data/vel_norm.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Velocity normalization\n",
    "\n",
    "We normalize velocities in the Bodhidharma dataset to get rid of a mismatch with respect to the training distribution.\n",
    "\n",
    "The histogram plotting takes a bit long and may be skipped."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import itertools\n",
    "import os\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from museflow.io.note_sequence_io import NoteSequenceDB\n",
    "from museflow.note_sequence_utils import filter_sequence\n",
    "import numpy as np\n",
    "import scipy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATASET_DIR = 'bodhidharma/final'\n",
    "TRAIN_DB = 'synth/train/final/all.db'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compute velocity statistics on training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "velocities_train = []\n",
    "with NoteSequenceDB(TRAIN_DB) as db, db.begin() as txn:\n",
    "    for _, seq in txn.items():\n",
    "        velocities_train.extend(n.velocity for n in seq.notes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DescribeResult(nobs=71799353, minmax=(1, 127), mean=64.25399574283072, variance=432.9165195560093, skewness=0.19585816884222212, kurtosis=-0.06955028101708605)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_stats = scipy.stats.describe(velocities_train)\n",
    "train_stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(velocities_train, bins=np.linspace(0, 128, 129))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compute the same statistics on the Bodhidharma dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "velocities_bodh = []\n",
    "with NoteSequenceDB(os.path.join(DATASET_DIR, 'all.db')) as db, db.begin() as txn:\n",
    "    for _, seq in txn.items():\n",
    "        velocities_bodh.extend(n.velocity for n in seq.notes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DescribeResult(nobs=3270166, minmax=(1, 127), mean=91.25231685486303, variance=637.4622812064066, skewness=-0.4970251410175147, kurtosis=-0.23826480021904972)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bodh_stats = scipy.stats.describe(velocities_bodh)\n",
    "bodh_stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(velocities_bodh, bins=np.linspace(0, 128, 129))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Normalize velocities and check the new statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.makedirs(os.path.join(DATASET_DIR, 'vel_norm_biab'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "velocities_norm = []\n",
    "with NoteSequenceDB(os.path.join(DATASET_DIR, 'all.db')) as db, db.begin() as txn, \\\n",
    "     NoteSequenceDB(os.path.join(DATASET_DIR, 'vel_norm_biab', 'all.db'), write=True,\n",
    "                    map_size=2 * db.db.info()['map_size']) as db_out, db_out.begin() as txn_out:\n",
    "    for key, seq in txn.items():\n",
    "        seq_velocities = np.array([n.velocity for n in seq.notes])\n",
    "        seq_velocities = seq_velocities - np.mean(seq_velocities)\n",
    "        if np.std(seq_velocities) > 0:\n",
    "            seq_velocities *= np.sqrt(train_stats.variance) / np.std(seq_velocities)\n",
    "        seq_velocities += train_stats.mean\n",
    "        seq_velocities = np.rint(seq_velocities).astype(np.int32)\n",
    "        np.clip(seq_velocities, 1, 127, out=seq_velocities)\n",
    "        for note, velocity in zip(seq.notes, seq_velocities):\n",
    "            note.velocity = velocity\n",
    "        txn_out.put(key, seq)\n",
    "        velocities_norm.extend(seq_velocities)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DescribeResult(nobs=3270166, minmax=(1, 127), mean=64.28374706360472, variance=389.6695570711818, skewness=-0.29765833233658223, kurtosis=0.3096898322707906)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scipy.stats.describe(velocities_norm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(velocities_train, density=True, bins=np.linspace(0, 128, 129), alpha=0.7)\n",
    "plt.hist(velocities_norm, density=True, bins=np.linspace(0, 128, 129), alpha=0.7)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(velocities_train, density=True, bins=np.linspace(0, 128, 9), alpha=0.7)\n",
    "plt.hist(velocities_norm, density=True, bins=np.linspace(0, 128, 9), alpha=0.7)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Filter out drums"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "with NoteSequenceDB(os.path.join(DATASET_DIR, 'vel_norm_biab', 'all.db')) as db, db.begin() as txn, \\\n",
    "     NoteSequenceDB(os.path.join(DATASET_DIR, 'vel_norm_biab', 'all_except_drums.db'), write=True,\n",
    "                    map_size=2 * db.db.info()['map_size']) as db_out, db_out.begin() as txn_out:\n",
    "    for key, seq in txn.items():\n",
    "        filter_sequence(seq, drums=False)\n",
    "        txn_out.put(key, seq)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}