  09_The_Word_all_tracks.mid\n",
      "[info] file 10:  10_Free_As_A_Bird_all_tracks.mid\n",
      "[info] file 11:  11_Hey_Jude_all_tracks.mid\n",
      "[info] file 12:  12_Little_Child_all_tracks.mid\n",
      "[info] file 13:  13_Hey_Bulldog_all_tracks.mid\n",
      "[info] file 14:  14_Lovely_Rita_all_tracks.mid\n",
      "[info] file 15:  15_The_Night_Before_all_tracks.mid\n",
      "[info] file 16:  16_From_Me_To_You_all_tracks.mid\n",
      "[info] file 17:  17_Roll_Over_Beethoven_all_tracks.mid\n",
      "[info] file 18:  18_Come_Together_all_tracks.mid\n",
      "[info] file 19:  19_Babys_In_Black_all_tracks.mid\n",
      "[info] file 20:  21_Michael_Jackson_-_Bad_all_tracks.mid\n",
      "[info] file 21:  22_Michael_Jackson_-_Billie_Jean_all_tracks.mid\n",
      "[info] file 22:  23_Michael_Jackson_-_Man_In_The_Mirror_all_tracks.mid\n",
      "[info] file 23:  24_Michael_Jackson_-_Smooth_Criminal_all_tracks.mid\n",
      "[info] file 24:  25_Michael_Jackson_-_Beat_It_all_tracks.mid\n",
      "\n",
      "[info] All 24 files are processed.\n"
     ]
    }
   ],
   "source": [
    "midi_obj_list = []\n",
    "\n",
    "total_files = len(all_tracks_mid_flist)\n",
    "#total_files = 1\n",
    "\n",
    "for file_idx in range(0, total_files):\n",
    "\n",
    "    midi_fname = all_tracks_mid_flist[file_idx]\n",
    "    print ('[info] file {:2d}:  {}'.format(file_idx+1, midi_fname.split('/')[-1]))\n",
    "    \n",
    "    midi_data1_mt = get_merged_mt(Multitrack(midi_fname))\n",
    "    midi_data1 = midi_data1_mt.to_pretty_midi()\n",
    "    \n",
    "    midi_fname = no_drum_mid_flist[file_idx]\n",
    "    #print (midi_fname)\n",
    "    midi_data2_mt = get_merged_mt(Multitrack(midi_fname))\n",
    "    midi_data2 = midi_data2_mt.to_pretty_midi()\n",
    "    \n",
    "    midi_fname = drum_only_mid_flist[file_idx]\n",
    "    #print (midi_fname)\n",
    "    midi_data3_mt = get_merged_mt(Multitrack(midi_fname))\n",
    "    midi_data3 = midi_data3_mt.to_pretty_midi()\n",
    "\n",
    "    # make sure all beat is the same\n",
    "    all_beat_period = np.sort(midi_data1.get_downbeats()[1:] - midi_data1.get_downbeats()[:-1])\n",
    "    all_beat_period_avg = np.mean(all_beat_period)\n",
    "    assert abs((all_beat_period[0] - all_beat_period_avg)/all_beat_period_avg) < 0.05\n",
    "    assert abs((all_beat_period[-1] - all_beat_period_avg)/all_beat_period_avg) < 0.05\n",
    "    \n",
    "    # make sure all tracks are the longest\n",
    "    assert (midi_data1.get_end_time()>=midi_data2.get_end_time())\n",
    "    assert (midi_data1.get_end_time()>=midi_data3.get_end_time())\n",
    "    \n",
    "    \n",
    "    # init midi object\n",
    "    midi_track_obj = midi_track()\n",
    "    midi_track_obj.file_name = all_tracks_mid_flist[file_idx]\n",
    "    midi_track_obj.pmidi_data = copy.deepcopy(midi_data1)    \n",
    "    midi_track_obj.pmidi_all_tracks_data = copy.deepcopy(midi_data1)\n",
    "    midi_track_obj.pmidi_no_drum_data = copy.deepcopy(midi_data1)\n",
    "    midi_track_obj.pmidi_drum_only_data = copy.deepcopy(midi_data1)\n",
    "    \n",
    "    # get song tempo\n",
    "    beats_list = midi_track_obj.pmidi_data.get_beats().tolist()\n",
    "    midi_track_obj.tempo = np.round(60./np.mean(np.array(beats_list[1:]) - np.array(beats_list[:-1])), 2)\n",
    "    \n",
    "    # get fixed downbeats data\n",
    "    downbeats_list = midi_track_obj.pmidi_data.get_downbeats().tolist()\n",
    "    \n",
    "    bar_portion = 96\n",
    "    basic_time_unit = np.mean(np.array(downbeats_list[1:]) - np.array(downbeats_list[:-1]))/bar_portion\n",
    "    \n",
    "    # calculate downbeat time range\n",
    "    downbeats_range_p = np.array(copy.deepcopy(downbeats_list)) + basic_time_unit*0.5\n",
    "    downbeats_range_n = np.array(copy.deepcopy(downbeats_list)) - basic_time_unit*0.5\n",
    "    downbeats_range = np.vstack([downbeats_range_n, downbeats_range_p])\n",
    "    downbeats_range[downbeats_range<0] = 0 \n",
    "    downbeats_range_cnt = np.zeros(downbeats_range.shape[1])\n",
    "    downbeats_range_p1_cnt = np.zeros(downbeats_range.shape[1])\n",
    "    downbeats_range_n1_cnt = np.zeros(downbeats_range.shape[1])\n",
    "\n",
    "    downbeats_range_num = len(downbeats_list)\n",
    "    \n",
    "    for inst_idx in range(0, len(midi_track_obj.pmidi_data.instruments)):\n",
    "        for note_data in midi_track_obj.pmidi_data.instruments[inst_idx].notes:\n",
    "            note_onset = note_data.start\n",
    "            range_found = 0\n",
    "            # check note in standard downbeats_range\n",
    "            for beat_range_idx in range(0, downbeats_range_num):\n",
    "                range_start = downbeats_range[0, beat_range_idx]\n",
    "                range_end = downbeats_range[1, beat_range_idx]\n",
    "                if note_onset>=range_start and note_onset<range_end:\n",
    "                    downbeats_range_cnt[beat_range_idx] += 1\n",
    "                    range_found = 1\n",
    "                    break\n",
    "\n",
    "            if (range_found==1):\n",
    "                continue\n",
    "                \n",
    "            # check note in standard downbeats_range + 1 basic time unit\n",
    "            for beat_range_idx in range(0, downbeats_range_num):\n",
    "                range_start = downbeats_range[0, beat_range_idx] + basic_time_unit\n",
    "                range_end = downbeats_range[1, beat_range_idx] + basic_time_unit\n",
    "                if note_onset>=range_start and note_onset<range_end:\n",
    "                    downbeats_range_p1_cnt[beat_range_idx] += 1\n",
    "                    range_found = 1\n",
    "                    break         \n",
    "            \n",
    "            if (range_found==1):\n",
    "                continue        \n",
    "        \n",
    "            # check note in standard downbeats_range - 1 basic time unit\n",
    "            for beat_range_idx in range(0, downbeats_range_num):\n",
    "                range_start = downbeats_range[0, beat_range_idx] - basic_time_unit\n",
    "                range_end = downbeats_range[1, beat_range_idx] - basic_time_unit\n",
    "                if note_onset>=range_start and note_onset<range_end:\n",
    "                    downbeats_range_n1_cnt[beat_range_idx] += 1\n",
    "                    break \n",
    "    \n",
    "    # calculate downbeat shift\n",
    "    downbeats_range_shift = np.zeros([downbeats_range_n1_cnt.shape[0]])\n",
    "\n",
    "    for x in range(downbeats_range_n1_cnt.shape[0]):\n",
    "        if max([downbeats_range_p1_cnt[x], downbeats_range_n1_cnt[x]]) > downbeats_range_cnt[x]:\n",
    "            if downbeats_range_p1_cnt[x] > downbeats_range_n1_cnt[x]:\n",
    "                downbeats_range_shift[x] = 1\n",
    "            else:\n",
    "                downbeats_range_shift[x] = -1\n",
    "                \n",
    "    # fix downbeats_list\n",
    "    downbeats_list_fixed = np.array(copy.deepcopy(downbeats_list)) + downbeats_range_shift*basic_time_unit\n",
    "    downbeats_list_fixed[downbeats_list_fixed<0] = 0\n",
    "    downbeats_list_fixed = downbeats_list_fixed.tolist()\n",
    "    \n",
    "\n",
    "    \n",
    "    # calculate bar_range_list using downbeats_list_fixed\n",
    "    bar_range_list_fixed = []\n",
    "    for beat_idx in range(0, len(downbeats_list_fixed)-1):\n",
    "        beat_period = downbeats_list_fixed[beat_idx+1] - downbeats_list_fixed[beat_idx]\n",
    "        \n",
    "        basic_unit = beat_period/bar_portion\n",
    "        bar_start = max(downbeats_list_fixed[beat_idx] - basic_unit/2., 0)\n",
    "        bar_end = min(downbeats_list_fixed[beat_idx+1] - basic_unit/2., downbeats_list_fixed[len(downbeats_list_fixed)-1])\n",
    "        \n",
    "        bar_range_list_fixed.append([bar_start, bar_end])\n",
    "    \n",
    "    #define function to process note\n",
    "    def check_note_in_bar(cnib_note_time, cnib_bar_range_list):\n",
    "        if cnib_note_time>=cnib_bar_range_list[0][0] and cnib_note_time<cnib_bar_range_list[-1][-1]:\n",
    "            return(True)\n",
    "        else:\n",
    "            return(False)\n",
    "\n",
    "    def get_bar_num(gbn_note_time, gbn_bar_range_list):\n",
    "        for bar_idx in range(0, len(gbn_bar_range_list)):\n",
    "            if gbn_note_time>=gbn_bar_range_list[bar_idx][0] and gbn_note_time<gbn_bar_range_list[bar_idx][1]:\n",
    "                return(bar_idx)\n",
    "            \n",
    "    def get_beat_num(gbtn_note_time, gbtn_bar_num, gbtn_downbeats_list):\n",
    "        time_plus_in_bar = gbtn_note_time - gbtn_downbeats_list[gbtn_bar_num]\n",
    "        time_plus_in_bar_q = time_plus_in_bar/((gbtn_downbeats_list[gbtn_bar_num+1] - gbtn_downbeats_list[gbtn_bar_num])/bar_portion)\n",
    "        time_plus_in_bar_q = np.round(time_plus_in_bar_q).astype(int)\n",
    "        return (time_plus_in_bar_q)\n",
    "        \n",
    "    def get_bar_beat(gbb_note_time, gbb_downbeats_list, gbb_bar_range_list):\n",
    "        note_bar_tmp = get_bar_num(gbb_note_time, gbb_bar_range_list)\n",
    "        note_beat_tmp = get_beat_num(gbb_note_time, note_bar_tmp, gbb_downbeats_list)\n",
    "        return(note_bar_tmp, note_beat_tmp)\n",
    "    \n",
    "    # put all drum note into bar list\n",
    "    # create empty list[ ary, ary, ary... ]\n",
    "    drum_bar_list = []\n",
    "    for drum_bar_idx in range(0, len(bar_range_list_fixed)):\n",
    "        bar_note_ary = np.zeros([128, bar_portion])\n",
    "        drum_bar_list.append(bar_note_ary)\n",
    "        \n",
    "    \n",
    "    # get drum instrument index\n",
    "    durm_inst_idx_list = []\n",
    "    for Instrument_idx in range(len(midi_data1.instruments)):\n",
    "        if midi_data1.instruments[Instrument_idx].is_drum==True:\n",
    "            durm_inst_idx_list.append(Instrument_idx)       \n",
    "    assert (len(durm_inst_idx_list)==1)\n",
    "    durm_inst_idx = durm_inst_idx_list[0]\n",
    "    \n",
    "    # put note into drum_bar_list\n",
    "    for note_data in midi_track_obj.pmidi_data.instruments[durm_inst_idx].notes:\n",
    "        \n",
    "        note_onset = note_data.start\n",
    "        note_pitch = note_data.pitch\n",
    "        note_velocity = note_data.velocity\n",
    "        \n",
    "        if check_note_in_bar(note_onset, bar_range_list_fixed):\n",
    "            note_bar_0, note_beat_0 = get_bar_beat(note_onset, downbeats_list_fixed, bar_range_list_fixed)        \n",
    "            \n",
    "            drum_bar_list[note_bar_0][note_pitch, note_beat_0] = note_velocity\n",
    "\n",
    "    # save downbeats_list_fixed\n",
    "    midi_track_obj.downbeats_list_fixed = copy.deepcopy(downbeats_list_fixed)\n",
    "    \n",
    "    # save bar_range_list_fixed\n",
    "    midi_track_obj.bar_range_list_fixed = copy.deepcopy(bar_range_list_fixed)\n",
    "    \n",
    "    # save drum bar and notes\n",
    "    midi_track_obj.drum_bar_list = copy.deepcopy(drum_bar_list)\n",
    "    midi_track_obj.drum_bar_list_bin = copy.deepcopy(drum_bar_list)\n",
    "    \n",
    "    # make note with > 0 value = 1.0\n",
    "    for bar_idx in range(0, len(midi_track_obj.drum_bar_list_bin)):\n",
    "        midi_track_obj.drum_bar_list_bin[bar_idx][midi_track_obj.drum_bar_list_bin[bar_idx]>=0.5] = 1.0\n",
    "        midi_track_obj.drum_bar_list_bin[bar_idx][midi_track_obj.drum_bar_list_bin[bar_idx]<0.5] = 0.0\n",
    "    \n",
    "        # save note count\n",
    "        midi_track_obj.drum_bar_note_num.append(np.round(np.sum(midi_track_obj.drum_bar_list_bin[bar_idx])).astype(int))\n",
    "    \n",
    "    \n",
    "    # save all_tracks_file_name\n",
    "    midi_fname_nodrum = './pre_processed_data/proc_all_tracks_mid/{}_all_tracks.mid'.format(midi_track_obj.file_name.split('/')[-1][:-15])\n",
    "    ensure_dir(midi_fname_nodrum)\n",
    "    midi_track_obj.pmidi_all_tracks_data.write(midi_fname_nodrum)\n",
    "    #os.system('chown -R sma1033:sma1033 {}'.format(midi_fname_nodrum))\n",
    "    \n",
    "    \n",
    "    # set drum track note velocity=0\n",
    "    for inst_idx in range(len(midi_track_obj.pmidi_no_drum_data.instruments)):\n",
    "        if inst_idx==durm_inst_idx:\n",
    "            for note in midi_track_obj.pmidi_no_drum_data.instruments[inst_idx].notes:        \n",
    "                note.velocity = 0\n",
    "                                    \n",
    "    # save no_drum_midi_file_name\n",
    "    midi_fname_nodrum = './pre_processed_data/proc_no_drum_mid/{}_no_drum.mid'.format(midi_track_obj.file_name.split('/')[-1][:-15])\n",
    "    ensure_dir(midi_fname_nodrum)\n",
    "    midi_track_obj.pmidi_no_drum_data.write(midi_fname_nodrum)\n",
    "    #os.system('chown -R sma1033:sma1033 {}'.format(midi_fname_nodrum))\n",
    "    \n",
    "    wav_fname_nodrum = './pre_processed_data/proc_no_drum_wav/{}_no_drum.wav'.format(midi_track_obj.file_name.split('/')[-1][:-15])\n",
    "    ensure_dir(wav_fname_nodrum)\n",
    "    syn_midi(midi_fname_nodrum, wav_fname_nodrum, 44100)\n",
    "    #os.system('chown -R sma1033:sma1033 {}'.format(wav_fname_nodrum))\n",
    "    \n",
    "    \n",
    "    # set Not drum track note velocity=0\n",
    "    for inst_idx in range(len(midi_track_obj.pmidi_drum_only_data.instruments)):\n",
    "        if inst_idx!=durm_inst_idx:\n",
    "            for note in midi_track_obj.pmidi_drum_only_data.instruments[inst_idx].notes:        \n",
    "                note.velocity = 0\n",
    "                \n",
    "    # save no_drum_midi_file_name\n",
    "    midi_fname_nodrum = './pre_processed_data/proc_drum_only_mid/{}_drum_only.mid'.format(midi_track_obj.file_name.split('/')[-1][:-15])\n",
    "    ensure_dir(midi_fname_nodrum)\n",
    "    midi_track_obj.pmidi_drum_only_data.write(midi_fname_nodrum)\n",
    "    #os.system('chown -R sma1033:sma1033 {}'.format(midi_fname_nodrum))\n",
    "\n",
    "\n",
    "    # save midi obj into a list\n",
    "    midi_obj_list.append(midi_track_obj)\n",
    "    \n",
    "    \n",
    "print ('\\n[info] All {} files are processed.'.format(total_files))\n",
    "\n",
    "obj_file_name = './pre_processed_data/proc_midi_object.pkl'\n",
    "ensure_dir(obj_file_name)\n",
    "with open(obj_file_name, 'wb') as pkl_file:\n",
    "    pickle.dump(midi_obj_list, pkl_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "scrolled": true
   },
   "source": [
    "# Extract MIDI file beat/bar information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Song: 0 done, total bars: 101\n",
      "Song: 1 done, total bars: 81\n",
      "Song: 2 done, total bars: 91\n",
      "Song: 3 done, total bars: 94\n",
      "Song: 4 done, total bars: 86\n",
      "Song: 5 done, total bars: 84\n",
      "Song: 6 done, total bars: 101\n",
      "Song: 7 done, total bars: 91\n",
      "Song: 8 done, total bars: 88\n",
      "Song: 9 done, total bars: 79\n",
      "Song: 10 done, total bars: 140\n",
      "Song: 11 done, total bars: 68\n",
      "Song: 12 done, total bars: 74\n",
      "Song: 13 done, total bars: 62\n",
      "Song: 14 done, total bars: 109\n",
      "Song: 15 done, total bars: 67\n",
      "Song: 16 done, total bars: 111\n",
      "Song: 17 done, total bars: 95\n",
      "Song: 18 done, total bars: 52\n",
      "Song: 19 done, total bars: 118\n",
      "Song: 20 done, total bars: 140\n",
      "Song: 21 done, total bars: 133\n",
      "Song: 22 done, total bars: 100\n",
      "Song: 23 done, total bars: 146\n",
      "\n",
      "All 24 songs process done.\n"
     ]
    }
   ],
   "source": [
    "# reload processed MIDI object\n",
    "obj_file_name = './pre_processed_data/proc_midi_object.pkl'\n",
    "with open(obj_file_name, 'rb') as pkl_file:\n",
    "    midi_obj_list_reload = pickle.load(pkl_file)\n",
    "\n",
    "\n",
    "# create empty list to store beat/bar info.\n",
    "song_bar_grid_range_list = []\n",
    "\n",
    "total_songs = len(midi_obj_list_reload)\n",
    "\n",
    "for file_idx in range(0, total_songs):\n",
    "        \n",
    "    # load downbeat data\n",
    "    song_downbeats_list_fixed = midi_obj_list_reload[file_idx].downbeats_list_fixed\n",
    "    song_bar_range_list_fixed = midi_obj_list_reload[file_idx].bar_range_list_fixed\n",
    "    \n",
    "    bar_resolution = 96\n",
    "\n",
    "    bar_grid_range_list = []\n",
    "\n",
    "    beat_num = len(song_downbeats_list_fixed) - 1\n",
    "\n",
    "    for beat_idx in range(0, beat_num):\n",
    "        bar_start_edge = song_downbeats_list_fixed[beat_idx]\n",
    "        bar_end_edge = song_downbeats_list_fixed[beat_idx+1]\n",
    "        time_grid = np.linspace(bar_start_edge, bar_end_edge, bar_resolution+1)\n",
    "        \n",
    "        bar_grid_range = np.zeros([bar_resolution, 2])\n",
    "\n",
    "        bar_grid_range[:, 0] = time_grid[ :-1]\n",
    "        bar_grid_range[:, 1] = time_grid[ 1: ]\n",
    "        bar_grid_range -= ((time_grid[-1] - time_grid[0])/bar_resolution)*0.5\n",
    "        bar_grid_range[bar_grid_range<0] = 0\n",
    "        \n",
    "        bar_grid_range_list.append(copy.copy(bar_grid_range))\n",
    "        \n",
    "    print ('Song: {} done, total bars: {}'.format(file_idx, len(bar_grid_range_list)))\n",
    "    \n",
    "    song_bar_grid_range_list.append(copy.copy(bar_grid_range_list))\n",
    "        \n",
    "print ('\\nAll {} songs process done.'.format(total_songs))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Check Processed beat/bar data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 99.98958333 100.01041667]\n",
      " [100.01041667 100.03125   ]\n",
      " [100.03125    100.05208333]\n",
      " [100.05208333 100.07291667]\n",
      " [100.07291667 100.09375   ]\n",
      " [100.09375    100.11458333]\n",
      " [100.11458333 100.13541667]\n",
      " [100.13541667 100.15625   ]\n",
      " [100.15625    100.17708333]\n",
      " [100.17708333 100.19791667]\n",
      " [100.19791667 100.21875   ]\n",
      " [100.21875    100.23958333]\n",
      " [100.23958333 100.26041667]\n",
      " [100.26041667 100.28125   ]\n",
      " [100.28125    100.30208333]\n",
      " [100.30208333 100.32291667]\n",
      " [100.32291667 100.34375   ]\n",
      " [100.34375    100.36458333]\n",
      " [100.36458333 100.38541667]\n",
      " [100.38541667 100.40625   ]\n",
      " [100.40625    100.42708333]\n",
      " [100.42708333 100.44791667]\n",
      " [100.44791667 100.46875   ]\n",
      " [100.46875    100.48958333]\n",
      " [100.48958333 100.51041667]\n",
      " [100.51041667 100.53125   ]\n",
      " [100.53125    100.55208333]\n",
      " [100.55208333 100.57291667]\n",
      " [100.57291667 100.59375   ]\n",
      " [100.59375    100.61458333]\n",
      " [100.61458333 100.63541667]\n",
      " [100.63541667 100.65625   ]\n",
      " [100.65625    100.67708333]\n",
      " [100.67708333 100.69791667]\n",
      " [100.69791667 100.71875   ]\n",
      " [100.71875    100.73958333]\n",
      " [100.73958333 100.76041667]\n",
      " [100.76041667 100.78125   ]\n",
      " [100.78125    100.80208333]\n",
      " [100.80208333 100.82291667]\n",
      " [100.82291667 100.84375   ]\n",
      " [100.84375    100.86458333]\n",
      " [100.86458333 100.88541667]\n",
      " [100.88541667 100.90625   ]\n",
      " [100.90625    100.92708333]\n",
      " [100.92708333 100.94791667]\n",
      " [100.94791667 100.96875   ]\n",
      " [100.96875    100.98958333]\n",
      " [100.98958333 101.01041667]\n",
      " [101.01041667 101.03125   ]\n",
      " [101.03125    101.05208333]\n",
      " [101.05208333 101.07291667]\n",
      " [101.07291667 101.09375   ]\n",
      " [101.09375    101.11458333]\n",
      " [101.11458333 101.13541667]\n",
      " [101.13541667 101.15625   ]\n",
      " [101.15625    101.17708333]\n",
      " [101.17708333 101.19791667]\n",
      " [101.19791667 101.21875   ]\n",
      " [101.21875    101.23958333]\n",
      " [101.23958333 101.26041667]\n",
      " [101.26041667 101.28125   ]\n",
      " [101.28125    101.30208333]\n",
      " [101.30208333 101.32291667]\n",
      " [101.32291667 101.34375   ]\n",
      " [101.34375    101.36458333]\n",
      " [101.36458333 101.38541667]\n",
      " [101.38541667 101.40625   ]\n",
      " [101.40625    101.42708333]\n",
      " [101.42708333 101.44791667]\n",
      " [101.44791667 101.46875   ]\n",
      " [101.46875    101.48958333]\n",
      " [101.48958333 101.51041667]\n",
      " [101.51041667 101.53125   ]\n",
      " [101.53125    101.55208333]\n",
      " [101.55208333 101.57291667]\n",
      " [101.57291667 101.59375   ]\n",
      " [101.59375    101.61458333]\n",
      " [101.61458333 101.63541667]\n",
      " [101.63541667 101.65625   ]\n",
      " [101.65625    101.67708333]\n",
      " [101.67708333 101.69791667]\n",
      " [101.69791667 101.71875   ]\n",
      " [101.71875    101.73958333]\n",
      " [101.73958333 101.76041667]\n",
      " [101.76041667 101.78125   ]\n",
      " [101.78125    101.80208333]\n",
      " [101.80208333 101.82291667]\n",
      " [101.82291667 101.84375   ]\n",
      " [101.84375    101.86458333]\n",
      " [101.86458333 101.88541667]\n",
      " [101.88541667 101.90625   ]\n",
      " [101.90625    101.92708333]\n",
      " [101.92708333 101.94791667]\n",
      " [101.94791667 101.96875   ]\n",
      " [101.96875    101.98958333]]\n"
     ]
    }
   ],
   "source": [
    "file_idx = 10\n",
    "bar_idx = 50\n",
    "bar_grid_range = copy.copy(song_bar_grid_range_list[file_idx][bar_idx])\n",
    "print(bar_grid_range)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# save processed bar range information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File is saved.\n"
     ]
    }
   ],
   "source": [
    "file_name = './pre_processed_data/proc_midi_beat_bar_info_list.pkl'\n",
    "with open(file_name, 'wb') as pkl_file:\n",
    "    pickle.dump(song_bar_grid_range_list, pkl_file)\n",
    "print ('File is saved.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# read wav file names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total files: 24\n",
      "./pre_processed_data/proc_no_drum_wav/01_A_Hard_Days_Night_no_drum.wav\n",
      "./pre_processed_data/proc_no_drum_wav/02_Anna_no_drum.wav\n",
      "./pre_processed_data/proc_no_drum_wav/03_Back_In_The_USSR_no_drum.wav\n"
     ]
    }
   ],
   "source": [
    "no_drum_wav_flist = np.sort(glob.glob('./pre_processed_data/proc_no_drum_wav/*.wav', recursive=True)).tolist()\n",
    "print ('Total files: {}'.format(len(no_drum_wav_flist)))\n",
    "for x in no_drum_wav_flist[:3]: print (x)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define function to get CQT data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_cqt_data(input_wav_file):\n",
    "    \n",
    "    audio_data, samp_rate = librosa.load(input_wav_file, sr=44100, mono=True)\n",
    "    audio_len = float(len(audio_data)) / float(samp_rate)\n",
    "    audio_data_cqt = librosa.amplitude_to_db(np.abs(librosa.cqt(audio_data, sr=samp_rate,  hop_length=256)), ref=np.max)\n",
    "    audio_cqt_fps = audio_data_cqt.shape[1] / audio_len\n",
    "    \n",
    "    return (audio_data_cqt, audio_cqt_fps)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# loop over files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Conversion starts...\n",
      "[info] Current Time:     2019/10/18  17:54:57\n",
      "\n",
      "[info] MIDI file: 01_A_Hard_Days_Night_no_drum.mid\n",
      "[info] song: 01_A_Hard_Days_Night_no_drum done, total bars: 101\n",
      "[info] File [ 1 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:03\n",
      "\n",
      "[info] MIDI file: 02_Anna_no_drum.mid\n",
      "[info] song: 02_Anna_no_drum done, total bars: 81\n",
      "[info] File [ 2 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:08\n",
      "\n",
      "[info] MIDI file: 03_Back_In_The_USSR_no_drum.mid\n",
      "[info] song: 03_Back_In_The_USSR_no_drum done, total bars: 91\n",
      "[info] File [ 3 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:14\n",
      "\n",
      "[info] MIDI file: 04_Cant_Buy_Me_Love_no_drum.mid\n",
      "[info] song: 04_Cant_Buy_Me_Love_no_drum done, total bars: 94\n",
      "[info] File [ 4 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:20\n",
      "\n",
      "[info] MIDI file: 05_Hold_Me_Tight_no_drum.mid\n",
      "[info] song: 05_Hold_Me_Tight_no_drum done, total bars: 86\n",
      "[info] File [ 5 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:25\n",
      "\n",
      "[info] MIDI file: 06_I_Call_Your_Name_no_drum.mid\n",
      "[info] song: 06_I_Call_Your_Name_no_drum done, total bars: 84\n",
      "[info] File [ 6 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:30\n",
      "\n",
      "[info] MIDI file: 07_I_Wanna_Be_Your_Man_no_drum.mid\n",
      "[info] song: 07_I_Wanna_Be_Your_Man_no_drum done, total bars: 101\n",
      "[info] File [ 7 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:37\n",
      "\n",
      "[info] MIDI file: 08_Money_no_drum.mid\n",
      "[info] song: 08_Money_no_drum done, total bars: 91\n",
      "[info] File [ 8 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:42\n",
      "\n",
      "[info] MIDI file: 09_The_Word_no_drum.mid\n",
      "[info] song: 09_The_Word_no_drum done, total bars: 88\n",
      "[info] File [ 9 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:48\n",
      "\n",
      "[info] MIDI file: 10_Free_As_A_Bird_no_drum.mid\n",
      "[info] song: 10_Free_As_A_Bird_no_drum done, total bars: 79\n",
      "[info] File [ 10 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:55:52\n",
      "\n",
      "[info] MIDI file: 11_Hey_Jude_no_drum.mid\n",
      "[info] song: 11_Hey_Jude_no_drum done, total bars: 140\n",
      "[info] File [ 11 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:01\n",
      "\n",
      "[info] MIDI file: 12_Little_Child_no_drum.mid\n",
      "[info] song: 12_Little_Child_no_drum done, total bars: 68\n",
      "[info] File [ 12 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:05\n",
      "\n",
      "[info] MIDI file: 13_Hey_Bulldog_no_drum.mid\n",
      "[info] song: 13_Hey_Bulldog_no_drum done, total bars: 74\n",
      "[info] File [ 13 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:10\n",
      "\n",
      "[info] MIDI file: 14_Lovely_Rita_no_drum.mid\n",
      "[info] song: 14_Lovely_Rita_no_drum done, total bars: 62\n",
      "[info] File [ 14 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:14\n",
      "\n",
      "[info] MIDI file: 15_The_Night_Before_no_drum.mid\n",
      "[info] song: 15_The_Night_Before_no_drum done, total bars: 109\n",
      "[info] File [ 15 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:20\n",
      "\n",
      "[info] MIDI file: 16_From_Me_To_You_no_drum.mid\n",
      "[info] song: 16_From_Me_To_You_no_drum done, total bars: 67\n",
      "[info] File [ 16 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:24\n",
      "\n",
      "[info] MIDI file: 17_Roll_Over_Beethoven_no_drum.mid\n",
      "[info] song: 17_Roll_Over_Beethoven_no_drum done, total bars: 111\n",
      "[info] File [ 17 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:31\n",
      "\n",
      "[info] MIDI file: 18_Come_Together_no_drum.mid\n",
      "[info] song: 18_Come_Together_no_drum done, total bars: 95\n",
      "[info] File [ 18 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:37\n",
      "\n",
      "[info] MIDI file: 19_Babys_In_Black_no_drum.mid\n",
      "[info] song: 19_Babys_In_Black_no_drum done, total bars: 52\n",
      "[info] File [ 19 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:40\n",
      "\n",
      "[info] MIDI file: 21_Michael_Jackson_-_Bad_no_drum.mid\n",
      "[info] song: 21_Michael_Jackson_-_Bad_no_drum done, total bars: 118\n",
      "[info] File [ 20 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:47\n",
      "\n",
      "[info] MIDI file: 22_Michael_Jackson_-_Billie_Jean_no_drum.mid\n",
      "[info] song: 22_Michael_Jackson_-_Billie_Jean_no_drum done, total bars: 140\n",
      "[info] File [ 21 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:56:56\n",
      "\n",
      "[info] MIDI file: 23_Michael_Jackson_-_Man_In_The_Mirror_no_drum.mid\n",
      "[info] song: 23_Michael_Jackson_-_Man_In_The_Mirror_no_drum done, total bars: 133\n",
      "[info] File [ 22 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:57:04\n",
      "\n",
      "[info] MIDI file: 24_Michael_Jackson_-_Smooth_Criminal_no_drum.mid\n",
      "[info] song: 24_Michael_Jackson_-_Smooth_Criminal_no_drum done, total bars: 100\n",
      "[info] File [ 23 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:57:10\n",
      "\n",
      "[info] MIDI file: 25_Michael_Jackson_-_Beat_It_no_drum.mid\n",
      "[info] song: 25_Michael_Jackson_-_Beat_It_no_drum done, total bars: 146\n",
      "[info] File [ 24 / 24 ] done.\n",
      "[info] Current Time:     2019/10/18  17:57:19\n",
      "\n",
      "\n",
      "[info] Conversion ends.\n",
      "[info] Current Time:     2019/10/18  17:57:19\n"
     ]
    }
   ],
   "source": [
    "print ('[info] Conversion starts...')\n",
    "print (\"[info] Current Time:     \" + datetime.datetime.now().strftime('%Y/%m/%d  %H:%M:%S') + '\\n')\n",
    "\n",
    "total_midi_files = len(song_bar_grid_range_list)\n",
    "\n",
    "bad_file_note_counter = np.zeros(total_midi_files).astype(np.int)\n",
    "\n",
    "midi_index_start = 0\n",
    "midi_index_end = total_midi_files\n",
    "\n",
    "for midi_index in range(midi_index_start, midi_index_end):\n",
    "    \n",
    "    #full_file_name = midi_obj_list_reload[midi_index].file_name\n",
    "    \n",
    "    base_file_name = no_drum_mid_flist[midi_index].split('/')[-1][:-4]\n",
    "\n",
    "    print('[info] MIDI file: {}'.format(no_drum_mid_flist[midi_index].split('/')[-1]))\n",
    "\n",
    "    song_cqt_data, song_cqt_fps = get_cqt_data(no_drum_wav_flist[midi_index])\n",
    "    \n",
    "    file_bar_num = len(song_bar_grid_range_list[midi_index])\n",
    "    \n",
    "    bars_cqt_list = []\n",
    "    \n",
    "    # loop all bars\n",
    "    for bar_idx in range(0, file_bar_num):\n",
    "        single_bar_grid_range = song_bar_grid_range_list[midi_index][bar_idx]\n",
    "    \n",
    "        bar_cqt_notes = []\n",
    "    \n",
    "        # loop all notes in bars\n",
    "        for note_idx in range(0, 96):\n",
    "            note_time_start = single_bar_grid_range[note_idx,0]\n",
    "            note_time_end = single_bar_grid_range[note_idx,1]\n",
    "            \n",
    "            note_frame_start = np.round(note_time_start * song_cqt_fps, 0).astype(np.int)\n",
    "            note_frame_end = np.round(note_time_end * song_cqt_fps, 0).astype(np.int)\n",
    "            \n",
    "            song_cqt_data_slice = copy.copy(song_cqt_data[:, note_frame_start:note_frame_end])\n",
    "            \n",
    "            # filter out zero case\n",
    "            if song_cqt_data_slice.shape[1]<1:\n",
    "                print ('catch a zero slice!')\n",
    "                print (midi_index, bar_idx, note_idx)\n",
    "                #song_cqt_data_slice = np.zeros([song_cqt_data.shape[0], 1])\n",
    "                #song_cqt_data_slice[:] = -80.0                \n",
    "                cqt_note_84 = cqt_note_84\n",
    "                \n",
    "                bad_file_note_counter[midi_index] += 1\n",
    "                \n",
    "            else:\n",
    "                cqt_note_84 = np.mean(song_cqt_data_slice, axis=1)\n",
    "            \n",
    "            bar_cqt_notes.append(cqt_note_84)\n",
    "            \n",
    "        bar_cqt_notes_ary = np.array(bar_cqt_notes).transpose()\n",
    "            \n",
    "        bars_cqt_list.append(bar_cqt_notes_ary)\n",
    "            \n",
    "    file_name = './pre_processed_data/cqt_pooled_data/{}.pkl'.format(base_file_name)\n",
    "    ensure_dir(file_name)\n",
    "    with open(file_name, 'wb') as pkl_file:\n",
    "        pickle.dump(bars_cqt_list, pkl_file)\n",
    "    \n",
    "    print ('[info] song: {} done, total bars: {}'.format(base_file_name, len(bars_cqt_list)))\n",
    "    \n",
    "    print('[info] File [ {} / {} ] done.'.format(midi_index-midi_index_start+1, midi_index_end - midi_index_start))\n",
    "    \n",
    "    print (\"[info] Current Time:     \" + datetime.datetime.now().strftime('%Y/%m/%d  %H:%M:%S') + '\\n')\n",
    "\n",
    "print ('\\n[info] Conversion ends.')\n",
    "print (\"[info] Current Time:     \" + datetime.datetime.now().strftime('%Y/%m/%d  %H:%M:%S'))   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define function to read CQT data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Total songs: 24\n",
      "./pre_processed_data/cqt_pooled_data/01_A_Hard_Days_Night_no_drum.pkl\n",
      "./pre_processed_data/cqt_pooled_data/02_Anna_no_drum.pkl\n",
      "./pre_processed_data/cqt_pooled_data/03_Back_In_The_USSR_no_drum.pkl\n",
      "\n",
      "[info] # of bars in song 0: 101\n",
      "[info] Spectrogram shape: (84, 96)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "read_pooled_cqt_flist = np.sort(glob.glob('./pre_processed_data/cqt_pooled_data/*.pkl', recursive=True)).tolist()\n",
    "print ('[info] Total songs: {}'.format(len(read_pooled_cqt_flist)))\n",
    "for x in read_pooled_cqt_flist[:3]: print (x)\n",
    "\n",
    "def read_pooled_cqt(file_idx,\n",
    "                    read_pooled_cqt_flist=read_pooled_cqt_flist):\n",
    "    \n",
    "    file_name = read_pooled_cqt_flist[file_idx]\n",
    "    with open(file_name, 'rb') as pkl_file:\n",
    "        pooled_cqt_data = pickle.load(pkl_file)\n",
    "        \n",
    "    return (pooled_cqt_data)\n",
    "\n",
    "print ('')\n",
    "print('[info] # of bars in song 0: {}'.format(len(read_pooled_cqt(0))))\n",
    "print('[info] Spectrogram shape: {}'.format(read_pooled_cqt(0)[0].shape))\n",
    "\n",
    "plt.imshow(read_pooled_cqt(1)[2]); plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Define CQT SSM calculation function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] SSM calculation function defined.\n"
     ]
    }
   ],
   "source": [
    "def get_song_ssm(gss_cqt_spec):\n",
    "    \n",
    "    print ('[Info] Start SSM Transfer')\n",
    "    print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))\n",
    "    \n",
    "    audio_spec = np.transpose(gss_cqt_spec)\n",
    "    \n",
    "    audio_df = pd.DataFrame(audio_spec)\n",
    "\n",
    "    print('[Info] frame length: {}'.format(len(audio_df.iloc[:,0])))\n",
    "    print('[Info] pitch class: {}'.format(len(audio_df.iloc[0,:])))\n",
    "\n",
    "    ssm_data_full_pix = squareform(pdist(audio_df, metric='euclidean'))\n",
    "\n",
    "    print ('[Info] SSM data shape: {}'.format(ssm_data_full_pix.shape))    \n",
    "    print ('[Info] SSM transfer done.')\n",
    "    print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))\n",
    "    \n",
    "    return (ssm_data_full_pix)\n",
    "\n",
    "\n",
    "def get_ssm_pix_start_end(gsp_abs_bar, gsp_bar_in_row):\n",
    "    row = gsp_abs_bar // gsp_bar_in_row\n",
    "    col = gsp_abs_bar - (gsp_bar_in_row * row)\n",
    "\n",
    "    row_start = row * 96\n",
    "    row_end = (row+1) * 96\n",
    "\n",
    "    col_start = col * 96\n",
    "    col_end = (col+1) * 96\n",
    "    \n",
    "    return([row_start, row_end, col_start, col_end])\n",
    "\n",
    "\n",
    "def get_ssm_pix_data(gspd_abs_bar, gspd_total_bars_1d, gspd_ssm_data):\n",
    "        \n",
    "    xxyy = get_ssm_pix_start_end(gspd_abs_bar, gspd_total_bars_1d)\n",
    "    \n",
    "    x_start = xxyy[0]\n",
    "    x_end = xxyy[1]\n",
    "    y_start = xxyy[2]\n",
    "    y_end = xxyy[3]\n",
    "    \n",
    "    return(np.mean(gspd_ssm_data[x_start:x_end, y_start:y_end]))\n",
    "\n",
    "\n",
    "def get_bar_lev_ssm(gbls_ssm):\n",
    "    \n",
    "    total_bars_1d = int(np.round(gbls_ssm.shape[1]/96))    \n",
    "    all_ssm_abs_bar_list = list(range(total_bars_1d**2))\n",
    "    bar_mean_ssm_data = [get_ssm_pix_data(abs_bar_idx, total_bars_1d, gbls_ssm) for abs_bar_idx in all_ssm_abs_bar_list]\n",
    "    \n",
    "    return(np.array(bar_mean_ssm_data).reshape(total_bars_1d, total_bars_1d))\n",
    "\n",
    "\n",
    "print ('[info] SSM calculation function defined.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Loop all songs for CQT SSM calculation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Info] start SSM calculation...\n",
      "[Info] 2019-10-18 17:57:20\n",
      "\n",
      "[Info] Song Index: 0, 01_A_Hard_Days_Night_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:20\n",
      "[Info] frame length: 9696\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (9696, 9696)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:23\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:24\n",
      "\n",
      "[Info] Song Index: 1, 02_Anna_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:24\n",
      "[Info] frame length: 7776\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (7776, 7776)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:26\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:26\n",
      "\n",
      "[Info] Song Index: 2, 03_Back_In_The_USSR_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:26\n",
      "[Info] frame length: 8736\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (8736, 8736)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:29\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:30\n",
      "\n",
      "[Info] Song Index: 3, 04_Cant_Buy_Me_Love_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:30\n",
      "[Info] frame length: 9024\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (9024, 9024)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:33\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:33\n",
      "\n",
      "[Info] Song Index: 4, 05_Hold_Me_Tight_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:33\n",
      "[Info] frame length: 8256\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (8256, 8256)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:36\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:36\n",
      "\n",
      "[Info] Song Index: 5, 06_I_Call_Your_Name_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:36\n",
      "[Info] frame length: 8064\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (8064, 8064)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:39\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:39\n",
      "\n",
      "[Info] Song Index: 6, 07_I_Wanna_Be_Your_Man_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:39\n",
      "[Info] frame length: 9696\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (9696, 9696)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:43\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:43\n",
      "\n",
      "[Info] Song Index: 7, 08_Money_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:43\n",
      "[Info] frame length: 8736\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (8736, 8736)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:46\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:47\n",
      "\n",
      "[Info] Song Index: 8, 09_The_Word_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:47\n",
      "[Info] frame length: 8448\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (8448, 8448)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:49\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:50\n",
      "\n",
      "[Info] Song Index: 9, 10_Free_As_A_Bird_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:50\n",
      "[Info] frame length: 7584\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (7584, 7584)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:52\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:57:52\n",
      "\n",
      "[Info] Song Index: 10, 11_Hey_Jude_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:57:52\n",
      "[Info] frame length: 13440\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (13440, 13440)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:57:59\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:00\n",
      "\n",
      "[Info] Song Index: 11, 12_Little_Child_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:00\n",
      "[Info] frame length: 6528\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (6528, 6528)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:02\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:02\n",
      "\n",
      "[Info] Song Index: 12, 13_Hey_Bulldog_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:02\n",
      "[Info] frame length: 7104\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (7104, 7104)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:04\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:04\n",
      "\n",
      "[Info] Song Index: 13, 14_Lovely_Rita_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:04\n",
      "[Info] frame length: 5952\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (5952, 5952)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:06\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:06\n",
      "\n",
      "[Info] Song Index: 14, 15_The_Night_Before_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:06\n",
      "[Info] frame length: 10464\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (10464, 10464)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:10\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:10\n",
      "\n",
      "[Info] Song Index: 15, 16_From_Me_To_You_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:10\n",
      "[Info] frame length: 6432\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (6432, 6432)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:12\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:12\n",
      "\n",
      "[Info] Song Index: 16, 17_Roll_Over_Beethoven_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:12\n",
      "[Info] frame length: 10656\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (10656, 10656)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:17\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:17\n",
      "\n",
      "[Info] Song Index: 17, 18_Come_Together_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:17\n",
      "[Info] frame length: 9120\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (9120, 9120)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:20\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:21\n",
      "\n",
      "[Info] Song Index: 18, 19_Babys_In_Black_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:21\n",
      "[Info] frame length: 4992\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (4992, 4992)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:22\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:22\n",
      "\n",
      "[Info] Song Index: 19, 21_Michael_Jackson_-_Bad_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:22\n",
      "[Info] frame length: 11328\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (11328, 11328)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:27\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:27\n",
      "\n",
      "[Info] Song Index: 20, 22_Michael_Jackson_-_Billie_Jean_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:27\n",
      "[Info] frame length: 13440\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (13440, 13440)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:35\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:35\n",
      "\n",
      "[Info] Song Index: 21, 23_Michael_Jackson_-_Man_In_The_Mirror_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:35\n",
      "[Info] frame length: 12768\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (12768, 12768)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:42\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:43\n",
      "\n",
      "[Info] Song Index: 22, 24_Michael_Jackson_-_Smooth_Criminal_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:43\n",
      "[Info] frame length: 9600\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (9600, 9600)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:46\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:47\n",
      "\n",
      "[Info] Song Index: 23, 25_Michael_Jackson_-_Beat_It_no_drum\n",
      "[Info] Start SSM Transfer\n",
      "[Info] 2019-10-18 17:58:47\n",
      "[Info] frame length: 14016\n",
      "[Info] pitch class: 84\n",
      "[Info] SSM data shape: (14016, 14016)\n",
      "[Info] SSM transfer done.\n",
      "[Info] 2019-10-18 17:58:54\n",
      "[Info] batch SSMs calculation done.\n",
      "[Info] 2019-10-18 17:58:55\n",
      "\n",
      "\n",
      "[info] files write done.\n",
      "[Info] 2019-10-18 17:58:55\n"
     ]
    }
   ],
   "source": [
    "print ('[Info] start SSM calculation...')\n",
    "print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + '\\n')\n",
    "\n",
    "song_idx_idx_start = 0\n",
    "song_idx_idx_end = len(read_pooled_cqt_flist)\n",
    "\n",
    "for song_idx in range(song_idx_idx_start, song_idx_idx_end):\n",
    "\n",
    "    base_file_name = no_drum_mid_flist[song_idx].split('/')[-1][:-4]\n",
    "        \n",
    "    print ('[Info] Song Index: {}, {}'.format(song_idx, base_file_name))\n",
    "\n",
    "    full_song_spec = np.concatenate(read_pooled_cqt(song_idx), axis=1)\n",
    "\n",
    "    ssm_result = get_song_ssm(full_song_spec)\n",
    "\n",
    "    bar_ssm_data_tmp = get_bar_lev_ssm(ssm_result)\n",
    "        \n",
    "    file_name = 'song_barlv_ssm_{:0>5}.pkl'.format(base_file_name)\n",
    "    folder_name = './pre_processed_data/bar_level_cqt_ssm/'\n",
    "    full_file_name = folder_name + file_name\n",
    "    ensure_dir(full_file_name)\n",
    "        \n",
    "    with open(full_file_name, 'wb') as pkl_file:\n",
    "        pickle.dump(bar_ssm_data_tmp, pkl_file)\n",
    "            \n",
    "    print ('[Info] batch SSMs calculation done.')\n",
    "    print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + '\\n')\n",
    "\n",
    "#print(bar_ssm_data_tmp.shape)\n",
    "print ('\\n[info] files write done.')\n",
    "print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Check CQT SSM data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Info] Song Index: 1, ./pre_processed_data/bar_level_cqt_ssm/song_barlv_ssm_02_Anna_no_drum.pkl\n",
      "[info] SSM shape: (81, 81)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def read_cqt_ssm_data(song_idx_in):\n",
    "    cqt_ssm_flist = np.sort(glob.glob('./pre_processed_data/bar_level_cqt_ssm/*.pkl', recursive=True)).tolist()\n",
    "    print ('[Info] Song Index: {}, {}'.format(song_idx, cqt_ssm_flist[song_idx_in]))\n",
    "    with open(cqt_ssm_flist[song_idx_in], 'rb') as pkl_file:\n",
    "        cqt_ssm_data_reload = pickle.load(pkl_file)\n",
    "    return (cqt_ssm_data_reload)\n",
    "\n",
    "# show CQT SSM data\n",
    "song_idx = 1\n",
    "plot_data1 = copy.deepcopy(read_cqt_ssm_data(song_idx))\n",
    "print ('[info] SSM shape: {}'.format(plot_data1.shape))\n",
    "plt.figure(figsize=(8, 8))\n",
    "normal_data = plot_data1\n",
    "abnormal_data = (np.max(normal_data) - normal_data)\n",
    "plt.imshow(abnormal_data, origin='lower', cmap='hot')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Condense drum note time step from 96 to 16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## reload midi object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Total songs: 24\n"
     ]
    }
   ],
   "source": [
    "obj_file_name = './pre_processed_data/proc_midi_object.pkl'\n",
    "with open(obj_file_name, 'rb') as pkl_file:\n",
    "    midi_track_obj_list = pickle.load(pkl_file)\n",
    "print ('[info] Total songs: {}'.format(len(midi_track_obj_list)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## plot data format before condese (128 x 96)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABBgAAAFrCAYAAAB7bzqhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvFvnyVgAAIABJREFUeJzs3XuYnWV5qPH7gRAQoglIjJwsYFoBQZAgtCjKQRGViqDi1mqroNZqAQ8o1MPG1qoUbCtCtVJBqyhg2ZwU5SAV0MopEJQgIGcbBAFNhCCnwLP/+D5wOZkwy/lm1nrnzf27rlxm1qx1r/U4L9j97TXPisxEkiRJkiSpi1WG/QIkSZIkSdLU5wUGSZIkSZLUmRcYJEmSJElSZ15gkCRJkiRJnXmBQZIkSZIkdeYFBkmSJEmS1JkXGCRJkiRJUmdeYJAkSZIkSZ15gUGSJEmSJHWXmUP/A+Qqo/xZf731Rr19su9nc+Vs1jaPzfKbtc1j0/Nh0/Nhs4xmbfPY9HzYHP75AO7u5/9tv8pYFyAi4viIuCsiFvbcdmREXBcRP4mI0yJiVs/3nhcRF0fENRFxdUSsMdZzrAKsMcqfgz/wgVFvn+z72Vw5m7XNY7P8Zm3z2PR82PR82CyjWds8Nj0fNod/PoDb6MOYFxiArwB7jLjtPGDLzHwe8DPg7wAiYhpwAvCuzHwusDPwSD8vRJIkSZIkTV1jXmDIzIuAX4+47dzMXNZ+eQmwYfv33YGfZOaP2/v9KjMfncDXK0mSJEmSChTtDoQnv1PExsC3M3PLUb73LeDkzDwhIt4LzAOeAcwGTsrMI1bQfCfwToCZM2fOO+xjH1vuPhtsuCG3L1o05uub6PvZXDmbtc1js/xmbfPY9HzYHFyztnlsej5sDq5Z2zw2B3M+3n/wwVdk5nZjBvpcwrgxsHCU2z8CnMbvLlQcDNwCrAusCVwM7NbHIohcc5Q///KZz4x6+2Tfz+bK2axtHpvlN2ubx6bnw6bnw2YZzdrmsen5sDn88wHMn7Alj8B8YG7PbUdGxB3Ah4FVgZnttxYBF7UXF+4CfgNsO+ZVDkmSJEmSNKX1u+Txr0bctrj9szFwLe2SR+AcYCvgKOBsYFPgpxPwOiVJkiRJUsGm9XGfvwF2A1aPiEXAYcB+wOo0nyYxE1gGHJKZiyPiQuAtwGPANZl51qS8ckmSJEmSVIyJXvI4g+aiw8to9jEszczPrKDpkkebRTVrm8dm+c3a5rHp+bA5uGZt89j0fNgcXLO2eWzWveTxM8C+7d8/Dhzc5yKIoS6ssGlz2M9tc+Vs1jaPTc+HTc+HzTKatc1j0/Nhc/jngz6XPI75KxLtksdXAzN6bjsSeDPwNOB8ml+TWAK8HDggIk6g2e/w24h4MDOPGfNKhyRJkiRJmrImesnjXwKbZOZ04FggvbggSZIkSVL9JnrJ44Kex90FrBYRq2fmQxP7siVJkiRJUkkmdMnjiNtfB7wrM1+6gqZLHm0W1axtHpvlN2ubx6bnw+bgmrXNY9PzYXNwzdrmsVnxksee258L3AQ82yWPNqdKs7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB30ueRxzB0O75HE+MLfntiMj4g7gw8CqNL8m8fj3Pg0soPn1i7lIkiRJkqTqTeiSx4jYAXgvzSdM7Ax8PiJWnZiXKkmSJEmSSjWhSx6BT9C8o+HD7WNn03x05Xcm9mVLkiRJkqSSTOiSx4g4Brjk8YWPEXEc8N3MPGWUx7nk0WZRzdrmsVl+s7Z5bHo+bA6uWds8Nj0fNgfXrG0emxUveQSOAd7c8/3jgNe55NHmVGjWNo/N8pu1zWPT82HT82GzjGZt89j0fNgc/vmgzyWPY/6KRLvk8dXAjJ7bXg98FlgfeFH+7m0QdwDvi4hDaH79YlXgS2Ne5ZAkSZIkSVPaeJc8zgIeAH4EPNRz+8PApsB2wD7AJsAvO79KSZIkSZJUtPEueTyEZsnj1sDXI+L7mfkuYBFwJ/BT4DGaiwtLJuOFS5IkSZKkcnRa8hgRFwAHZ+b89uvVgK/RXJBYE3hfZh67gqZLHm0W1axtHpvlN2ubx6bnw+bgmrXNY9PzYXNwzdrmsVnXkscLgO16vn4h8HVgNeAZwPXApi55tDkVmrXNY7P8Zm3z2PR82PR82CyjWds8Nj0fNod/PuhzyeOYOxjaJY/zgbk9t70+Iq4BXgJs3nP3NwHXABcB3wdmA3825lUOSZIkSZI0pY13yeNCmiWOvxlx+yLgYOBdwPbA3TQXHCRJkiRJUsX6ucDwN8CXaZc8RsT+wGbA+cDTgM9GxDntfa8H7gO+AVwOfCkzr5r4ly1JkiRJkkoy0Use3wvMo9m/MBs4KTOPWEHTJY82i2rWNo/N8pu1zWPT82FzcM3a5rHp+bA5uGZt89ise8njwcAtwLo0nyJxMbCbSx5tToVmbfPYLL9Z2zw2PR82PR82y2jWNo9Nz4fN4Z8PhrTkcRHNgsc1gbtodjRsO+ZVDkmSJEmSNKVN9JLHc4CtgKOAs4FNgZ92e4mSJEmSJKl0E7rkMTMXAxcCOwEvBm7PzLMm5ZVLkiRJkqRiTPSSxxnAecDLaPYxLM3Mz6yg6ZJHm0U1a5vHZvnN2uax6fmwObhmbfPY9HzYHFyztnls1r3k8TPAvu3fP05z8aGfRRBDXVhh0+awn9vmytmsbR6bng+bng+bZTRrm8em58Pm8M8HfS55nDbWBYh2yeOrgRk9t72+vYCwBc2Sx/ntt14OHBARJ9D8+sVvI+LBzDxmzCsdkiRJkiRpyproJY9/CWySmdOBY4H04oIkSZIkSfWb6CWPCzLzF+3j7gJWi4jVJ+F1S5IkSZKkgkzokscR33sd8K7MfOkKmi55tFlUs7Z5bJbfrG0em54Pm4Nr1jaPTc+HzcE1a5vHZsVLHntufy5wE/BslzzanCrN2uaxWX6ztnlsej5sej5sltGsbR6bng+bwz8f9LnkccxfkWiXPM4H5vbc9vqIuAZ4Cc2Sx977fxpYAEzrfYwkSZIkSarXhC55jIgdgPcCbwZ2Bj4fEat2fpWSJEmSJKloE7rkEfgEsCrwYeA0YDbNR1dKkiRJkqSKTeiSx4g4BrgkM09ovz4O+G5mnjJK0yWPNotq1jaPzfKbtc1j0/Nhc3DN2uax6fmwObhmbfPYrHjJI3AM8Oaer48DXueSR5tToVnbPDbLb9Y2j03Ph03Ph80ymrXNY9PzYXP454OJWvLYOgLYLCIWPn5DRKwDPA84NSLOi4i1gduBuRHxrYj4MfAGYJs+n0OSJEmSJE1R/V5gOAW4dcRthwJLaJY9nt9+fSbw18D1wGuAe4B3R8T0iXixkiRJkiSpTNPGukNEnEjziRDrtl/vD/waeB/wGHAW8FNg/cw8JCJ+ArwD+HPg48BHgWWT8NolSZIkSVIhxrzAkJlvhN9b9Hhc+/X9mTmr/XsAi9uHvI7mnQyb0exkeENmPjbhr1ySJEmSJBWjr0+RgOU/SSIiljx+gaH9enFmrh0RrwNeCLwfeDZwHrB1Zt47ouenSNgsqlnbPDbLb9Y2j03Ph83BNWubx6bnw+bgmrXNY3OKfYoEcDxwF81ehYXtbesA9wM301xA2Ay4vv3excANwDXAhcB/A9v7KRI2S2/WNo/N8pu1zWPT82HT82GzjGZt89j0fNgc/vlgAj9F4ivAHiNuOxSYDxxLs+Dx34EzImIW8Byadzo8F3h3+/XNfTyPJEmSJEmaovrZwXBRRJwJbAoQEYvax+0KHA1sAsyh+TSJNwFfBZ4bEVcDARySmfdMzsuXJEmSJEkl6PdjKg+k+RWI1TJzQ2CNzPxpZu5Gs2fhkcz8NfAnwGrAdOBh4IjMPGEyXrgkSZIkSSpHX0se/4AFj8cA2wG7AU+h2cfwqsz82ShNlzzaLKpZ2zw2y2/WNo9Nz4fNwTVrm8em58Pm4Jq1zWNzii15bC9AbEy74LH9+npgvfbv6/G7BY+HAn/fc7/jgNf3sQhiqAsrbNoc9nPbXDmbtc1j0/Nh0/Nhs4xmbfPY9HzYHP75YAKXPAIcAWwWEQvbr88E3hUR5wFXA6tHxNrAGcCLImJaRLwI2A/YsM/nkCRJkiRJU9SYSx4j4kSaX3lI4DkRsT9wOHAlsBZwFfA/wKGZeUhEnA38BHgWcC3wv5P02iVJkiRJUiH6+RSJN8Lv7WE4rv36QeBPM/OOiFgPuIDmEyOOjIhHgEeAF0zS65YkSZIkSQXpa8kjPPmix4gIYHFmzoqIDYBvALsAx7ePOWWUnksebRbVrG0em+U3a5vHpufD5uCatc1j0/Nhc3DN2uaxOQWXPK5g0eOSEd9f3P7nf9G8swHgK8DrXPJocyo0a5vHZvnN2uax6fmw6fmwWUaztnlsej5sDv980OeSx352MBwP7Akspvm1ByJiHWC1iLgZuAk4ALgrIv4C2AvYKyIeAx4DXhkRyzLz9DGvdkiSJEmSpCmpn0+R+Aqwx4jbDgXmA8cC5wP/TvMJErcAczJzOrA3cD/wbi8uSJIkSZJUt36WPF4UEWcCmwJExKL2cbsCRwObAHOAfTLz1z0PvQRYc8JfsSRJkiRJKk5fSx77XfA44jEHA5tl5ttX0HTJo82imrXNY7P8Zm3z2PR82Bxcs7Z5bHo+bA6uWds8Nqfgkkf6XPDY8/UuwLXA0/vpu+TRZgnN2uaxWX6ztnlsej5sej5sltGsbR6bng+bwz8f9LnksZ8dDABHAJtFxML2619GxOYRcV676HF6RKwNEBHPA06j+fWI70fEtn0+hyRJkiRJmqL6vcBwCnBrz9dn0ix2PJ9m0eN84NCIeBZwNs27Fzam+RWIL0zQa5UkSZIkSYXq52MqTwR2BtZtv94fOBz4ObARcDPwLuB04Ontn4eBBcAy4KkRsV5m3jEJr1+SJEmSJBWgryWP0P+ix4j4NnB4Zv6w/d75wCGZOX9EzyWPNotq1jaPzfKbtc1j0/Nhc3DN2uax6fmwObhmbfPYnIJLHv+QRY/At4EX9dx+PrCdSx5tlt6sbR6b5Tdrm8em58Om58NmGc3a5rHp+bA5/PNBn0sex/wViSfxy4j4GPAGYDXgkfb2B4EvR8T9NL8iMRu4vcPzSJIkSZKkwvW75HE0PwLeA2wP/Cfw24iYC2wKLAGeD3wdWNv9C5IkSZIk1a2vdzD0LnqMiEXAYcBFwMuBHwO3Ad8A9gHubh92I807G340sS9ZkiRJkiSVpu8lj8s9MGJz4Azgz4AHaHYtzAc+D5wDBM07JHbMzNtGebxLHm0W1axtHpvlN2ubx6bnw+bgmrXNY9PzYXNwzdrmsTlFlzyuYPHj/sAVNO9m+ALwWeBzwGvb7+8LfG+sjksebZbQrG0em+U3a5vHpufDpufDZhnN2uax6fmwOfzzwQCWPALMAFYH1gDmAqcBnwZuiIjrgEfb2yVJkiRJUsXGveQxIrYE3kWz5HFPYEfgYpoFj28FtgYOBK7r/ColSZIkSVLRuryDYXNgbZq9C48A3wReBtwEbAhcRvORlft1fI2SJEmSJKlwk7Hkcaf29j1oLjAcnJmXj/J4lzzaLKpZ2zw2y2/WNo9Nz4fNwTVrm8em58Pm4Jq1zWOz/iWPC4GjaT5FYnvgFtoLGS55tFlys7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB30ueRz3DobW40sen06zzPFnwCLgVOD9wKXt/dbt+DySJEmSJKlg497B0LPkcR4wh+adCx8CHgP2otnR8AtgOnBP51cqSZIkSZKKNRlLHj8L3NDeti7whhzvogdJkiRJkjQlTMaSx+8Bu2bmQRFxK7BdZi73DgaXPNosrVnbPDbLb9Y2j03Ph83BNWubx6bnw+bgmrXNY7PuJY/H0uxdmNl+/1Zg3T4WQQx1YYVNm8N+bpsrZ7O2eWx6Pmx6PmyW0axtHpueD5vDPx8MacnjNcCWwD0R8TDwLGBBRDyz4/NIkiRJkqSCjfsCQ8+Sx+2BPYEdad7JsDfwlMycDtwHnJaZd07Aa5UkSZIkSYWa8CWPmXlEz30eAtbv8BySJEmSJGkKmPAlj5l5QM99vgWcnJknjPJ4lzzaLKpZ2zw2y2/WNo9Nz4fNwTVrm8em58Pm4Jq1zWOz7iWPn+353keA02gvYrjk0WbpzdrmsVl+s7Z5bHo+bHo+bJbRrG0em54Pm8M/HwxpyePPImKdiFgIfBR4GjCr43NIkiRJkqTCTfSSx4uBLwKzaT5B4jzg0O4vU5IkSZIklWzClzwCewF301xcmAZsCBzS7WVKkiRJkqSSTfiSR+AtmTmrvU8Aix//esTjXfJos6hmbfPYLL9Z2zw2PR82B9esbR6bng+bg2vWNo/Nypc8AktG3GexSx5tToVmbfPYLL9Z2zw2PR82PR82y2jWNo9Nz4fN4Z8P+lzy2OVXJKBZ4rgGzaLHF9PsX7gvIn4CPAY8BCzu+BySJEmSJKlwXZY8bgC8F9gOeCWwEbAMWAs4LzO3Ae4A7p+A1ylJkiRJkgrW9R0MzwQW0LxT4RrgRuBXwIsj4gYggDM7PockSZIkSSrcuN/BkJm3Ax+i+ZSI9YGbM/NcYD9gU+ApwMPAx7u/TEmSJEmSVLIunyKxNvD/gDcAS4D/Ak4B9gH+KTMvjYgPAs/JzLeP8ng/RcJmUc3a5rFZfrO2eWx6PmwOrlnbPDY9HzYH16xtHpuVfIoE8HrgEppfjVgIXAwcC9wEfBL4GXADcKefImFzKjRrm8dm+c3a5rHp+bDp+bBZRrO2eWx6PmwO/3wwgE+ReADYlmYPw2LglvbiwjOALYDNgLcBe3d4DkmSJEmSNAV0ucCwgOYiwyU0uxamAScCO9BcYFhAc+Fhv46vUZIkSZIkFa7rksf/S7PgcQ5wYWaeBawDfBV4hOYCxKoT8DolSZIkSVLBJmPJ478Dh2XmP0fEPsD7MnOnUR7vkkebRTVrm8dm+c3a5rHp+bA5uGZt89j0fNgcXLO2eWzWteTxuJ6v/xL4PHAdsEl7WwC/ccmjzanQrG0em+U3a5vHpufDpufDZhnN2uax6fmwOfzzwQCWPP4ceGVE/BR4DFgTOBq4F9glIt4HvIPmEyYkSZIkSVLFulxgWASsQbNj4TGafQ73AV8Gvg08F5gOvL3ja5QkSZIkSYUb95LH1lLgz4BtaN6p8HOaiwyP0HySxAOZ+eOOzyFJkiRJkgo37iWPABFxEPBJmk+LODcz/6K9bZXM/NeIWJqZM1bwWJc82iyqWds8Nstv1jaPTc+HzcE1a5vHpufD5uCatc1js54lj2sD/w3MBlYDTqdZ9PhDYFp7n6X9tFzyaLOEZm3z2Cy/Wds8Nj0fNj0fNsto1jaPTc+HzeGfDwaw5PGlNIsdL6B5wvuAfwDmAA9ERAKrRcSNmTm3w/NIkiRJkqTCddnB8ACwLbATsBWwHs27F15Ls9xxdWAZ8M8dX6MkSZIkSSpcl3cwLKC5yHAJ8HDbOjEzv/P4HSLiMWDDTq9QkiRJkiQVb8KXPPZ8bzXgUuCgzPzBKI91yaPNopq1zWOz/GZt89j0fNgcXLO2eWx6PmwOrlnbPDbrXvL45p7v/wfwWZc82pwqzdrmsVl+s7Z5bHo+bHo+bJbRrG0em54Pm8M/HwxpyeNOEfE/wA+AWcBZETE9Mx/u8DySJEmSJKlwE73kcS3gm8D9NO9sWAzs3/E1SpIkSZKkwnW5wNC75PFqmiWPJwHbtX+/GNgVeH/H1yhJkiRJkgo3oUsegYOASzJzbvv9jYDvZuaWozzWJY82i2rWNo/N8pu1zWPT82FzcM3a5rHp+bA5uGZt89isfMkjcGPPfTYCFrrk0eZUaNY2j83ym7XNY9PzYdPzYbOMZm3z2PR82Bz++WBASx5/DZzXfr028Erg4YiYB/wbsA4wJyK2z8zLOjyXJEmSJEkqWJcdDD8HNgd2BJ4PXAg82P7nl4G/p3mHwwnAEd1epiRJkiRJKtm4LzBk5qXAKcCVNEse1wcWAgcAf0RzkeHpNEsgf9H5lUqSJEmSpGJ1WvL4e6GI44ErM/OYiNgcOAcImosYO2bmbSPu75JHm0U1a5vHZvnN2uax6fmwObhmbfPY9HzYHFyztnlsVrLkccTCx+nAPcCc9uvPAa9t/74v8D2XPNosvVnbPDbLb9Y2j03Ph03Ph80ymrXNY9PzYXP454M+lzyO+1ckIuI5EXFVRFwFXAfMAt7YfvsdwCcj4hpgO2D78T6PJEmSJEkqX5cdDNdn5jaZuQ1wOXA/cFpE7AI8BhyQmc8FLgVumJBXK0mSJEmSitTlYyoBiIi1gFcAP83M2yLiSOAjwOER8RmaT5Z4Z9fnkSRJkiRJ5ZqQJY8jFjxeBZwB7EFzceHgzLx8lMe45NFmUc3a5rFZfrO2eWx6PmwOrlnbPDY9HzYH16xtHpuVLXlk+QWPC4GjaT5BYnvgFtoLGS55tFlys7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB5O95BGaRY80Cx6nA+dExL3AqsCp2Vxt2AnYGPiTLs8jSZIkSZLK1ukCQ2ZeD1wGHATMA34LfA3YJSI2Al4DPAr8quPrlCRJkiRJBev6Doa1gJcBpwK7ATcBnwE2pflViXVpfn1CkiRJkiRVbEKWPMJyix73AnbNzIMi4lZgu8y8Z8T9XfJos6hmbfPYLL9Z2zw2PR82B9esbR6bng+bg2vWNo/NypY8jlz0CKwJXArMbL93K7CuSx5tlt6sbR6b5Tdrm8em58Om58NmGc3a5rHp+bA5/PNBn0sep415BWIF2gWPJ7dfPg2YBbwR2JpmH8PdEfEIsDpwZURsn5l3jvf5JEmSJElSuca9gyEzr8/MbTJzG+By4H7gNOBEYI3MnA4c096+rRcXJEmSJEmqV6clj/DEosdXANdm5m2ZeW5mLmu/fQnNx1ZKkiRJkqSKTciSx94FjyNu/xZwcmaeMMpjXPJos6hmbfPYLL9Z2zw2PR82B9esbR6bng+bg2vWNo/NypY80rPgccTtH6H5lYkYq+GSR5slNGubx2b5zdrmsen5sOn5sFlGs7Z5bHo+bA7/fNDnksdOvyLRLnq8rr3IcE5E3BsR742I9wAfBLYCzo2Itbs8jyRJkiRJKlunCwyZeT1wGXAQzSdH/Bb4NfD3wNGZORc4Hzi04+uUJEmSJEkF6/oOhrWAlwGnArsBNwH/F5gJ7B0RVwFbAK/p+DolSZIkSVLBJmTJI/z+oseIWJKZs9rbA1j8+Nc993fJo82imrXNY7P8Zm3z2PR82Bxcs7Z5bHo+bA6uWds8Nitb8jjaokdgyYjvL3bJo83Sm7XNY7P8Zm3z2PR82PR82CyjWds8Nj0fNod/PhjQksdZEXEKcDOwBrBpRGwDrB4RCyNifkS8Ariry/NIkiRJkqSyTev4+KOAs4FlwPeAa4FvAmcB84GfAP8GnNLxeSRJkiRJUsHG/Q6GiJgJvBg4kWbR439l5hKat098t73tOGBN4PDuL1WSJEmSJJWqy69IbALcTfMOhZ8D/9x+qsR7gcOAPwEeA3bIzF93faGSJEmSJKlc4/4UiYjYDrgEeGFmXhoRRwH30nxE5YWZ+f8iYl/gnZn50lEe76dI2CyqWds8Nstv1jaPTc+HzcE1a5vHpufD5uCatc1js5JPkQCeSfPOhVOA64Bbgf8BfgMc0N52DfBQH5smh7oR06bNYT+3zZWzWds8Nj0fNj0fNsto1jaPTc+HzeGfDyb7UyQy805gdWBBZm4GfA24AlgCvBXYGjiwvdAgSZIkSZIqNu5PkWiXPC4D9o6IN9B8VOXbgC2BDYHLgAeB/SbgdUqSJEmSpIJ1XfJ4O7AQeBS4B3gYWIfmkyUebP90eQ5JkiRJkjQFTMaSx72B79P8esQLgJOBTXPEE7nk0WZpzdrmsVl+s7Z5bHo+bA6uWds8Nj0fNgfXrG0em3Uteby15+udgLOAs4Fdem6/CZjtkkebpTdrm8dm+c3a5rHp+bDp+bBZRrO2eWx6PmwO/3wwoCWPv4iIsyPiOuBUmgWPpwO7RMQHIiKBNWh+fUKSJEmSJFVq3EseW/cAz6P51YgfAR8G7gBOAl5Gs5PhPSN/PUKSJEmSJNWl66dIbAVsMsp+hceAFwFnAD/s9AolSZIkSVLxuix53AY4FvgpsDVwBXAQ8FJg18w8KCJuBbbLzOV+RcIljzZLa9Y2j83ym7XNY9PzYXNwzdrmsen5sDm4Zm3z2KxnyeN2wDJgh/bro4AjgUuBme1ttwLr9rEIYqgLK2zaHPZz21w5m7XNY9PzYdPzYbOMZm3z2PR82Bz++aDPJY9ddjAsAn4BfDAitqRZ5ng3sAVwT7vgcRqwICJe0C6FlCRJkiRJFer6KRKrAwsyczPga8DFwN7AUzJzOnAfcJoXFyRJkiRJqlvXJY/LgL0j4g3AzcDbMnNxz90eAtbv9hIlSZIkSVLpJnzJY2be33OfbwEnZ+YJozzeJY82i2rWNo/N8pu1zWPT82FzcM3a5rHp+bA5uGZt89ise8njJ3q+/xHgNNqLGC55tFl6s7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB0Na8nh7RPwrcBHwbOBHwCxg8QorkiRJkiRpypvoJY9XAF8EZgPPAs4DDp2A1ylJkiRJkgo24UsegV/SfFzleW1/Q+CQ7i9VkiRJkiSVasKXPAK3Z+as9j4BLH786xGPd8mjzaKatc1js/xmbfPY9HzYHFyztnlsej5sDq5Z2zw2K1/yCCwZcb/FLnm0ORWatc1js/xmbfPY9HzY9HzYLKNZ2zw2PR82h38+GNCSx0XAyRFxH82Sx9nALyPipcA/AjOAp0TE9pl5WYfnkiRJkiRJBRv3BYbMvDMi/hfYBNgF+Ftgrfbbn6f5dYmtad7pcASwc6dXKkmSJEmSitXlHQwABwCXAhcCN9AseVwF2A84HrgGOAnYtePzSJIkSZKkgo17yeMTgYhbgMU0v5fxxcw8NiI2B84BguaCw46ZeduIx7nk0WZRzdrmsVl+s7Z5bHo+bA6uWds8Nj0fNgfXrG0em5UseexZ4rhB+5/PAH55jmo1AAAgAElEQVQMvBj4HPDa9vZ9ge+55NFm6c3a5rFZfrO2eWx6Pmx6PmyW0axtHpueD5vDPx8MYMkjEXErcF9EPErziRKnAdsD7wB2j4h/AM5qb5MkSZIkSZVapePjA9gzM7cBXgLs3jYfAw7IzOfS7Gi4oePzSJIkSZKkgnVd8rgqcFb7DoZpwDdoPjXiI8DhEfEZ4EHaXQuSJEmSJKlOnZY8rmDB41XAGcAeNBcXDs7My0d5rEsebRbVrG0em+U3a5vHpufD5uCatc1j0/Nhc3DN2uaxWdGSR0Zf8LgQOJrm1ye2B26hvZDhkkebJTdrm8dm+c3a5rHp+bDp+bBZRrO2eWx6PmwO/3zQ55LHTjsYMvP2dtHj+cAc4OvAIuDUbK487ARsDPxJl+eRJEmSJEllG/cOhohYi98tiXwVcBLwDzQXFHaJiBuB1wCPAr96stbz581j/vz5y91+wQUXcH/Pr3CsFTHel6sC9f5sHzfyZw7+3Gsy2s8c/LnXzn/WVz7+s75y6vefdfDnXhP/Hb/y8Z91PZku72CYA/wQWB84FzgrM88Gjgc2pflViXWBe7q+SEmSJEmSVLZOSx5hhYse9wJ2zcyD2l+h2C4z7xnxuCeWPM6ZM2feSSedtFx76dKlzJgx44mvF1xxxaivYSosy7C5/H2fP2/ecvcb+TOHwf3cS/zvqLbmaD9z6P/nPpVnL+25B9ks7Z/1qdKcyvP4z/rkN0ucp99/1qHbz73E2UtrDvK5u/w7vrb/3qdK03/WbY7nfgNZ8vgkix4vBWa2t98KrOuSR5ulN2ubx2b5zdrmsen5sOn5sFlGs7Z5bHo+bA7/fNDnksdx72AAaN+dcF9EPAosA04DPgXMA+6OiEeA1YErI2L7zLyzy/NJkiRJkqQydfoUCZqPotwzM7cBXgLsDnwHWCMzpwPHAPcD23pxQZIkSZKkenW9wLAqcFZE/Bi4jGbR46cyc1n7/Uva+0iSJEmSpIp1WvI42oLHEd//FnByZp4wymOfWPI4c+bMeYd97GPL9UtbaGKz7mZt89gsv1nbPDY9HzYH16xtHpueD5uDa9Y2j82KljwyyoLHnu99hGYnQ4zVccmjzRKatc1js/xmbfPY9HzY9HzYLKNZ2zw2PR82h38+6HPJY6dfkcjM29tFj+cDc4CvA0TEe4APAlsB50bE2l2eR5IkSZIklW3cFxgiYq2IeGr75auAm4F3RMQewN8DR2fmXJqLD4d2fqWSJEmSJKlYXd7BMAf4IbA+cC7NgsezaT45Yiawd0RcBWwBvKbrC5UkSZIkSeXqtOQRRl/0GBFLMnNW+/0AFj/+dc/jXPJos6hmbfPYLL9Z2zw2PR82B9esbR6bng+bg2vWNo/NipY8rmjRI7BkxH0Wu+TRZunN2uaxWX6ztnlsej5sej5sltGsbR6bng+bwz8f9LnkcdqYVyDGdmdELABup/nUiDcBa0TEQmAJzbLHuybgeSRJkiRJUqE6XWCIiLWAA4FrgbWB3YFn0XyaxA3AvTQ7Gc7o9jIlSZIkSVLJOn1MJbA18GFgB+BFwFnA/cBJwMuAfwDWAQ7v+DySJEmSJKlgnZY8RsQpwKeBpwIHZ+aeEbETcDrwAM07GP40M+8d5bEuebRZVLO2eWyW36xtHpueD5uDa9Y2j03Ph83BNWubx2YlSx6BPYHPt3/fGfh2+/dTgR3av38Q+NJYLZc82iyhWds8Nstv1jaPTc+HTc+HzTKatc1j0/Nhc/jngwEseXwh8OqIeCWwAbBKRJwFbNbe/jUgaN7dIEmSJEmSKjbuHQyZ+XeZuSHwOeBC4B5gL5qPq9yC5kLD4cD8CXidkiRJkiSpYF0/RWJD4FXAt4GtMnNZRNxJc4FhAbAY2K/zq5QkSZIkSUXr+ikSnwU+RHMx4fL2tnWArwKP0Cx6XLXjc0iSJEmSpMKN+1MkImJP4JWZ+e6I2JnffYrEUuCwzPzniNgHeF9m7jTK4/0UCZtFNWubx2b5zdrmsen5sDm4Zm3z2PR82Bxcs7Z5bNbzKRKfBhYBt9K8W+FR4ATgOmATmt0MS4Hf+CkSNqdCs7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB31+isSELnnMzDcDpwNvA9am+fWIn433OSRJkiRJ0tQwoUse25uPoLmocDcwHXh7l+eQJEmSJEnlm4wlj28BPpmZmwMPZOaPOz6HJEmSJEkq3IQueaRZ2vhNYOf2IyuXZuaMFTzeJY82i2rWNo/N8pu1zWPT82FzcM3a5rHp+bA5uGZt89isc8njncBvgcXt329t/zwG3OiSR5tToVnbPDbLb9Y2j03Ph03Ph80ymrXNY9PzYXP454MBLnl8NnAfcF9mrg2cDzxE8wkSjwKbj/c5JEmSJEnS1NB1BwPAQcBtPV9/HdiM3y19HHPJ4/PnzeP+zOX+jLxddennZ+7PvS6j/cz9udfPn/nKx3/WpZWH/6yvfPr9mftzXzl1usDQ8ykSn6Jd8piZ38kWzQLIDTu/SkmSJEmSVLRxL3kEiIhTaHYxPBU4ODP37PneasClwEGZ+YNRHvvEksc5c+bMO+mkk5brL126lBkzfrcjcsEVV4z6OqbCsgyby9/3+fPmLXe/kT9zGNzPvcT/jmprjvYzh/5/7lN59tKee5DN0v5ZnyrNqTyP/6xPfrO2eWxO3fPR5d/xtf33PlWag/rfdej2c6/tv/ep0hzmksc9gc+3f98Z+PaI7/8H8Nk+F0EMdWGFTZvDfm6bK2eztnlsej5sej5sltGsbR6bng+bwz8fTPaSR+CFwKsj4lbgPOAVEXFCRGwSEYuANwLrRcT0Ds8hSZIkSZKmgIn4FInPARcC92Tmm4FvAvcDs2k+tnL/iXihkqY+F79JklQX/3d95eOSRz2ZiVryeFb7dQDbAdOAi4Fdgfd3fI2SJEmSJKlwE7rkEXgrcElmzm2/vxHw3czccpTHPrHkcebMmfMO+9jHluuXttDEZt3N2uYpsenit3Ke22b5zdrmsen5sDm4pksebU7mc7vkse5mUUsegXWBG3vusxGw0CWPNqdCs7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPB30ueZw25hWIFXt8yeP+NL9qsSrw38CsiHgZ8E/ADGB2RMzNzBs7PJckSZIkSSpYpyWPNO9QeDrwMuC7wAPAj4GvAn9Bc8HhHOCjnV+pJEmSJEkqVqclj9lY2n4ZwGrAMcAs4Fyaiw8LgV90eR5JkiRJklS2TkseASJiVeAKYC7wb5l5SETsBJxO846Ge4E/zcx7RzzOJY82i2rWNo/N8pu1zWPT82FzcM3a5rHp+bA5uGZt89isZMnjKEsfZwHfB7YETgV2aG//IPAllzzaLL1Z2zw2y2/WNo9Nz4dNz4fNMpq1zWPT82Fz+OeDyV7yGBFrABcBqwPTgFPaCwyvAHYBtoyIR4GTgB3H+zySJEmSJKl8XXYwPATsC7wE2IbmwsI+wAbAU4A/z8zNgd8A1z5Z6Pnz5nF/5nJ/Rt4uSZp6+vn3u/+OlyRJmvrG/Q6GzMyIeBpwWtuZS/PpETsC7wNOiYjHgMXAfhPwWiVJkiRJUqE6LXlcwYLHXwH/AuwN3A0cmJk3jPLYJ5Y8zpkzZ95JJ520XH/p0qXMmDHjia8XXHHFqK9jKizLsFl+s7Z5bJbfrG2eFd33+fPmLXe/kf9+B/8dX9Jz2yy/Wds8Nj0fNgfXrG0emxUueeT3FzwuBT7Q3r4P8IOxHu+SR5slNGubx2b5zdrmsen5sOn5sFlGs7Z5bHo+bA7/fDDZSx5huUWP69G8c2ERzadIAOwMvKjLc0iSJEmSpPJ1+RSJ2cAjwK7Ao8B5wB8BlwK7RMTTgc2AxybgdUqSJEmSpIJ1+RSJ9Wh+LeJHwOXt3+8H/hN4LXABsA7Np01IkiRJkqSKdVryCCtc9HgQsEpm/mtELM3MGaM87okljzNnzpx32Mc+tlzbhSY2B9msbR6b5Tdrm8em58Pm4Jq1zWPT82FzcM3a5rFZ4ZLHEYseXwz8EJjW3r7UJY82p0Kztnlslt+sbR6bng+bng+bZTRrm8em58Pm8M8Hk73kccSCx2nAKe0FhmOALYDfRsRDwJoRcWNmzh3vc0mSJEmSpLJ1+RSJh4B9gSU0uxf+B5hO8wkSW7f3+QawtxcXJEmSJEmq27iXPGbz+w9Po3nXwpXAVsClmfnxbAGXdXkOSZIkSZI0NXRa8jjagsee761G85GVB2XmD0Z5rEsebRbVrG0em+U3a5vHpufD5uCatc1j0/Nhc3DN2uaxWeGSR3634HHLntv+A/hsP493yaPNEpq1zWOz/GZt89j0fNj0fNgso1nbPDY9HzaHfz6Y7CWPsNyix/WAfwF2j4h/BV4L/Doi1gPekpkPd3kuSZIkSZJUrnHvR4iI2cAawK7AnwI/A/4oIj4FvIXmVyPmAouB/SfgtUqSJEmSpEJ1WcC4Hs2vRfwIuLz9+/3Ah4C1gQ9ExFXtc7ym4+uUJEmSJEkF67TkEZZf9AgcCVzy+EdTRsRGwHczc8sRj3PJo82imrXNY7P8Zm3z2PR82Bxcs7Z5bHo+bA6uWds8Nitc8jhi0eOLgBt7bt8IWOiSR5ulN2ubx2b5zdrmsen5sOn5sFlGs7Z5bHo+bA7/fDDZSx7bdyZ8FZhD84THthcYXkuzi+EqYBnweeD28T6PJEmSJEkqX5dPkVgGfAK4Eni0/c97gQAuofl1iXuBLwH/2O1lSpIkSZKkko37AkNm3tF+ksT3gVVp3snwI+CZwLeB9wPPAh4Gjuv+UiVJkiRJUqk6L3kEiIiNgYuALYENgHNo3smwCrBjZt42ymNc8mizqGZt89gsv1nbPDY9HzYH16xtHpueD5uDa9Y2j83KljwCM2g+RWKf9uvPAa9t/74v8L2xGi55tFlCs7Z5bJbfrG0em54Pm54Pm2U0a5vHpufD5vDPBwNa8vg1YDtgKc2nRQD8FXBDRFxHs5th7nifQ5IkSZIkTQ2rdHjsMpoLC18C/hh4T0RsASwB3gpsDRwIXNfxNUqSJEmSpMJ1+RSJZwOvAq4GdqZZ8rg3cBOwIXAZ8CCwX7eXKEmSJEmSSjfudzBk5g8zMzLzecBrgN8ARwPrACfSXFx4sMtzSJIkSZKkqaHzp0hExAzgQuCTmXlqRCyk+ejKA4EXACcDm+aIJ/JTJGyW1qxtHpvlN2ubx6bnw+bgmrXNY9PzYXNwzdrmsVnJp0jQLHW8gGYPw53AQe3tZwO7AB+g2TZ5CzDbT5GwWXqztnlslt+sbR6bng+bng+bZTRrm8em58Pm8M8HfX6KxGQseTwd2AvYHfgFMB24p8PzSJIkSZKkwk3GkscjgRuAR4B1gTeM/PUISZIkSZJUl8lY8vgK4PTMnAvcAfxwQl6pJEmSJEkq1oQueaTZv/B9YPfM/E1E3Apsl5nL/YqESx5tltasbR6b5Tdrm8em58Pm4Jq1zWPT82FzcM3a5rFZyZLH9sLEasA5wPvbr7cC7gJubf8sA34OPNMljzZLb9Y2j83ym7XNY9PzYdPzYbOMZm3z2PR82Bz++aDPJY/j3sEQERsBl7cXGTaMiEcz86iI+E/gz4GHgYeAnTPzzvE+jyRJkiRJKl+XT5HYlmax4+3Ao8A/RcRfA+cBW7a7GZYB7+38KiVJkiRJUtHG/Q6GzDwDiMe/jogzgJsz87yeu70VeN24X50kSZIkSZoSOi95BIiIjYGLaN65cG/P7d8CTs7ME0Z5jEsebRbVrG0em+U3a5vHpufD5uCatc1j0/Nhc3DN2uaxWdGSx/bixAzgCmCfEbd/BDiN9iKGSx5tlt6sbR6b5Tdrm8em58Om58NmGc3a5rHp+bA5/PNBn0sex72DISI2iogLgDuBDYCN2tvXiYiFwEeBpwGzxvsckiRJkiRpauiy5HEZsBT4EvDHwHsiYgvgi8Bs4Fk0Cx8P7foiJUmSJElS2ca95BF4NvAq4GpgZ5pPlNgb2Au4m+biwjRgQ+CQTq9SkiRJkiQVbcKXPAI/z8xZ7e0BLH786xGPccmjzaKatc1js/xmbfPY9HzYHFyztnlsej5sDq5Z2zw2K1/yCCwZ8f3FLnm0ORWatc1js/xmbfPY9HzY9HzYLKNZ2zw2PR82h38+6HPJY5dfkSAivgy8Cfh1Zp7a3rwkIq4AVgUCuHdFj5ckSZIkSXXo8ikSAawHnAL8asS3f5yZ2wBX0lxkkCRJkiRJFevyDoYXAi8HrgM2joirgA8DNwMviIgbgAeByzq/SkmSJEmSVLTOSx7bBY/fzswt2683B86heefCKsCOmXnbKI9zyaPNopq1zWOz/GZt89j0fNgcXLO2eWx6PmwOrlnbPDbrW/K4MbCw5+vPAa9t/74v8D2XPNqcCs3a5rFZfrO2eWx6Pmx6PmyW0axtHpueD5vDPx8MaMnj8cCraT5J4nF/BdwQEdcBjwJzuzyHJEmSJEkq37iXPLa+QnNBodcS4K3A1sCBNDsaJEmSJElSxTq9gwH4G2A3YPWIWAQcBtwEbEiz3PFBYL+OzyFJkiRJkgo3GUserwLOAPagucBwcGZePsrjXPJos6hmbfPYLL9Z2zw2PR82B9esbR6bng+bg2vWNo/N+pc8LgSOpvkUie2BW2gvZLjk0WbJzdrmsVl+s7Z5bHo+bHo+bJbRrG0em54Pm8M/H/S55LHTDoZ2yeN8fn+R4yLgVOD9wKXtbet2eR5JkiRJklS2yVjyeDqwF7A78AtgOnBPx+eRJEmSJEkFm4wlj8cDNwCP0Lxz4Q3ZddGDJEmSJEkq2mQsedwL2DUzD4qIW4HtMnO5dzC45NFmac3a5rFZfrO2eWx6PmwOrlnbPDY9HzYH16xtHpsVL3kE1qTZuzCz/fpWYN0+FkEMdWGFTZvDfm6bK2eztnlsej5sej5sltGsbR6bng+bwz8fDGnJ47OBLYF7IuJh4FnAgoh4ZpfnkSRJkiRJZZvQJY+ZeTWwN/CUzJwO3Aeclpl3dnweSZIkSZJUsK4XGP4G+DLtkseI2D8zz83MZe33HwLW7/gckiRJkiSpcBO+5HHE974FnJyZJ4zyPZc82iyqWds8Nstv1jaPTc+HzcE1a5vHpufD5uCatc1js+IljyNu/whwGu1FDJc82iy9Wds8Nstv1jaPTc+HTc+HzTKatc1j0/Nhc/jngz6XPE4b8wrEk2iXPL4amNFz2zrARTQLH38EzAIWd3keSZIkSZJUtgld8tj6IjCb5hMkzgMO7fgckiRJkiSpcJ3ewUCz5HE32iWPwGHAXsDdNBcXpgEbAod0fB5JkiRJklSwCV/yGBFLMnNW+/cAFj/+9YjHueTRZlHN2uaxWX6ztnlsej5sDq5Z2zw2PR82B9esbR6blS95BJaM+P5ilzzanArN2uaxWX6ztnlsej5sej5sltGsbR6bng+bwz8f9LnksesOhtH8MiLWi4j3RcT1wJoRcWJErDEJzyVJkiRJkgowGRcYzgQOAA4EvgYcBawK/J9JeC5JkiRJklSAThcYIuJE4GLgORGxKCL2Bw4HXkyz3PGlwJHAmsAvOr5WSZIkSZJUqM5LHlcYjjgI+CTwAHBuZv7FiO+75NFmUc3a5rFZfrO2eWx6PmwOrlnbPDY9HzYH16xtHpuVLXlcweLHtYH/BmYDqwGnA292yaPNkpu1zWOz/GZt89j0fNj0fNgso1nbPDY9HzaHfz7oc8njtDGvQIzPS4HbgS8AWwJPAx4FTpik55MkSZIkSUM0GUseAX4O/DnNuxg2B84HLp2k55IkSZIkSUM2WRcYrqN5G8WBwNXtbUdN0nNJkiRJkqQhm5QljxGxDXAs8FNga+AK4KDMvL/nPi55tFlUs7Z5bJbfrG0em54Pm4Nr1jaPTc+HzcE1a5vH5sqx5HE7YBmwQ/v1UcAnXPJos+RmbfPYLL9Z2zw2PR82PR82y2jWNo9Nz4fN4Z8PhrzkcVH7Z35ELKD5qMrFk/RckiRJkiRpyCZlB0Nm3gn8L/CPwLXAujS/LiFJkiRJkio0We9gAPgEcApwN7AW8KlJfC5JkiRJkjREk3mB4Z3ALsBTgYMz01+RkCRJkiSpUpP1KRJ7Aq/MzHdHxM40Fxj2HHEfP0XCZlHN2uaxWX6ztnlsej5sDq5Z2zw2PR82B9esbR6bK8enSBwDPAg8DDzS/ucJfoqEzZKbtc1js/xmbfPY9HzY9HzYLKNZ2zw2PR82h38+6PNTJCZlySPwSWDHzJwOvLq9wOAOBkmSJEmSKjVZnyJxR2Ze2X75ALAU2GAynkuSJEmSJA3fZL2DodetNO9guHQAzyVJkiRJkoZgUpY8PhGPmAFcCHwyM08d8T2XPNosqlnbPDbLb9Y2j03Ph83BNWubx6bnw+bgmrXNY3MlWPLYXrRYDTgHeH8fiyCGurDCps1hP7fNlbNZ2zw2PR82PR82y2jWNo9Nz4fN4Z8P+lzyOG3MKxDjEBEBfBfYFnh2REzPzMMn47kkSZIkSdLwTdYOhp2A3YC7gPuBwyLiryfpuSRJkiRJ0pBNyjsYgEeAczPz5QAR8XfAOpP0XJIkSZIkacgmZcljRLwO2CMz395+/RZgh8z82577uOTRZlHN2uaxWX6ztnlsej5sDq5Z2zw2PR82B9esbR6bZS15HNoFhhH3vxu4bZRvrQvc08dTTvT9bK6czdrmsVl+s7Z5bHo+bA6uWds8Nj0fNgfXrG0em4M5H3+UmbPHfPQkfYLEnwHn9Hz9d8DfjaPT16bKib6fzZWzWds8Nstv1jaPTc+HTc+HzTKatc1j0/Nhs8zzMdqfyVryeDnwxxGxSURMB/4PcOYkPZckSZIkSRqySVnymJnLIuJvgXOAVYHjM/OayXguSZIkSZI0fJP1KRJk5neA73TMHDuk+9lcOZu1zWOz/GZt89is57ltlt+sbR6b9Ty3zfKbtc1jc3jPvZxJWfIoSZIkSZJWLpO1g0GSJEmSJK1EvMAgSZIkSZI68wKDJEmSJEnqbNKWPP6hImIzYC9gg/am24EzM/Pajs0NgEszc2nP7Xtk5tkj7rs9kJl5eURsAewBXNcuq3yy5/hqZv5lH6/lRcD2wMLMPLfn9h2AazPz3oh4CnAosC3wU+BTmfmbnvseCJyWmf87xnM9/tGgv8jM70XEm4AdgWuBYzPzkZ77bgrsA2wEPAr8DPhGZt471kzSRIiIZ2TmXcN+HZI0Wfz3XPn8GZXPn5HUzaD+GSriHQwRcQhwEhDAZe2fAE6MiEP/gM7bev5+IHAGcACwMCL26rnrp0Y87jDgc8AXIuLTwDHAWsChEfGRnvudOeLPt4B9Hv96RPOynr+/o20+FThsxEzHA79t/34UMBP4p/a2L48Y8RPApRHxg4h4d0TMXsF/FV8GXgUcFBFfA14PXAq8APjSiP+O/h1Yo/3e6jQXGi6JiJ1X0J5yIuIZw34NgxARMyPi8Ii4LiJ+HRG/iohr29tmTfJzPzMivhAR/xYRT4+Ij0fE1RHxzYhYr+d+64z483TgsohYOyLWmcDX8/SOj98uIr4fESdExEYRcV5E/CYiLo+I54+zOS0i/joizo6In7R/vhsR74qI1f6AzrE9f1+1bX4iIl444n4fHfH1mhHxoYj4YESsERFvbf/ddUREzBjjOX82ym3P6/n7ahHx0bb3qYhYc8R9/zYi1m3/PjciLoqIJRFxaURs1XO/UyPizWO9nva+m0bE8RHxjxExIyL+IyIWRsR/RcTGPfdbJSL2i4izIuLHEXFlRJw08t9xtf182tv7+hn1+/Npv9/Xz6jfn097335/RkP7d1z7/P57buxm8f871N7Xn5E/o8efZ9w/o0n6+VT1v0VR2f+t0N7Xf4bGkplD/0Pz/2u+2ii3Twdu+AM6P+/5+9XAjPbvGwPzgYParxeMeNzVwKrAmsC9wNPa258C/KTnflcCJwA7Ay9p//OO9u8vGdFc0PP3y4HZ7d/XAq7u+d61vf0RjatGNmkuCu0OHAfcDf+/vTMPlquo4vDXIRB5IiEEKoFiXyKIyJ6ghDICKpsYtlKgIIkGSoIEERVUJEJBiYhEULEEYhCQXQyLBJBNZAmrkAWSAGEJCjHsCJZs7R/njOnX7943p2fmZpJ5faq65t6+vzlLn+6+PT3dfbgJGAN8LMDN1M/+wCJgBb13kT2zgmddwJ16vV5BGQ0ETgfmAq8CryArIk4HVlsKdWQo8Bvg18Bg4Meq/5XAWgFu9SgNBp4FBgGrt1CfwU1+f3vgDq1P6wJ/Ad7QurJNgzxvBo4Hhkbldjxwi5HH9OB6VeAnwMXAwRHu3Oj+JmQy7wRgpspcV/OuDXAfAs9E6T39XBDx3D2qf1OU96XAkODZ6cAaQbkuAJ4CnqNnu3wEOBHYuE45PADsARwELAQO0Pxdgfsi7CrAKcAc9eFiYAYwNsJdpnV4R2AdTTtq3hURNq7HYX1+IcBdoOXxLeBh4KzQ1ojnlcDPgXOB25BJz52BnwEXB7i3kH7wTb1+C1nd9BbwZhF/5Xsh0hdOBi6KZM8Jrv8M7KvXo4B7gmf/AK5G+pgrgX2BlUp8dBdwpNa52cBxWue+Dtwe4KYi/cVI4Bfqq88DtwJHd6p/Unxk9U+Kj6z+SfRR031c7ucq7+fa5iOrf1J8ZPVPio+s/knxkdU/rfJR6J++7COrfxLbUEe9i+iwsUJfb0PWlPyFKhLyo3X9gvz1gXlR3sySNAv4b1ElDRr2TcBZFPxwL7rW+0eD637AsciPwa01b0GJTY8hP2oHAw/1Iu8qYFxQubfX62HAg9H34k5gRWAfpDNaHOTPRiZnBiENfXXN/wjdJzRmAQP0elCoJ7KVo6WNKW5Q1saU0qCsjSmlQdHGgR1pg4Z5vcibF1xvW5K2A14McH9U20cD1+l9rb7EdTGs089Hz8I2dIyYZCUAAAymSURBVJz6cssg75kSncOX0gXAqUifcCwwLazHwfUdwA5BG4rb3jPAmcDz6oNjgbULZPdmT9xHXAuMRQYB3wZ+BGwK/B7Z5lTDze/FP/Oj+w+0noX1uHb/boALJwz7I3GLr0FWIxX2ZchE40ssCVMcTzyeA1xE9zbQw0dRGT2KThLH/ArqX9yvzYx5In3DocCNSJ2fCnyhER8V6DJDPwfQvT/sKP+k+MjqnxQfWf2T6CNTH6f3uZ9rTz/XNh9Z/ZPiI6t/Unxk9U+Kj6z+SfGR1T992UdW/yS2oY56F9FhY4W+3oasKfkLVSTkvIOngOla6c/TAnmK4EegYhcBW6vhYdoAOXOghrsdnQQI8vprY/ggyr8f6NLrfkH+QKIBhuavg0wM/Cp2boB5liWNfAH6Lzvyg/HRSMaFwNOqx3uK/yuwVVlDKJDXFVwfqzyeAyYiM5DnIxMKkwLcMcgP6vORSZ7aRMeawF2NNKaUBmVtTCkNytqYUhoUbRzYkTZouAX4Ht07+yHIZMytQd4HSPu4oyD9p6hc9f6HwD3IpFnsn8eC61OjZ7Oi+1r7OQvZNlQ2SfdIL7qEPn8C6K/XM+rIDnnujMzQv6S2HxE8uw9ZKXQg0o5Ga/5nC3z+WHT/oH72Q85xqeXPUH5hH9MP+ApyTkzI40lgvZJyWRhczy14Pkn99GSUH5bZ7+rYsJ3WkYmqYw8fIX3MfsD+9Hz5xvxOQ/q5jYAfIP+irA+MA24o8k+QNxj4Bj3/9X4YaYPDgZdZMjm7Cd0HIg+jE35IP3RX8OzxTvVP4KN96/nI6p8UHwX+2aE3/yT6yNTHaf7S7udim/pqP9c2H6X4x+ojq39SfGT1T4qPrP5J8ZHVP33ZR1b/JLahjnoXseyOFTalgbHCctyGJtOCNmRNyV+oKmnF3FEr4P56vUIBbgowsoTHpVGBDi3B7RTdDyjBrUHwQ7Xg+V5EP/IMdnYBGxbkrwpspY21cDkKMCxBztroD1tgNeAAYHgBbgt9tlkdfsvywK5jf8CSNmgYhJzfMRd4DVk69oTmrR7gZgOblpTJwsjuftHzschqiuei/FPQLUlR/ibA1SWy9kFepC+VPH8BmVQ5DnlBueBZ+FI4WuvnLsjytrO1HE8mWM4X+yfIWwGZ5Jwa5G2FrNqZDmymPF9X2z8Tff9etE9Sm24OnoUz2RsAVwD/QraFzdfrK4j6BOAoognG0N7g+hKiSVjNHw+8F+VdUOKjjYG7C/L7IYOGvxFM3gbPp0ZpiOYPBW4rwI9FJlFfRlZWPY6chzMwwNwVf68sIat95mk9HYlMUD6pZfrlALcLMuH3JDIBOELz1wTOKPDPYvVNjddy6R/FXGj1ETKA69U/KT6q45/REbbmo6fURzuW+MjUxyk293Pd84r6ua3p2c+9prbH4yRrP9c2HzXin3o+svonxUcU/FlU5B/NL3oX9fCR1T8pPrL6py/7qMQ/eazQ/flU5Ly5jhgrRG3oCaT99Jk2ZC73Rr6UU99LdH8hxY1pUITNA7vueQ2/lEgYNGjeZsBucVnRfTvIAcDHS8pkdHB9BrBbAWZ3Cs5GUdm79iY7xiHnnHyyBDcpSrVzTIbSc9/eKOTl+3dkpc6NwBFEZ7sAlyfU+c2N9myFrFZ5Dbi7VrbIS2lihB2BzKIPBnYCvgPsWSJ/OEtWy3xC62oPrBXXC3YvgrpfgNsZOKlE9ogGZW+BtLtm7RkR8SwsT+DTVp6KGazpkoT6clErcSE29k+EWQt4pQI9L66A5w1EfX7wzKFb0aw8tW4eR7QkVp9V0s9FmJFal3rIL9DzxBI929rPJZRn3M8N0/xu/Zy2yYF63YW8u29AxgrxZFVLfaSywzO0TgauL5EdYrtUxq0xNsU/Vh+l+Ccqz5XLyhP4lMU/mjcRWNcg2+SfRB+ZZBdhCcYLjbYhzf9cq3yUaE/NR6/35iNki/MYZP//YOAQ5M+qo+jZ1gcAh9XKHjgYWV19VGRPb7iVGuS5UoQ7FBlLTyiQPaYevxKehyBnr6XYU8RzTMSzR3kqbhxwYD2e+nxj4LvIVpHJyOqJVatuQ5q3ETLeORv5M7WH7BLsb5HfNLGeYfs5iTptyJJq+2oyZWqYnHPjvPdTg/sDkH/25xVgR3vvp+n1Gcj5DbdGmN2BX3rvN21EfvRsZWTJ0+wCPSdF8HO994udc0OR2crDAuwo5JCYYchWm4XANGT52PsB7nLv/Vctelvs0dN3L0CWcs0Bvua9n+8kgshB3vtzgu9NRDrCJ5B/po7x3l+rzx7x3m8bYE0hXHvB7eG9nx7cHw18s57sFB1bpGdRSNq6WNVzAjKhZtFzc+U5oxeek5BzN/oj57gMB+5EBhE3e+9PC74XY0cgq2C6Ya24JnkW6tmkPVaeKfY0paeLIgEp7YKsxsJ7v08gO8Y6ZNDaDWvFNckzRc9C7FKyvRV6PuC9H67X45G+ZBqyGux67/3pBXxqMgpDRafgIvmHq/w/xfILcBMq0HNnxc3qDdcbzyr0dM7NQf5Rfd/JKfZvI/8Y7qr5+wXYOEz394FtiMJ0K26u9/6NOrhY9jvIQXBFsk16RrK7kHOgesgusKeG3RZ5d8f21LW7ST2Ppzzk+RvK52nkzKmrvPcvF/jYFBo9BRvJvkxlLzZir1yKelZhj5XnH5D31crI2VsfRfqZXZHJ5jEF2C5k4mIV5GyFbtg6OLz3Y408/48twBXq2aTsRniWlVGv5Zmo50Rgb+QAyT2RyarXkS2JE7z3dwbYjZAtIusiK7znIyvu3yQiCzZR9kTgS8i2+3rYjVX2OvX0NFHqjEROOcWJknMoSrDjWolLkb8M6GnlabVnXHRvipyCzLjPQwaSz9J9iVi4xeNoCy5Rdkp0F5N8qz2JPFP0nIhMRFh41o1Wk4JtJ88OtCcpSpAFa8Ul8qxCz+XFdlN0Js17ILg+HDlcbBKyDe+EVFyK/Cb0HG/Uc7yWWZmeVtur0DMlMtYclmw/PA85uX2k8r2mAVyKbBPWKrsEO7kZeyrU0xqR7A3gn8jS9wm1+lGUrFir7KWk55EEq6aWgj2x7DKepkhwKdh28uw0ezTPFIUPGR/egqxguxdZjXEaMvE3KuJpwlplV6WnNSV/Iae+mTBG7zDwaWgywCq/3Xpasa3Qs6CMTJFTqGYywCo7JbpLFXpaeVahpylaTQq2nTw70J6UKEEmbObZcp6m6EwFfm/VZIA1OlS79bTyrELPlMhY1h/PVlyKbBPWKrsKeyrU0xqRrIrJAJPsZUDPKuyx8jRFgkvBtpNnp9mjeaYofFQzGZASAbDlelpTfzJlstEQ4IvI/r6QHDLbtSTDuZklPJzyScIlym+rnglYk56JZbTIObe19/5RAO/9v51zeyOH62wZ4Pp5XcbvvX9Wt39c7ZxbX/mm4lJkW3FV6WnFVqHnu865Lu/9O8hhrgA45wYi4VVpANtOnh1lj/f+Q2Cyc+4q/VwExe9IKzbzbC1PJOrSw0i78s65tbz3LzrnVqGgrTvnBiEDeud12bL3/m3n3PsN4FLkt1tPK7YKPccDZzvnTkQOabvPObcQ2VY4PuIZbl18zDm3vff+IefcMCSiViouRbYVa5VdhT1V6dnNt97795BIXtc52V4RPPIfIv9s3uKcW5El4bXPRM4OSMVaZbdbzyrssfKcgqyKXAE5/Pwq59wC5PD7yyOeVmw7eXaaPSBblx90zt2PnF3zUwAn25dfjbD9kS0HA5A/q/DeP6/+j8mCTZFdlZ71qZFZiZz6XsIYvUPvraFETbgU+cuAnlaeVntSZJsip2AM4WrFJcpOie5ShZ5WnlXoaY5WY8W2k2en2VPw3BwlyIrNPFvLM8D3iM6EPVS0CZcqv516NmtTM3oGeEtkLFOYbisuRbYVmyK7Knsq0NMUkQxjaPQUrFX2MqBnFfak6GmKBJeCbSfPTrNHn9eNwgccg6xOPh+ZvKitRlqTKBJGItYUAbAKPa0pH/KYqeXknJuCREy4u+DZpd77g1Nwy5OerbapijJyzq0DvO+9f6ng2U7e+3tScFVRFXpWYVO7yylTpkx20n8Vh3jvn2kFriqqQs8qbGoFT+fcqsCGyMTsC977Rc3gqqAU2e20p5U8nXPDvPfzW41tNVWhZxX2tLOMMrWPnHNbIFHJZnvv57YK22pqtew8wZApU6ZMmTJlypQpU6ZMmTJlapr6tVuBTJkyZcqUKVOmTJkyZcqUKdPyT3mCIVOmTJkyZcqUKVOmTJkyZcrUNOUJhkyZMmXKlClTpkyZMmXKlClT05QnGDJlypQpU6ZMmTJlypQpU6ZMTdP/AOtjt/jJQBwcAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1296x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "song_idx = 1; bar_idx = 80;\n",
    "\n",
    "plot_data1 = copy.deepcopy(midi_track_obj_list[song_idx].drum_bar_list_bin[bar_idx])\n",
    "\n",
    "plt.figure(figsize=(18, 6))\n",
    "plt.imshow(plot_data1, interpolation='nearest', origin='lower', aspect='auto', cmap='hot')\n",
    "plt.xticks(np.arange(0, plot_data1.shape[1], 1), rotation=90)\n",
    "plt.yticks(np.arange(0, plot_data1.shape[0], 2))\n",
    "#plt.ylim(20,80)\n",
    "plt.grid() ; plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Select instruments to keep in drum channel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "keeped Insts: 28\n",
      "keeped Insts: 46\n"
     ]
    }
   ],
   "source": [
    "# keep 95 % of all instrument count (total 28 insts)\n",
    "selected_inst_list_28 = [28, 35, 36, 37, 38, 39, 40, 42, 43, 44, 45, 46, 49, 51, 53, 54, 56, 57, \\\n",
    "                         59, 60, 61, 62, 63, 64, 69, 70, 80, 82]\n",
    "print ('keeped Insts: {}'.format(len(selected_inst_list_28)))\n",
    "\n",
    "# keep 99 % of all instrument count (total 46 insts)\n",
    "selected_inst_list_46 = [27, 28, 33, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, \\\n",
    "                         51, 53, 54, 55, 56, 57, 59, 60, 61, 62, 63, 64, 65, 67, 68, 69, 70, 73, \\\n",
    "                         74, 75, 76, 77, 80, 81, 82, 83, 85, 87]\n",
    "print ('keeped Insts: {}'.format(len(selected_inst_list_46)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## loop all files to compress drum [pitch / time] data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] file idx: [ 0 / 23 ], total bars: 101\n",
      "[info] file idx: [ 1 / 23 ], total bars: 81\n",
      "[info] file idx: [ 2 / 23 ], total bars: 91\n",
      "[info] file idx: [ 3 / 23 ], total bars: 94\n",
      "[info] file idx: [ 4 / 23 ], total bars: 86\n",
      "[info] file idx: [ 5 / 23 ], total bars: 84\n",
      "[info] file idx: [ 6 / 23 ], total bars: 101\n",
      "[info] file idx: [ 7 / 23 ], total bars: 91\n",
      "[info] file idx: [ 8 / 23 ], total bars: 88\n",
      "[info] file idx: [ 9 / 23 ], total bars: 79\n",
      "[info] file idx: [ 10 / 23 ], total bars: 140\n",
      "[info] file idx: [ 11 / 23 ], total bars: 68\n",
      "[info] file idx: [ 12 / 23 ], total bars: 74\n",
      "[info] file idx: [ 13 / 23 ], total bars: 62\n",
      "[info] file idx: [ 14 / 23 ], total bars: 109\n",
      "[info] file idx: [ 15 / 23 ], total bars: 67\n",
      "[info] file idx: [ 16 / 23 ], total bars: 111\n",
      "[info] file idx: [ 17 / 23 ], total bars: 95\n",
      "[info] file idx: [ 18 / 23 ], total bars: 52\n",
      "[info] file idx: [ 19 / 23 ], total bars: 118\n",
      "[info] file idx: [ 20 / 23 ], total bars: 140\n",
      "[info] file idx: [ 21 / 23 ], total bars: 133\n",
      "[info] file idx: [ 22 / 23 ], total bars: 100\n",
      "[info] file idx: [ 23 / 23 ], total bars: 146\n",
      "\n",
      "[info] All songs are processed.\n"
     ]
    }
   ],
   "source": [
    "song_drum_bar_list_28 = []\n",
    "song_drum_bar_list_46 = []\n",
    "\n",
    "total_files = len(midi_track_obj_list)\n",
    "\n",
    "for file_idx in range(total_files):\n",
    "\n",
    "    single_song_drum_bar_list = copy.deepcopy(midi_track_obj_list[file_idx].drum_bar_list_bin)\n",
    "    concat_bars_array = np.concatenate(copy.deepcopy(single_song_drum_bar_list), axis=1)\n",
    "    #print (concat_bars_array.shape)\n",
    "\n",
    "    drum_bars_list_28 = []\n",
    "    drum_bars_list_46 = []    \n",
    "\n",
    "    total_bars_num = len(single_song_drum_bar_list)\n",
    "\n",
    "    time_step_in_bar = single_song_drum_bar_list[0].shape[1]\n",
    "\n",
    "    t_range_base = np.array(list(range(0, time_step_in_bar, int(time_step_in_bar/16))))\n",
    "\n",
    "    for bar_idx in range(0, total_bars_num):\n",
    "\n",
    "        t_range = t_range_base + bar_idx*time_step_in_bar\n",
    "        t_range_n1 = copy.deepcopy(t_range_base + bar_idx*time_step_in_bar -1)\n",
    "        t_range_n1[t_range_n1<0] = 0\n",
    "        t_range_p1 = copy.deepcopy(t_range_base + bar_idx*time_step_in_bar +1)\n",
    "        t_range_p1[t_range_n1>concat_bars_array.shape[1]] = concat_bars_array.shape[1]\n",
    "        \n",
    "        \n",
    "        select_bar = concat_bars_array[:, t_range]\n",
    "        select_bar_n1 = concat_bars_array[:, t_range_n1]\n",
    "        select_bar_p1 = concat_bars_array[:, t_range_p1]\n",
    "        \n",
    "        bar_mixed = select_bar + select_bar_n1 + select_bar_p1\n",
    "        bar_mixed[bar_mixed>=0.5] = 1\n",
    "        bar_mixed[bar_mixed<0.5] = 0\n",
    "        bar_mixed_select_28 = copy.deepcopy(bar_mixed[selected_inst_list_28, :])        \n",
    "        bar_mixed_select_46 = copy.deepcopy(bar_mixed[selected_inst_list_46, :])\n",
    "\n",
    "        drum_bars_list_28.append(copy.deepcopy(bar_mixed_select_28))        \n",
    "        drum_bars_list_46.append(copy.deepcopy(bar_mixed_select_46))\n",
    "\n",
    "        \n",
    "    print ('[info] file idx: [ {} / {} ], total bars: {}'.format(file_idx, \n",
    "                                                                 total_files-1, \n",
    "                                                                 len(drum_bars_list_28)))\n",
    "    \n",
    "    song_drum_bar_list_28.append(drum_bars_list_28)\n",
    "    song_drum_bar_list_46.append(drum_bars_list_46)\n",
    "    \n",
    "print ('\\n[info] All songs are processed.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## save condensed drum midi data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] ./pre_processed_data/cdsed_drum_bar_list_28_46/song_drum_bar_list_28.pkl is saved.\n",
      "[info] ./pre_processed_data/cdsed_drum_bar_list_28_46/song_drum_bar_list_46.pkl is saved.\n"
     ]
    }
   ],
   "source": [
    "file_name = 'song_drum_bar_list_28.pkl'\n",
    "folder_name = './pre_processed_data/cdsed_drum_bar_list_28_46/'\n",
    "full_file_name = folder_name + file_name\n",
    "ensure_dir(full_file_name)\n",
    "    \n",
    "with open(full_file_name, 'wb') as pkl_file:\n",
    "    pickle.dump(song_drum_bar_list_28, pkl_file)\n",
    "    print ('[info] {} is saved.'.format(full_file_name))\n",
    "    \n",
    "file_name = 'song_drum_bar_list_46.pkl'\n",
    "folder_name = './pre_processed_data/cdsed_drum_bar_list_28_46/'\n",
    "full_file_name = folder_name + file_name\n",
    "ensure_dir(full_file_name)\n",
    "    \n",
    "with open(full_file_name, 'wb') as pkl_file:\n",
    "    pickle.dump(song_drum_bar_list_46, pkl_file)\n",
    "    print ('[info] {} is saved.'.format(full_file_name))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## check total note numbers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Total full-inst drum note num: 39283\n",
      "[info] Total 28-inst drum note num: 38490\n",
      "[info] Total 46-inst drum note num: 37864\n"
     ]
    }
   ],
   "source": [
    "song_note_num = 0\n",
    "for file_idx in range(len(midi_track_obj_list)):\n",
    "    single_song_drum_bar_list = copy.deepcopy(midi_track_obj_list[file_idx].drum_bar_list_bin)\n",
    "    concat_bars_array = np.concatenate(copy.deepcopy(single_song_drum_bar_list), axis=1)\n",
    "    song_note_num += np.sum(concat_bars_array).astype(int)\n",
    "print ('[info] Total full-inst drum note num: {}'.format(song_note_num))\n",
    "\n",
    "song_note_num = 0\n",
    "for song_idx in range(0, len(song_drum_bar_list_46)):\n",
    "    song_note_num += np.sum(np.array(song_drum_bar_list_46[song_idx])).astype(int)\n",
    "print ('[info] Total 28-inst drum note num: {}'.format(song_note_num))\n",
    "\n",
    "song_note_num = 0\n",
    "for song_idx in range(0, len(song_drum_bar_list_28)):\n",
    "    song_note_num += np.sum(np.array(song_drum_bar_list_28[song_idx])).astype(int)\n",
    "print ('[info] Total 46-inst drum note num: {}'.format(song_note_num))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show simplified drum data (28/46 insts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x324 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "song_idx = 22; bar_idx = 50;\n",
    "\n",
    "# show 28 instruments data\n",
    "plot_data1 = copy.deepcopy(song_drum_bar_list_28[song_idx][bar_idx])\n",
    "plt.figure(figsize=(18, 4))\n",
    "plt.imshow(plot_data1, interpolation='nearest', origin='lower', aspect='auto', cmap='hot')\n",
    "plt.xticks(np.arange(0, plot_data1.shape[1], 1), rotation=90)\n",
    "plt.yticks(np.arange(0, plot_data1.shape[0], 1))\n",
    "#plt.ylim(20,80)\n",
    "plt.title('28 insts bar data'); plt.grid() ; plt.show();\n",
    "\n",
    "# show 46 instruments data\n",
    "plot_data2 = copy.deepcopy(song_drum_bar_list_46[song_idx][bar_idx])\n",
    "plt.figure(figsize=(18, 4.5))\n",
    "plt.imshow(plot_data2, interpolation='nearest', origin='lower', aspect='auto', cmap='hot')\n",
    "plt.xticks(np.arange(0, plot_data2.shape[1], 1), rotation=90)\n",
    "plt.yticks(np.arange(0, plot_data2.shape[0], 2))\n",
    "#plt.ylim(20,80)\n",
    "plt.title('46 insts bar data'); plt.grid() ; plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## loop over files to calculate 46-inst drum SSM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Song idx: 0 , Total bars: 101\n",
      "SSM data shape: (101, 101)\n",
      "SSM data mean: 3.11\n",
      "SSM data std: 2.13\n",
      "SSM data max: 5.74\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 1 , Total bars: 81\n",
      "SSM data shape: (81, 81)\n",
      "SSM data mean: 3.03\n",
      "SSM data std: 1.78\n",
      "SSM data max: 5.83\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 2 , Total bars: 91\n",
      "SSM data shape: (91, 91)\n",
      "SSM data mean: 3.14\n",
      "SSM data std: 1.25\n",
      "SSM data max: 5.57\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 3 , Total bars: 94\n",
      "SSM data shape: (94, 94)\n",
      "SSM data mean: 1.42\n",
      "SSM data std: 1.48\n",
      "SSM data max: 3.46\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 4 , Total bars: 86\n",
      "SSM data shape: (86, 86)\n",
      "SSM data mean: 2.14\n",
      "SSM data std: 2.21\n",
      "SSM data max: 5.20\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 5 , Total bars: 84\n",
      "SSM data shape: (84, 84)\n",
      "SSM data mean: 1.65\n",
      "SSM data std: 1.24\n",
      "SSM data max: 4.80\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 6 , Total bars: 101\n",
      "SSM data shape: (101, 101)\n",
      "SSM data mean: 1.12\n",
      "SSM data std: 1.55\n",
      "SSM data max: 4.24\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 7 , Total bars: 91\n",
      "SSM data shape: (91, 91)\n",
      "SSM data mean: 2.99\n",
      "SSM data std: 1.75\n",
      "SSM data max: 5.74\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 8 , Total bars: 88\n",
      "SSM data shape: (88, 88)\n",
      "SSM data mean: 2.39\n",
      "SSM data std: 1.68\n",
      "SSM data max: 5.48\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 9 , Total bars: 79\n",
      "SSM data shape: (79, 79)\n",
      "SSM data mean: 1.41\n",
      "SSM data std: 1.20\n",
      "SSM data max: 4.58\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 10 , Total bars: 140\n",
      "SSM data shape: (140, 140)\n",
      "SSM data mean: 3.76\n",
      "SSM data std: 1.91\n",
      "SSM data max: 6.63\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 11 , Total bars: 68\n",
      "SSM data shape: (68, 68)\n",
      "SSM data mean: 0.90\n",
      "SSM data std: 1.37\n",
      "SSM data max: 4.12\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 12 , Total bars: 74\n",
      "SSM data shape: (74, 74)\n",
      "SSM data mean: 2.53\n",
      "SSM data std: 1.69\n",
      "SSM data max: 5.00\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 13 , Total bars: 62\n",
      "SSM data shape: (62, 62)\n",
      "SSM data mean: 2.37\n",
      "SSM data std: 1.69\n",
      "SSM data max: 6.00\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 14 , Total bars: 109\n",
      "SSM data shape: (109, 109)\n",
      "SSM data mean: 1.56\n",
      "SSM data std: 2.13\n",
      "SSM data max: 5.10\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 15 , Total bars: 67\n",
      "SSM data shape: (67, 67)\n",
      "SSM data mean: 1.89\n",
      "SSM data std: 1.81\n",
      "SSM data max: 5.29\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 16 , Total bars: 111\n",
      "SSM data shape: (111, 111)\n",
      "SSM data mean: 2.12\n",
      "SSM data std: 1.82\n",
      "SSM data max: 4.90\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 17 , Total bars: 95\n",
      "SSM data shape: (95, 95)\n",
      "SSM data mean: 3.65\n",
      "SSM data std: 1.51\n",
      "SSM data max: 5.57\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 18 , Total bars: 52\n",
      "SSM data shape: (52, 52)\n",
      "SSM data mean: 2.48\n",
      "SSM data std: 1.32\n",
      "SSM data max: 4.24\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 19 , Total bars: 118\n",
      "SSM data shape: (118, 118)\n",
      "SSM data mean: 2.03\n",
      "SSM data std: 1.37\n",
      "SSM data max: 6.00\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 20 , Total bars: 140\n",
      "SSM data shape: (140, 140)\n",
      "SSM data mean: 1.35\n",
      "SSM data std: 0.98\n",
      "SSM data max: 5.00\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 21 , Total bars: 133\n",
      "SSM data shape: (133, 133)\n",
      "SSM data mean: 2.99\n",
      "SSM data std: 1.43\n",
      "SSM data max: 5.20\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 22 , Total bars: 100\n",
      "SSM data shape: (100, 100)\n",
      "SSM data mean: 3.36\n",
      "SSM data std: 1.27\n",
      "SSM data max: 6.71\n",
      "SSM data min: 0.00\n",
      "\n",
      "[info] Song idx: 23 , Total bars: 146\n",
      "SSM data shape: (146, 146)\n",
      "SSM data mean: 1.70\n",
      "SSM data std: 1.29\n",
      "SSM data max: 4.47\n",
      "SSM data min: 0.00\n",
      "\n",
      "\n",
      "\n",
      "[info] Process done, total songs: 24\n"
     ]
    }
   ],
   "source": [
    "total_files = len(midi_track_obj_list)\n",
    "\n",
    "drum_ssm_data_list = []\n",
    "\n",
    "for song_idx in range(0, total_files):\n",
    "    \n",
    "    #if song_idx > 10: break\n",
    "\n",
    "    drum_data = copy.deepcopy(song_drum_bar_list_46[song_idx])\n",
    "\n",
    "    print ('[info] Song idx: {} , Total bars: {}'.format(song_idx, len(drum_data)))\n",
    "    #print ('Bars data shape: {}'.format(drum_data[0].shape))\n",
    "\n",
    "    drum_data_flat = np.array([np.reshape(x, np.prod(x.shape)) for x in drum_data])\n",
    "    #print (drum_data_flat.shape)\n",
    "\n",
    "    drum_df = pd.DataFrame(drum_data_flat)\n",
    "\n",
    "    #print('frame length: {}'.format(len(drum_df.iloc[:,0])))\n",
    "    #print('pitch class: {}'.format(len(drum_df.iloc[0,:])))\n",
    "\n",
    "    drum_ssm_data = squareform(pdist(drum_df, metric='euclidean'))\n",
    "\n",
    "    print ('SSM data shape: {}'.format(drum_ssm_data.shape))\n",
    "    print ('SSM data mean: {:.2f}'.format(np.mean(drum_ssm_data)))\n",
    "    print ('SSM data std: {:.2f}'.format(np.std(drum_ssm_data)))\n",
    "    print ('SSM data max: {:.2f}'.format(np.max(drum_ssm_data)))\n",
    "    print ('SSM data min: {:.2f}'.format(np.min(drum_ssm_data)))\n",
    "    print ('')\n",
    "    \n",
    "    drum_ssm_data_list.append(copy.copy(drum_ssm_data))\n",
    "    \n",
    "print ('\\n\\n[info] Process done, total songs: {}'.format(len(drum_ssm_data_list)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check drum SSM data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show data 1\n",
    "song_idx = 5\n",
    "plot_data1 = copy.deepcopy(drum_ssm_data_list[song_idx])\n",
    "plt.figure(figsize=(8, 8))\n",
    "normal_data = plot_data1\n",
    "abnormal_data = (np.max(normal_data) - normal_data)\n",
    "plt.imshow(abnormal_data, origin='lower', cmap='hot')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save all drum SSM data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Info] start SSM saving...\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 0, 01_A_Hard_Days_Night_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 1, 02_Anna_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 2, 03_Back_In_The_USSR_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 3, 04_Cant_Buy_Me_Love_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 4, 05_Hold_Me_Tight_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 5, 06_I_Call_Your_Name_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 6, 07_I_Wanna_Be_Your_Man_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 7, 08_Money_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 8, 09_The_Word_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 9, 10_Free_As_A_Bird_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 10, 11_Hey_Jude_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 11, 12_Little_Child_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 12, 13_Hey_Bulldog_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 13, 14_Lovely_Rita_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 14, 15_The_Night_Before_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 15, 16_From_Me_To_You_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 16, 17_Roll_Over_Beethoven_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 17, 18_Come_Together_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 18, 19_Babys_In_Black_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 19, 21_Michael_Jackson_-_Bad_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 20, 22_Michael_Jackson_-_Billie_Jean_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 21, 23_Michael_Jackson_-_Man_In_The_Mirror_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 22, 24_Michael_Jackson_-_Smooth_Criminal_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "[Info] Song Index: 23, 25_Michael_Jackson_-_Beat_It_drum_only\n",
      "[Info] drum SSMs save done.\n",
      "[Info] 2019-10-18 17:59:02\n",
      "\n",
      "\n",
      "[info] files write done.\n",
      "[Info] 2019-10-18 17:59:02\n"
     ]
    }
   ],
   "source": [
    "print ('[Info] start SSM saving...')\n",
    "print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + '\\n')\n",
    "\n",
    "song_idx_idx_start = 0\n",
    "song_idx_idx_end = len(drum_ssm_data_list)\n",
    "\n",
    "for song_idx in range(song_idx_idx_start, song_idx_idx_end):\n",
    "    \n",
    "    base_file_name = drum_only_mid_flist[song_idx].split('/')[-1][:-4]\n",
    "    print ('[Info] Song Index: {}, {}'.format(song_idx, base_file_name))\n",
    "        \n",
    "    file_name = 'song_barlv_ssm_{:0>5}.pkl'.format(base_file_name)\n",
    "    folder_name = './pre_processed_data/bar_level_drum_ssm/'\n",
    "    full_file_name = folder_name + file_name\n",
    "    ensure_dir(full_file_name)\n",
    "        \n",
    "    with open(full_file_name, 'wb') as pkl_file:\n",
    "        pickle.dump(drum_ssm_data_list[song_idx], pkl_file)\n",
    "            \n",
    "    print ('[Info] drum SSMs save done.')\n",
    "    print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + '\\n')\n",
    "\n",
    "#print(bar_ssm_data_tmp.shape)\n",
    "print ('\\n[info] files write done.')\n",
    "print ('[Info] ' + datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define read CQT/drum SSM pkl file function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total CQT SSM files: 24\n",
      "Total Drum SSM files: 24\n",
      "function define done.\n"
     ]
    }
   ],
   "source": [
    "# load CQT SSM file list\n",
    "cqt_ssm_flist = np.sort(glob.glob('./pre_processed_data/bar_level_cqt_ssm/*.pkl', recursive=True)).tolist()\n",
    "print ('Total CQT SSM files: {}'.format(len(cqt_ssm_flist)))\n",
    "\n",
    "# load drum SSM file list\n",
    "drum_ssm_flist = np.sort(glob.glob('./pre_processed_data/bar_level_drum_ssm/*.pkl', recursive=True)).tolist()\n",
    "print ('Total Drum SSM files: {}'.format(len(drum_ssm_flist)))\n",
    "\n",
    "# define read CQT SSM function\n",
    "def get_cqt_ssm_data(cqt_file_index, \n",
    "                     cqt_ssm_flist_in=cqt_ssm_flist):\n",
    "    \n",
    "    cqt_file_name = cqt_ssm_flist_in[cqt_file_index]\n",
    "    with open(cqt_file_name, 'rb') as pkl_file:        \n",
    "        load_cqt_data = pickle.load(pkl_file)\n",
    "    \n",
    "    # make data value in range [0 ~ 10.5]\n",
    "    return(load_cqt_data/50.0)\n",
    "\n",
    "# define read drum SSM function\n",
    "def get_drum_ssm_data(drum_file_index, \n",
    "                      drum_ssm_flist_in=drum_ssm_flist):\n",
    "    \n",
    "    drum_file_name = drum_ssm_flist_in[drum_file_index]\n",
    "    with open(drum_file_name, 'rb') as pkl_file:\n",
    "        load_drum_data = pickle.load(pkl_file)\n",
    "    \n",
    "    # make data value in range [0 ~ 1.0]\n",
    "    return(load_drum_data/15.0)\n",
    "\n",
    "print ('function define done.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define CQT padding Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Padding CQT SSM shape: (256, 256)\n",
      "Padding CQT SSM (reflective) shape: (256, 256)\n"
     ]
    }
   ],
   "source": [
    "# output data shape = (256, 256)\n",
    "def get_cqt_pad(input_ssm_data):\n",
    "    \n",
    "    #ssm_with_pad = np.zeros([260, 260])\n",
    "    ssm_with_pad = np.zeros([256, 256])\n",
    "    \n",
    "    input_ssm_h = input_ssm_data.shape[0]\n",
    "    input_ssm_w = input_ssm_data.shape[1]\n",
    "    \n",
    "    if input_ssm_h > 256:\n",
    "        ssm_data_crop = input_ssm_data[0:256, 0:256]\n",
    "    else:\n",
    "        ssm_data_crop = input_ssm_data\n",
    "    \n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(input_ssm_data)\n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(np.rot90(np.triu(np.transpose(input_ssm_data), k=1)))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.rot90(np.triu(np.transpose(ssm_data_crop), k=1)))\n",
    "    ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(ssm_data_crop))\n",
    "    \n",
    "    # get value=1 mask as the ssm shape\n",
    "    #ssm_mask = np.zeros([256, 256])\n",
    "    #ssm_mask[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(np.ones_like(ssm_data_crop)))\n",
    "    \n",
    "    \n",
    "    # get original ssm size\n",
    "    get_original_song_len = input_ssm_data.shape[1]\n",
    "    \n",
    "    #return (ssm_with_pad, ssm_mask, get_original_song_len)\n",
    "    #return (ssm_with_pad, get_original_song_len)\n",
    "    return (ssm_with_pad)\n",
    "\n",
    "# show padding SSM shape\n",
    "file_idx = 0\n",
    "print('Padding CQT SSM shape: {}'.format(get_cqt_pad(get_cqt_ssm_data(file_idx)).shape))\n",
    "#print('padding cqt SSM mask shape: {}'.format(get_cqt_pad(cqt_ssm_list[song_index])[1].shape))\n",
    "#print('padding drum SSM shape: {}'.format(get_padding_ssm(drum_ssm_list[song_index])[0].shape))\n",
    "#print('padding drum SSM mask shape: {}'.format(get_padding_ssm(drum_ssm_list[song_index])[1].shape))\n",
    "\n",
    "\n",
    "def get_cqt_pad_refltv(input_ssm_data):\n",
    "    \n",
    "    #ssm_with_pad = np.zeros([260, 260])\n",
    "    ssm_with_pad = np.zeros([256, 256])\n",
    "    \n",
    "    input_ssm_h = input_ssm_data.shape[0]\n",
    "    input_ssm_w = input_ssm_data.shape[1]\n",
    "    \n",
    "    repeat_ratio = (256//input_ssm_h) + 2\n",
    "    \n",
    "    ssm_htemp = copy.copy(input_ssm_data)\n",
    "    \n",
    "    #print ('ratio: {}'.format(repeat_ratio))\n",
    "    \n",
    "    for repead_idx in range(0, repeat_ratio):\n",
    "        ssm_htemp = np.hstack([ssm_htemp, copy.copy(input_ssm_data)])\n",
    "        #print ('stack: {}'.format(repead_idx))\n",
    "        \n",
    "    ssm_vtemp = copy.copy(ssm_htemp)\n",
    "        \n",
    "    for repead_idx in range(0, repeat_ratio):\n",
    "        ssm_vtemp = np.vstack([ssm_vtemp, copy.copy(ssm_htemp)])\n",
    "    \n",
    "    \n",
    "    ssm_full_pad_out = ssm_vtemp[0:256, 0:256]\n",
    "    \n",
    "    #if input_ssm_h > 256:\n",
    "    #    ssm_data_crop = input_ssm_data[0:256, 0:256]\n",
    "    #else:\n",
    "    #    ssm_data_crop = input_ssm_data\n",
    "    \n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(input_ssm_data)\n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(np.rot90(np.triu(np.transpose(input_ssm_data), k=1)))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.rot90(np.triu(np.transpose(ssm_data_crop), k=1)))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(ssm_data_crop))\n",
    "    \n",
    "    # get value=1 mask as the ssm shape\n",
    "    #ssm_mask = np.zeros([256, 256])\n",
    "    #ssm_mask[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(np.ones_like(ssm_data_crop)))\n",
    "    \n",
    "    \n",
    "    # get original ssm size\n",
    "    get_original_song_len = input_ssm_data.shape[1]\n",
    "\n",
    "    #return (ssm_with_pad, ssm_mask, get_original_song_len)    \n",
    "    #return (ssm_full_pad_out, get_original_song_len)\n",
    "    return (ssm_full_pad_out)\n",
    "\n",
    "\n",
    "# show padding SSM shape\n",
    "file_idx = 0\n",
    "print('Padding CQT SSM (reflective) shape: {}'.format(get_cqt_pad_refltv(get_cqt_ssm_data(file_idx)).shape))\n",
    "#print('padding cqt SSM mask shape: {}'.format(get_cqt_pad(cqt_ssm_list[song_index])[1].shape))\n",
    "#print('padding drum SSM shape: {}'.format(get_padding_ssm(drum_ssm_list[song_index])[0].shape))\n",
    "#print('padding drum SSM mask shape: {}'.format(get_padding_ssm(drum_ssm_list[song_index])[1].shape))\n",
    "\n",
    "\n",
    "\n",
    "def get_ssm_vdiff(input_ssm_vdata):\n",
    "    gsh_ssm_vout_tmp = np.vstack([np.zeros([1, input_ssm_vdata.shape[1]]), input_ssm_vdata])[:input_ssm_vdata.shape[0],:]    \n",
    "    gsh_ssm_vout = copy.copy(input_ssm_vdata - gsh_ssm_vout_tmp)\n",
    "    return(gsh_ssm_vout)\n",
    "\n",
    "\n",
    "def get_ssm_hdiff(input_ssm_hdata):\n",
    "    gsh_ssm_hout_tmp = np.hstack([np.zeros([input_ssm_hdata.shape[0], 1]), input_ssm_hdata])[:, :input_ssm_hdata.shape[1]]\n",
    "    gsh_ssm_hout = copy.copy(input_ssm_hdata - gsh_ssm_hout_tmp)\n",
    "    return(gsh_ssm_hout)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Check CQT SSM data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "song_length: 81\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# show cqt ssm data\n",
    "file_idx = 1\n",
    "\n",
    "song_length = get_cqt_ssm_data(file_idx).shape[0]\n",
    "print ('song_length: {}'.format(song_length))\n",
    "\n",
    "plot_data_0 = get_cqt_ssm_data(file_idx)\n",
    "plot_data_1 = get_cqt_pad(get_cqt_ssm_data(file_idx))\n",
    "plot_data_2 = get_cqt_pad_refltv(get_cqt_ssm_data(file_idx))\n",
    "plot_data_3 = get_ssm_vdiff(get_cqt_ssm_data(file_idx))\n",
    "plot_data_4 = get_ssm_hdiff(get_cqt_ssm_data(file_idx))\n",
    "\n",
    "\n",
    "# plot data0 (original SSM)\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.imshow(plot_data_0, origin='lower', cmap='hot')\n",
    "plt.show()\n",
    "\n",
    "# plot data1 (pad 256 SSM)\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.imshow(plot_data_1, origin='lower', cmap='hot')\n",
    "plt.show()\n",
    "\n",
    "# plot data2 (reflactive padding)\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.imshow(plot_data_2, origin='lower', cmap='hot')\n",
    "plt.show()\n",
    "\n",
    "# plot data3 (h axis diff)\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.imshow(plot_data_3, origin='lower', cmap='hot')\n",
    "plt.show()\n",
    "\n",
    "# plot data4 (v axis diff)\n",
    "plt.figure(figsize=(6, 6))\n",
    "plt.imshow(plot_data_4, origin='lower', cmap='hot')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define drum SSM padding Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Padding drum SSM shape: (256, 256)\n",
      "Padding drum SSM mask shape: (256, 256)\n"
     ]
    }
   ],
   "source": [
    "# output data shape = (256, 256)\n",
    "def get_drum_pad(input_ssm_data):\n",
    "    \n",
    "    canvas_size = 256\n",
    "    \n",
    "    ssm_with_pad = np.zeros([canvas_size, canvas_size])\n",
    "    \n",
    "    input_ssm_h = input_ssm_data.shape[0]\n",
    "    input_ssm_w = input_ssm_data.shape[1]\n",
    "    \n",
    "    if input_ssm_h > canvas_size:\n",
    "        ssm_data_crop = input_ssm_data[0:canvas_size, 0:canvas_size]\n",
    "    else:\n",
    "        ssm_data_crop = input_ssm_data\n",
    "    \n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(input_ssm_data)\n",
    "    #ssm_with_pad[1:input_ssm_h+1, 1:input_ssm_w+1] = copy.copy(np.rot90(np.triu(np.transpose(input_ssm_data), k=1)))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.rot90(np.triu(np.transpose(ssm_data_crop), k=1)))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(ssm_data_crop))\n",
    "    #ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.triu(np.transpose(ssm_data_crop)))\n",
    "    ssm_with_pad[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.transpose(ssm_data_crop))\n",
    "    \n",
    "    # get value=1 mask as the ssm shape\n",
    "    ssm_mask = np.zeros([canvas_size, canvas_size])\n",
    "    ssm_mask[0:input_ssm_h, 0:input_ssm_w] = copy.copy(np.triu(np.transpose(np.ones_like(ssm_data_crop))))\n",
    "    \n",
    "    \n",
    "    # get original ssm size\n",
    "    get_original_song_len = input_ssm_data.shape[1]\n",
    "    \n",
    "    #return (ssm_with_pad, ssm_mask, get_original_song_len)\n",
    "    return (ssm_with_pad, ssm_mask)\n",
    "\n",
    "\n",
    "\n",
    "def get_ssm_vdiff(input_ssm_vdata):\n",
    "    \n",
    "    gsh_ssm_vout_tmp = np.vstack([np.zeros([1, input_ssm_vdata.shape[1]]), input_ssm_vdata])[:input_ssm_vdata.shape[0],:]\n",
    "    \n",
    "    gsh_ssm_vout = copy.copy(input_ssm_vdata - gsh_ssm_vout_tmp)\n",
    "    \n",
    "    #return(gsh_ssm_vout_tmp)\n",
    "    return(gsh_ssm_vout)\n",
    "\n",
    "\n",
    "def get_ssm_hdiff(input_ssm_hdata):\n",
    "    \n",
    "    gsh_ssm_hout_tmp = np.hstack([np.zeros([input_ssm_hdata.shape[0], 1]), input_ssm_hdata])[:, :input_ssm_hdata.shape[1]]\n",
    "    \n",
    "    gsh_ssm_hout = copy.copy(input_ssm_hdata - gsh_ssm_hout_tmp)\n",
    "    \n",
    "    #return(gsh_ssm_hout_tmp)\n",
    "    return(gsh_ssm_hout)\n",
    "\n",
    "\n",
    "# show data shape\n",
    "song_idx = 20\n",
    "#print (get_drum_pad(get_drum_ssm_data(file_idx))[2])\n",
    "print ('Padding drum SSM shape: {}'.format(get_drum_pad(get_drum_ssm_data(file_idx))[0].shape))\n",
    "print ('Padding drum SSM mask shape: {}'.format(get_drum_pad(get_drum_ssm_data(file_idx))[1].shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "scrolled": true
   },
   "source": [
    "# Check drum SSM padding data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "song_idx = 5\n",
    "plot_data0 = get_drum_pad(get_drum_ssm_data(file_idx))[0]\n",
    "plot_data1 = get_drum_pad(get_drum_ssm_data(file_idx))[1]\n",
    "\n",
    "plt.figure(figsize=(8, 8))\n",
    "plt.imshow(plot_data0, origin='lower', cmap='hot')\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(8, 8))\n",
    "plt.imshow(plot_data1, origin='lower', cmap='hot')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# loop over files to add pad "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File processed: 1\n",
      "File processed: 2\n",
      "File processed: 3\n",
      "File processed: 4\n",
      "File processed: 5\n",
      "File processed: 6\n",
      "File processed: 7\n",
      "File processed: 8\n",
      "File processed: 9\n",
      "File processed: 10\n",
      "File processed: 11\n",
      "File processed: 12\n",
      "File processed: 13\n",
      "File processed: 14\n",
      "File processed: 15\n",
      "File processed: 16\n",
      "File processed: 17\n",
      "File processed: 18\n",
      "File processed: 19\n",
      "File processed: 20\n",
      "File processed: 21\n",
      "File processed: 22\n",
      "File processed: 23\n",
      "File processed: 24\n",
      "\n",
      "All files are processed: 24\n"
     ]
    }
   ],
   "source": [
    "total_good_files = len(cqt_ssm_flist)\n",
    "\n",
    "cqt_drum_ssm_data_list = []\n",
    "\n",
    "for file_idx in range(0, total_good_files):\n",
    "    \n",
    "    song_len = get_cqt_ssm_data(file_idx).shape[0]\n",
    "    \n",
    "    # get cqt SSM data\n",
    "    cqt_ssm_data_loaded = get_cqt_ssm_data(file_idx)\n",
    "    \n",
    "    cqt_ssm_reflect_pad_no_diff = get_cqt_pad_refltv(cqt_ssm_data_loaded)\n",
    "    \n",
    "    cqt_ssm_reflect_pad_hdiff = get_ssm_hdiff(cqt_ssm_reflect_pad_no_diff)\n",
    "    \n",
    "    cqt_ssm_reflect_pad_vdiff = get_ssm_vdiff(cqt_ssm_reflect_pad_no_diff)\n",
    "    \n",
    "    cqt_ssm_reflect_pad = np.concatenate((cqt_ssm_reflect_pad_no_diff[:,:,np.newaxis],\n",
    "                                          cqt_ssm_reflect_pad_hdiff[:,:,np.newaxis],\n",
    "                                          cqt_ssm_reflect_pad_vdiff[:,:,np.newaxis]), \n",
    "                                         axis=-1)      \n",
    "    \n",
    "    \n",
    "    \n",
    "    # get drum SSM data\n",
    "    drum_ssm_data_loaded = get_drum_ssm_data(file_idx)\n",
    "    \n",
    "    drum_ssm_pad_no_diff = get_drum_pad(drum_ssm_data_loaded)[0]\n",
    "    \n",
    "    drum_ssm_pad_hdiff = get_ssm_hdiff(drum_ssm_pad_no_diff)\n",
    "    \n",
    "    drum_ssm_pad_vdiff = get_ssm_vdiff(drum_ssm_pad_no_diff)\n",
    "\n",
    "    drum_ssm_pad = np.concatenate((drum_ssm_pad_no_diff[:,:,np.newaxis],\n",
    "                                   drum_ssm_pad_hdiff[:,:,np.newaxis],\n",
    "                                   drum_ssm_pad_vdiff[:,:,np.newaxis]), \n",
    "                                  axis=-1)     \n",
    "\n",
    "    drum_ssm_pad_mask = get_drum_pad(drum_ssm_data_loaded)[1]\n",
    "    \n",
    "    drum_ssm_pad_mask_all_layer = np.concatenate((drum_ssm_pad_mask[:,:,np.newaxis],\n",
    "                                                  drum_ssm_pad_mask[:,:,np.newaxis],\n",
    "                                                  drum_ssm_pad_mask[:,:,np.newaxis]), \n",
    "                                                 axis=-1)    \n",
    "    \n",
    "    ssm_valid_pix = np.mean(drum_ssm_pad_mask)\n",
    "    \n",
    "    \n",
    "    # save data into list\n",
    "    cqt_drum_ssm_data_list.append([cqt_ssm_reflect_pad,\n",
    "                                   drum_ssm_pad,\n",
    "                                   drum_ssm_pad_mask_all_layer,\n",
    "                                   ssm_valid_pix,\n",
    "                                   song_len])\n",
    "\n",
    "    if (file_idx+1)%1==0:\n",
    "        print ('File processed: {}'.format(file_idx+1))\n",
    "\n",
    "\n",
    "print ('\\nAll files are processed: {}'.format(len(cqt_drum_ssm_data_list)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# check data format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] Total songs: 24\n",
      "(256, 256, 3)\n",
      "(256, 256, 3)\n",
      "(256, 256, 3)\n",
      "0.0506744384765625\n",
      "81\n"
     ]
    }
   ],
   "source": [
    "print('[info] Total songs: {}'.format(len(cqt_drum_ssm_data_list)))\n",
    "print(cqt_drum_ssm_data_list[1][0].shape)\n",
    "print(cqt_drum_ssm_data_list[1][1].shape)\n",
    "print(cqt_drum_ssm_data_list[1][2].shape)\n",
    "print(cqt_drum_ssm_data_list[1][3])\n",
    "print(cqt_drum_ssm_data_list[1][4])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# save all SSM data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_00.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_01.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_02.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_03.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_04.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_05.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_06.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_07.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_08.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_09.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_10.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_11.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_12.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_13.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_14.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_15.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_16.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_17.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_18.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_19.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_20.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_21.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_22.pkl\" is saved.\n",
      "[info] File: \"./pre_processed_data/merged_ssm_pkg/test_midi_song_23.pkl\" is saved.\n",
      "\n",
      "[info] All SSM pkg files are saved.\n"
     ]
    }
   ],
   "source": [
    "total_files = len(cqt_drum_ssm_data_list)\n",
    "\n",
    "for file_idx in range(0, total_files):\n",
    "    \n",
    "    cqt_ssm_data_ary = cqt_drum_ssm_data_list[file_idx][0]\n",
    "    drum_ssm_data_ary = cqt_drum_ssm_data_list[file_idx][1]\n",
    "    drum_ssm_mask_ary = cqt_drum_ssm_data_list[file_idx][2]\n",
    "    valid_pix_mean_ary = cqt_drum_ssm_data_list[file_idx][3]\n",
    "    song_len_data_ary = cqt_drum_ssm_data_list[file_idx][4]\n",
    "    \n",
    "    dump_data = [copy.deepcopy(cqt_ssm_data_ary),\n",
    "                 copy.deepcopy(drum_ssm_data_ary),\n",
    "                 copy.deepcopy(drum_ssm_mask_ary),\n",
    "                 copy.deepcopy(valid_pix_mean_ary),\n",
    "                 copy.deepcopy(song_len_data_ary)\n",
    "                 ]\n",
    "        \n",
    "    file_name = 'test_midi_song_{:0>2}.pkl'.format(file_idx)\n",
    "    folder_name = './pre_processed_data/merged_ssm_pkg/'\n",
    "    full_file_name = folder_name + file_name    \n",
    "    ensure_dir(full_file_name)\n",
    "        \n",
    "    with open(full_file_name, 'wb') as pkl_file:\n",
    "        pickle.dump(dump_data, pkl_file)\n",
    "        print ('[info] File: \\\"{}\\\" is saved.'.format(full_file_name))\n",
    "        \n",
    "print ('\\n[info] All SSM pkg files are saved.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}