coati_interactive_prep.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "import plotly.graph_objs as go\n",
    "import plotly\n",
    "import os\n",
    "import seaborn as sns\n",
    "from plotly.graph_objs import *\n",
    "from ipywidgets import Image, HTML, Layout, Output, VBox, HBox\n",
    "from IPython.display import Audio, display"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import imageio\n",
    "#import PIL\n",
    "from PIL import Image, ImageEnhance, ImageFilter\n",
    "#!pip install librosa==0.7.1\n",
    "import librosa\n",
    "import librosa.display\n",
    "import pylab"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "PROJECT_HOME = '/Users/marathomas/Documents/Bioinformatik/Masterarbeit/Unsup_clustering/coati_viz/'\n",
    "IMAGES_FOLDER = PROJECT_HOME+'spec_images/'\n",
    "DATA_FOLDER = PROJECT_HOME+'data/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data/ds1.json\n",
      "data/ds2.json\n",
      "data/ds3.json\n"
     ]
    }
   ],
   "source": [
    "os.chdir(PROJECT_HOME)\n",
    "\n",
    "df_list=[]\n",
    "\n",
    "for i in range(1,4):\n",
    "    check = pd.read_json('data/ds'+str(i)+'.json', orient ='split', compression = 'infer')\n",
    "    print('data/ds'+str(i)+'.json')\n",
    "    df_list.append(check)\n",
    "\n",
    "df = pd.concat(df_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2033, 13)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>Name</th>\n",
       "      <th>Start</th>\n",
       "      <th>samplerate_hz</th>\n",
       "      <th>duration_s</th>\n",
       "      <th>start_s</th>\n",
       "      <th>raw_audio</th>\n",
       "      <th>denoised_spectrograms</th>\n",
       "      <th>indv</th>\n",
       "      <th>label</th>\n",
       "      <th>index_id</th>\n",
       "      <th>UMAP1</th>\n",
       "      <th>UMAP2</th>\n",
       "      <th>UMAP3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>109</td>\n",
       "      <td>grunt</td>\n",
       "      <td>2:48:58.745</td>\n",
       "      <td>22050</td>\n",
       "      <td>0.080</td>\n",
       "      <td>10138.745</td>\n",
       "      <td>[-0.00390625, -0.00390625, -0.005859375, -0.00...</td>\n",
       "      <td>[[21.1363868713, 20.6111793518, 23.7496910095,...</td>\n",
       "      <td>6887</td>\n",
       "      <td>grunt</td>\n",
       "      <td>109</td>\n",
       "      <td>9.670739</td>\n",
       "      <td>4.418707</td>\n",
       "      <td>2.416416</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>110</td>\n",
       "      <td>grunt</td>\n",
       "      <td>3:18:50.647</td>\n",
       "      <td>22050</td>\n",
       "      <td>0.031</td>\n",
       "      <td>11930.647</td>\n",
       "      <td>[0.013671875, 0.01171875, 0.0078125, 0.015625,...</td>\n",
       "      <td>[[9.191781044, 7.8991680145, 5.7343931198, 2.7...</td>\n",
       "      <td>6887</td>\n",
       "      <td>grunt</td>\n",
       "      <td>110</td>\n",
       "      <td>11.735151</td>\n",
       "      <td>5.220063</td>\n",
       "      <td>4.854806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>609</td>\n",
       "      <td>chirp</td>\n",
       "      <td>5:48.635</td>\n",
       "      <td>22050</td>\n",
       "      <td>0.079</td>\n",
       "      <td>348.635</td>\n",
       "      <td>[0.017578125, 0.017578125, 0.005859375, 0.0136...</td>\n",
       "      <td>[[16.8022613525, 15.9717550278, 14.6228752136,...</td>\n",
       "      <td>6887</td>\n",
       "      <td>chirp</td>\n",
       "      <td>609</td>\n",
       "      <td>9.091060</td>\n",
       "      <td>4.764462</td>\n",
       "      <td>2.464877</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>610</td>\n",
       "      <td>chirp</td>\n",
       "      <td>5:53.450</td>\n",
       "      <td>22050</td>\n",
       "      <td>0.079</td>\n",
       "      <td>353.450</td>\n",
       "      <td>[-0.0078125, -0.02734375, 0.0078125, -0.041015...</td>\n",
       "      <td>[[5.6005115509, 6.254076004, 6.4096355438, 5.6...</td>\n",
       "      <td>6887</td>\n",
       "      <td>chirp</td>\n",
       "      <td>610</td>\n",
       "      <td>9.580486</td>\n",
       "      <td>5.062622</td>\n",
       "      <td>3.355001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>611</td>\n",
       "      <td>chirp</td>\n",
       "      <td>5:56.069</td>\n",
       "      <td>22050</td>\n",
       "      <td>0.135</td>\n",
       "      <td>356.069</td>\n",
       "      <td>[-0.044921875, -0.029296875000000003, -0.02148...</td>\n",
       "      <td>[[15.1517496109, 11.6087598801, 8.7584705353, ...</td>\n",
       "      <td>6887</td>\n",
       "      <td>chirp</td>\n",
       "      <td>611</td>\n",
       "      <td>10.634771</td>\n",
       "      <td>4.980813</td>\n",
       "      <td>5.306195</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      Name        Start  samplerate_hz  duration_s    start_s  \\\n",
       "109  grunt  2:48:58.745          22050       0.080  10138.745   \n",
       "110  grunt  3:18:50.647          22050       0.031  11930.647   \n",
       "609  chirp     5:48.635          22050       0.079    348.635   \n",
       "610  chirp     5:53.450          22050       0.079    353.450   \n",
       "611  chirp     5:56.069          22050       0.135    356.069   \n",
       "\n",
       "                                             raw_audio  \\\n",
       "109  [-0.00390625, -0.00390625, -0.005859375, -0.00...   \n",
       "110  [0.013671875, 0.01171875, 0.0078125, 0.015625,...   \n",
       "609  [0.017578125, 0.017578125, 0.005859375, 0.0136...   \n",
       "610  [-0.0078125, -0.02734375, 0.0078125, -0.041015...   \n",
       "611  [-0.044921875, -0.029296875000000003, -0.02148...   \n",
       "\n",
       "                                 denoised_spectrograms  indv  label  index_id  \\\n",
       "109  [[21.1363868713, 20.6111793518, 23.7496910095,...  6887  grunt       109   \n",
       "110  [[9.191781044, 7.8991680145, 5.7343931198, 2.7...  6887  grunt       110   \n",
       "609  [[16.8022613525, 15.9717550278, 14.6228752136,...  6887  chirp       609   \n",
       "610  [[5.6005115509, 6.254076004, 6.4096355438, 5.6...  6887  chirp       610   \n",
       "611  [[15.1517496109, 11.6087598801, 8.7584705353, ...  6887  chirp       611   \n",
       "\n",
       "         UMAP1     UMAP2     UMAP3  \n",
       "109   9.670739  4.418707  2.416416  \n",
       "110  11.735151  5.220063  4.854806  \n",
       "609   9.091060  4.764462  2.464877  \n",
       "610   9.580486  5.062622  3.355001  \n",
       "611  10.634771  4.980813  5.306195  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGES_FOLDER = 'spec_images/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "SR = 22050\n",
    "FFT_WIN = 0.03 # FFT_WIN*samplerate = length of fft/n_fft (number of audio frames that go in one fft)\n",
    "FFT_HOP = FFT_WIN/8 # FFT_HOP*samplerate = n of audio frames between successive ffts\n",
    "FMAX = SR/2\n",
    "\n",
    "def make_img(spec,id):\n",
    "    outname = str(id)+\".jpg\"    \n",
    "    librosa.display.specshow(spec,sr=SR, hop_length=int(FFT_HOP * SR) , fmax=FMAX, y_axis='mel', x_axis='s',cmap='inferno')\n",
    "    pylab.savefig(outname, bbox_inches=None, pad_inches=0)\n",
    "    pylab.close()\n",
    "    #plt.imshow(spec, interpolation='nearest', aspect='equal')\n",
    "    #plt.savefig(outname)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pad_spectro(spec,maxlen):\n",
    "    padding = maxlen - spec.shape[1]\n",
    "    z = np.zeros((spec.shape[0],padding))\n",
    "    padded_spec=np.append(spec, z, axis=1)\n",
    "    return padded_spec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "specs = [np.asarray(x) for x in df.denoised_spectrograms.values]\n",
    "maxlen= np.max([spec.shape[1] for spec in specs])\n",
    "padded_specs = [pad_spectro(spec, maxlen) for spec in specs]\n",
    "df['padded_specs'] = padded_specs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "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": [
    "#os.chdir(IMAGES_FOLDER)\n",
    "spec = np.asarray(df.iloc[2,6])\n",
    "#spec = np.asarray(df.iloc[2,13])  # padded\n",
    "librosa.display.specshow(spec,sr=SR, hop_length=int(FFT_HOP * SR) , fmax=FMAX, y_axis='mel', x_axis='s',cmap='inferno')\n",
    "pylab.savefig('example.jpg', bbox_inches=None, pad_inches=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.chdir(IMAGES_FOLDER)\n",
    "x = df.apply(lambda row: make_img(np.asarray(row['denoised_spectrograms']),row['index_id']), axis=1)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}