DataAugmentation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Augmentation "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import modules and implementation of data augmentation functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import librosa\n",
    "import csv\n",
    "from sklearn.model_selection import train_test_split\n",
    "import tensorflow as tf\n",
    "from pandarallel import pandarallel\n",
    "import numpy as np\n",
    "import cv2\n",
    "import h5py\n",
    "\n",
    "pd.options.mode.chained_assignment = None  # default='warn'\n",
    "\n",
    "def load_soundfile_from_drive(filename, path='public_dataset_cleaned_segmented//'):\n",
    "    x , sr = librosa.load(path + filename)\n",
    "    return x, sr\n",
    "\n",
    "def addNoise(data):\n",
    "    noise_factor = 0.05*np.random.rand()\n",
    "    sampling_rate = 22050\n",
    "    noise = np.random.randn(len(data))\n",
    "    augmented_data = data + noise_factor * noise\n",
    "    # Cast back to same data type\n",
    "    augmented_data = augmented_data.astype(type(data[0]))\n",
    "    return augmented_data\n",
    "\n",
    "def shiftTime(data):\n",
    "    sampling_rate = 22050\n",
    "    shift_max = 8\n",
    "    shift = np.random.randint(sampling_rate * shift_max)\n",
    "    direction = np.random.randint(0, 2)\n",
    "    if direction == 1:\n",
    "        shift = -shift\n",
    "    augmented_data = np.roll(data, shift)\n",
    "    # Set to silence for heading/ tailing\n",
    "    if shift > 0:\n",
    "        augmented_data[:shift] = 0\n",
    "    else:\n",
    "        augmented_data[shift:] = 0\n",
    "    return augmented_data\n",
    "\n",
    "def changeSpeed(data):\n",
    "    sampling_rate = 22050\n",
    "    speed_factor = 1 + 0.4*(np.random.rand() - 0.5)\n",
    "    augmented_data = librosa.effects.time_stretch(data, speed_factor)\n",
    "    return augmented_data\n",
    "\n",
    "def changePitch(data):\n",
    "    pitch_factor = 1 + 0.4 * (np.random.rand() - 0.5)\n",
    "    sampling_rate = 22050\n",
    "    augmented_data = librosa.effects.pitch_shift(data, sampling_rate, pitch_factor)\n",
    "    return augmented_data\n",
    "\n",
    "def generateMelSpectrogram(data):\n",
    "    x = data\n",
    "    sr = 22050\n",
    "    M = librosa.feature.melspectrogram(y=x, sr=sr, n_mels=128)\n",
    "    M_db = librosa.power_to_db(M, ref=np.max) * -1\n",
    "    dim = (320,240) # resize to dimensions from https://www.sciencedirect.com/science/article/pii/S2352914820303026\n",
    "    resized = cv2.resize(M_db, dim, interpolation=cv2.INTER_LINEAR)\n",
    "    return resized\n",
    "\n",
    "def expandDimensions(data):\n",
    "    return np.expand_dims(data, axis=2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Use of a pandas extension for parallel processing in order to decrease waiting times"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO: Pandarallel will run on 8 workers.\n",
      "INFO: Pandarallel will use standard multiprocessing data transfer (pipe) to transfer data between the main process and workers.\n"
     ]
    }
   ],
   "source": [
    "pandarallel.initialize(use_memory_fs=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reading in data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv(\"public_dataset_cleaned_segmented//labels.csv\", names=[\"soundfilename\",\"status\",\"cough_value\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data.dropna(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data[\"sounddata\"] = data[\"soundfilename\"].apply(lambda x: load_soundfile_from_drive(x,\"public_dataset_cleaned_segmented//\")[0]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>soundfilename</th>\n",
       "      <th>status</th>\n",
       "      <th>cough_value</th>\n",
       "      <th>sounddata</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>28632a0e-f38e-4195-a566-1cc3ad08e5b90.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9967</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>28632a0e-f38e-4195-a566-1cc3ad08e5b91.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9967</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>28632a0e-f38e-4195-a566-1cc3ad08e5b92.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9967</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>082991fb-ee66-4970-91dd-50890f4cc9b10.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9431</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>082991fb-ee66-4970-91dd-50890f4cc9b11.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9431</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46970</th>\n",
       "      <td>e30be719-4bdc-46c5-b515-7549d06f12a20.wav</td>\n",
       "      <td>symptomatic</td>\n",
       "      <td>0.9984</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46972</th>\n",
       "      <td>e30be719-4bdc-46c5-b515-7549d06f12a21.wav</td>\n",
       "      <td>symptomatic</td>\n",
       "      <td>0.9984</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46974</th>\n",
       "      <td>e30be719-4bdc-46c5-b515-7549d06f12a22.wav</td>\n",
       "      <td>symptomatic</td>\n",
       "      <td>0.9984</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46976</th>\n",
       "      <td>e3157d31-7cc6-4a85-a041-793bbe4cc0ce0.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9890</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46978</th>\n",
       "      <td>e3157d31-7cc6-4a85-a041-793bbe4cc0ce1.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9890</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>23490 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   soundfilename       status  cough_value  \\\n",
       "0      28632a0e-f38e-4195-a566-1cc3ad08e5b90.wav      healthy       0.9967   \n",
       "2      28632a0e-f38e-4195-a566-1cc3ad08e5b91.wav      healthy       0.9967   \n",
       "4      28632a0e-f38e-4195-a566-1cc3ad08e5b92.wav      healthy       0.9967   \n",
       "6      082991fb-ee66-4970-91dd-50890f4cc9b10.wav     COVID-19       0.9431   \n",
       "8      082991fb-ee66-4970-91dd-50890f4cc9b11.wav     COVID-19       0.9431   \n",
       "...                                          ...          ...          ...   \n",
       "46970  e30be719-4bdc-46c5-b515-7549d06f12a20.wav  symptomatic       0.9984   \n",
       "46972  e30be719-4bdc-46c5-b515-7549d06f12a21.wav  symptomatic       0.9984   \n",
       "46974  e30be719-4bdc-46c5-b515-7549d06f12a22.wav  symptomatic       0.9984   \n",
       "46976  e3157d31-7cc6-4a85-a041-793bbe4cc0ce0.wav      healthy       0.9890   \n",
       "46978  e3157d31-7cc6-4a85-a041-793bbe4cc0ce1.wav      healthy       0.9890   \n",
       "\n",
       "                                               sounddata  \n",
       "0      [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2      [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "4      [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "6      [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "8      [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "...                                                  ...  \n",
       "46970  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "46972  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "46974  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "46976  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "46978  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "\n",
       "[23490 rows x 4 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Covid positive test data:  582\n",
      "Covid negative test data:  2156\n"
     ]
    }
   ],
   "source": [
    "pos_data = data[data[\"status\"]==\"COVID-19\"]\n",
    "neg_data = data[data[\"status\"]!=\"COVID-19\"]\n",
    "\n",
    "pos_data_train, pos_data_test = train_test_split(pos_data, test_size=0.3)\n",
    "print(\"Covid positive test data: \",len(pos_data_test))\n",
    "neg_data_train, neg_data_test = train_test_split(neg_data, test_size=0.1)\n",
    "print(\"Covid negative test data: \",len(neg_data_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data augmentation part"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos_data_train_noised = pos_data_train\n",
    "pos_data_train_noised[\"sounddata\"] = pos_data_train_noised[\"sounddata\"].parallel_apply(lambda x: addNoise(x))\n",
    "#pos_data_train_noised[\"augmentation\"] = \"Noise\"\n",
    "pos_data_train_aug = pos_data_train.append(pos_data_train_noised)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos_data_train_aug = pos_data_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos_data_train_pitched = pos_data_train_aug\n",
    "pos_data_train_pitched[\"sounddata\"] = pos_data_train_pitched[\"sounddata\"].parallel_apply(lambda x: changePitch(x))\n",
    "#pos_data_train_pitched[\"augmentation\"] = \"pitch\"\n",
    "pos_data_train_aug = pos_data_train_aug.append(pos_data_train_pitched)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos_data_train_shifted = pos_data_train_aug\n",
    "pos_data_train_shifted[\"sounddata\"] = pos_data_train_shifted[\"sounddata\"].parallel_apply(lambda x: shiftTime(x))\n",
    "#pos_data_train_shifted[\"augmentation\"] = \"shift\"\n",
    "pos_data_train_aug = pos_data_train_aug.append(pos_data_train_shifted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "pos_data_train_speed = pos_data_train_aug\n",
    "pos_data_train_speed[\"sounddata\"] = pos_data_train_speed[\"sounddata\"].parallel_apply(lambda x: changeSpeed(x))\n",
    "#pos_data_train_speed[\"augmentation\"] = \"speed\"\n",
    "pos_data_train_aug = pos_data_train_aug.append(pos_data_train_speed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>soundfilename</th>\n",
       "      <th>status</th>\n",
       "      <th>cough_value</th>\n",
       "      <th>sounddata</th>\n",
       "      <th>augmentation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>a7758e1f-e7b5-42cc-a66d-99966ea8a44e1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.8515</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>7f8e60f4-5cbd-4ec6-bc73-e8ab2b6c3abd0.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9958</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3be23405-2584-49fd-8f74-a22fbeeea0440.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.0511</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2c38b55e-7a52-4b23-a4a5-f0da83cd2cba1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9377</td>\n",
       "      <td>[-0.00036621094, 0.0006713867, -0.0040893555, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>95e40d65-5398-4dd8-bda8-5da98d06ed800.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.4218</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2705</th>\n",
       "      <td>ab228702-1c84-4514-8ea9-7a207b1919c91.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9911</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2706</th>\n",
       "      <td>41a9aa47-9814-4a0f-a0a3-f3d919bb893d1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.1800</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2707</th>\n",
       "      <td>0ed846f2-5ce3-4f7f-9f60-b0258ab2047e1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.0352</td>\n",
       "      <td>[0.0017089844, 0.0062561035, 0.009246826, 0.01...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2708</th>\n",
       "      <td>36e4f76f-ebb6-45e4-b246-64c8fbb9e7db2.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.5202</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2709</th>\n",
       "      <td>b3add103-8854-4aba-8448-18361df4dac24.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9869</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>shift</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2710 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  soundfilename    status  cough_value  \\\n",
       "0     a7758e1f-e7b5-42cc-a66d-99966ea8a44e1.wav  COVID-19       0.8515   \n",
       "1     7f8e60f4-5cbd-4ec6-bc73-e8ab2b6c3abd0.wav  COVID-19       0.9958   \n",
       "2     3be23405-2584-49fd-8f74-a22fbeeea0440.wav  COVID-19       0.0511   \n",
       "3     2c38b55e-7a52-4b23-a4a5-f0da83cd2cba1.wav  COVID-19       0.9377   \n",
       "4     95e40d65-5398-4dd8-bda8-5da98d06ed800.wav  COVID-19       0.4218   \n",
       "...                                         ...       ...          ...   \n",
       "2705  ab228702-1c84-4514-8ea9-7a207b1919c91.wav  COVID-19       0.9911   \n",
       "2706  41a9aa47-9814-4a0f-a0a3-f3d919bb893d1.wav  COVID-19       0.1800   \n",
       "2707  0ed846f2-5ce3-4f7f-9f60-b0258ab2047e1.wav  COVID-19       0.0352   \n",
       "2708  36e4f76f-ebb6-45e4-b246-64c8fbb9e7db2.wav  COVID-19       0.5202   \n",
       "2709  b3add103-8854-4aba-8448-18361df4dac24.wav  COVID-19       0.9869   \n",
       "\n",
       "                                              sounddata augmentation  \n",
       "0     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "1     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "2     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "3     [-0.00036621094, 0.0006713867, -0.0040893555, ...        shift  \n",
       "4     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "...                                                 ...          ...  \n",
       "2705  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "2706  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "2707  [0.0017089844, 0.0062561035, 0.009246826, 0.01...        shift  \n",
       "2708  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "2709  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...        shift  \n",
       "\n",
       "[2710 rows x 5 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pos_data_train_aug.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set = pos_data_train_aug.append(neg_data_train, ignore_index=True)\n",
    "test_set = pos_data_test.append(neg_data_test, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>soundfilename</th>\n",
       "      <th>status</th>\n",
       "      <th>cough_value</th>\n",
       "      <th>sounddata</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>e136f725-f618-40e9-82e9-d4712a5479dd1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.7330</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>d5b05c83-3dab-4fc9-8db7-9f46bcb06cac1.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.7904</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>920ae37e-af2e-4e43-a80e-2150a66db6b10.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9949</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>e8537e2a-98b1-4ddd-8237-3835215218b43.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9956</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>814a5c9e-6174-4efa-a378-a990ac566ae21.wav</td>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9930</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2733</th>\n",
       "      <td>dd3e8166-0375-4882-b037-add3a813f84f1.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.8056</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2734</th>\n",
       "      <td>11324b9a-7584-42eb-95af-fbb39dcfc1054.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9949</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2735</th>\n",
       "      <td>ff18f5dd-fc75-4f22-be2e-374eb3e2372d4.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.4838</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2736</th>\n",
       "      <td>b17e7db8-041b-40e9-81f3-8771bafc010a4.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.6983</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2737</th>\n",
       "      <td>ce29aef9-0a86-42e6-ab49-258b8a6b1a5e0.wav</td>\n",
       "      <td>healthy</td>\n",
       "      <td>0.8687</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2738 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  soundfilename    status  cough_value  \\\n",
       "0     e136f725-f618-40e9-82e9-d4712a5479dd1.wav  COVID-19       0.7330   \n",
       "1     d5b05c83-3dab-4fc9-8db7-9f46bcb06cac1.wav  COVID-19       0.7904   \n",
       "2     920ae37e-af2e-4e43-a80e-2150a66db6b10.wav  COVID-19       0.9949   \n",
       "3     e8537e2a-98b1-4ddd-8237-3835215218b43.wav  COVID-19       0.9956   \n",
       "4     814a5c9e-6174-4efa-a378-a990ac566ae21.wav  COVID-19       0.9930   \n",
       "...                                         ...       ...          ...   \n",
       "2733  dd3e8166-0375-4882-b037-add3a813f84f1.wav   healthy       0.8056   \n",
       "2734  11324b9a-7584-42eb-95af-fbb39dcfc1054.wav   healthy       0.9949   \n",
       "2735  ff18f5dd-fc75-4f22-be2e-374eb3e2372d4.wav   healthy       0.4838   \n",
       "2736  b17e7db8-041b-40e9-81f3-8771bafc010a4.wav   healthy       0.6983   \n",
       "2737  ce29aef9-0a86-42e6-ab49-258b8a6b1a5e0.wav   healthy       0.8687   \n",
       "\n",
       "                                              sounddata  \n",
       "0     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "1     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "3     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "4     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "...                                                 ...  \n",
       "2733  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2734  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2735  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2736  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2737  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "\n",
       "[2738 rows x 4 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_set.reset_index(drop=True)\n",
    "test_set.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generating mel spectrograms from augmented sound data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set[\"Melspecs\"] = train_set[\"sounddata\"].parallel_apply(lambda x: generateMelSpectrogram(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set[\"Melspecs\"] = test_set[\"sounddata\"].parallel_apply(lambda x: generateMelSpectrogram(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set[\"Melspecs\"] = train_set[\"Melspecs\"].parallel_apply(lambda x: expandDimensions(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set[\"Melspecs\"] = test_set[\"Melspecs\"].parallel_apply(lambda x: expandDimensions(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set.drop([\"sounddata\",\"soundfilename\",\"augmentation\"],axis = 1, inplace=True)\n",
    "test_set.drop([\"sounddata\",\"soundfilename\",\"augmentation\"],axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>status</th>\n",
       "      <th>cough_value</th>\n",
       "      <th>augmentation</th>\n",
       "      <th>Melspecs</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.8515</td>\n",
       "      <td>shift</td>\n",
       "      <td>[[[0.0], [0.0], [0.0], [0.0], [0.0], [0.0], [0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9958</td>\n",
       "      <td>shift</td>\n",
       "      <td>[[[0.0], [0.0], [0.0], [0.0], [0.0], [0.0], [0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.0511</td>\n",
       "      <td>shift</td>\n",
       "      <td>[[[72.59998], [72.59998], [72.59998], [72.5999...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.9377</td>\n",
       "      <td>shift</td>\n",
       "      <td>[[[7.0643253], [7.7245693], [10.176903], [12.6...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>COVID-19</td>\n",
       "      <td>0.4218</td>\n",
       "      <td>shift</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22102</th>\n",
       "      <td>symptomatic</td>\n",
       "      <td>0.9926</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22103</th>\n",
       "      <td>healthy</td>\n",
       "      <td>0.3102</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22104</th>\n",
       "      <td>healthy</td>\n",
       "      <td>0.4330</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22105</th>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9961</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22106</th>\n",
       "      <td>healthy</td>\n",
       "      <td>0.9705</td>\n",
       "      <td>NaN</td>\n",
       "      <td>[[[80.0], [80.0], [80.0], [80.0], [80.0], [80....</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>22107 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            status  cough_value augmentation  \\\n",
       "0         COVID-19       0.8515        shift   \n",
       "1         COVID-19       0.9958        shift   \n",
       "2         COVID-19       0.0511        shift   \n",
       "3         COVID-19       0.9377        shift   \n",
       "4         COVID-19       0.4218        shift   \n",
       "...            ...          ...          ...   \n",
       "22102  symptomatic       0.9926          NaN   \n",
       "22103      healthy       0.3102          NaN   \n",
       "22104      healthy       0.4330          NaN   \n",
       "22105      healthy       0.9961          NaN   \n",
       "22106      healthy       0.9705          NaN   \n",
       "\n",
       "                                                Melspecs  \n",
       "0      [[[0.0], [0.0], [0.0], [0.0], [0.0], [0.0], [0...  \n",
       "1      [[[0.0], [0.0], [0.0], [0.0], [0.0], [0.0], [0...  \n",
       "2      [[[72.59998], [72.59998], [72.59998], [72.5999...  \n",
       "3      [[[7.0643253], [7.7245693], [10.176903], [12.6...  \n",
       "4      [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "...                                                  ...  \n",
       "22102  [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "22103  [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "22104  [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "22105  [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "22106  [[[80.0], [80.0], [80.0], [80.0], [80.0], [80....  \n",
       "\n",
       "[22107 rows x 4 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save datasets to pickle file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set.to_pickle(\"trainshift.pkl\")\n",
    "test_set.to_pickle(\"testshift.pkl\")"
   ]
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}