{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extracting information from MIDI files\n",
    "\n",
    "In our first lesson we want to get to know how to work with datasets so we can\n",
    "\n",
    "* parse a dataset\n",
    "* analyse the dataset and make sure it matches our expectations\n",
    "* check if errors occured during parsing\n",
    "\n",
    "After this is done we want to find out how we can generate new drum patterns from the existing one.\n",
    "But in order to do this we need to take a look at the quantisation (?) of our patterns.\n",
    "\n",
    "In such experiments, the cleaning of the dataset and setting the data up properly takes most of the time. But if we make mistakes here those mistakes will propagate through our system – so it's a good idea to spend some time with this task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import glob\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "np.random.seed(42)  # makes the randomness deterministic\n",
    "\n",
    "%matplotlib inline\n",
    "# todo: try %matplotlib widget\n",
    "plt.rcParams['figure.figsize'] = (15, 5)\n",
    "plt.rcParams['axes.grid'] = True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting the dataset\n",
    "\n",
    "As our methods in this machine learning based workshop rely on data, we need a way to obtain such data. This will be on of our first endeavours.\n",
    "Thankfully there are search engines which help us to find data in the internet easily.\n",
    "When searching for *\"midi dataset\"*, one search result is [https://colinraffel.com/projects/lmd/](https://colinraffel.com/projects/lmd/).\n",
    "\n",
    "> The Lakh MIDI dataset is a collection of 176,581 unique MIDI files, 45,129 of which have been matched and aligned to entries in the Million Song Dataset. [...]\n",
    "\n",
    "Of course the cultural bias of such a dataset is a thing we need to be aware of.\n",
    "What kind of music is transcribable into the MIDI format and what kind of music is transcribed as MIDI at all?\n",
    "Maybe we can shed some light on the last question by inspecting the dataset.\n",
    "But before we can do this, we need to download and understand the data.\n",
    "\n",
    "We use a function which will check if the files are already downloaded and if not will download and extract the directory."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "midi already downloaded to ../datasets/lmd/lmd_full.tap.gz\n",
      "json already downloaded to ../datasets/lmd/md5_to_path.json\n"
     ]
    }
   ],
   "source": [
    "import urllib.request\n",
    "import subprocess\n",
    "\n",
    "def download_dataset(download_path: str = \"../datasets/lmd\"):\n",
    "    os.makedirs(download_path, exist_ok=True)\n",
    "    archive_dir = os.path.join(download_path, \"lmd_full\")\n",
    "    \n",
    "    dl_files = {\n",
    "        \"midi\": {\n",
    "            \"path\": os.path.join(download_path, \"lmd_full.tap.gz\"),\n",
    "            \"url\": \"http://hog.ee.columbia.edu/craffel/lmd/lmd_full.tar.gz\"\n",
    "        },\n",
    "        \"json\": {\n",
    "            \"path\": os.path.join(download_path, \"md5_to_path.json\"),\n",
    "            \"url\": \"http://hog.ee.columbia.edu/craffel/lmd/md5_to_paths.json\"\n",
    "        },\n",
    "    }\n",
    "    \n",
    "    for dl_name, dl in dl_files.items():\n",
    "        if os.path.isfile(dl[\"path\"]) or (dl_name == \"midi\" and os.path.isdir(archive_dir)):\n",
    "            print(f\"{dl_name} already downloaded to {dl['path']}\")\n",
    "            continue\n",
    "        print(f\"Start downloading {dl_name} to {dl['path']} - this can take multiple minutes!\")\n",
    "        urllib.request.urlretrieve(dl['url'], dl['path'])\n",
    "        print(f\"Finished downloading\")\n",
    "    \n",
    "    if not os.path.isdir(archive_dir):\n",
    "        print(\"Start extracting the files of archive - this will take some minutes\")\n",
    "        # todo: windows has no tar\n",
    "        subprocess.check_output([\n",
    "            'tar', '-xzf', dl_files[\"midi\"][\"path\"],\n",
    "            '-C', os.path.join(download_path)\n",
    "        ])\n",
    "        print(\"Finished extracting\")\n",
    "    \n",
    "    if os.path.isfile(dl_files[\"midi\"][\"path\"]):\n",
    "        print(\"Remove archive\")\n",
    "        os.remove(dl_files[\"midi\"][\"path\"])\n",
    "\n",
    "download_dataset()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Parsing the dataset\n",
    "\n",
    "When working with large sets of files the unix utility [glob](https://en.wikipedia.org/wiki/Glob_(programming)) comes in handy as we can describe the pattern of the file paths we want to match instead of listing all files.\n",
    "\n",
    "When we take a quick look at the pattern it seems they all follow a structure like\n",
    "\n",
    "```\n",
    "../datasets/lmd/lmd_full/2/4a0cbb3f083d14d57858c87b26f85873.mid\n",
    "```\n",
    "\n",
    "Soon we will understand why the filename has this cryptic format, but for now we simply want to parse all available files into an array."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 178561 midi files in dataset\n"
     ]
    }
   ],
   "source": [
    "midi_files = glob.glob('../datasets/lmd/lmd_full/*/*.mid')\n",
    "print(f'Found {len(midi_files)} midi files in dataset')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['../datasets/lmd/lmd_full/4/4a69b1a10e4dbe409ac922de4f6256d8.mid',\n",
       "       '../datasets/lmd/lmd_full/b/b102261b4c27ea58bad4777b5df5be5e.mid',\n",
       "       '../datasets/lmd/lmd_full/3/3a8ccaab480919c35f37fdf08c238e29.mid',\n",
       "       '../datasets/lmd/lmd_full/d/dc6416232e05b3d44fb62007ca84b474.mid',\n",
       "       '../datasets/lmd/lmd_full/4/4fb744a04c1afbb0309b581410b33363.mid'],\n",
       "      dtype='<U63')"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# select 5 random files\n",
    "np.random.choice(midi_files, 5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "But how are sure that we matched every midi file in our folder?\n",
    "We can simply match every file in our dataset directory and show us the differences by using [sets](https://realpython.com/python-sets/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Did not match 0 files\n"
     ]
    }
   ],
   "source": [
    "missed_files = set(glob.glob('../datasets/lmd/lmd_full/**/*.*', recursive=True)) - set(midi_files)\n",
    "print(f'Did not match {len(missed_files)} files')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Thankfully we have a rather clean dataset here which does not force us to do file acrobatics.\n",
    "\n",
    "But we did not just download MIDI files, we also downloaded a JSON file which we will now take a look into."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Choosen ../datasets/lmd/lmd_full/e/ed05335f8f6c273f506a997137b2805b.mid as example MIDI file\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['QUALITY MIDI/EClapton-Promises.mid']"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import json\n",
    "\n",
    "with open(\"../datasets/lmd/md5_to_path.json\") as f:\n",
    "    md5_filenames = json.load(f)\n",
    "\n",
    "example_midi_file = '../datasets/lmd/lmd_full/e/ed05335f8f6c273f506a997137b2805b.mid'\n",
    "print(f\"Choosen {example_midi_file} as example MIDI file\")\n",
    "# with split we remove the folder name and the file extension\n",
    "md5_filenames[example_midi_file.split('/')[-1].split('.')[0]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So - why do we have an additional file which tells us the original filename?\n",
    "This is actually a good trick in order to avoid duplicates in the dataset - we use a so called *hash function* which will return a cryptic string depending on the data that we *hash*.\n",
    "Two files will result in the same hash output - so if we simple rename each file to its hash output, we will automatically delete and detect all duplicates in our dataset.\n",
    "This is a common practice if the data was obtained by scraping multiple sources for data.\n",
    "\n",
    "Let's just take a look if we can verify this claim."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hash for ../datasets/lmd/lmd_full/e/ed05335f8f6c273f506a997137b2805b.mid is ed05335f8f6c273f506a997137b2805b\n"
     ]
    }
   ],
   "source": [
    "import hashlib\n",
    "\n",
    "with open(example_midi_file, 'rb') as f:\n",
    "    file_hash = hashlib.md5()\n",
    "    while chunk := f.read(8192):\n",
    "        file_hash.update(chunk)\n",
    "        \n",
    "print(f\"Hash for {example_midi_file} is {file_hash.hexdigest()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This also gives us a nice additional information - a common MIDI file will probably have lots of different names as it was scraped from multiple files which probably changed the filenames a bit."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading MIDI files in Python\n",
    "\n",
    "Now as we have all file paths available it is a good practice to take a first look at the data.\n",
    "We need to understand how the information we are interested in is accessible. Also it is probably not clean and standardized.\n",
    "\n",
    "As there is no build in functionality in *Python* to work with MIDI data we will need to rely on a library.\n",
    "[There are a numerous libraries](https://pypi.org/search/?q=midi) for working with MIDI data in Python, but we will rely on [pretty_midi](https://github.com/craffel/pretty-midi) to inspect the MIDI files.\n",
    "\n",
    "As we have quite a lot of data to process, it is worth some research on how we can reduce the amount of data before we process the whole dataset and bring it into a good format we can work with.\n",
    "Here, we will concentrate on *percussion tracks*. From the [wikipedia article on General MIDI (GM)](https://en.wikipedia.org/wiki/General_MIDI#Percussive) we can see that they should be on channel 10 - so for now we will simply ignore the other tracks.\n",
    "\n",
    "In a former version of this document we used the library [music21](http://web.mit.edu/music21/) for parsing and loading of our MIDI files but it turns out that [music21 is too slow for our big dataset](https://github.com/capital-G/musikinformatik-sose2021/issues/24#issuecomment-823625482).\n",
    "\n",
    "We also start annotating our dataset using [pandas](https://pandas.pydata.org/), which is a library for data analysis and manipulation. We can use it to keep track of any metadata the files. It will also help us on the analysis of the metadafiles.\n",
    "\n",
    "Additionally we will use [caching](https://en.wikipedia.org/wiki/Cache_(computing)) because the meta-analyis of our MIDI files takes about 2 hours on a recent MacBook.\n",
    "Caching means that we will store the results on the hard drive so once we calculated them we do not need to re-calculate them because we will just load the files from the hard drive.\n",
    "It is a common practice in data science to write a function which either loads the data from a file or generates the file we are using as generating or calculating the data is a step which takes often a long time.\n",
    "\n",
    "We will start by inspecting how we can access MIDI information from within *pretty_midi*."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<pretty_midi.pretty_midi.PrettyMIDI at 0x159b7d760>"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pretty_midi as pm\n",
    "\n",
    "midi_stream = pm.PrettyMIDI(example_midi_file)\n",
    "midi_stream"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After we have now loaded a MIDI file with *pretty_midi* we now take a look at how we can access the data.\n",
    "Therefore the [documentation of pretty_midi](https://craffel.github.io/pretty-midi/) will help us to know how to access the MIDI information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "189.431068"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.get_end_time()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "will give us the time in seconds of the MIDI file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.0000000e+00, 1.2500000e-02, 4.4107100e-01, 8.6964200e-01,\n",
       "       1.2982130e+00, 1.7267840e+00, 2.1553550e+00, 2.5839260e+00,\n",
       "       3.0124970e+00, 3.4410680e+00, 3.8410680e+00, 4.2410680e+00,\n",
       "       4.6410680e+00, 5.0410680e+00, 5.4410680e+00, 5.8410680e+00,\n",
       "       6.2410680e+00, 6.6410680e+00, 7.0410680e+00, 7.4410680e+00,\n",
       "       7.8410680e+00, 8.2410680e+00, 8.6410680e+00, 9.0410680e+00,\n",
       "       9.4410680e+00, 9.8410680e+00, 1.0241068e+01, 1.0641068e+01,\n",
       "       1.1041068e+01, 1.1441068e+01, 1.1841068e+01, 1.2241068e+01,\n",
       "       1.2641068e+01, 1.3041068e+01, 1.3441068e+01, 1.3841068e+01,\n",
       "       1.4241068e+01, 1.4641068e+01, 1.5041068e+01, 1.5441068e+01,\n",
       "       1.5841068e+01, 1.6241068e+01, 1.6641068e+01, 1.7041068e+01,\n",
       "       1.7441068e+01, 1.7841068e+01, 1.8241068e+01, 1.8641068e+01,\n",
       "       1.9041068e+01, 1.9441068e+01])"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.get_beats()[0:50]  # limit to 50 entries for printing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "will give us the position in secounds of all beats (so any action) in a MIDI file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0.      , 0.0125  , 3.441068]),\n",
       " array([120.     , 140.00014, 150.     ]))"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.get_tempo_changes()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "will tell us any changes in the tempo in bpm.\n",
    "Keep in mind that we do not have two tempo changes here but that the function returns two arrays - the first one with the mark in seconds where the tempo gets changed and another one to what bpm it changes - for now we are only interested in the second array.\n",
    "\n",
    "We should also prepare for the case that the MIDI file does not provide any tempo information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[TimeSignature(numerator=4, denominator=4, time=0.0125),\n",
       " TimeSignature(numerator=2, denominator=4, time=21.041068000000003),\n",
       " TimeSignature(numerator=4, denominator=4, time=21.841068000000003),\n",
       " TimeSignature(numerator=2, denominator=4, time=36.241068000000006),\n",
       " TimeSignature(numerator=4, denominator=4, time=37.041068),\n",
       " TimeSignature(numerator=2, denominator=4, time=51.441068),\n",
       " TimeSignature(numerator=4, denominator=4, time=52.241068000000006),\n",
       " TimeSignature(numerator=2, denominator=4, time=66.641068),\n",
       " TimeSignature(numerator=4, denominator=4, time=67.441068),\n",
       " TimeSignature(numerator=2, denominator=4, time=107.441068),\n",
       " TimeSignature(numerator=4, denominator=4, time=108.24106800000001),\n",
       " TimeSignature(numerator=2, denominator=4, time=122.641068),\n",
       " TimeSignature(numerator=4, denominator=4, time=123.44106800000002)]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.time_signature_changes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "tells us the time signatures of the MIDI file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4\n"
     ]
    }
   ],
   "source": [
    "time_signature = midi_stream.time_signature_changes[0]\n",
    "print(f'{time_signature.numerator}/{time_signature.denominator}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As with the tempo we should also prepare for the case that no time signature information is given on the MIDI file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Instrument(program=62, is_drum=False, name=\"\"),\n",
       " Instrument(program=34, is_drum=False, name=\"\"),\n",
       " Instrument(program=25, is_drum=False, name=\"\"),\n",
       " Instrument(program=28, is_drum=False, name=\"\"),\n",
       " Instrument(program=25, is_drum=False, name=\"\"),\n",
       " Instrument(program=30, is_drum=False, name=\"\"),\n",
       " Instrument(program=26, is_drum=False, name=\"\"),\n",
       " Instrument(program=18, is_drum=False, name=\"\"),\n",
       " Instrument(program=0, is_drum=True, name=\"\"),\n",
       " Instrument(program=0, is_drum=True, name=\"\"),\n",
       " Instrument(program=0, is_drum=True, name=\"\"),\n",
       " Instrument(program=0, is_drum=True, name=\"\"),\n",
       " Instrument(program=0, is_drum=True, name=\"\")]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.instruments"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "gives us access to all instruments with its names and if this is a drum track or not."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can also take a look at the notes of each instrument"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Note(start=9.834401, end=9.937735, pitch=62, velocity=109),\n",
       " Note(start=10.037735, end=10.311068, pitch=62, velocity=107),\n",
       " Note(start=10.437735, end=10.847735, pitch=62, velocity=113),\n",
       " Note(start=10.834401, end=11.057735, pitch=60, velocity=109),\n",
       " Note(start=11.034401, end=11.237735, pitch=59, velocity=106),\n",
       " Note(start=11.237735, end=11.594401, pitch=62, velocity=108),\n",
       " Note(start=11.627735, end=11.834401, pitch=60, velocity=101),\n",
       " Note(start=11.827735, end=12.247735, pitch=59, velocity=108),\n",
       " Note(start=12.224401, end=12.544401, pitch=62, velocity=106),\n",
       " Note(start=13.024401, end=13.131068, pitch=64, velocity=102)]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_stream.instruments[0].notes[0:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Performance\n",
    "\n",
    "Now we know how we can use *pretty_midi* to load a MIDI file and access the information of a single MIDI file.\n",
    "But the problem is that a loading of a MIDI file is pretty slow in Python - we will do some measurements now to have a quick way to access the information we are interested in.\n",
    "\n",
    "We can use a built in tool of *Jupyter* which is `%%timeit` which will do multiple runs of a cell and measure the time it takes for each run."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "186 ms ± 6.59 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "\n",
    "pm.PrettyMIDI(example_midi_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Of course this time to parse the MIDI file depends on the complexity of our MIDI file. But now its time to do some maths.\n",
    "For the sake of simplicity we will assume that it takes around 100ms to load a MIDI file.\n",
    "\n",
    "$$ 0.1 \\frac{\\text{sec}}{\\text{file}} * 178000 \\text{ files} = 17800 \\text{ sec} \\approx 296 \\text{ minutes} \\approx 4 \\text{ hours}$$\n",
    "\n",
    "Note that this assumes that we do not process multiple files in paralell, which is possible but also tricky.\n",
    "Also this is only for the loading of the data, we not have acessed yet any data.\n",
    "\n",
    "Maybe there is a file format to which we can store the MIDI files and load them quicker whenever we need them.\n",
    "[*note_seq*](https://github.com/magenta/note-seq) from the [magenata project](https://magenta.tensorflow.org/) provide such a format called *note sequence* which nicely interacts with *pretty_midi*."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import note_seq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14.4 ms ± 381 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "\n",
    "notes = note_seq.midi_to_note_sequence(midi_stream)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So the conversion of this file is fast compared to the parsing of our MIDI file.\n",
    "The problem is that we still need the MIDI file parsed and we have not yet taken a look at the saving and loading of a note sequence.\n",
    "\n",
    "*note_seq* is using a special kind of format for this which is called [protobuf](https://developers.google.com/protocol-buffers/docs/pythontutorial) - a binary file format to store and exchange files developed by Google.\n",
    "The advantage is that this is really fast and efficient as it is a binary format and we can exchange the files to other programming languages - think of it like a binary version of JSON but with type safety included.\n",
    "\n",
    "We start by saving the note sequence into a protobuf and try loading it again from our harddrive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "notes = note_seq.midi_to_note_sequence(midi_stream)\n",
    "\n",
    "with open('note_seq_test.protobuf', 'wb') as f:\n",
    "    f.write(notes.SerializeToString())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3.47 ms ± 97.9 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "\n",
    "with open('note_seq_test.protobuf', 'rb') as f:\n",
    "    notes_loaded = note_seq.NoteSequence.FromString(f.read())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets say it takes about 4 ms to load a protobuf file - now we will do the calculation from before again.\n",
    "\n",
    "$$ 0.004 \\frac{\\text{sec}}{\\text{file}} * 178000 \\text{ files} = 712 \\text{ sec} \\approx 11 \\text{ minutes}$$\n",
    "\n",
    "which is much more acceptable.\n",
    "And we have not yet considered any parallelisation of the code."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "But before we convert everything to a note sequence as a protobuf we should compare the file size of the two file formats."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File size of ../datasets/lmd/lmd_full/e/ed05335f8f6c273f506a997137b2805b.mid: 46.6513671875 kbytes\n",
      "File size of note_seq_test.protobuf: 195.958984375 kbytes\n"
     ]
    }
   ],
   "source": [
    "for f in [example_midi_file, 'note_seq_test.protobuf']:\n",
    "    print(f'File size of {f}: {os.path.getsize(f)/1024} kbytes')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It turns out that we are trading storage in favour of computation power.\n",
    "\n",
    "Before we can decide if we really want to trade, we may need to take a look at the file size of our MIDI dataset.\n",
    "For this we can use a functionality of Jupyter to execute commands in a shell from within Jupyter by appending a `!` in front of the command.\n",
    "In order to get the size of a dictionary we will use the unix tool [du](https://linux.die.net/man/1/du) with the arguments `s` (sum everything within the path) and `h` (convert number of bytes to a human readable format)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5.9G\t../datasets/lmd/lmd_full\n"
     ]
    }
   ],
   "source": [
    "!du -sh ../datasets/lmd/lmd_full"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have around 6 gigabytes of files here - when we convert them to protobuf we will blow them up by a factor of $\\approx 4$ so we will have about 24 GB of protobuf files. This is still acceptable somehow."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Processing all files\n",
    "\n",
    "So after we now know how to load, access and convert the files to a nicer format, we still need to take a look at how we can do this on our 180k files and maybe save some time here by a little bit of programming effort.\n",
    "\n",
    "First we will start by programming a function in which we combine everything we did before and return to us all necessary information as a dictionary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "PROTO_SAVE_PATH = '../datasets/lmd/proto/'\n",
    "\n",
    "# make sure that the folder we want to save to actually exists\n",
    "os.makedirs(PROTO_SAVE_PATH, exist_ok=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import Dict\n",
    "\n",
    "from mido import KeySignatureError\n",
    "\n",
    "def parse_midi_file(midi_file_path: str, proto_save_path: str=PROTO_SAVE_PATH) -> Dict:\n",
    "    r = {\n",
    "        'midi_path': midi_file_path,\n",
    "        'midi_error': True,\n",
    "    }\n",
    "    midi_name = os.path.splitext(midi_file_path.split(os.sep)[-1])[0]\n",
    "    proto_path = os.path.join(PROTO_SAVE_PATH, f'{midi_name}.protobuf')\n",
    "    \n",
    "    r['original_names'] = md5_filenames[midi_name]\n",
    "    \n",
    "    try:\n",
    "        stream = pm.PrettyMIDI(midi_file_path)\n",
    "        r['midi_error'] = False\n",
    "    except (OSError, ValueError, IndexError, KeySignatureError, EOFError, ZeroDivisionError):\n",
    "        return r\n",
    "    \n",
    "    try:\n",
    "        # r['beat_start'] = stream.estimate_beat_start() # omitted b/c adds 200ms to parsing\n",
    "        r['estimate_tempo'] = stream.estimate_tempo()\n",
    "        r['tempi_sec'], r['tempi'] = stream.get_tempo_changes() \n",
    "        r['end_time'] = stream.get_end_time()\n",
    "        r['drums'] = any([i.is_drum for i in stream.instruments])\n",
    "        r['resolution'] = stream.resolution\n",
    "        r['instrument_names'] = [i.name.strip() for i in stream.instruments]\n",
    "        r['num_time_signature_changes'] = len(stream.time_signature_changes)\n",
    "    except ValueError as e:\n",
    "        # ValueError: Can't estimate beat start when there are no notes.\n",
    "        # ValueError: Can't provide a global tempo estimate when there are fewer than two notes.\n",
    "        print(f\"Could not parse MIDI file {midi_file_path}: {e}\")\n",
    "    \n",
    "    notes = note_seq.midi_to_note_sequence(stream)\n",
    "    with open(proto_path, 'wb') as f:\n",
    "        f.write(notes.SerializeToString())\n",
    "    r['proto_path'] = proto_path\n",
    "    \n",
    "    return r"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we want to try out the function on a single MIDI file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'midi_path': '../datasets/lmd/lmd_full/e/ed05335f8f6c273f506a997137b2805b.mid',\n",
       " 'midi_error': False,\n",
       " 'original_names': ['QUALITY MIDI/EClapton-Promises.mid'],\n",
       " 'estimate_tempo': 159.6553471870248,\n",
       " 'tempi_sec': array([0.      , 0.0125  , 3.441068]),\n",
       " 'tempi': array([120.     , 140.00014, 150.     ]),\n",
       " 'end_time': 189.431068,\n",
       " 'drums': True,\n",
       " 'resolution': 120,\n",
       " 'instrument_names': ['', '', '', '', '', '', '', '', '', '', '', '', ''],\n",
       " 'num_time_signature_changes': 13,\n",
       " 'proto_path': '../datasets/lmd/proto/ed05335f8f6c273f506a997137b2805b.protobuf'}"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "parse_midi_file(example_midi_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So this works as expected - time to time it as well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "233 ms ± 10.1 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "\n",
    "parse_midi_file(example_midi_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It seems that accessing the MIDI data via *pretty_midi* does not come for free.\n",
    "Repeating the calculation from before:\n",
    "\n",
    "$$ 0.2 \\frac{\\text{sec}}{\\text{file}} * 178000 \\text{ files} = 35600 \\text{ sec} \\approx 593 \\text{ minutes} \\approx 10 \\text{ hours}$$\n",
    "\n",
    "This is not unheard of but we can reduce the time by parallelizing it for which we need to do some tricks into which we will not go detail.\n",
    "On a recent i9 processor the next step will take around 90 minutes if executed for the first time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Use cached parquet file midi_files_full.parquet\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>81923</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f4b72836981feedb9e1...</td>\n",
       "      <td>False</td>\n",
       "      <td>175.232953</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[82.00003553334874]</td>\n",
       "      <td>128.780432</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[harmnica, accordn2, drum, acoubass, pianogrd,...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/f4b72836981feedb9e1046f1...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>171833</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5f3eaf75cbc1b2554df...</td>\n",
       "      <td>False</td>\n",
       "      <td>213.043443</td>\n",
       "      <td>[0.0, 297.04290665, 297.1153704, 297.188899816...</td>\n",
       "      <td>[69.99998833333528, 69.00001725000433, 67.9999...</td>\n",
       "      <td>302.411045</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[YO TE AMO;CHAYANE, YO TE AMO;CHAYANE, YO TE A...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5f3eaf75cbc1b2554dfa60bd...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161501</th>\n",
       "      <td>../datasets/lmd/lmd_full/2/2f02103cfb1764e661d...</td>\n",
       "      <td>False</td>\n",
       "      <td>170.691458</td>\n",
       "      <td>[0.0, 0.004166666666666667, 2.0015625]</td>\n",
       "      <td>[75.0, 120.0, 75.0]</td>\n",
       "      <td>174.718229</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[, , , , , , , , , , , , , , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/2f02103cfb1764e661d54151...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39942</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/609b19dde84d2374b7d...</td>\n",
       "      <td>False</td>\n",
       "      <td>237.529957</td>\n",
       "      <td>[0.0, 0.0026041666666666665]</td>\n",
       "      <td>[120.0, 110.00091667430561]</td>\n",
       "      <td>254.776049</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[S-File, S-File, S-File, S-File, S-File, S-Fil...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/609b19dde84d2374b7d19124...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>100593</th>\n",
       "      <td>../datasets/lmd/lmd_full/d/de4dde0f1e960e41360...</td>\n",
       "      <td>False</td>\n",
       "      <td>170.384180</td>\n",
       "      <td>[0.0, 3.0, 183.33574275, 183.66386775]</td>\n",
       "      <td>[120.0, 67.00002903334591, 60.0, 49.0000318500...</td>\n",
       "      <td>190.137078</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[, , , , , ]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/de4dde0f1e960e413602c493...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44272</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/6f74186f4c163863630...</td>\n",
       "      <td>False</td>\n",
       "      <td>242.119770</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[122.00006913337249]</td>\n",
       "      <td>228.161756</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[drummix, bass, ac piano, Vocal, back voc, ac ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/6f74186f4c163863630e8baa...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15131</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/0d8e9c4eb761a3e632b...</td>\n",
       "      <td>False</td>\n",
       "      <td>245.724188</td>\n",
       "      <td>[0.0, 3.9344240000000004, 14.843497999999999, ...</td>\n",
       "      <td>[122.00006913337249, 121.00018755029072, 123.0...</td>\n",
       "      <td>251.814031</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[TAKE ME HOME                    Bextor/Aller/...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/0d8e9c4eb761a3e632ba4f10...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34735</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/6d3149b41d088c63627...</td>\n",
       "      <td>False</td>\n",
       "      <td>84.000025</td>\n",
       "      <td>[0.0, 89.2856875, 89.88351320833333, 90.410473...</td>\n",
       "      <td>[42.00001260000378, 46.00002913335179, 51.0000...</td>\n",
       "      <td>340.282369</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Voice (Alto), Voice (Alto), Voice (Alto), Pia...</td>\n",
       "      <td>6.0</td>\n",
       "      <td>../datasets/lmd/proto/6d3149b41d088c6362792faf...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45905</th>\n",
       "      <td>../datasets/lmd/lmd_full/1/1a931072e21340f256b...</td>\n",
       "      <td>False</td>\n",
       "      <td>207.446994</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>280.114583</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[A.PIANO 1, FINGERDBAS, A.PIANO 1, PAN FLUTE, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/1a931072e21340f256b96104...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12844</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/0a7c38887ba4882921e...</td>\n",
       "      <td>False</td>\n",
       "      <td>224.582071</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[127.96996971377385]</td>\n",
       "      <td>193.497545</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[CANDOMBE, PARA, GARDEL, http://fberni.tripod....</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/0a7c38887ba4882921ee2c82...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>114123</th>\n",
       "      <td>../datasets/lmd/lmd_full/4/455fcb4f4c5ad4d10da...</td>\n",
       "      <td>False</td>\n",
       "      <td>153.020774</td>\n",
       "      <td>[0.0, 7.15475625, 7.3302190000000005, 7.506771...</td>\n",
       "      <td>[85.95680670463092, 85.48823040787859, 84.9608...</td>\n",
       "      <td>100.750126</td>\n",
       "      <td>False</td>\n",
       "      <td>1024.0</td>\n",
       "      <td>[]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/455fcb4f4c5ad4d10da384b6...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14339</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/029c4baa3089eca3817...</td>\n",
       "      <td>False</td>\n",
       "      <td>199.999987</td>\n",
       "      <td>[0.0, 6.461539, 7.961539, 9.790110733333334, 9...</td>\n",
       "      <td>[64.99999458333379, 40.0, 69.99998833333528, 6...</td>\n",
       "      <td>330.377789</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[, Bass, Bass, Strings, Melodia, Slowstrings, ...</td>\n",
       "      <td>5.0</td>\n",
       "      <td>../datasets/lmd/proto/029c4baa3089eca381772e2b...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>131099</th>\n",
       "      <td>../datasets/lmd/lmd_full/3/353116ea8d2eb4041c9...</td>\n",
       "      <td>False</td>\n",
       "      <td>260.000260</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[130.00013000013]</td>\n",
       "      <td>51.692256</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[MIDI Ch. 1, MIDI Ch. 2, MIDI Drums]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/353116ea8d2eb4041c9f4622...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>124606</th>\n",
       "      <td>../datasets/lmd/lmd_full/3/3574baf9f9cdfa4c3e1...</td>\n",
       "      <td>False</td>\n",
       "      <td>207.262310</td>\n",
       "      <td>[0.0, 0.16129033333333334, 0.9425403333333333,...</td>\n",
       "      <td>[30.999997933333468, 32.0, 32.99999670000033, ...</td>\n",
       "      <td>273.626126</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Rhodes Piano, Synth Bass H, Synth Bass L, Aco...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/3574baf9f9cdfa4c3e1da456...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136078</th>\n",
       "      <td>../datasets/lmd/lmd_full/e/eb59c9f6854d897a03a...</td>\n",
       "      <td>False</td>\n",
       "      <td>193.775528</td>\n",
       "      <td>[0.0, 3.5416666666666665]</td>\n",
       "      <td>[120.0, 109.99990833340973]</td>\n",
       "      <td>217.860027</td>\n",
       "      <td>True</td>\n",
       "      <td>240.0</td>\n",
       "      <td>[Bass / Acoustic bass, Piano / Rhodes, Melody ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/eb59c9f6854d897a03a3ed78...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "81923   ../datasets/lmd/lmd_full/f/f4b72836981feedb9e1...       False   \n",
       "171833  ../datasets/lmd/lmd_full/5/5f3eaf75cbc1b2554df...       False   \n",
       "161501  ../datasets/lmd/lmd_full/2/2f02103cfb1764e661d...       False   \n",
       "39942   ../datasets/lmd/lmd_full/6/609b19dde84d2374b7d...       False   \n",
       "100593  ../datasets/lmd/lmd_full/d/de4dde0f1e960e41360...       False   \n",
       "44272   ../datasets/lmd/lmd_full/6/6f74186f4c163863630...       False   \n",
       "15131   ../datasets/lmd/lmd_full/0/0d8e9c4eb761a3e632b...       False   \n",
       "34735   ../datasets/lmd/lmd_full/6/6d3149b41d088c63627...       False   \n",
       "45905   ../datasets/lmd/lmd_full/1/1a931072e21340f256b...       False   \n",
       "12844   ../datasets/lmd/lmd_full/0/0a7c38887ba4882921e...       False   \n",
       "114123  ../datasets/lmd/lmd_full/4/455fcb4f4c5ad4d10da...       False   \n",
       "14339   ../datasets/lmd/lmd_full/0/029c4baa3089eca3817...       False   \n",
       "131099  ../datasets/lmd/lmd_full/3/353116ea8d2eb4041c9...       False   \n",
       "124606  ../datasets/lmd/lmd_full/3/3574baf9f9cdfa4c3e1...       False   \n",
       "136078  ../datasets/lmd/lmd_full/e/eb59c9f6854d897a03a...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "81923       175.232953                                              [0.0]   \n",
       "171833      213.043443  [0.0, 297.04290665, 297.1153704, 297.188899816...   \n",
       "161501      170.691458             [0.0, 0.004166666666666667, 2.0015625]   \n",
       "39942       237.529957                       [0.0, 0.0026041666666666665]   \n",
       "100593      170.384180             [0.0, 3.0, 183.33574275, 183.66386775]   \n",
       "44272       242.119770                                              [0.0]   \n",
       "15131       245.724188  [0.0, 3.9344240000000004, 14.843497999999999, ...   \n",
       "34735        84.000025  [0.0, 89.2856875, 89.88351320833333, 90.410473...   \n",
       "45905       207.446994                                              [0.0]   \n",
       "12844       224.582071                                              [0.0]   \n",
       "114123      153.020774  [0.0, 7.15475625, 7.3302190000000005, 7.506771...   \n",
       "14339       199.999987  [0.0, 6.461539, 7.961539, 9.790110733333334, 9...   \n",
       "131099      260.000260                                              [0.0]   \n",
       "124606      207.262310  [0.0, 0.16129033333333334, 0.9425403333333333,...   \n",
       "136078      193.775528                          [0.0, 3.5416666666666665]   \n",
       "\n",
       "                                                    tempi    end_time  drums  \\\n",
       "81923                                 [82.00003553334874]  128.780432   True   \n",
       "171833  [69.99998833333528, 69.00001725000433, 67.9999...  302.411045   True   \n",
       "161501                                [75.0, 120.0, 75.0]  174.718229   True   \n",
       "39942                         [120.0, 110.00091667430561]  254.776049   True   \n",
       "100593  [120.0, 67.00002903334591, 60.0, 49.0000318500...  190.137078  False   \n",
       "44272                                [122.00006913337249]  228.161756   True   \n",
       "15131   [122.00006913337249, 121.00018755029072, 123.0...  251.814031   True   \n",
       "34735   [42.00001260000378, 46.00002913335179, 51.0000...  340.282369  False   \n",
       "45905                                             [120.0]  280.114583   True   \n",
       "12844                                [127.96996971377385]  193.497545   True   \n",
       "114123  [85.95680670463092, 85.48823040787859, 84.9608...  100.750126  False   \n",
       "14339   [64.99999458333379, 40.0, 69.99998833333528, 6...  330.377789   True   \n",
       "131099                                  [130.00013000013]   51.692256   True   \n",
       "124606  [30.999997933333468, 32.0, 32.99999670000033, ...  273.626126   True   \n",
       "136078                        [120.0, 109.99990833340973]  217.860027   True   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "81923        384.0  [harmnica, accordn2, drum, acoubass, pianogrd,...   \n",
       "171833       120.0  [YO TE AMO;CHAYANE, YO TE AMO;CHAYANE, YO TE A...   \n",
       "161501       384.0                   [, , , , , , , , , , , , , , , ]   \n",
       "39942        384.0  [S-File, S-File, S-File, S-File, S-File, S-Fil...   \n",
       "100593       192.0                                       [, , , , , ]   \n",
       "44272        480.0  [drummix, bass, ac piano, Vocal, back voc, ac ...   \n",
       "15131         96.0  [TAKE ME HOME                    Bextor/Aller/...   \n",
       "34735         96.0  [Voice (Alto), Voice (Alto), Voice (Alto), Pia...   \n",
       "45905        384.0  [A.PIANO 1, FINGERDBAS, A.PIANO 1, PAN FLUTE, ...   \n",
       "12844        480.0  [CANDOMBE, PARA, GARDEL, http://fberni.tripod....   \n",
       "114123      1024.0                                                 []   \n",
       "14339        120.0  [, Bass, Bass, Strings, Melodia, Slowstrings, ...   \n",
       "131099        96.0               [MIDI Ch. 1, MIDI Ch. 2, MIDI Drums]   \n",
       "124606       120.0  [Rhodes Piano, Synth Bass H, Synth Bass L, Aco...   \n",
       "136078       240.0  [Bass / Acoustic bass, Piano / Rhodes, Melody ...   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "81923                          1.0   \n",
       "171833                         1.0   \n",
       "161501                         1.0   \n",
       "39942                          1.0   \n",
       "100593                         0.0   \n",
       "44272                          1.0   \n",
       "15131                          1.0   \n",
       "34735                          6.0   \n",
       "45905                          1.0   \n",
       "12844                          1.0   \n",
       "114123                         1.0   \n",
       "14339                          5.0   \n",
       "131099                         0.0   \n",
       "124606                         1.0   \n",
       "136078                         1.0   \n",
       "\n",
       "                                               proto_path  \n",
       "81923   ../datasets/lmd/proto/f4b72836981feedb9e1046f1...  \n",
       "171833  ../datasets/lmd/proto/5f3eaf75cbc1b2554dfa60bd...  \n",
       "161501  ../datasets/lmd/proto/2f02103cfb1764e661d54151...  \n",
       "39942   ../datasets/lmd/proto/609b19dde84d2374b7d19124...  \n",
       "100593  ../datasets/lmd/proto/de4dde0f1e960e413602c493...  \n",
       "44272   ../datasets/lmd/proto/6f74186f4c163863630e8baa...  \n",
       "15131   ../datasets/lmd/proto/0d8e9c4eb761a3e632ba4f10...  \n",
       "34735   ../datasets/lmd/proto/6d3149b41d088c6362792faf...  \n",
       "45905   ../datasets/lmd/proto/1a931072e21340f256b96104...  \n",
       "12844   ../datasets/lmd/proto/0a7c38887ba4882921ee2c82...  \n",
       "114123  ../datasets/lmd/proto/455fcb4f4c5ad4d10da384b6...  \n",
       "14339   ../datasets/lmd/proto/029c4baa3089eca381772e2b...  \n",
       "131099  ../datasets/lmd/proto/353116ea8d2eb4041c9f4622...  \n",
       "124606  ../datasets/lmd/proto/3574baf9f9cdfa4c3e1da456...  \n",
       "136078  ../datasets/lmd/proto/eb59c9f6854d897a03a3ed78...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import math\n",
    "import multiprocessing\n",
    "from typing import Optional\n",
    "\n",
    "from helpers.helpers import *\n",
    "\n",
    "def load_midi_files(parquet_path: str = 'midi_files_{}.parquet', limit: Optional[int] = None) -> pd.DataFrame:\n",
    "    parquet_path = parquet_path.format(limit if limit else 'full')\n",
    "    \n",
    "    if os.path.isfile(parquet_path):\n",
    "        print(f'Use cached parquet file {parquet_path}')\n",
    "        return pd.read_parquet(parquet_path)\n",
    "    \n",
    "    midi_files = glob.glob('../datasets/lmd/lmd_full/*/*.mid')\n",
    "    if limit:\n",
    "        print(f'Limit to {limit} files')\n",
    "        midi_files = np.random.choice(midi_files, limit)\n",
    "    print(f'Parse {len(midi_files)} midi files')\n",
    "    \n",
    "    cpu_count = math.ceil(multiprocessing.cpu_count()*3/4)\n",
    "    with multiprocessing.Pool(cpu_count, maxtasksperchild=5) as p:\n",
    "        midi_meta = p.map(parse_midi_file_async, midi_files, chunksize=255)\n",
    "    print(f'Finished parsing midi files')\n",
    "    \n",
    "    midi_meta = pd.DataFrame(midi_meta)\n",
    "    midi_meta.to_parquet(parquet_path)\n",
    "    \n",
    "    return midi_meta\n",
    "\n",
    "# if you want to work with a random subset set the limit argument to num of files as a limit\n",
    "midi_df = load_midi_files(limit=None)\n",
    "\n",
    "midi_df.sample(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now also add the original file names from the md5 json file to our dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>31900</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/745f66980de9ca27fe0...</td>\n",
       "      <td>False</td>\n",
       "      <td>169.981663</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[151.00037750094376]</td>\n",
       "      <td>141.733917</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Acc Bass, Piano, Melody, Sax, Solo, E.Guitar,...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/745f66980de9ca27fe055cc6...</td>\n",
       "      <td>[BonJovi/Runaway3.mid, MikeDoyle/MyLittleRunAw...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33585</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/68b145d7ffdf4c43e1f...</td>\n",
       "      <td>False</td>\n",
       "      <td>204.472843</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[99.99999999999999]</td>\n",
       "      <td>163.656250</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Remelexo, Remelexo, Remelexo, Remelexo, Remel...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>../datasets/lmd/proto/68b145d7ffdf4c43e1f5f59c...</td>\n",
       "      <td>[C/Cesar E Paulinho - Remelexo.mid, Midis Româ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67436</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/a373ed7cd11fb80fac3...</td>\n",
       "      <td>False</td>\n",
       "      <td>159.432272</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[159.0002067002687]</td>\n",
       "      <td>154.150743</td>\n",
       "      <td>True</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[Crash Cymbal, Ride Cymbal, Open HiHat, Closed...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/a373ed7cd11fb80fac3dca06...</td>\n",
       "      <td>[t/trust_me.mid, T/trust_me.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>142153</th>\n",
       "      <td>../datasets/lmd/lmd_full/e/ec2fff2c295943aedf4...</td>\n",
       "      <td>False</td>\n",
       "      <td>194.483086</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[88.70005632453577]</td>\n",
       "      <td>259.977287</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[You Can Leave Your Hat On, You Can Leave Your...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/ec2fff2c295943aedf4797c3...</td>\n",
       "      <td>[Y/You Can Leave Your Hat on Pm L.mid, Y/You C...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>127493</th>\n",
       "      <td>../datasets/lmd/lmd_full/3/3fae950c18f739286ec...</td>\n",
       "      <td>False</td>\n",
       "      <td>160.190137</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[160.0]</td>\n",
       "      <td>100.425000</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Sax, Sax, French Horn, Piano, Bass, Drums]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/3fae950c18f739286ec42107...</td>\n",
       "      <td>[2009 MIDI/barbra_Ann2-Regents-F160.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138301</th>\n",
       "      <td>../datasets/lmd/lmd_full/e/ec16ec22a846e4d4d91...</td>\n",
       "      <td>False</td>\n",
       "      <td>164.382517</td>\n",
       "      <td>[0.0, 186.66648]</td>\n",
       "      <td>[162.000162000162, 120.0]</td>\n",
       "      <td>186.666480</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Vocal, ElecGtr 1, ElecGtr 2, PickedBass, Orga...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/ec16ec22a846e4d4d91959cc...</td>\n",
       "      <td>[c/chaingm.mid, C/chaingm.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>83006</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/fd8a6fcf912a7ddc602...</td>\n",
       "      <td>False</td>\n",
       "      <td>160.171123</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[80.0]</td>\n",
       "      <td>274.525000</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Corazon Partio, Corazon Partio, Corazon Parti...</td>\n",
       "      <td>7.0</td>\n",
       "      <td>../datasets/lmd/proto/fd8a6fcf912a7ddc60242146...</td>\n",
       "      <td>[C/Corazon Partio Pm L.mid, Midis Latinas/Cora...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13575</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/0dff276c070ccc16e2b...</td>\n",
       "      <td>False</td>\n",
       "      <td>203.014003</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[190.0002850004275]</td>\n",
       "      <td>48.976900</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Bop on the Rocks, Bop on the Rocks, Bop on th...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>../datasets/lmd/proto/0dff276c070ccc16e2bd6fa1...</td>\n",
       "      <td>[Jazz/Bop on.mid, b/bop-on.mid, Jazz/BOP_ON.MI...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>134536</th>\n",
       "      <td>../datasets/lmd/lmd_full/e/ed793dadf58b913ea24...</td>\n",
       "      <td>False</td>\n",
       "      <td>192.000192</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[140.00014000014]</td>\n",
       "      <td>6.861600</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[z3ta+ (MIDI)]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/ed793dadf58b913ea2460ad1...</td>\n",
       "      <td>[M/Machinehead - Headwave (Zatox Mix).mid, M/M...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>116286</th>\n",
       "      <td>../datasets/lmd/lmd_full/4/4c4d48e046e6f5a2fc2...</td>\n",
       "      <td>False</td>\n",
       "      <td>255.499940</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[145.9999659333413]</td>\n",
       "      <td>6.575344</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[MIDI out]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/4c4d48e046e6f5a2fc2f89cb...</td>\n",
       "      <td>[J/jan_johnston__flesh__tilt_remix__bamford.mi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>126553</th>\n",
       "      <td>../datasets/lmd/lmd_full/3/312b4e4b6f9bbd3abe3...</td>\n",
       "      <td>False</td>\n",
       "      <td>192.617363</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[96.0]</td>\n",
       "      <td>87.500000</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[ACOU BASS, A.PIANO 1, A.PIANO 1, JAZZ GTR, SL...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>../datasets/lmd/proto/312b4e4b6f9bbd3abe3af2bb...</td>\n",
       "      <td>[Diversen/Strike-Up-The-Band.mid, DIVERSEN/STR...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>150997</th>\n",
       "      <td>../datasets/lmd/lmd_full/b/bf3f61d4bf0c7a99c7c...</td>\n",
       "      <td>False</td>\n",
       "      <td>196.007331</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[98.0036653370836]</td>\n",
       "      <td>364.744011</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[SHOUT, SHOUT, SHOUT, SHOUT, SHOUT, SHOUT, SHO...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/bf3f61d4bf0c7a99c7c5dded...</td>\n",
       "      <td>[080/Shout.mid, 080/Shout.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66108</th>\n",
       "      <td>../datasets/lmd/lmd_full/8/80fd25fb1ff70896ed8...</td>\n",
       "      <td>False</td>\n",
       "      <td>180.717774</td>\n",
       "      <td>[0.0, 0.008333333333333333, 11.500865666666666...</td>\n",
       "      <td>[60.0, 67.00002903334591, 65.99999340000066, 6...</td>\n",
       "      <td>109.561609</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[English (open lyrics window), French (open ly...</td>\n",
       "      <td>16.0</td>\n",
       "      <td>../datasets/lmd/proto/80fd25fb1ff70896ed8b1115...</td>\n",
       "      <td>[R/Ravel Maurice Chanson Des Cueilleuses De Le...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5427</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/91100d1888462c59031...</td>\n",
       "      <td>False</td>\n",
       "      <td>246.623850</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>32.005208</td>\n",
       "      <td>False</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Nylon-Str. Gt., Bass, *Piano, Warm Pad, Strings]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/91100d1888462c5903132be8...</td>\n",
       "      <td>[COREL COLLECTION/MOOD_06.MID]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78360</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f577f833d678672ef3b...</td>\n",
       "      <td>False</td>\n",
       "      <td>194.594595</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>200.916667</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[Snare / Tom/cymbal, Snare / Tom/cymbal, Snare...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>../datasets/lmd/proto/f577f833d678672ef3b2787c...</td>\n",
       "      <td>[D/Double Vision.mid, D/Double Vision.mid, D/D...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "31900   ../datasets/lmd/lmd_full/7/745f66980de9ca27fe0...       False   \n",
       "33585   ../datasets/lmd/lmd_full/6/68b145d7ffdf4c43e1f...       False   \n",
       "67436   ../datasets/lmd/lmd_full/a/a373ed7cd11fb80fac3...       False   \n",
       "142153  ../datasets/lmd/lmd_full/e/ec2fff2c295943aedf4...       False   \n",
       "127493  ../datasets/lmd/lmd_full/3/3fae950c18f739286ec...       False   \n",
       "138301  ../datasets/lmd/lmd_full/e/ec16ec22a846e4d4d91...       False   \n",
       "83006   ../datasets/lmd/lmd_full/f/fd8a6fcf912a7ddc602...       False   \n",
       "13575   ../datasets/lmd/lmd_full/0/0dff276c070ccc16e2b...       False   \n",
       "134536  ../datasets/lmd/lmd_full/e/ed793dadf58b913ea24...       False   \n",
       "116286  ../datasets/lmd/lmd_full/4/4c4d48e046e6f5a2fc2...       False   \n",
       "126553  ../datasets/lmd/lmd_full/3/312b4e4b6f9bbd3abe3...       False   \n",
       "150997  ../datasets/lmd/lmd_full/b/bf3f61d4bf0c7a99c7c...       False   \n",
       "66108   ../datasets/lmd/lmd_full/8/80fd25fb1ff70896ed8...       False   \n",
       "5427    ../datasets/lmd/lmd_full/9/91100d1888462c59031...       False   \n",
       "78360   ../datasets/lmd/lmd_full/f/f577f833d678672ef3b...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "31900       169.981663                                              [0.0]   \n",
       "33585       204.472843                                              [0.0]   \n",
       "67436       159.432272                                              [0.0]   \n",
       "142153      194.483086                                              [0.0]   \n",
       "127493      160.190137                                              [0.0]   \n",
       "138301      164.382517                                   [0.0, 186.66648]   \n",
       "83006       160.171123                                              [0.0]   \n",
       "13575       203.014003                                              [0.0]   \n",
       "134536      192.000192                                              [0.0]   \n",
       "116286      255.499940                                              [0.0]   \n",
       "126553      192.617363                                              [0.0]   \n",
       "150997      196.007331                                              [0.0]   \n",
       "66108       180.717774  [0.0, 0.008333333333333333, 11.500865666666666...   \n",
       "5427        246.623850                                              [0.0]   \n",
       "78360       194.594595                                              [0.0]   \n",
       "\n",
       "                                                    tempi    end_time  drums  \\\n",
       "31900                                [151.00037750094376]  141.733917   True   \n",
       "33585                                 [99.99999999999999]  163.656250   True   \n",
       "67436                                 [159.0002067002687]  154.150743   True   \n",
       "142153                                [88.70005632453577]  259.977287   True   \n",
       "127493                                            [160.0]  100.425000   True   \n",
       "138301                          [162.000162000162, 120.0]  186.666480   True   \n",
       "83006                                              [80.0]  274.525000   True   \n",
       "13575                                 [190.0002850004275]   48.976900   True   \n",
       "134536                                  [140.00014000014]    6.861600  False   \n",
       "116286                                [145.9999659333413]    6.575344  False   \n",
       "126553                                             [96.0]   87.500000   True   \n",
       "150997                                 [98.0036653370836]  364.744011   True   \n",
       "66108   [60.0, 67.00002903334591, 65.99999340000066, 6...  109.561609  False   \n",
       "5427                                              [120.0]   32.005208  False   \n",
       "78360                                             [120.0]  200.916667   True   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "31900         96.0  [Acc Bass, Piano, Melody, Sax, Solo, E.Guitar,...   \n",
       "33585         96.0  [Remelexo, Remelexo, Remelexo, Remelexo, Remel...   \n",
       "67436        192.0  [Crash Cymbal, Ride Cymbal, Open HiHat, Closed...   \n",
       "142153       120.0  [You Can Leave Your Hat On, You Can Leave Your...   \n",
       "127493       480.0        [Sax, Sax, French Horn, Piano, Bass, Drums]   \n",
       "138301       120.0  [Vocal, ElecGtr 1, ElecGtr 2, PickedBass, Orga...   \n",
       "83006        120.0  [Corazon Partio, Corazon Partio, Corazon Parti...   \n",
       "13575         96.0  [Bop on the Rocks, Bop on the Rocks, Bop on th...   \n",
       "134536        96.0                                     [z3ta+ (MIDI)]   \n",
       "116286        96.0                                         [MIDI out]   \n",
       "126553       384.0  [ACOU BASS, A.PIANO 1, A.PIANO 1, JAZZ GTR, SL...   \n",
       "150997       384.0  [SHOUT, SHOUT, SHOUT, SHOUT, SHOUT, SHOUT, SHO...   \n",
       "66108        120.0  [English (open lyrics window), French (open ly...   \n",
       "5427         480.0  [Nylon-Str. Gt., Bass, *Piano, Warm Pad, Strings]   \n",
       "78360        384.0  [Snare / Tom/cymbal, Snare / Tom/cymbal, Snare...   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "31900                          1.0   \n",
       "33585                          2.0   \n",
       "67436                          1.0   \n",
       "142153                         1.0   \n",
       "127493                         1.0   \n",
       "138301                         1.0   \n",
       "83006                          7.0   \n",
       "13575                          3.0   \n",
       "134536                         0.0   \n",
       "116286                         0.0   \n",
       "126553                         2.0   \n",
       "150997                         1.0   \n",
       "66108                         16.0   \n",
       "5427                           1.0   \n",
       "78360                          3.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "31900   ../datasets/lmd/proto/745f66980de9ca27fe055cc6...   \n",
       "33585   ../datasets/lmd/proto/68b145d7ffdf4c43e1f5f59c...   \n",
       "67436   ../datasets/lmd/proto/a373ed7cd11fb80fac3dca06...   \n",
       "142153  ../datasets/lmd/proto/ec2fff2c295943aedf4797c3...   \n",
       "127493  ../datasets/lmd/proto/3fae950c18f739286ec42107...   \n",
       "138301  ../datasets/lmd/proto/ec16ec22a846e4d4d91959cc...   \n",
       "83006   ../datasets/lmd/proto/fd8a6fcf912a7ddc60242146...   \n",
       "13575   ../datasets/lmd/proto/0dff276c070ccc16e2bd6fa1...   \n",
       "134536  ../datasets/lmd/proto/ed793dadf58b913ea2460ad1...   \n",
       "116286  ../datasets/lmd/proto/4c4d48e046e6f5a2fc2f89cb...   \n",
       "126553  ../datasets/lmd/proto/312b4e4b6f9bbd3abe3af2bb...   \n",
       "150997  ../datasets/lmd/proto/bf3f61d4bf0c7a99c7c5dded...   \n",
       "66108   ../datasets/lmd/proto/80fd25fb1ff70896ed8b1115...   \n",
       "5427    ../datasets/lmd/proto/91100d1888462c5903132be8...   \n",
       "78360   ../datasets/lmd/proto/f577f833d678672ef3b2787c...   \n",
       "\n",
       "                                           original_files  \n",
       "31900   [BonJovi/Runaway3.mid, MikeDoyle/MyLittleRunAw...  \n",
       "33585   [C/Cesar E Paulinho - Remelexo.mid, Midis Româ...  \n",
       "67436                    [t/trust_me.mid, T/trust_me.mid]  \n",
       "142153  [Y/You Can Leave Your Hat on Pm L.mid, Y/You C...  \n",
       "127493           [2009 MIDI/barbra_Ann2-Regents-F160.mid]  \n",
       "138301                     [c/chaingm.mid, C/chaingm.mid]  \n",
       "83006   [C/Corazon Partio Pm L.mid, Midis Latinas/Cora...  \n",
       "13575   [Jazz/Bop on.mid, b/bop-on.mid, Jazz/BOP_ON.MI...  \n",
       "134536  [M/Machinehead - Headwave (Zatox Mix).mid, M/M...  \n",
       "116286  [J/jan_johnston__flesh__tilt_remix__bamford.mi...  \n",
       "126553  [Diversen/Strike-Up-The-Band.mid, DIVERSEN/STR...  \n",
       "150997                     [080/Shout.mid, 080/Shout.mid]  \n",
       "66108   [R/Ravel Maurice Chanson Des Cueilleuses De Le...  \n",
       "5427                       [COREL COLLECTION/MOOD_06.MID]  \n",
       "78360   [D/Double Vision.mid, D/Double Vision.mid, D/D...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "original_files = []\n",
    "for _, midi_path in midi_df['midi_path'].iteritems():\n",
    "    midi_name = midi_path.split(os.sep)[-1].split('.')[0]\n",
    "    original_files.append(md5_filenames[midi_name])\n",
    "\n",
    "midi_df['original_files'] = original_files\n",
    "\n",
    "midi_df.sample(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analysis of metadata\n",
    "\n",
    "Let's examine the extracted metadata of those files.\n",
    "Before we start plotting it is always a good idea to take a look a the dataframe and its description."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>160813</th>\n",
       "      <td>../datasets/lmd/lmd_full/2/2b03576c2eac6b20c55...</td>\n",
       "      <td>False</td>\n",
       "      <td>220.504202</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[99.99999999999999]</td>\n",
       "      <td>200.200000</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[Lead, Bass, Guitar, Rhythm guitar, Voice oohs...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/2b03576c2eac6b20c5537514...</td>\n",
       "      <td>[M/Monday Monday.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>83189</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/fc7280752baa136165c...</td>\n",
       "      <td>False</td>\n",
       "      <td>149.380190</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[160.01024065540196]</td>\n",
       "      <td>150.326316</td>\n",
       "      <td>True</td>\n",
       "      <td>48.0</td>\n",
       "      <td>[SYNTH, SYNTH, SYNTH, SYNTH, SYNTH, SYNTH, SYN...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/fc7280752baa136165c911e0...</td>\n",
       "      <td>[e/entreaty.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96858</th>\n",
       "      <td>../datasets/lmd/lmd_full/c/cb226a46053507ae8a4...</td>\n",
       "      <td>False</td>\n",
       "      <td>170.000085</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[170.0000850000425]</td>\n",
       "      <td>11.294112</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Neo Cortex - Elements (Styles &amp; Breeze Remix)]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/cb226a46053507ae8a47a16c...</td>\n",
       "      <td>[N/Neo Cortex - Elements (Breeze &amp; Styles Remi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34038</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/6d12ce4d832f9c39ca6...</td>\n",
       "      <td>False</td>\n",
       "      <td>231.914925</td>\n",
       "      <td>[0.0, 2.368422, 3.493422, 11.1907935, 11.19735...</td>\n",
       "      <td>[151.99993920002433, 80.0, 151.99993920002433,...</td>\n",
       "      <td>503.284151</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Primo RH, Primo LH G, Primo LH F, Secondo RH ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/6d12ce4d832f9c39ca683d53...</td>\n",
       "      <td>[Mendelsonn/Allegro Brillant for Piano 4-hands...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>171000</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/586c46e8f1c65beff59...</td>\n",
       "      <td>False</td>\n",
       "      <td>236.159582</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[130.0023833770286]</td>\n",
       "      <td>91.378132</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[ACCORDION, A.PIANO 1, FINGERDBAS, CRYSTAL, DR...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/586c46e8f1c65beff59afe0b...</td>\n",
       "      <td>[Byrd, Charlie/The-Girl-From-Ipanema.mid, BYRD...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103720</th>\n",
       "      <td>../datasets/lmd/lmd_full/d/dabbeaf50e6bad0b54d...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>[Sure.Polyphone.Midi/Entertainer.mid, Various ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71588</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/aaf2e0398a707b34bbe...</td>\n",
       "      <td>False</td>\n",
       "      <td>174.928646</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[124.000248000496]</td>\n",
       "      <td>131.612640</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[ESTCE PR HASARD, ESTCE PR HASARD, ESTCE PR HA...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/aaf2e0398a707b34bbeaefb9...</td>\n",
       "      <td>[E/Est Ce Par Hasard Dave.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>139917</th>\n",
       "      <td>../datasets/lmd/lmd_full/e/ebc5edd3c533092d324...</td>\n",
       "      <td>False</td>\n",
       "      <td>196.431974</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[85.00028333427778]</td>\n",
       "      <td>276.574446</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[TORN, TORN, TORN, TORN, TORN, TORN, TORN, TOR...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/ebc5edd3c533092d32461eee...</td>\n",
       "      <td>[X/Xgtorn.mid, 097/XGtorn.mid, 097/XGtorn.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45209</th>\n",
       "      <td>../datasets/lmd/lmd_full/1/1450ab4b1e52e25438d...</td>\n",
       "      <td>False</td>\n",
       "      <td>195.096618</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[107.9999136000691]</td>\n",
       "      <td>224.166846</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Soprano, Soprano, Soprano, Soprano, Soprano, ...</td>\n",
       "      <td>38.0</td>\n",
       "      <td>../datasets/lmd/proto/1450ab4b1e52e25438de8e7e...</td>\n",
       "      <td>[civilwar2/61sfttobfa.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>74319</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/aae666057efdbc7268d...</td>\n",
       "      <td>False</td>\n",
       "      <td>240.000000</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>66.010417</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[3xOsc (MIDI), 3xOsc #2 (MIDI), 3xOsc #3 (MIDI...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/aae666057efdbc7268d62b8f...</td>\n",
       "      <td>[K/Kara_Sun_-_Into_The_Sun_(Airbase_Dub_Mix)__...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>90199</th>\n",
       "      <td>../datasets/lmd/lmd_full/c/c750e5a017155f06e63...</td>\n",
       "      <td>False</td>\n",
       "      <td>256.000000</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[128.0]</td>\n",
       "      <td>15.117188</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Sylenth1 4, Sylenth1 3, Sylenth1 1]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/c750e5a017155f06e63ae69c...</td>\n",
       "      <td>[C/Cascada_-_BadBoy__DjMixdOut_20130211014527....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99470</th>\n",
       "      <td>../datasets/lmd/lmd_full/c/cc6b412aeeb834e8fec...</td>\n",
       "      <td>False</td>\n",
       "      <td>275.408685</td>\n",
       "      <td>[0.0, 182.14267500000003]</td>\n",
       "      <td>[140.00014000014, 108.99994368336242]</td>\n",
       "      <td>185.995888</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[, , , , , , , , , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/cc6b412aeeb834e8fecfd308...</td>\n",
       "      <td>[VerucaSalt/Seether.mid, VerucaSalt/Seether.mi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>175912</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/500a1212bae787f7ef8...</td>\n",
       "      <td>False</td>\n",
       "      <td>240.000000</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>98.494792</td>\n",
       "      <td>False</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[The Flowers, Of The Heath, Sequenced By, Barr...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/500a1212bae787f7ef8fb099...</td>\n",
       "      <td>[h/heath.mid, H/heath.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30403</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7dc2e8000f630ab4e57...</td>\n",
       "      <td>False</td>\n",
       "      <td>200.644292</td>\n",
       "      <td>[0.0, 9.69696, 38.49696, 48.193920000000006, 5...</td>\n",
       "      <td>[198.000198000198, 199.99999999999997, 198.000...</td>\n",
       "      <td>144.326112</td>\n",
       "      <td>True</td>\n",
       "      <td>384.0</td>\n",
       "      <td>[Drums - Travis, Bass - Mark, Guitar 1 - Tom, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/7dc2e8000f630ab4e57eebd5...</td>\n",
       "      <td>[B/blink_182-dumpweed.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>108557</th>\n",
       "      <td>../datasets/lmd/lmd_full/d/d864a2adc13ad8b9797...</td>\n",
       "      <td>False</td>\n",
       "      <td>251.560269</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[126.99998518333504]</td>\n",
       "      <td>205.980339</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[The Corrs, \"Breathless\", , ****************, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/d864a2adc13ad8b9797d6be2...</td>\n",
       "      <td>[Corrs/The Corrs - Breathless.mid, Various/The...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "160813  ../datasets/lmd/lmd_full/2/2b03576c2eac6b20c55...       False   \n",
       "83189   ../datasets/lmd/lmd_full/f/fc7280752baa136165c...       False   \n",
       "96858   ../datasets/lmd/lmd_full/c/cb226a46053507ae8a4...       False   \n",
       "34038   ../datasets/lmd/lmd_full/6/6d12ce4d832f9c39ca6...       False   \n",
       "171000  ../datasets/lmd/lmd_full/5/586c46e8f1c65beff59...       False   \n",
       "103720  ../datasets/lmd/lmd_full/d/dabbeaf50e6bad0b54d...        True   \n",
       "71588   ../datasets/lmd/lmd_full/a/aaf2e0398a707b34bbe...       False   \n",
       "139917  ../datasets/lmd/lmd_full/e/ebc5edd3c533092d324...       False   \n",
       "45209   ../datasets/lmd/lmd_full/1/1450ab4b1e52e25438d...       False   \n",
       "74319   ../datasets/lmd/lmd_full/a/aae666057efdbc7268d...       False   \n",
       "90199   ../datasets/lmd/lmd_full/c/c750e5a017155f06e63...       False   \n",
       "99470   ../datasets/lmd/lmd_full/c/cc6b412aeeb834e8fec...       False   \n",
       "175912  ../datasets/lmd/lmd_full/5/500a1212bae787f7ef8...       False   \n",
       "30403   ../datasets/lmd/lmd_full/7/7dc2e8000f630ab4e57...       False   \n",
       "108557  ../datasets/lmd/lmd_full/d/d864a2adc13ad8b9797...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "160813      220.504202                                              [0.0]   \n",
       "83189       149.380190                                              [0.0]   \n",
       "96858       170.000085                                              [0.0]   \n",
       "34038       231.914925  [0.0, 2.368422, 3.493422, 11.1907935, 11.19735...   \n",
       "171000      236.159582                                              [0.0]   \n",
       "103720             NaN                                               None   \n",
       "71588       174.928646                                              [0.0]   \n",
       "139917      196.431974                                              [0.0]   \n",
       "45209       195.096618                                              [0.0]   \n",
       "74319       240.000000                                              [0.0]   \n",
       "90199       256.000000                                              [0.0]   \n",
       "99470       275.408685                          [0.0, 182.14267500000003]   \n",
       "175912      240.000000                                              [0.0]   \n",
       "30403       200.644292  [0.0, 9.69696, 38.49696, 48.193920000000006, 5...   \n",
       "108557      251.560269                                              [0.0]   \n",
       "\n",
       "                                                    tempi    end_time  drums  \\\n",
       "160813                                [99.99999999999999]  200.200000   True   \n",
       "83189                                [160.01024065540196]  150.326316   True   \n",
       "96858                                 [170.0000850000425]   11.294112  False   \n",
       "34038   [151.99993920002433, 80.0, 151.99993920002433,...  503.284151  False   \n",
       "171000                                [130.0023833770286]   91.378132   True   \n",
       "103720                                               None         NaN   None   \n",
       "71588                                  [124.000248000496]  131.612640   True   \n",
       "139917                                [85.00028333427778]  276.574446   True   \n",
       "45209                                 [107.9999136000691]  224.166846  False   \n",
       "74319                                             [120.0]   66.010417  False   \n",
       "90199                                             [128.0]   15.117188  False   \n",
       "99470               [140.00014000014, 108.99994368336242]  185.995888   True   \n",
       "175912                                            [120.0]   98.494792  False   \n",
       "30403   [198.000198000198, 199.99999999999997, 198.000...  144.326112   True   \n",
       "108557                               [126.99998518333504]  205.980339   True   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "160813       384.0  [Lead, Bass, Guitar, Rhythm guitar, Voice oohs...   \n",
       "83189         48.0  [SYNTH, SYNTH, SYNTH, SYNTH, SYNTH, SYNTH, SYN...   \n",
       "96858         96.0    [Neo Cortex - Elements (Styles & Breeze Remix)]   \n",
       "34038        120.0  [Primo RH, Primo LH G, Primo LH F, Secondo RH ...   \n",
       "171000       384.0  [ACCORDION, A.PIANO 1, FINGERDBAS, CRYSTAL, DR...   \n",
       "103720         NaN                                               None   \n",
       "71588        384.0  [ESTCE PR HASARD, ESTCE PR HASARD, ESTCE PR HA...   \n",
       "139917       384.0  [TORN, TORN, TORN, TORN, TORN, TORN, TORN, TOR...   \n",
       "45209         96.0  [Soprano, Soprano, Soprano, Soprano, Soprano, ...   \n",
       "74319         96.0  [3xOsc (MIDI), 3xOsc #2 (MIDI), 3xOsc #3 (MIDI...   \n",
       "90199         96.0               [Sylenth1 4, Sylenth1 3, Sylenth1 1]   \n",
       "99470        120.0                             [, , , , , , , , , , ]   \n",
       "175912       384.0  [The Flowers, Of The Heath, Sequenced By, Barr...   \n",
       "30403        384.0  [Drums - Travis, Bass - Mark, Guitar 1 - Tom, ...   \n",
       "108557       120.0  [The Corrs, \"Breathless\", , ****************, ...   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "160813                         1.0   \n",
       "83189                          1.0   \n",
       "96858                          0.0   \n",
       "34038                          1.0   \n",
       "171000                         1.0   \n",
       "103720                         NaN   \n",
       "71588                          1.0   \n",
       "139917                         1.0   \n",
       "45209                         38.0   \n",
       "74319                          1.0   \n",
       "90199                          1.0   \n",
       "99470                          1.0   \n",
       "175912                         1.0   \n",
       "30403                          1.0   \n",
       "108557                         1.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "160813  ../datasets/lmd/proto/2b03576c2eac6b20c5537514...   \n",
       "83189   ../datasets/lmd/proto/fc7280752baa136165c911e0...   \n",
       "96858   ../datasets/lmd/proto/cb226a46053507ae8a47a16c...   \n",
       "34038   ../datasets/lmd/proto/6d12ce4d832f9c39ca683d53...   \n",
       "171000  ../datasets/lmd/proto/586c46e8f1c65beff59afe0b...   \n",
       "103720                                               None   \n",
       "71588   ../datasets/lmd/proto/aaf2e0398a707b34bbeaefb9...   \n",
       "139917  ../datasets/lmd/proto/ebc5edd3c533092d32461eee...   \n",
       "45209   ../datasets/lmd/proto/1450ab4b1e52e25438de8e7e...   \n",
       "74319   ../datasets/lmd/proto/aae666057efdbc7268d62b8f...   \n",
       "90199   ../datasets/lmd/proto/c750e5a017155f06e63ae69c...   \n",
       "99470   ../datasets/lmd/proto/cc6b412aeeb834e8fecfd308...   \n",
       "175912  ../datasets/lmd/proto/500a1212bae787f7ef8fb099...   \n",
       "30403   ../datasets/lmd/proto/7dc2e8000f630ab4e57eebd5...   \n",
       "108557  ../datasets/lmd/proto/d864a2adc13ad8b9797d6be2...   \n",
       "\n",
       "                                           original_files  \n",
       "160813                              [M/Monday Monday.mid]  \n",
       "83189                                    [e/entreaty.mid]  \n",
       "96858   [N/Neo Cortex - Elements (Breeze & Styles Remi...  \n",
       "34038   [Mendelsonn/Allegro Brillant for Piano 4-hands...  \n",
       "171000  [Byrd, Charlie/The-Girl-From-Ipanema.mid, BYRD...  \n",
       "103720  [Sure.Polyphone.Midi/Entertainer.mid, Various ...  \n",
       "71588                      [E/Est Ce Par Hasard Dave.mid]  \n",
       "139917     [X/Xgtorn.mid, 097/XGtorn.mid, 097/XGtorn.mid]  \n",
       "45209                          [civilwar2/61sfttobfa.mid]  \n",
       "74319   [K/Kara_Sun_-_Into_The_Sun_(Airbase_Dub_Mix)__...  \n",
       "90199   [C/Cascada_-_BadBoy__DjMixdOut_20130211014527....  \n",
       "99470   [VerucaSalt/Seether.mid, VerucaSalt/Seether.mi...  \n",
       "175912                         [h/heath.mid, H/heath.mid]  \n",
       "30403                          [B/blink_182-dumpweed.mid]  \n",
       "108557  [Corrs/The Corrs - Breathless.mid, Various/The...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.sample(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 178561 entries, 0 to 178560\n",
      "Data columns (total 12 columns):\n",
      " #   Column                      Non-Null Count   Dtype  \n",
      "---  ------                      --------------   -----  \n",
      " 0   midi_path                   178561 non-null  object \n",
      " 1   midi_error                  178561 non-null  bool   \n",
      " 2   estimate_tempo              173984 non-null  float64\n",
      " 3   tempi_sec                   173984 non-null  object \n",
      " 4   tempi                       173984 non-null  object \n",
      " 5   end_time                    173984 non-null  float64\n",
      " 6   drums                       173984 non-null  object \n",
      " 7   resolution                  173984 non-null  float64\n",
      " 8   instrument_names            173984 non-null  object \n",
      " 9   num_time_signature_changes  173984 non-null  float64\n",
      " 10  proto_path                  174476 non-null  object \n",
      " 11  original_files              178561 non-null  object \n",
      "dtypes: bool(1), float64(4), object(7)\n",
      "memory usage: 15.2+ MB\n"
     ]
    }
   ],
   "source": [
    "midi_df.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With a pandas dataframe we can simply aggregate and plot data into different formats.\n",
    "Let's start by taking a look at our success rate of parsing the MIDI files."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df['midi_error'].value_counts().plot.pie()\n",
    "plt.title('Faulty MIDI files');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.midi_error.value_counts().plot.bar()\n",
    "plt.title(\"Number of faulty MIDI files\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After we know the amount of failed MIDI files we could inspect those further to see if they are indeed corrupted or if our library has a bug."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/911cd08fa1fae36e5e0...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[s/sou.mid, S/sou.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/94862530febd2b295b9...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[e/ELVIS_PRESLEY__Dont_Be_Cruel.mid, ElvisPres...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/906e72809900e01f919...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[dream theater/uagm1.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/99e40264f321a4bfc5d...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[V/vong_tay_cau_hon.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/9f22ccab9572cafafce...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[W/Wippenberg - Pong (Tocadisco Remix).mid, W/...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178315</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5063a2b400597a372e7...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[f/faded.mid, 2009 MIDI/faded_love1-D120patsy_...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178377</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/523ce5adc656c872a69...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[l/laputa.mid, L/laputa.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178419</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/546df09d78a32141369...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[h/heraldic.mid, H/heraldic.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178484</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5a4e49112c6cf832341...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[w/WESTERN.MID, 2009 MIDI/good_bad_and_ugly4-C...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178525</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5cae276ae00fb1fd464...</td>\n",
       "      <td>True</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[soad/Chic_N_Stu.mid]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4084 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "3       ../datasets/lmd/lmd_full/9/911cd08fa1fae36e5e0...        True   \n",
       "24      ../datasets/lmd/lmd_full/9/94862530febd2b295b9...        True   \n",
       "25      ../datasets/lmd/lmd_full/9/906e72809900e01f919...        True   \n",
       "98      ../datasets/lmd/lmd_full/9/99e40264f321a4bfc5d...        True   \n",
       "106     ../datasets/lmd/lmd_full/9/9f22ccab9572cafafce...        True   \n",
       "...                                                   ...         ...   \n",
       "178315  ../datasets/lmd/lmd_full/5/5063a2b400597a372e7...        True   \n",
       "178377  ../datasets/lmd/lmd_full/5/523ce5adc656c872a69...        True   \n",
       "178419  ../datasets/lmd/lmd_full/5/546df09d78a32141369...        True   \n",
       "178484  ../datasets/lmd/lmd_full/5/5a4e49112c6cf832341...        True   \n",
       "178525  ../datasets/lmd/lmd_full/5/5cae276ae00fb1fd464...        True   \n",
       "\n",
       "        estimate_tempo tempi_sec tempi  end_time drums  resolution  \\\n",
       "3                  NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "24                 NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "25                 NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "98                 NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "106                NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "...                ...       ...   ...       ...   ...         ...   \n",
       "178315             NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "178377             NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "178419             NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "178484             NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "178525             NaN       NaN   NaN       NaN   NaN         NaN   \n",
       "\n",
       "       instrument_names  num_time_signature_changes proto_path  \\\n",
       "3                   NaN                         NaN        NaN   \n",
       "24                  NaN                         NaN        NaN   \n",
       "25                  NaN                         NaN        NaN   \n",
       "98                  NaN                         NaN        NaN   \n",
       "106                 NaN                         NaN        NaN   \n",
       "...                 ...                         ...        ...   \n",
       "178315              NaN                         NaN        NaN   \n",
       "178377              NaN                         NaN        NaN   \n",
       "178419              NaN                         NaN        NaN   \n",
       "178484              NaN                         NaN        NaN   \n",
       "178525              NaN                         NaN        NaN   \n",
       "\n",
       "                                           original_files  \n",
       "3                                  [s/sou.mid, S/sou.mid]  \n",
       "24      [e/ELVIS_PRESLEY__Dont_Be_Cruel.mid, ElvisPres...  \n",
       "25                              [dream theater/uagm1.mid]  \n",
       "98                               [V/vong_tay_cau_hon.mid]  \n",
       "106     [W/Wippenberg - Pong (Tocadisco Remix).mid, W/...  \n",
       "...                                                   ...  \n",
       "178315  [f/faded.mid, 2009 MIDI/faded_love1-D120patsy_...  \n",
       "178377                       [l/laputa.mid, L/laputa.mid]  \n",
       "178419                   [h/heraldic.mid, H/heraldic.mid]  \n",
       "178484  [w/WESTERN.MID, 2009 MIDI/good_bad_and_ugly4-C...  \n",
       "178525                              [soad/Chic_N_Stu.mid]  \n",
       "\n",
       "[4084 rows x 12 columns]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df[midi_df['midi_error'] == True]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For now, we will simply ignore those files and focus on the ones that we could sucessfully parse.\n",
    "\n",
    "For our experiments it is important that the MIDI files contain a drum track - only those files that we could parse succssfully and that contain a drum track can be used for training of our model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.drums.value_counts().plot.pie()\n",
    "plt.title(\"MIDI files that contain drums\");"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA4AAAAFOCAYAAADXSXO6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Z1A+gAAAACXBIWXMAAAsTAAALEwEAmpwYAAAiBklEQVR4nO3dfdRlZX0f/O9PRhRfQU3mUQaFKE2KkrQ6QRKTdiqJgkmDa9UYLAYwVJoVjXlaUoNNGp74kmhTSvSJ+iwaUEiNaIyJtJAgVe/Y2KLiS0SkxgmCDL6gzKCi8WX09/xxromH8Z63+57hzMz+fNY6a/a+9rX3/p1z7jnn/t7XPtep7g4AAAAHv3stugAAAADuGQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAATUFWvr6qXLujcVVWvq6otVfW+RdSwI1X1pKr6RFXdVVVPr6o/r6ozx7azquqv7oEaFvbc3FOq6ser6uN76VhHV1VX1Zq9cTyAqREAARagqm6uqtur6v5zbf+qqpYWWNa+8mNJfjLJuu4+YfuNI2h1VV24Xfupo/31Y/1uv/iP4PSNqvryuH20qn6nqh683bF3FuJenOT3u/sB3f1n3X1Kd1+6F+7zsvZ2qKyq/6eq/uveOt6+Om93/8/u/v59WRMAu0cABFicQ5L8yqKL2FNVdcge7vKoJDd391d20udvkzxzu1GdM5P8zS6O/R+7+4FJvifJc5KcmOQ988F6N2q7YTf7sp8bo81+twHYCS+SAIvzu0l+taoO337Dcpe5VdVSVf2rsXxWVb2nqi6sqjur6qaq+tHRfusYXTxzu8M+rKquGaNlf1lVj5o79g+MbZur6uNV9cy5ba+vqtdW1VVV9ZUk/2yZeh9RVVeM/TdW1XNH+9lJ/iDJj4zLLH9rB4/FZ5Ncn+SpY7+HJPnRJFfsxuOY7v5ad78/yc8keWhmYXCnqupvk3xfkv82arvP/GO8TP+dPUZPq6qPjcf2tqr61WX2/4dJ/r9857G4c27zEVV15dj/vVX16Ln9Xjme0y9V1Qeq6sdH+8lJ/n2SnxvH++sd1H1UVb21qj5fVXdU1e+P9ntV1W9U1S3j5+WybaOncz9/Z1bVp6rqC1X16zs7b1U9p6puHPfhpqr613M1bKiqTXPrN1fVr1bVR6rqi1X1pqq67w7qP6Sq/tOo4aYkP7Xd9qWqellVvSfJV5N83zj+T8z1+fsRy7n79pzxuG6pql+sqh8e9dy57TEa/R8z/r98cdTwpuXqBDhQCIAAi3NdkqUk3xUWdtMTk3wks8DzR0kuT/LDSR6T5NlJfr+qHjDX//QkL0nysCQfTvKGJKnZaNk14xjfm+S0JK+pquPm9v2XSV6W5IFJlruE8fIkm5I8Iskzkvx2VT25uy9O8otJ/ve4zPL8ndyfy5KcMZZPS/K2JF/f5aMwp7u/PO7Lj+9G30cn+VSSfz5q2+G5duMxujjJvx6jkY9L8s5lzndj7v5YHD63+bQkv5XkiCQbM3ust3l/kn+U5CHj/H9cVfft7r9I8ttJ3jSO90PL1H1Ikv+e5JYkRyc5MrPnKknOGrd/llkQfkCS39/uED+W5PuTnJTkN6vqH+7kvLcn+ekkD8osgF9YVY/fvqY5z0xycpJjkvzgqGU5zx3H/cdJ1mf287W9n09yTmY/n7fs5Jzznpjk2CQ/l+T3kvx6kp9I8tjMRqP/6ej3kiRvz+y5WZfk/93N4wPslwRAgMX6zSS/XFXfs4J9P9ndr+vubyV5U5Kjkry4u7/e3W9P8o3MwuA2V3b3u0fQ+fXMRqKOyuyX65vHsbZ294eS/EmSn53b923d/Z7u/nZ3f22+iHGMJyX5tTES9+HMRv3OyJ750yQbxijUGZkFwpX4dGZhaW/a1WP0zSTHVdWDuntLd39wD4//p939vu7emlkw/0fbNnT3f+3uO8Z5L0hyn8xC2e44IbNQ/u+6+yvj+dkW4E9P8p+7+6buvivJi5KcVne/DPe3uvvvuvuvk/x1ku8KmXN1Xtndf9szf5lZaNpZEH9Vd3+6uzcn+W/z93k7z0zye9196+j7O8v0eX133zAeo2/u5JzzXjIej7cn+UqSN3b37d19W5L/mVngTGbP7aOSPGK7xw/ggCQAAixQd380sxGa81aw++fmlv9uHG/7tvkRwFvnzntXks2ZhYNHJXniuPTtznFp4ulJ/q/l9l3GI5JsHqNv29yS2WjTbuvuv0tyZZLfSPLQ7n7Pnuw/58jM7tvetKvH6F8keVqSW8blgj+yh8f/7NzyVzP3vI1LJW8clyDemeTBmY3i7o6jktwyguX2HpG7j5bdkmRNkrW7U9f2quqUqrp2XCJ7Z2aPx87q3N1jPyJ3//lbboRvZz+fO7L9/5Ud/d95YZJK8r6quqGqfmEF5wLYb5hCGWDxzk/ywSQXzLVtmzDlfkm+NJbnA9lKHLVtYVwa+pDMRstuTfKX3f2TO9m3d7Lt00keUlUPnAuBj0xy2wpqvCyzyyd39FnBnRr36ydy90so94adPkbj84enVtW9kzw/yZsz93jPd92Tk47P+70ws0swb+jub1fVlswCye4c79Ykj6yqNcuEwE9nFmy3eWSSrZkFoXW7OO7dzltV98lsRPSMzEaLv1lVfzZX52p8Jnd/LB+5q3oy+/9zv7n1Ff/f6e7PZnYZaqrqx5L8j6p6d3dvXOkxARbJCCDAgo1fJN+U5AVzbZ/PLEA9e0yC8QtJHr2DQ+yup1XVj1XVoZl9runa7r41sxHIf1BVP19V9x63H67ZpCW7U/+tSf5Xkt+pqvtW1Q8mOTvJSr6e4C8z+8qIPfqcVc0mcHlCkj9LsiXJ61Zw7p3Z4WNUVYdW1elV9eBx+eGXknx7B8f5XJJ14znYHQ/MLJR9PsmaqvrNzD5jN3+8o2vHM1++L7MA9fKquv94fp40tr0xyb+pqmNGcN72ub7lRguXux/z5z00s0tTP59ka1WdkuQpu3kfd+XNSV5QVeuq6ojs3mj5hzO7nPXeVbWjzw3ulqr62araFoi3ZBY2d/T8Auz3BECA/cOLk2z/1QXPTfLvktyR2cQU/2uV5/ijzEYbNyd5QmYTxWybOOUpmU1E8unMLs17RWa/0O+uZ2U2ycinM/ss3/nd/T/2tMDx+bF3jM967Y4XVtWXM3uMLkvygSQ/uouvnNhju/EY/XySm6vqS5lN9HL6Dg71zsy+duKzVfWF3Tj11Un+IrOvw7glyddy98sd/3j8e0dVfdfnDsfnQ/95Zp8F/VRmE/X83Nh8SZI/TPLuJJ8cx/7l3ajpu847Hp8XZBbWtmQ2adBuzeC6G/5LZo/DX2c2Uv7W3djnP2T2B5MtmY0m/9Eqzv/DSd5bVXdldp9+pbtvWsXxABaquvfoahQAAAAOUEYAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACbioPsi+Ic97GF99NFHL7oMOGB95Stfyf3vv/23EQDAPcd7EazOBz7wgS909/cst+2gC4BHH310rrvuukWXAQespaWlbNiwYdFlADBh3otgdarqlh1tcwkoAADARAiAAAAAEyEAAgAATMQuA2BVXVJVt1fVR+fafreq/k9VfaSq/rSqDp/b9qKq2lhVH6+qp861nzzaNlbVeXPtx1TVe0f7m6rq0NF+n7G+cWw/em/daQAAgCnanRHA1yc5ebu2a5I8rrt/MMnfJHlRklTVcUlOS/LYsc9rquqQqjokyauTnJLkuCTPGn2T5BVJLuzuxyTZkuTs0X52ki2j/cLRDwAAgBXaZQDs7ncn2bxd29u7e+tYvTbJurF8apLLu/vr3f3JJBuTnDBuG7v7pu7+RpLLk5xaVZXkyUneMva/NMnT54516Vh+S5KTRn8AAABWYG98DcQvJHnTWD4ys0C4zabRliS3btf+xCQPTXLnXJic73/ktn26e2tVfXH0/8L2BVTVOUnOSZK1a9dmaWlpdfcIJuyuu+7yfwiAhfJeBPvOqgJgVf16kq1J3rB3ylmZ7r4oyUVJsn79+va9MbByvnsJgEXzXgT7zooDYFWdleSnk5zU3T2ab0ty1Fy3daMtO2i/I8nhVbVmjALO9992rE1VtSbJg0d/AAAAVmBFXwNRVScneWGSn+nur85tuiLJaWMGz2OSHJvkfUnen+TYMePnoZlNFHPFCI7vSvKMsf+ZSd42d6wzx/IzkrxzLmgCAACwh3Y5AlhVb0yyIcnDqmpTkvMzm/XzPkmuGfOyXNvdv9jdN1TVm5N8LLNLQ5/X3d8ax3l+kquTHJLkku6+YZzi15JcXlUvTfKhJBeP9ouT/GFVbcxsEprT9sL9BQAAmKw62AbV1q9f39ddd92iy2Anjj7vykWXwE6ce/zWXHD93pgfin3l5pf/1KJLANinfAYQVqeqPtDd65fbtqJLQAEAADjwCIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAAAABMhAAIAAEyEAAgAADARAiAAAMBECIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAAAABMhAAIAAEyEAAgAADARAiAAAMBECIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAAAABMhAAIAAEzELgNgVV1SVbdX1Ufn2h5SVddU1SfGv0eM9qqqV1XVxqr6SFU9fm6fM0f/T1TVmXPtT6iq68c+r6qq2tk5AAAAWJndGQF8fZKTt2s7L8k7uvvYJO8Y60lySpJjx+2cJK9NZmEuyflJnpjkhCTnzwW61yZ57tx+J+/iHAAAAKzALgNgd787yebtmk9NculYvjTJ0+faL+uZa5McXlUPT/LUJNd09+bu3pLkmiQnj20P6u5ru7uTXLbdsZY7BwAAACuwZoX7re3uz4zlzyZZO5aPTHLrXL9No21n7ZuWad/ZOb5LVZ2T2Yhj1q5dm6WlpT28O9yTzj1+66JLYCfWHuY52t95jQMOdnfddZfXOthHVhoA/153d1X13ihmpefo7ouSXJQk69ev7w0bNuzLclils867ctElsBPnHr81F1y/6pcG9qGbT9+w6BIA9qmlpaX4fQ72jZXOAvq5cflmxr+3j/bbkhw112/daNtZ+7pl2nd2DgAAAFZgpQHwiiTbZvI8M8nb5trPGLOBnpjki+MyzquTPKWqjhiTvzwlydVj25eq6sQx++cZ2x1ruXMAAACwAru8zquq3phkQ5KHVdWmzGbzfHmSN1fV2UluSfLM0f2qJE9LsjHJV5M8J0m6e3NVvSTJ+0e/F3f3tollfimzmUYPS/Ln45adnAMAAIAV2GUA7O5n7WDTScv07STP28FxLklyyTLt1yV53DLtdyx3DgAAAFZmpZeAAgAAcIARAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJmJVAbCq/k1V3VBVH62qN1bVfavqmKp6b1VtrKo3VdWho+99xvrGsf3oueO8aLR/vKqeOtd+8mjbWFXnraZWAACAqVtxAKyqI5O8IMn67n5ckkOSnJbkFUku7O7HJNmS5Oyxy9lJtoz2C0e/VNVxY7/HJjk5yWuq6pCqOiTJq5OckuS4JM8afQEAAFiB1V4CuibJYVW1Jsn9knwmyZOTvGVsvzTJ08fyqWM9Y/tJVVWj/fLu/np3fzLJxiQnjNvG7r6pu7+R5PLRFwAAgBVYcQDs7tuS/Kckn8os+H0xyQeS3NndW0e3TUmOHMtHJrl17Lt19H/ofPt2++yoHQAAgBVYs9Idq+qIzEbkjklyZ5I/zuwSzntcVZ2T5JwkWbt2bZaWlhZRBrvp3OO37roTC7P2MM/R/s5rHHCwu+uuu7zWwT6y4gCY5CeSfLK7P58kVfXWJE9KcnhVrRmjfOuS3Db635bkqCSbxiWjD05yx1z7NvP77Kj9brr7oiQXJcn69et7w4YNq7hb7GtnnXfloktgJ849fmsuuH41Lw3sazefvmHRJQDsU0tLS/H7HOwbq/kM4KeSnFhV9xuf5TspyceSvCvJM0afM5O8bSxfMdYztr+zu3u0nzZmCT0mybFJ3pfk/UmOHbOKHprZRDFXrKJeAACASVvxn/m7+71V9ZYkH0yyNcmHMhuFuzLJ5VX10tF28djl4iR/WFUbk2zOLNClu2+oqjdnFh63Jnled38rSarq+UmuzmyG0Uu6+4aV1gsAADB1q7rOq7vPT3L+ds03ZTaD5/Z9v5bkZ3dwnJcledky7VcluWo1NQIAADCz2q+BAAAA4AAhAAIAAEyEAAgAADARAiAAAMBECIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAAAABMhAAIAAEyEAAgAADARAiAAAMBECIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBECIAAAwEQIgAAAABMhAAIAAEyEAAgAADARAiAAAMBECIAAAAATIQACAABMhAAIAAAwEQIgAADARAiAAAAAEyEAAgAATIQACAAAMBGrCoBVdXhVvaWq/k9V3VhVP1JVD6mqa6rqE+PfI0bfqqpXVdXGqvpIVT1+7jhnjv6fqKoz59qfUFXXj31eVVW1mnoBAACmbLUjgK9M8hfd/QNJfijJjUnOS/KO7j42yTvGepKckuTYcTsnyWuTpKoekuT8JE9MckKS87eFxtHnuXP7nbzKegEAACZrxQGwqh6c5J8kuThJuvsb3X1nklOTXDq6XZrk6WP51CSX9cy1SQ6vqocneWqSa7p7c3dvSXJNkpPHtgd197Xd3UkumzsWAAAAe2g1I4DHJPl8ktdV1Yeq6g+q6v5J1nb3Z0afzyZZO5aPTHLr3P6bRtvO2jct0w4AAMAKrFnlvo9P8svd/d6qemW+c7lnkqS7u6p6NQXujqo6J7PLSrN27dosLS3t61OyCucev3XRJbATaw/zHO3vvMYBB7u77rrLax3sI6sJgJuSbOru9471t2QWAD9XVQ/v7s+MyzhvH9tvS3LU3P7rRtttSTZs17402tct0/+7dPdFSS5KkvXr1/eGDRuW68Z+4qzzrlx0CezEucdvzQXXr+algX3t5tM3LLoEgH1qaWkpfp+DfWPFl4B292eT3FpV3z+aTkrysSRXJNk2k+eZSd42lq9IcsaYDfTEJF8cl4peneQpVXXEmPzlKUmuHtu+VFUnjtk/z5g7FgAAAHtotX/m/+Ukb6iqQ5PclOQ5mYXKN1fV2UluSfLM0feqJE9LsjHJV0ffdPfmqnpJkvePfi/u7s1j+ZeSvD7JYUn+fNwAAABYgVUFwO7+cJL1y2w6aZm+neR5OzjOJUkuWab9uiSPW02NAAAAzKz2ewABAAA4QAiAAAAAEyEAAgAATIQACAAAMBG+7AsAmJSjfR/tfu/c47f63uD92M0v/6lFl8AqGAEEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCJWHQCr6pCq+lBV/fexfkxVvbeqNlbVm6rq0NF+n7G+cWw/eu4YLxrtH6+qp861nzzaNlbVeautFQAAYMr2xgjgryS5cW79FUku7O7HJNmS5OzRfnaSLaP9wtEvVXVcktOSPDbJyUleM0LlIUleneSUJMcledboCwAAwAqsKgBW1bokP5XkD8Z6JXlykreMLpcmefpYPnWsZ2w/afQ/Ncnl3f317v5kko1JThi3jd19U3d/I8nloy8AAAArsNoRwN9L8sIk3x7rD01yZ3dvHeubkhw5lo9McmuSjO1fHP3/vn27fXbUDgAAwAqsWemOVfXTSW7v7g9U1Ya9VtHKajknyTlJsnbt2iwtLS2yHHbh3OO37roTC7P2MM/R/s5rHKyO17j9n/ei/Zv3oQPbigNgkicl+ZmqelqS+yZ5UJJXJjm8qtaMUb51SW4b/W9LclSSTVW1JsmDk9wx177N/D47ar+b7r4oyUVJsn79+t6wYcMq7hb72lnnXbnoEtiJc4/fmguuX81LA/vazadvWHQJcEDzPrT/8160f/M+dGBb8SWg3f2i7l7X3UdnNonLO7v79CTvSvKM0e3MJG8by1eM9Yzt7+zuHu2njVlCj0lybJL3JXl/kmPHrKKHjnNcsdJ6AQAApm5f/Gnl15JcXlUvTfKhJBeP9ouT/GFVbUyyObNAl+6+oarenORjSbYmeV53fytJqur5Sa5OckiSS7r7hn1QLwAAwCTslQDY3UtJlsbyTZnN4Ll9n68l+dkd7P+yJC9bpv2qJFftjRoBAACmbm98DyAAAAAHAAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJWHEArKqjqupdVfWxqrqhqn5ltD+kqq6pqk+Mf48Y7VVVr6qqjVX1kap6/Nyxzhz9P1FVZ861P6Gqrh/7vKqqajV3FgAAYMpWMwK4Ncm53X1ckhOTPK+qjktyXpJ3dPexSd4x1pPklCTHjts5SV6bzAJjkvOTPDHJCUnO3xYaR5/nzu138irqBQAAmLQVB8Du/kx3f3AsfznJjUmOTHJqkktHt0uTPH0sn5rksp65NsnhVfXwJE9Nck13b+7uLUmuSXLy2Pag7r62uzvJZXPHAgAAYA/tlc8AVtXRSf5xkvcmWdvdnxmbPptk7Vg+Msmtc7ttGm07a9+0TDsAAAArsGa1B6iqByT5kyT/d3d/af5jet3dVdWrPcdu1HBOZpeVZu3atVlaWtrXp2QVzj1+66JLYCfWHuY52t95jYPV8Rq3//NetH/zPnRgW1UArKp7Zxb+3tDdbx3Nn6uqh3f3Z8ZlnLeP9tuSHDW3+7rRdluSDdu1L432dcv0/y7dfVGSi5Jk/fr1vWHDhuW6sZ8467wrF10CO3Hu8VtzwfWr/tsQ+9DNp29YdAlwQPM+tP/zXrR/8z50YFvNLKCV5OIkN3b3f57bdEWSbTN5npnkbXPtZ4zZQE9M8sVxqejVSZ5SVUeMyV+ekuTqse1LVXXiONcZc8cCAABgD63mTytPSvLzSa6vqg+Ptn+f5OVJ3lxVZye5Jckzx7arkjwtycYkX03ynCTp7s1V9ZIk7x/9Xtzdm8fyLyV5fZLDkvz5uAEAALACKw6A3f1XSXb0vXwnLdO/kzxvB8e6JMkly7Rfl+RxK60RAACA79grs4ACAACw/xMAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYCAEQAABgIgRAAACAiRAAAQAAJkIABAAAmAgBEAAAYCIEQAAAgIkQAAEAACZCAAQAAJgIARAAAGAiBEAAAICJEAABAAAmQgAEAACYiP0+AFbVyVX18araWFXnLboeAACAA9V+HQCr6pAkr05ySpLjkjyrqo5bbFUAAAAHpv06ACY5IcnG7r6pu7+R5PIkpy64JgAAgAPS/h4Aj0xy69z6ptEGAADAHlqz6AL2hqo6J8k5Y/Wuqvr4IuuBA9kLkocl+cKi62DH6hWLrgBg3/JetH/zPnRAeNSONuzvAfC2JEfNra8bbXfT3RclueieKgoOZlV1XXevX3QdAEyX9yLYd/b3S0Dfn+TYqjqmqg5NclqSKxZcEwAAwAFpvx4B7O6tVfX8JFcnOSTJJd19w4LLAgAAOCDt1wEwSbr7qiRXLboOmBCXUwOwaN6LYB+p7l50DQAAANwD9vfPAAIAALCXCIAAAAATIQDCxNXMs6vqN8f6I6vqhEXXBQDA3icAAq9J8iNJnjXWv5zk1YsrB4Apqqr7VdV/qKr/MtaPraqfXnRdcLARAIEndvfzknwtSbp7S5JDF1sSABP0uiRfz+yPkklyW5KXLq4cODgJgMA3q+qQJJ0kVfU9Sb692JIAmKBHd/d/TPLNJOnuryapxZYEBx8BEHhVkj9N8r1V9bIkf5XktxdbEgAT9I2qOizf+YPkozMbEQT2It8DCKSqfiDJSZn9pfUd3X3jgksCYGKq6ieT/EaS45K8PcmTkpzV3UuLrAsONgIgTFxVPXK59u7+1D1dCwDTVlUPTXJiZn+QvLa7v7DgkuCgIwDCxFXV9ZldblNJ7pvkmCQf7+7HLrQwACalqp6U5MPd/ZWqenaSxyd5ZXffsuDS4KDiM4Awcd19fHf/4Pj32CQnJPnfi64LgMl5bZKvVtUPJfm3Sf42yWWLLQkOPgIgcDfd/cEkT1x0HQBMztaeXZp2apJXd/erkzxwwTXBQWfNogsAFquq/u3c6r0yu+Tm0wsqB4Dp+nJVvSjJs5P8k6q6V5J7L7gmOOgYAQQeOHe7T5IrM/vrKwDck34us699OLu7P5tkXZLfXWxJcPAxCQxM2PgC+Fd0968uuhYAAPY9l4DCRFXVmu7eOmZdA4CFqKovZ3z5+/abknR3P+geLgkOakYAYaKq6oPd/fiqem2SI5P8cZKvbNve3W9dWHEAAOwTRgCB+ya5I8mT853vA+wkAiAA97iq+t7M3puSJN39qQWWAwcdARCm63vHDKAfzXeC3zYuDQDgHlVVP5PkgiSPSHJ7kkcluTHJYxdZFxxszAIK03VIkgeM2wPnlrfdAOCe9JIkJyb5m+4+JslJSa5dbElw8DECCNP1me5+8aKLAIDhm919R1Xdq6ru1d3vqqrfW3RRcLARAGG6atddAOAec2dVPSDJu5O8oapuz9zkZMDeYRZQmKiqekh3b150HQBMW1U9srs/VVX3T/J3mX1E6fQkD07yhu6+Y6EFwkFGAAQAYGG2fS3RWP6T7v4Xi64JDmYmgQEAYJHmP5LwfQurAiZCAAQAYJF6B8vAPuASUAAAFqaqvpXZZC+V5LAkX922KUl394MWVRscjARAAACAiXAJKAAAwEQIgAAAABMhAAIAAEyEAAgAADARAiAAAMBE/P8c/QLybsAmUAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.drums.value_counts().plot.bar();\n",
    "plt.title(\"Number of MIDI files that contain drums\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We should verify that the files in which we did not detect any drums have indeed no drums by listening to a random subset of them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>170937</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5c7f00cfbfb1c7a4c8d...</td>\n",
       "      <td>False</td>\n",
       "      <td>219.517230</td>\n",
       "      <td>[0.0, 216.0]</td>\n",
       "      <td>[80.0, 82.00003553334875]</td>\n",
       "      <td>422.539545</td>\n",
       "      <td>False</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Piano]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5c7f00cfbfb1c7a4c8df5cc0...</td>\n",
       "      <td>[C/Casablan.MID, C/CASABLAN.MID, C/CASABLAN.MID]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>173537</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5ccabab69a29eb49782...</td>\n",
       "      <td>False</td>\n",
       "      <td>134.749985</td>\n",
       "      <td>[0.0, 0.4225350000000001, 2.9225340000000006, ...</td>\n",
       "      <td>[71.00003550001773, 72.0000288000115, 68.00007...</td>\n",
       "      <td>271.779287</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[Flute, Celeste, Harp, Vibes]</td>\n",
       "      <td>10.0</td>\n",
       "      <td>../datasets/lmd/proto/5ccabab69a29eb49782f7ac2...</td>\n",
       "      <td>[A/AS_OP19.MID, A/AS_OP19.MID]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>160054</th>\n",
       "      <td>../datasets/lmd/lmd_full/2/248de65be7464e72e13...</td>\n",
       "      <td>False</td>\n",
       "      <td>288.075739</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[143.99988480009216]</td>\n",
       "      <td>13.333344</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[MIDI out]</td>\n",
       "      <td>0.0</td>\n",
       "      <td>../datasets/lmd/proto/248de65be7464e72e13026ee...</td>\n",
       "      <td>[H/Heaven's Cry - I Dont Need This No More.mid...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>104986</th>\n",
       "      <td>../datasets/lmd/lmd_full/d/d62040c47cc8ccc1256...</td>\n",
       "      <td>False</td>\n",
       "      <td>230.565249</td>\n",
       "      <td>[0.0, 1.1940279999999999, 1.6062414999999999, ...</td>\n",
       "      <td>[201.00031155048293, 131.00007641671124, 18.00...</td>\n",
       "      <td>418.198189</td>\n",
       "      <td>False</td>\n",
       "      <td>240.0</td>\n",
       "      <td>[Diskant]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/d62040c47cc8ccc12566150b...</td>\n",
       "      <td>[r/raffsonatacl.mid, R/raffsonatacl.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>177415</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5d3a4dcffc36dc21507...</td>\n",
       "      <td>False</td>\n",
       "      <td>183.485497</td>\n",
       "      <td>[0.0, 26.052641999999995, 26.252641999999994, ...</td>\n",
       "      <td>[75.99996960001218, 75.0, 73.99998273333736, 7...</td>\n",
       "      <td>355.954855</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[PRIMI MANDOLINI, SECONDI MANDOLINI, MANDOLE, ...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>../datasets/lmd/proto/5d3a4dcffc36dc21507ef683...</td>\n",
       "      <td>[p/primiracconti.mid, P/primiracconti.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25978</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7a6107b2ebe71033dd6...</td>\n",
       "      <td>False</td>\n",
       "      <td>133.333333</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[100.0]</td>\n",
       "      <td>46.500000</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[, , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/7a6107b2ebe71033dd6d7a5b...</td>\n",
       "      <td>[h/himno442.mid, H/himno442.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17961</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/09634cc18ee4b3e3d8e...</td>\n",
       "      <td>False</td>\n",
       "      <td>224.868701</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[132.000132000132]</td>\n",
       "      <td>198.167415</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[Clarinet, Trumpet, Tuba, Strings, Strings]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/09634cc18ee4b3e3d8e84072...</td>\n",
       "      <td>[a/Airship_Remix.mid, A/Airship_Remix.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34639</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/603e957bc609b539948...</td>\n",
       "      <td>False</td>\n",
       "      <td>100.000000</td>\n",
       "      <td>[0.0, 62.5, 65.5, 86.5, 91.5, 112.5, 113.5, 11...</td>\n",
       "      <td>[60.0, 30.0, 60.0, 30.0, 60.0, 30.0, 60.0, 30....</td>\n",
       "      <td>191.763885</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Soprano, Alto, Tenor, Bass, Piano (hi), Piano...</td>\n",
       "      <td>5.0</td>\n",
       "      <td>../datasets/lmd/proto/603e957bc609b5399481f26f...</td>\n",
       "      <td>[bliss/ppb74mhamt.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>161340</th>\n",
       "      <td>../datasets/lmd/lmd_full/2/24c247fe99ec2cfb909...</td>\n",
       "      <td>False</td>\n",
       "      <td>141.750142</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[126.00012600012599]</td>\n",
       "      <td>63.805492</td>\n",
       "      <td>False</td>\n",
       "      <td>240.0</td>\n",
       "      <td>[Piano, Piano]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/24c247fe99ec2cfb909c4aa4...</td>\n",
       "      <td>[W/waltz_05.mid, brahms/waltz_05.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29071</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7d3cac11d3db19406bb...</td>\n",
       "      <td>False</td>\n",
       "      <td>106.219130</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[91.04496862742124]</td>\n",
       "      <td>42.176960</td>\n",
       "      <td>False</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[, , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/7d3cac11d3db19406bb568a5...</td>\n",
       "      <td>[Christian/Inmyhart.mid, Various Artists/inmyh...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "170937  ../datasets/lmd/lmd_full/5/5c7f00cfbfb1c7a4c8d...       False   \n",
       "173537  ../datasets/lmd/lmd_full/5/5ccabab69a29eb49782...       False   \n",
       "160054  ../datasets/lmd/lmd_full/2/248de65be7464e72e13...       False   \n",
       "104986  ../datasets/lmd/lmd_full/d/d62040c47cc8ccc1256...       False   \n",
       "177415  ../datasets/lmd/lmd_full/5/5d3a4dcffc36dc21507...       False   \n",
       "25978   ../datasets/lmd/lmd_full/7/7a6107b2ebe71033dd6...       False   \n",
       "17961   ../datasets/lmd/lmd_full/0/09634cc18ee4b3e3d8e...       False   \n",
       "34639   ../datasets/lmd/lmd_full/6/603e957bc609b539948...       False   \n",
       "161340  ../datasets/lmd/lmd_full/2/24c247fe99ec2cfb909...       False   \n",
       "29071   ../datasets/lmd/lmd_full/7/7d3cac11d3db19406bb...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "170937      219.517230                                       [0.0, 216.0]   \n",
       "173537      134.749985  [0.0, 0.4225350000000001, 2.9225340000000006, ...   \n",
       "160054      288.075739                                              [0.0]   \n",
       "104986      230.565249  [0.0, 1.1940279999999999, 1.6062414999999999, ...   \n",
       "177415      183.485497  [0.0, 26.052641999999995, 26.252641999999994, ...   \n",
       "25978       133.333333                                              [0.0]   \n",
       "17961       224.868701                                              [0.0]   \n",
       "34639       100.000000  [0.0, 62.5, 65.5, 86.5, 91.5, 112.5, 113.5, 11...   \n",
       "161340      141.750142                                              [0.0]   \n",
       "29071       106.219130                                              [0.0]   \n",
       "\n",
       "                                                    tempi    end_time  drums  \\\n",
       "170937                          [80.0, 82.00003553334875]  422.539545  False   \n",
       "173537  [71.00003550001773, 72.0000288000115, 68.00007...  271.779287  False   \n",
       "160054                               [143.99988480009216]   13.333344  False   \n",
       "104986  [201.00031155048293, 131.00007641671124, 18.00...  418.198189  False   \n",
       "177415  [75.99996960001218, 75.0, 73.99998273333736, 7...  355.954855  False   \n",
       "25978                                             [100.0]   46.500000  False   \n",
       "17961                                  [132.000132000132]  198.167415  False   \n",
       "34639   [60.0, 30.0, 60.0, 30.0, 60.0, 30.0, 60.0, 30....  191.763885  False   \n",
       "161340                               [126.00012600012599]   63.805492  False   \n",
       "29071                                 [91.04496862742124]   42.176960  False   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "170937       480.0                                            [Piano]   \n",
       "173537       192.0                      [Flute, Celeste, Harp, Vibes]   \n",
       "160054        96.0                                         [MIDI out]   \n",
       "104986       240.0                                          [Diskant]   \n",
       "177415        96.0  [PRIMI MANDOLINI, SECONDI MANDOLINI, MANDOLE, ...   \n",
       "25978        120.0                                           [, , , ]   \n",
       "17961        192.0        [Clarinet, Trumpet, Tuba, Strings, Strings]   \n",
       "34639         96.0  [Soprano, Alto, Tenor, Bass, Piano (hi), Piano...   \n",
       "161340       240.0                                     [Piano, Piano]   \n",
       "29071         96.0                                           [, , , ]   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "170937                         1.0   \n",
       "173537                        10.0   \n",
       "160054                         0.0   \n",
       "104986                         1.0   \n",
       "177415                         3.0   \n",
       "25978                          1.0   \n",
       "17961                          1.0   \n",
       "34639                          5.0   \n",
       "161340                         1.0   \n",
       "29071                          1.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "170937  ../datasets/lmd/proto/5c7f00cfbfb1c7a4c8df5cc0...   \n",
       "173537  ../datasets/lmd/proto/5ccabab69a29eb49782f7ac2...   \n",
       "160054  ../datasets/lmd/proto/248de65be7464e72e13026ee...   \n",
       "104986  ../datasets/lmd/proto/d62040c47cc8ccc12566150b...   \n",
       "177415  ../datasets/lmd/proto/5d3a4dcffc36dc21507ef683...   \n",
       "25978   ../datasets/lmd/proto/7a6107b2ebe71033dd6d7a5b...   \n",
       "17961   ../datasets/lmd/proto/09634cc18ee4b3e3d8e84072...   \n",
       "34639   ../datasets/lmd/proto/603e957bc609b5399481f26f...   \n",
       "161340  ../datasets/lmd/proto/24c247fe99ec2cfb909c4aa4...   \n",
       "29071   ../datasets/lmd/proto/7d3cac11d3db19406bb568a5...   \n",
       "\n",
       "                                           original_files  \n",
       "170937   [C/Casablan.MID, C/CASABLAN.MID, C/CASABLAN.MID]  \n",
       "173537                     [A/AS_OP19.MID, A/AS_OP19.MID]  \n",
       "160054  [H/Heaven's Cry - I Dont Need This No More.mid...  \n",
       "104986           [r/raffsonatacl.mid, R/raffsonatacl.mid]  \n",
       "177415         [p/primiracconti.mid, P/primiracconti.mid]  \n",
       "25978                    [h/himno442.mid, H/himno442.mid]  \n",
       "17961          [a/Airship_Remix.mid, A/Airship_Remix.mid]  \n",
       "34639                              [bliss/ppb74mhamt.mid]  \n",
       "161340              [W/waltz_05.mid, brahms/waltz_05.mid]  \n",
       "29071   [Christian/Inmyhart.mid, Various Artists/inmyh...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df[midi_df['drums']==False].sample(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets take also a look at the other metadata we have extracted."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.explode('tempi')['tempi'].plot.box(vert=False);\n",
    "plt.title('Boxplot of tempi');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.explode('tempi')['tempi'].plot.hist(bins=5000, xlim=(0, 300));\n",
    "plt.title('Histogram of tempi');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We notice a big spork at around 120 bpm which is rather interesting and should be inspected what the exact casue of this spike is.\n",
    "\n",
    "It is important to get an understanding and feeling for the data.\n",
    "We have an expectation of our data which can help us to see if it makes sense and if we parsed it correctly. But that can also  be problematic because we project certain simplifying expectations on our data.\n",
    "\n",
    "Let's count the occurences of tempo changes in each song."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df.explode('tempi').groupby('midi_path').size().value_counts().plot.box(vert=False);\n",
    "plt.title('Boxplot of number of time chages per song');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "midi_path\n",
       "../datasets/lmd/lmd_full/9/95cf26881a51f376a54e4e3f049f2d48.mid    6597\n",
       "../datasets/lmd/lmd_full/3/34c04cd56b1087851f79780360f48225.mid    6096\n",
       "../datasets/lmd/lmd_full/2/2677cf785791e1cfca13a0566aafbc85.mid    6096\n",
       "../datasets/lmd/lmd_full/3/34a3df2a3a1e2267cf63657984f608cf.mid    5912\n",
       "../datasets/lmd/lmd_full/c/c615436a609bf4e82c102503ea01d855.mid    5758\n",
       "../datasets/lmd/lmd_full/1/1e40fd0edc293c8733a9c1b66517890b.mid    5758\n",
       "../datasets/lmd/lmd_full/6/6662d98153d0c84d93e794d0c2b3940f.mid    5758\n",
       "../datasets/lmd/lmd_full/8/8d956a0e5ee409a19adda9db287f7108.mid    4301\n",
       "../datasets/lmd/lmd_full/0/001549d8bc6ba6dc62ade443cf01a51e.mid    4058\n",
       "../datasets/lmd/lmd_full/0/0b0fdfe28eb0318ea3ac86046053724b.mid    3922\n",
       "dtype: int64"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.explode('tempi').groupby('midi_path').size().sort_values(ascending=False).head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Its worth to inspect those files in e.g. [MuseScore](https://musescore.org/de) and try to figure out what happens in them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df['end_time'].plot.box(vert=False);\n",
    "plt.title('Boxplot of end_time');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We see that the `end_time` has some extreme values that are probably wrong - let's inspect those examples closer to see if we can see a pattern."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>84737</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f7b41341a20201f860c...</td>\n",
       "      <td>False</td>\n",
       "      <td>225.041860</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[128.0]</td>\n",
       "      <td>31055.625000</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Melody, On the Sun Road, Vivian Lai, Joseph L...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>../datasets/lmd/proto/f7b41341a20201f860cb375e...</td>\n",
       "      <td>[s/sunroad.mid, S/sunroad.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28204</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7abab3566b77073f827...</td>\n",
       "      <td>False</td>\n",
       "      <td>220.109820</td>\n",
       "      <td>[0.0, 1.411764, 67.095876, 129.0154569375, 129...</td>\n",
       "      <td>[170.0000850000425, 95.00014250021376, 70.0000...</td>\n",
       "      <td>25714.187790</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[, , , , , , , ]</td>\n",
       "      <td>9.0</td>\n",
       "      <td>../datasets/lmd/proto/7abab3566b77073f8274c097...</td>\n",
       "      <td>[FErnszt/Olala.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>69901</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/af6b689cfbb13c302bd...</td>\n",
       "      <td>False</td>\n",
       "      <td>229.041171</td>\n",
       "      <td>[0.0, 5.052624, 40.420992, 94.420992, 118.4209...</td>\n",
       "      <td>[95.00014250021376, 190.0002850004275, 80.0, 1...</td>\n",
       "      <td>20384.791244</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Acoustic Bass, Brass, Grand piano, Rock Organ...</td>\n",
       "      <td>7.0</td>\n",
       "      <td>../datasets/lmd/proto/af6b689cfbb13c302bd41b86...</td>\n",
       "      <td>[FErnszt/Swingin.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26973</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/73f4f536d3d42293bd7...</td>\n",
       "      <td>False</td>\n",
       "      <td>144.311366</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[122.00006913337249]</td>\n",
       "      <td>19652.447880</td>\n",
       "      <td>True</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[tk1, tk2, tk3, tk4, tk10, tk11]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/73f4f536d3d42293bd789b34...</td>\n",
       "      <td>[Clapton Eric/Bellbottom Blues.mid, Various Ar...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24515</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7486b07d9a4060fa614...</td>\n",
       "      <td>False</td>\n",
       "      <td>181.059483</td>\n",
       "      <td>[0.0, 0.4829544166666666, 1.9996389166666664, ...</td>\n",
       "      <td>[88.00002346667293, 89.00994241056726, 90.0099...</td>\n",
       "      <td>19623.803343</td>\n",
       "      <td>True</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[Steel Str.Guitar, Fretless Bass, Piano, Violi...</td>\n",
       "      <td>13.0</td>\n",
       "      <td>../datasets/lmd/proto/7486b07d9a4060fa6142ac46...</td>\n",
       "      <td>[K/Kevin Parent Les Doigts 2.mid]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               midi_path  midi_error  \\\n",
       "84737  ../datasets/lmd/lmd_full/f/f7b41341a20201f860c...       False   \n",
       "28204  ../datasets/lmd/lmd_full/7/7abab3566b77073f827...       False   \n",
       "69901  ../datasets/lmd/lmd_full/a/af6b689cfbb13c302bd...       False   \n",
       "26973  ../datasets/lmd/lmd_full/7/73f4f536d3d42293bd7...       False   \n",
       "24515  ../datasets/lmd/lmd_full/7/7486b07d9a4060fa614...       False   \n",
       "\n",
       "       estimate_tempo                                          tempi_sec  \\\n",
       "84737      225.041860                                              [0.0]   \n",
       "28204      220.109820  [0.0, 1.411764, 67.095876, 129.0154569375, 129...   \n",
       "69901      229.041171  [0.0, 5.052624, 40.420992, 94.420992, 118.4209...   \n",
       "26973      144.311366                                              [0.0]   \n",
       "24515      181.059483  [0.0, 0.4829544166666666, 1.9996389166666664, ...   \n",
       "\n",
       "                                                   tempi      end_time  drums  \\\n",
       "84737                                            [128.0]  31055.625000  False   \n",
       "28204  [170.0000850000425, 95.00014250021376, 70.0000...  25714.187790   True   \n",
       "69901  [95.00014250021376, 190.0002850004275, 80.0, 1...  20384.791244   True   \n",
       "26973                               [122.00006913337249]  19652.447880   True   \n",
       "24515  [88.00002346667293, 89.00994241056726, 90.0099...  19623.803343   True   \n",
       "\n",
       "       resolution                                   instrument_names  \\\n",
       "84737       120.0  [Melody, On the Sun Road, Vivian Lai, Joseph L...   \n",
       "28204        96.0                                   [, , , , , , , ]   \n",
       "69901        96.0  [Acoustic Bass, Brass, Grand piano, Rock Organ...   \n",
       "26973       192.0                   [tk1, tk2, tk3, tk4, tk10, tk11]   \n",
       "24515       192.0  [Steel Str.Guitar, Fretless Bass, Piano, Violi...   \n",
       "\n",
       "       num_time_signature_changes  \\\n",
       "84737                         2.0   \n",
       "28204                         9.0   \n",
       "69901                         7.0   \n",
       "26973                         1.0   \n",
       "24515                        13.0   \n",
       "\n",
       "                                              proto_path  \\\n",
       "84737  ../datasets/lmd/proto/f7b41341a20201f860cb375e...   \n",
       "28204  ../datasets/lmd/proto/7abab3566b77073f8274c097...   \n",
       "69901  ../datasets/lmd/proto/af6b689cfbb13c302bd41b86...   \n",
       "26973  ../datasets/lmd/proto/73f4f536d3d42293bd789b34...   \n",
       "24515  ../datasets/lmd/proto/7486b07d9a4060fa6142ac46...   \n",
       "\n",
       "                                          original_files  \n",
       "84737                     [s/sunroad.mid, S/sunroad.mid]  \n",
       "28204                                [FErnszt/Olala.mid]  \n",
       "69901                              [FErnszt/Swingin.mid]  \n",
       "26973  [Clapton Eric/Bellbottom Blues.mid, Various Ar...  \n",
       "24515                  [K/Kevin Parent Les Doigts 2.mid]  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.sort_values('end_time', ascending=False).head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Regarding this, it seems we should filter out those really long pieces as those are most probably wrong files.\n",
    "\n",
    "Those errors could be caused by our parsing or by the MIDI files itself - but as we have a big enough corpora it is justifiable by not including those outliers."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Resolution corresponds to the [PPQN](https://en.wikipedia.org/wiki/Pulses_per_quarter_note) of the MIDI file.\n",
    "Everything above 1000 should be suspicious, so let's take a look at those examples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>33871</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/610987ebaf0a326db40...</td>\n",
       "      <td>False</td>\n",
       "      <td>239.221606</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>135.368840</td>\n",
       "      <td>False</td>\n",
       "      <td>25000.0</td>\n",
       "      <td>[, , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/610987ebaf0a326db4089867...</td>\n",
       "      <td>[b/beethovenalladanza.mid, B/beethovenalladanz...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45187</th>\n",
       "      <td>../datasets/lmd/lmd_full/1/1283eddbc4ab9042dcc...</td>\n",
       "      <td>False</td>\n",
       "      <td>225.608072</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[112.00005973336519]</td>\n",
       "      <td>79.280092</td>\n",
       "      <td>True</td>\n",
       "      <td>24576.0</td>\n",
       "      <td>[, , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/1283eddbc4ab9042dccaea09...</td>\n",
       "      <td>[Pop_and_Top40/2 Pac - Changes.mid, Pop_and_To...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>67351</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/a5c6eb45c94d84474cb...</td>\n",
       "      <td>False</td>\n",
       "      <td>235.213491</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>26.987488</td>\n",
       "      <td>False</td>\n",
       "      <td>16384.0</td>\n",
       "      <td>[, , ]</td>\n",
       "      <td>2.0</td>\n",
       "      <td>../datasets/lmd/proto/a5c6eb45c94d84474cbe00a3...</td>\n",
       "      <td>[a/amigo3.mid, A/amigo3.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>159033</th>\n",
       "      <td>../datasets/lmd/lmd_full/2/2076f0fa330be484c1c...</td>\n",
       "      <td>False</td>\n",
       "      <td>236.923314</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[140.00014000014]</td>\n",
       "      <td>26.999973</td>\n",
       "      <td>False</td>\n",
       "      <td>15360.0</td>\n",
       "      <td>[Main Lead]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/2076f0fa330be484c1c60242...</td>\n",
       "      <td>[M/M.I.D.O.R. - Far East.mid, M/m.i.d.o.r.__fa...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15472</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/0c3f038f4eaed7d7f8c...</td>\n",
       "      <td>False</td>\n",
       "      <td>272.000290</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[136.0001450668214]</td>\n",
       "      <td>28.014676</td>\n",
       "      <td>False</td>\n",
       "      <td>15360.0</td>\n",
       "      <td>[Melody]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/0c3f038f4eaed7d7f8c90478...</td>\n",
       "      <td>[W/whiteroom__someday_intro__ambia.mid, W/Whit...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "33871   ../datasets/lmd/lmd_full/6/610987ebaf0a326db40...       False   \n",
       "45187   ../datasets/lmd/lmd_full/1/1283eddbc4ab9042dcc...       False   \n",
       "67351   ../datasets/lmd/lmd_full/a/a5c6eb45c94d84474cb...       False   \n",
       "159033  ../datasets/lmd/lmd_full/2/2076f0fa330be484c1c...       False   \n",
       "15472   ../datasets/lmd/lmd_full/0/0c3f038f4eaed7d7f8c...       False   \n",
       "\n",
       "        estimate_tempo tempi_sec                 tempi    end_time  drums  \\\n",
       "33871       239.221606     [0.0]               [120.0]  135.368840  False   \n",
       "45187       225.608072     [0.0]  [112.00005973336519]   79.280092   True   \n",
       "67351       235.213491     [0.0]               [120.0]   26.987488  False   \n",
       "159033      236.923314     [0.0]     [140.00014000014]   26.999973  False   \n",
       "15472       272.000290     [0.0]   [136.0001450668214]   28.014676  False   \n",
       "\n",
       "        resolution instrument_names  num_time_signature_changes  \\\n",
       "33871      25000.0         [, , , ]                         1.0   \n",
       "45187      24576.0           [, , ]                         1.0   \n",
       "67351      16384.0           [, , ]                         2.0   \n",
       "159033     15360.0      [Main Lead]                         1.0   \n",
       "15472      15360.0         [Melody]                         1.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "33871   ../datasets/lmd/proto/610987ebaf0a326db4089867...   \n",
       "45187   ../datasets/lmd/proto/1283eddbc4ab9042dccaea09...   \n",
       "67351   ../datasets/lmd/proto/a5c6eb45c94d84474cbe00a3...   \n",
       "159033  ../datasets/lmd/proto/2076f0fa330be484c1c60242...   \n",
       "15472   ../datasets/lmd/proto/0c3f038f4eaed7d7f8c90478...   \n",
       "\n",
       "                                           original_files  \n",
       "33871   [b/beethovenalladanza.mid, B/beethovenalladanz...  \n",
       "45187   [Pop_and_Top40/2 Pac - Changes.mid, Pop_and_To...  \n",
       "67351                        [a/amigo3.mid, A/amigo3.mid]  \n",
       "159033  [M/M.I.D.O.R. - Far East.mid, M/m.i.d.o.r.__fa...  \n",
       "15472   [W/whiteroom__someday_intro__ambia.mid, W/Whit...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.sort_values('resolution', ascending=False).head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Another important metric is the number of time signatures.\n",
    "\n",
    "If we use a step-sequencer like approach, we will need our data to have one time signature only."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "midi_df['num_time_signature_changes'].plot.box(vert=False);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>72284</th>\n",
       "      <td>../datasets/lmd/lmd_full/a/a85612fb0992100bded...</td>\n",
       "      <td>False</td>\n",
       "      <td>222.143725</td>\n",
       "      <td>[0.0, 1.54624, 1.6513706666666668, 1.779370666...</td>\n",
       "      <td>[194.0190397350993, 190.23944805194805, 195.31...</td>\n",
       "      <td>204.310165</td>\n",
       "      <td>True</td>\n",
       "      <td>48.0</td>\n",
       "      <td>[STRAUSS, STRAUSS, STRAUSS, STRAUSS, STRAUSS, ...</td>\n",
       "      <td>959.0</td>\n",
       "      <td>../datasets/lmd/proto/a85612fb0992100bded76798...</td>\n",
       "      <td>[S/STRAUSS.MID, S/STRAUSS.MID]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26776</th>\n",
       "      <td>../datasets/lmd/lmd_full/7/7d42a95293cc47bcacf...</td>\n",
       "      <td>False</td>\n",
       "      <td>199.481296</td>\n",
       "      <td>[0.0, 1.6240640000000002, 2.071552, 2.22173866...</td>\n",
       "      <td>[160.0922131147541, 134.08180778032036, 133.16...</td>\n",
       "      <td>229.441685</td>\n",
       "      <td>True</td>\n",
       "      <td>48.0</td>\n",
       "      <td>[NACHT, NACHT, NACHT, NACHT, NACHT, NACHT, NACHT]</td>\n",
       "      <td>782.0</td>\n",
       "      <td>../datasets/lmd/proto/7d42a95293cc47bcacfaa568...</td>\n",
       "      <td>[Hollands/NACHT.MID]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>168045</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5693291bc82548fea37...</td>\n",
       "      <td>False</td>\n",
       "      <td>242.258977</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[136.0001450668214]</td>\n",
       "      <td>255.266272</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[VIBRA SLAP, VIBRA SLAP, VIBRA SLAP, VIBRA SLA...</td>\n",
       "      <td>479.0</td>\n",
       "      <td>../datasets/lmd/proto/5693291bc82548fea376ec42...</td>\n",
       "      <td>[S/Striving.MID, Midis Diversas/STRIVING.MID, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60293</th>\n",
       "      <td>../datasets/lmd/lmd_full/8/85b2d4e8637008f6b4e...</td>\n",
       "      <td>False</td>\n",
       "      <td>218.486679</td>\n",
       "      <td>[0.0, 2.5, 73.5, 75.586952, 89.884816, 131.884...</td>\n",
       "      <td>[96.0, 240.0, 230.00049833441304, 235.00013708...</td>\n",
       "      <td>283.476487</td>\n",
       "      <td>False</td>\n",
       "      <td>1024.0</td>\n",
       "      <td>[, ]</td>\n",
       "      <td>424.0</td>\n",
       "      <td>../datasets/lmd/proto/85b2d4e8637008f6b4e54e5c...</td>\n",
       "      <td>[B/beevar2.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44034</th>\n",
       "      <td>../datasets/lmd/lmd_full/6/69df67cd25dd33c1217...</td>\n",
       "      <td>False</td>\n",
       "      <td>228.828269</td>\n",
       "      <td>[0.0, 1.25, 3.65, 102.78041, 105.1804100000000...</td>\n",
       "      <td>[96.0, 100.0, 115.0000287500072, 100.0, 115.00...</td>\n",
       "      <td>407.079384</td>\n",
       "      <td>False</td>\n",
       "      <td>1024.0</td>\n",
       "      <td>[, ]</td>\n",
       "      <td>382.0</td>\n",
       "      <td>../datasets/lmd/proto/69df67cd25dd33c12177e159...</td>\n",
       "      <td>[Schumann/Schuman Toccata op7.mid, Schumann/Sc...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "72284   ../datasets/lmd/lmd_full/a/a85612fb0992100bded...       False   \n",
       "26776   ../datasets/lmd/lmd_full/7/7d42a95293cc47bcacf...       False   \n",
       "168045  ../datasets/lmd/lmd_full/5/5693291bc82548fea37...       False   \n",
       "60293   ../datasets/lmd/lmd_full/8/85b2d4e8637008f6b4e...       False   \n",
       "44034   ../datasets/lmd/lmd_full/6/69df67cd25dd33c1217...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "72284       222.143725  [0.0, 1.54624, 1.6513706666666668, 1.779370666...   \n",
       "26776       199.481296  [0.0, 1.6240640000000002, 2.071552, 2.22173866...   \n",
       "168045      242.258977                                              [0.0]   \n",
       "60293       218.486679  [0.0, 2.5, 73.5, 75.586952, 89.884816, 131.884...   \n",
       "44034       228.828269  [0.0, 1.25, 3.65, 102.78041, 105.1804100000000...   \n",
       "\n",
       "                                                    tempi    end_time  drums  \\\n",
       "72284   [194.0190397350993, 190.23944805194805, 195.31...  204.310165   True   \n",
       "26776   [160.0922131147541, 134.08180778032036, 133.16...  229.441685   True   \n",
       "168045                                [136.0001450668214]  255.266272   True   \n",
       "60293   [96.0, 240.0, 230.00049833441304, 235.00013708...  283.476487  False   \n",
       "44034   [96.0, 100.0, 115.0000287500072, 100.0, 115.00...  407.079384  False   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "72284         48.0  [STRAUSS, STRAUSS, STRAUSS, STRAUSS, STRAUSS, ...   \n",
       "26776         48.0  [NACHT, NACHT, NACHT, NACHT, NACHT, NACHT, NACHT]   \n",
       "168045        96.0  [VIBRA SLAP, VIBRA SLAP, VIBRA SLAP, VIBRA SLA...   \n",
       "60293       1024.0                                               [, ]   \n",
       "44034       1024.0                                               [, ]   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "72284                        959.0   \n",
       "26776                        782.0   \n",
       "168045                       479.0   \n",
       "60293                        424.0   \n",
       "44034                        382.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "72284   ../datasets/lmd/proto/a85612fb0992100bded76798...   \n",
       "26776   ../datasets/lmd/proto/7d42a95293cc47bcacfaa568...   \n",
       "168045  ../datasets/lmd/proto/5693291bc82548fea376ec42...   \n",
       "60293   ../datasets/lmd/proto/85b2d4e8637008f6b4e54e5c...   \n",
       "44034   ../datasets/lmd/proto/69df67cd25dd33c12177e159...   \n",
       "\n",
       "                                           original_files  \n",
       "72284                      [S/STRAUSS.MID, S/STRAUSS.MID]  \n",
       "26776                                [Hollands/NACHT.MID]  \n",
       "168045  [S/Striving.MID, Midis Diversas/STRIVING.MID, ...  \n",
       "60293                                     [B/beevar2.mid]  \n",
       "44034   [Schumann/Schuman Toccata op7.mid, Schumann/Sc...  "
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.sort_values('num_time_signature_changes', ascending=False).head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's check the amount of files we would loose if we limit our files to exactly one time signature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "(midi_df.num_time_signature_changes <= 1).value_counts().plot.pie();\n",
    "plt.title('Tracks without time signature changes');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can also take a look at the most common instrument names in our dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "                  354676\n",
       "Bass               29532\n",
       "Drums              19438\n",
       "untitled           18471\n",
       "Piano              16344\n",
       "Strings            10922\n",
       "Soprano            10478\n",
       "Guitar             10126\n",
       "Alto                9368\n",
       "Tenor               8987\n",
       "DRUMS               8013\n",
       "Melody              6693\n",
       "WinJammer Demo      6488\n",
       "Voice               6270\n",
       "Piano (hi)          4905\n",
       "Piano (lo)          4900\n",
       "Italian             4679\n",
       "STRINGS             4003\n",
       "bass                3785\n",
       "MELODY              3775\n",
       "Name: instrument_names, dtype: int64"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.explode('instrument_names').instrument_names.value_counts().head(20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And also take a look at the most common names of our original filenames."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-48-4f9f4be492b3>:1: FutureWarning: The default value of regex will change from True to False in a future version.\n",
      "  midi_df.explode('original_files').original_files.str.replace('[^a-zA-Z]', ' ').str.lower().str.split(' ').explode().value_counts().head(50)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "              1091324\n",
       "mid            570786\n",
       "midi            56765\n",
       "various         42479\n",
       "artists         36610\n",
       "a               34883\n",
       "the             34148\n",
       "l               29181\n",
       "s               28479\n",
       "polyphone       24771\n",
       "sure            23089\n",
       "i               22765\n",
       "m               20958\n",
       "e               20792\n",
       "t               19164\n",
       "midis           19014\n",
       "c               18162\n",
       "b               17950\n",
       "d               17712\n",
       "n               17012\n",
       "h               15452\n",
       "o               15425\n",
       "g               13303\n",
       "you             13228\n",
       "p               12267\n",
       "k               12032\n",
       "of              11890\n",
       "poly            11836\n",
       "in              11073\n",
       "f               10433\n",
       "beatles          9696\n",
       "love             9246\n",
       "r                8911\n",
       "me               8901\n",
       "j                8816\n",
       "diversen         8643\n",
       "my               8172\n",
       "to               7585\n",
       "divers           7025\n",
       "w                6894\n",
       "analisadas       6320\n",
       "no               6073\n",
       "de               5601\n",
       "on               5373\n",
       "and              5349\n",
       "midirip          5017\n",
       "it               4853\n",
       "bach             4692\n",
       "classical        4542\n",
       "unsorted         4419\n",
       "Name: original_files, dtype: int64"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "midi_df.explode('original_files').original_files.str.replace('[^a-zA-Z]', ' ').str.lower().str.split(' ').explode().value_counts().head(50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "An interesting aspect of this analysis reveals that the most common words in a text corpora do not often contain much information what the text is about. This is because the words that belong to the grammatical structure of the text introduce noise to the dataset.\n",
    "\n",
    "There are basic algorithms such as [tf-idf](https://en.wikipedia.org/wiki/Tf%E2%80%93idf) which allows us to filter out the noise by asserting that a word which is common in many documents has little no meaning (e.g. *the* or in our case *midi*) and therefore give it a low score.\n",
    "Although if a significant subset of our file names contain a word that is not so common in the other file names (e.g. *beatles*) this *token* will receive a high rating.\n",
    "\n",
    "For now we don't want to inspect this further, but its always good to know such kind of algorithms as it helps to filter out the relevant data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Its also good to make some sanity checks on our data.\n",
    "One assumption would be that every file in which we did not detect a drum track also does not contain an instrument with the name *drums*."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/0093ffc81428aabfa14...</td>\n",
       "      <td>False</td>\n",
       "      <td>190.000285</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[190.0002850004275]</td>\n",
       "      <td>137.151110</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/0093ffc81428aabfa14aaa13...</td>\n",
       "      <td>[M/Martha My Dear 3.mid, Beatles +GeorgeJohnPa...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/032e298d9560fdbe954...</td>\n",
       "      <td>False</td>\n",
       "      <td>120.000180</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[190.0002850004275]</td>\n",
       "      <td>114.947196</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/032e298d9560fdbe954d7551...</td>\n",
       "      <td>[s/smetmoth.mid, S/smetmoth.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/043e1c01e58a4bb4499...</td>\n",
       "      <td>False</td>\n",
       "      <td>139.999977</td>\n",
       "      <td>[0.0, 0.4285715, 7.071422000000001, 39.071422]</td>\n",
       "      <td>[69.99998833333528, 140.00014000014, 60.0, 69....</td>\n",
       "      <td>196.778591</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/043e1c01e58a4bb44995ee38...</td>\n",
       "      <td>[Songs+300/Tunnel_o.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/06f86377b1dbad0ebf0...</td>\n",
       "      <td>False</td>\n",
       "      <td>226.482465</td>\n",
       "      <td>[0.0, 1.0, 15.608691999999998]</td>\n",
       "      <td>[120.0, 115.0000287500072, 120.0]</td>\n",
       "      <td>59.483692</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>DRUMS</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/06f86377b1dbad0ebf08fdcd...</td>\n",
       "      <td>[s/seaquest2.mid, S/seaquest2.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>../datasets/lmd/lmd_full/0/078e2cc7c46eb5de726...</td>\n",
       "      <td>False</td>\n",
       "      <td>204.052854</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[98.00014373354414]</td>\n",
       "      <td>281.211322</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/078e2cc7c46eb5de72621329...</td>\n",
       "      <td>[Pop/SLEGDEHM.MID, PeterGabriel/Sledgehammer6....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>148</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f2a97c0f7fa26244a77...</td>\n",
       "      <td>False</td>\n",
       "      <td>218.459582</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[70.00007000007]</td>\n",
       "      <td>213.553358</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/f2a97c0f7fa26244a77a8d0f...</td>\n",
       "      <td>[u/unchainedmelody3.mid, U/UnchainedMelody3.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>149</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f6702b2445a6ce9aa8e...</td>\n",
       "      <td>False</td>\n",
       "      <td>211.143695</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[120.0]</td>\n",
       "      <td>21.241667</td>\n",
       "      <td>False</td>\n",
       "      <td>480.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/f6702b2445a6ce9aa8e6cad8...</td>\n",
       "      <td>[e/ElectroJam.Mid, E/ElectroJam.Mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>150</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/f8a56c3ebc00bfef70d...</td>\n",
       "      <td>False</td>\n",
       "      <td>157.793103</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[80.0]</td>\n",
       "      <td>263.847656</td>\n",
       "      <td>False</td>\n",
       "      <td>192.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/f8a56c3ebc00bfef70d033a8...</td>\n",
       "      <td>[t/tearinhand.mid, T/tearinhand.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>151</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/fa06799e6825c6a4095...</td>\n",
       "      <td>False</td>\n",
       "      <td>284.928064</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[148.99983858350816]</td>\n",
       "      <td>383.315852</td>\n",
       "      <td>False</td>\n",
       "      <td>240.0</td>\n",
       "      <td>drums</td>\n",
       "      <td>16.0</td>\n",
       "      <td>../datasets/lmd/proto/fa06799e6825c6a4095e2934...</td>\n",
       "      <td>[l/louder.mid, L/louder.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>152</th>\n",
       "      <td>../datasets/lmd/lmd_full/f/fd37e5d1258ba97fc0e...</td>\n",
       "      <td>False</td>\n",
       "      <td>235.685752</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[100.0]</td>\n",
       "      <td>191.795000</td>\n",
       "      <td>False</td>\n",
       "      <td>120.0</td>\n",
       "      <td>Drums</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/fd37e5d1258ba97fc0ef9779...</td>\n",
       "      <td>[Christian/Easyway.mid, Various Artists/easywa...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>153 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             midi_path  midi_error  \\\n",
       "0    ../datasets/lmd/lmd_full/0/0093ffc81428aabfa14...       False   \n",
       "1    ../datasets/lmd/lmd_full/0/032e298d9560fdbe954...       False   \n",
       "2    ../datasets/lmd/lmd_full/0/043e1c01e58a4bb4499...       False   \n",
       "3    ../datasets/lmd/lmd_full/0/06f86377b1dbad0ebf0...       False   \n",
       "4    ../datasets/lmd/lmd_full/0/078e2cc7c46eb5de726...       False   \n",
       "..                                                 ...         ...   \n",
       "148  ../datasets/lmd/lmd_full/f/f2a97c0f7fa26244a77...       False   \n",
       "149  ../datasets/lmd/lmd_full/f/f6702b2445a6ce9aa8e...       False   \n",
       "150  ../datasets/lmd/lmd_full/f/f8a56c3ebc00bfef70d...       False   \n",
       "151  ../datasets/lmd/lmd_full/f/fa06799e6825c6a4095...       False   \n",
       "152  ../datasets/lmd/lmd_full/f/fd37e5d1258ba97fc0e...       False   \n",
       "\n",
       "     estimate_tempo                                       tempi_sec  \\\n",
       "0        190.000285                                           [0.0]   \n",
       "1        120.000180                                           [0.0]   \n",
       "2        139.999977  [0.0, 0.4285715, 7.071422000000001, 39.071422]   \n",
       "3        226.482465                  [0.0, 1.0, 15.608691999999998]   \n",
       "4        204.052854                                           [0.0]   \n",
       "..              ...                                             ...   \n",
       "148      218.459582                                           [0.0]   \n",
       "149      211.143695                                           [0.0]   \n",
       "150      157.793103                                           [0.0]   \n",
       "151      284.928064                                           [0.0]   \n",
       "152      235.685752                                           [0.0]   \n",
       "\n",
       "                                                 tempi    end_time  drums  \\\n",
       "0                                  [190.0002850004275]  137.151110  False   \n",
       "1                                  [190.0002850004275]  114.947196  False   \n",
       "2    [69.99998833333528, 140.00014000014, 60.0, 69....  196.778591  False   \n",
       "3                    [120.0, 115.0000287500072, 120.0]   59.483692  False   \n",
       "4                                  [98.00014373354414]  281.211322  False   \n",
       "..                                                 ...         ...    ...   \n",
       "148                                   [70.00007000007]  213.553358  False   \n",
       "149                                            [120.0]   21.241667  False   \n",
       "150                                             [80.0]  263.847656  False   \n",
       "151                               [148.99983858350816]  383.315852  False   \n",
       "152                                            [100.0]  191.795000  False   \n",
       "\n",
       "     resolution instrument_names  num_time_signature_changes  \\\n",
       "0         192.0            Drums                         1.0   \n",
       "1         120.0            Drums                         1.0   \n",
       "2         120.0            Drums                         1.0   \n",
       "3         192.0            DRUMS                         1.0   \n",
       "4         192.0            Drums                         1.0   \n",
       "..          ...              ...                         ...   \n",
       "148       192.0            drums                         1.0   \n",
       "149       480.0            Drums                         1.0   \n",
       "150       192.0            Drums                         1.0   \n",
       "151       240.0            drums                        16.0   \n",
       "152       120.0            Drums                         1.0   \n",
       "\n",
       "                                            proto_path  \\\n",
       "0    ../datasets/lmd/proto/0093ffc81428aabfa14aaa13...   \n",
       "1    ../datasets/lmd/proto/032e298d9560fdbe954d7551...   \n",
       "2    ../datasets/lmd/proto/043e1c01e58a4bb44995ee38...   \n",
       "3    ../datasets/lmd/proto/06f86377b1dbad0ebf08fdcd...   \n",
       "4    ../datasets/lmd/proto/078e2cc7c46eb5de72621329...   \n",
       "..                                                 ...   \n",
       "148  ../datasets/lmd/proto/f2a97c0f7fa26244a77a8d0f...   \n",
       "149  ../datasets/lmd/proto/f6702b2445a6ce9aa8e6cad8...   \n",
       "150  ../datasets/lmd/proto/f8a56c3ebc00bfef70d033a8...   \n",
       "151  ../datasets/lmd/proto/fa06799e6825c6a4095e2934...   \n",
       "152  ../datasets/lmd/proto/fd37e5d1258ba97fc0ef9779...   \n",
       "\n",
       "                                        original_files  \n",
       "0    [M/Martha My Dear 3.mid, Beatles +GeorgeJohnPa...  \n",
       "1                     [s/smetmoth.mid, S/smetmoth.mid]  \n",
       "2                             [Songs+300/Tunnel_o.mid]  \n",
       "3                   [s/seaquest2.mid, S/seaquest2.mid]  \n",
       "4    [Pop/SLEGDEHM.MID, PeterGabriel/Sledgehammer6....  \n",
       "..                                                 ...  \n",
       "148   [u/unchainedmelody3.mid, U/UnchainedMelody3.mid]  \n",
       "149               [e/ElectroJam.Mid, E/ElectroJam.Mid]  \n",
       "150               [t/tearinhand.mid, T/tearinhand.mid]  \n",
       "151                       [l/louder.mid, L/louder.mid]  \n",
       "152  [Christian/Easyway.mid, Various Artists/easywa...  \n",
       "\n",
       "[153 rows x 12 columns]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "explodeded_df = midi_df.explode('instrument_names')\n",
    "explodeded_df[\n",
    "    (explodeded_df['instrument_names'].str.lower().isin(['drums', 'drum']))&\n",
    "    (explodeded_df['drums'] == False)\n",
    "].groupby('midi_path').first().reset_index()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Listening to those examples reveal that some of them use another instrument than drums for their drums - either by artistic choice or by mistake.\n",
    "As this *only* applies to 153 files we can live with this error and ignore the files were we did not detect the drum track."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Filter and extract the data\n",
    "\n",
    "After inspecting all availabe files we should filter out some *noisy* files that are either not interesting to us (they lack a drum track) or the parsing of the MIDI files did not work as expected.\n",
    "With the analysis above we can verify borders of data we want to allow.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>midi_path</th>\n",
       "      <th>midi_error</th>\n",
       "      <th>estimate_tempo</th>\n",
       "      <th>tempi_sec</th>\n",
       "      <th>tempi</th>\n",
       "      <th>end_time</th>\n",
       "      <th>drums</th>\n",
       "      <th>resolution</th>\n",
       "      <th>instrument_names</th>\n",
       "      <th>num_time_signature_changes</th>\n",
       "      <th>proto_path</th>\n",
       "      <th>original_files</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/9d30679480d0e55a07b...</td>\n",
       "      <td>False</td>\n",
       "      <td>224.502048</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[115.00002875000717]</td>\n",
       "      <td>296.347752</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[Piano, Organ, Big Brass, Horns, Horns2, Horns...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/9d30679480d0e55a07be1414...</td>\n",
       "      <td>[S/Smooth .mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/974f5b2466a5d5670df...</td>\n",
       "      <td>False</td>\n",
       "      <td>184.213982</td>\n",
       "      <td>[0.0, 2.714283, 170.86248999999998, 171.960050...</td>\n",
       "      <td>[210.00021000021, 80.99997165000993, 82.000035...</td>\n",
       "      <td>225.896259</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[, , , , , , , , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/974f5b2466a5d5670dfddee0...</td>\n",
       "      <td>[Sure.Polyphone.Midi/A Winter's Tale.mid, w/wi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/952df587b9fff5adbeb...</td>\n",
       "      <td>False</td>\n",
       "      <td>265.636110</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[144.0023040368646]</td>\n",
       "      <td>216.460946</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Track 2, Track 3, Track 4, Track 5, Track 6, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/952df587b9fff5adbeb5f40b...</td>\n",
       "      <td>[E/Enola Gay 63.mid, divers midi 3/OMD_-_Enola...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/93e2cb3a0c36af361bb...</td>\n",
       "      <td>False</td>\n",
       "      <td>165.000165</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[165.000165000165]</td>\n",
       "      <td>310.029993</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Chorus Guitar, Overdrive Guita, Overdrive Gui...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/93e2cb3a0c36af361bb49c5e...</td>\n",
       "      <td>[z/zombie05.mid, Z/zombie05.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>../datasets/lmd/lmd_full/9/97c1f795ffb676dfaa5...</td>\n",
       "      <td>False</td>\n",
       "      <td>212.877687</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[110.00011000011]</td>\n",
       "      <td>222.766823</td>\n",
       "      <td>True</td>\n",
       "      <td>192.0</td>\n",
       "      <td>[BASS, E.PIANO, GUITAR, STRINGS, PIANO, New Tr...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/97c1f795ffb676dfaa5a0202...</td>\n",
       "      <td>[J/J. Rivers Poor Side of Town.mid, divers mid...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178555</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5bd52f314b616bf50c4...</td>\n",
       "      <td>False</td>\n",
       "      <td>141.332152</td>\n",
       "      <td>[0.0, 183.428388, 186.85695800000002, 241.9283...</td>\n",
       "      <td>[70.00007000007, 35.00001458333941, 70.0000700...</td>\n",
       "      <td>250.221166</td>\n",
       "      <td>True</td>\n",
       "      <td>480.0</td>\n",
       "      <td>[Track 1, Track 2, Track 3, Track 4, Track 6, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5bd52f314b616bf50c40104c...</td>\n",
       "      <td>[T/taylor_swift-tim_mcgraw.mid]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178556</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5eea68ee369af3ee5cc...</td>\n",
       "      <td>False</td>\n",
       "      <td>184.252715</td>\n",
       "      <td>[0.0, 210.73867815, 210.75516165, 210.77763355...</td>\n",
       "      <td>[90.00009000009, 91.00009100009099, 89.0000400...</td>\n",
       "      <td>299.296908</td>\n",
       "      <td>True</td>\n",
       "      <td>240.0</td>\n",
       "      <td>[You've Made Me So Very Happy, You've Made Me ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5eea68ee369af3ee5cc577ba...</td>\n",
       "      <td>[Y/Youvemad L.mid, Y/YOUVEMAD L.mid, Y/YOUVEMA...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178557</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5309fb3f62bf73780e6...</td>\n",
       "      <td>False</td>\n",
       "      <td>183.172080</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[81.01003309259853]</td>\n",
       "      <td>139.390142</td>\n",
       "      <td>True</td>\n",
       "      <td>120.0</td>\n",
       "      <td>[, , , , , , , , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5309fb3f62bf73780e61f85b...</td>\n",
       "      <td>[R/Roberto Carlos - Propuesta L.mid, Midis Jov...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178559</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/5538b0174111bcef760...</td>\n",
       "      <td>False</td>\n",
       "      <td>274.353398</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[142.00007100003546]</td>\n",
       "      <td>172.389879</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[Std Drums, Jazz Gtr, Muted Gtr (Lead Dbl), Mu...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/5538b0174111bcef760a5f29...</td>\n",
       "      <td>[B/Boys 2.mid, b/boys_2.mid, Beatles +GeorgeJo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>178560</th>\n",
       "      <td>../datasets/lmd/lmd_full/5/51e756897765aed30af...</td>\n",
       "      <td>False</td>\n",
       "      <td>229.821409</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[64.99999458333379]</td>\n",
       "      <td>293.317332</td>\n",
       "      <td>True</td>\n",
       "      <td>96.0</td>\n",
       "      <td>[, , , , , ]</td>\n",
       "      <td>1.0</td>\n",
       "      <td>../datasets/lmd/proto/51e756897765aed30af232df...</td>\n",
       "      <td>[G/Gerry Boulet Deadline.mid]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>97848 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                midi_path  midi_error  \\\n",
       "5       ../datasets/lmd/lmd_full/9/9d30679480d0e55a07b...       False   \n",
       "9       ../datasets/lmd/lmd_full/9/974f5b2466a5d5670df...       False   \n",
       "10      ../datasets/lmd/lmd_full/9/952df587b9fff5adbeb...       False   \n",
       "11      ../datasets/lmd/lmd_full/9/93e2cb3a0c36af361bb...       False   \n",
       "12      ../datasets/lmd/lmd_full/9/97c1f795ffb676dfaa5...       False   \n",
       "...                                                   ...         ...   \n",
       "178555  ../datasets/lmd/lmd_full/5/5bd52f314b616bf50c4...       False   \n",
       "178556  ../datasets/lmd/lmd_full/5/5eea68ee369af3ee5cc...       False   \n",
       "178557  ../datasets/lmd/lmd_full/5/5309fb3f62bf73780e6...       False   \n",
       "178559  ../datasets/lmd/lmd_full/5/5538b0174111bcef760...       False   \n",
       "178560  ../datasets/lmd/lmd_full/5/51e756897765aed30af...       False   \n",
       "\n",
       "        estimate_tempo                                          tempi_sec  \\\n",
       "5           224.502048                                              [0.0]   \n",
       "9           184.213982  [0.0, 2.714283, 170.86248999999998, 171.960050...   \n",
       "10          265.636110                                              [0.0]   \n",
       "11          165.000165                                              [0.0]   \n",
       "12          212.877687                                              [0.0]   \n",
       "...                ...                                                ...   \n",
       "178555      141.332152  [0.0, 183.428388, 186.85695800000002, 241.9283...   \n",
       "178556      184.252715  [0.0, 210.73867815, 210.75516165, 210.77763355...   \n",
       "178557      183.172080                                              [0.0]   \n",
       "178559      274.353398                                              [0.0]   \n",
       "178560      229.821409                                              [0.0]   \n",
       "\n",
       "                                                    tempi    end_time drums  \\\n",
       "5                                    [115.00002875000717]  296.347752  True   \n",
       "9       [210.00021000021, 80.99997165000993, 82.000035...  225.896259  True   \n",
       "10                                    [144.0023040368646]  216.460946  True   \n",
       "11                                     [165.000165000165]  310.029993  True   \n",
       "12                                      [110.00011000011]  222.766823  True   \n",
       "...                                                   ...         ...   ...   \n",
       "178555  [70.00007000007, 35.00001458333941, 70.0000700...  250.221166  True   \n",
       "178556  [90.00009000009, 91.00009100009099, 89.0000400...  299.296908  True   \n",
       "178557                                [81.01003309259853]  139.390142  True   \n",
       "178559                               [142.00007100003546]  172.389879  True   \n",
       "178560                                [64.99999458333379]  293.317332  True   \n",
       "\n",
       "        resolution                                   instrument_names  \\\n",
       "5            120.0  [Piano, Organ, Big Brass, Horns, Horns2, Horns...   \n",
       "9            120.0                               [, , , , , , , , , ]   \n",
       "10           480.0  [Track 2, Track 3, Track 4, Track 5, Track 6, ...   \n",
       "11           480.0  [Chorus Guitar, Overdrive Guita, Overdrive Gui...   \n",
       "12           192.0  [BASS, E.PIANO, GUITAR, STRINGS, PIANO, New Tr...   \n",
       "...            ...                                                ...   \n",
       "178555       480.0  [Track 1, Track 2, Track 3, Track 4, Track 6, ...   \n",
       "178556       240.0  [You've Made Me So Very Happy, You've Made Me ...   \n",
       "178557       120.0                               [, , , , , , , , , ]   \n",
       "178559        96.0  [Std Drums, Jazz Gtr, Muted Gtr (Lead Dbl), Mu...   \n",
       "178560        96.0                                       [, , , , , ]   \n",
       "\n",
       "        num_time_signature_changes  \\\n",
       "5                              1.0   \n",
       "9                              1.0   \n",
       "10                             1.0   \n",
       "11                             1.0   \n",
       "12                             1.0   \n",
       "...                            ...   \n",
       "178555                         1.0   \n",
       "178556                         1.0   \n",
       "178557                         1.0   \n",
       "178559                         1.0   \n",
       "178560                         1.0   \n",
       "\n",
       "                                               proto_path  \\\n",
       "5       ../datasets/lmd/proto/9d30679480d0e55a07be1414...   \n",
       "9       ../datasets/lmd/proto/974f5b2466a5d5670dfddee0...   \n",
       "10      ../datasets/lmd/proto/952df587b9fff5adbeb5f40b...   \n",
       "11      ../datasets/lmd/proto/93e2cb3a0c36af361bb49c5e...   \n",
       "12      ../datasets/lmd/proto/97c1f795ffb676dfaa5a0202...   \n",
       "...                                                   ...   \n",
       "178555  ../datasets/lmd/proto/5bd52f314b616bf50c40104c...   \n",
       "178556  ../datasets/lmd/proto/5eea68ee369af3ee5cc577ba...   \n",
       "178557  ../datasets/lmd/proto/5309fb3f62bf73780e61f85b...   \n",
       "178559  ../datasets/lmd/proto/5538b0174111bcef760a5f29...   \n",
       "178560  ../datasets/lmd/proto/51e756897765aed30af232df...   \n",
       "\n",
       "                                           original_files  \n",
       "5                                         [S/Smooth .mid]  \n",
       "9       [Sure.Polyphone.Midi/A Winter's Tale.mid, w/wi...  \n",
       "10      [E/Enola Gay 63.mid, divers midi 3/OMD_-_Enola...  \n",
       "11                       [z/zombie05.mid, Z/zombie05.mid]  \n",
       "12      [J/J. Rivers Poor Side of Town.mid, divers mid...  \n",
       "...                                                   ...  \n",
       "178555                    [T/taylor_swift-tim_mcgraw.mid]  \n",
       "178556  [Y/Youvemad L.mid, Y/YOUVEMAD L.mid, Y/YOUVEMA...  \n",
       "178557  [R/Roberto Carlos - Propuesta L.mid, Midis Jov...  \n",
       "178559  [B/Boys 2.mid, b/boys_2.mid, Beatles +GeorgeJo...  \n",
       "178560                      [G/Gerry Boulet Deadline.mid]  \n",
       "\n",
       "[97848 rows x 12 columns]"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "filtered_midi_df = midi_df[\n",
    "    (midi_df.drums == True)\n",
    "    & (midi_df.midi_error == False)\n",
    "    & (midi_df.end_time.between(30, 800))\n",
    "    & (midi_df.num_time_signature_changes<=1)\n",
    "]\n",
    "filtered_midi_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5479696014247232"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(filtered_midi_df)/len(midi_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have now filtered half of our available examples, just by looking at the metadata of the MIDI files.\n",
    "The next step is to transform the drum track of each MIDI file into a mathematical representation. \n",
    "\n",
    "The notatation within the MIDI file is bound to time, but in music we often represent the time domain in relation to a tempo which remaps the time measured in seconds.\n",
    "A common unit for this is *beats per minute* (bpm) where 60 bpm maps each second to a *beat* - if we want to shorten the time between two beats (commonly refered to as *faster*) we increase the *bpm* and vice versa.\n",
    "\n",
    "Another instance for the segmentation of events in time is the *time signature* of a track.\n",
    "\n",
    "One notation which tries to simplify this notation is in the form of a *step seqencer* in which we snap notes to the nearest step on a grid.\n",
    "\n",
    "For this we can use *note_seq* with the function [midi_file_to_drum_track](https://github.com/magenta/note-seq/blob/6631a4299092cd52201e2f4a8e36c1223d76561f/note_seq/drums_lib.py#L269)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "\n",
    "We filtered out all noisy examples of our dataset and transformed the drum track into a mathematical, simplified notation.\n",
    "\n",
    "In the next step we will start to analyse the drum patterns that we extracted."
   ]
  }
 ],
 "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}