Breathing_model_implementation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(12, 156)\n"
     ]
    }
   ],
   "source": [
    "# THis is just for visualization\n",
    "\n",
    "import os \n",
    "import librosa\n",
    "import math\n",
    "import json\n",
    "import librosa.display\n",
    "import numpy as np\n",
    "\n",
    "path = \"./breath/pos_new/0pos-0426-126-breathe-m-46.wav\"\n",
    "sr = 22050\n",
    "\n",
    "\n",
    "# Zero crossing rate\n",
    "y, sr = librosa.load(path, sr = sr)\n",
    "zc = librosa.feature.chroma_stft(y=y, sr=sr,n_chroma=12, n_fft=4096)\n",
    "\n",
    "# mfcc = librosa.feature.mfcc(y,\n",
    "#                             sr =sr,\n",
    "#                             n_fft = 2048,\n",
    "#                             n_mfcc = 13,\n",
    "#                             hop_length = 512\n",
    "#                             )\n",
    "                    \n",
    "features = librosa.feature.chroma_cqt(y=y, sr=sr)\n",
    "\n",
    "# pitches, magnitudes = librosa.piptrack(y=y, sr=sr)\n",
    "# rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr, roll_percent=0.1)\n",
    "\n",
    "# features = np.concatenate((mfcc, chroma_cq, pitches, rolloff), axis = 0)\n",
    "\n",
    "# features = features.T\n",
    "# zc = librosa.feature.zero_crossing_rate(y)\n",
    "\n",
    "# print(spectral_entropy(zc, 100, method='welch', normalize=True))\n",
    "\n",
    "# librosa.display.waveplot(zc[0], sr)\n",
    "\n",
    "# print(mfcc.shape)\n",
    "# print(chroma_cq.shape)\n",
    "# print(pitches.shape)\n",
    "# print(rolloff.shape)\n",
    "print(features.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAoYAAAEYCAYAAADBDjfXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3deZicVZn+8fup7k53Ouns+wIJkLCviUQIuyAgyjKiAi6gjMw48hsdhREdBx1GXHAdFQczAyM7ggKCJKDsO7KDCRCSkD1k6SSddNJr1fP7o6qxk3qquwJdS3e+n+uqK913nT7vqXerk3c5r7m7AAAAgESpGwAAAIDyQMcQAAAAkugYAgAAIIOOIQAAACTRMQQAAEAGHUMAAABIomMI4D0ys2+b2Q2lbgcA4L2jYwggL2Z2jpk9Z2aNZrbKzOaY2RGlbhcAoOfQMQTQLTP7iqSfSfqupNGSdpH0K0mn7WA9lT3fup5Rzm0DgGKhYwigS2Y2WNJlkr7o7re7+xZ3b3P3u9394kyxfmZ2nZltNrO5Zja9098vNrOvmdkrkraYWaWZnZopt9HMHjazvbcrf7GZvWJmW8zsajMbnTlCudnM7jezoZ3K32Zmb5tZg5k9amb7dnpvuJndbWabzOxZM/uOmT3e6X03sy+a2ZuS3sxk/2VmyzJ/87yZHdmp/Lcz07sh05ZXzWyqmX3dzNZk/u6DBVgMAFAUdAwBdOcwSTWS7uiizKmSbpE0RNJdkn653ftnSzol8/5ukm6W9GVJIyXNlnS3mfXrVP6jkk6QNFXSRyTNkfSNTPmEpH/uVHaOpCmSRkl6QdKNnd67UtIWSWMknZt5be90STMk7ZP5/VlJB0kaJukmSbeZWU2n8h+RdL2koZJelHRfpk3jle5A/zqYBgD0CnQMAXRnuKR17t7eRZnH3X22uyeV7jQduN37P3f3Ze7eJOkTku5x9z+7e5ukH0nqL+nwTuV/4e6r3X2FpMckPePuL7p7s9Id1IM7Crr7Ne6+2d1bJH1b0oFmNtjMKpTuYH7L3be6+zxJ1wZt/567r8+0Te5+g7vXu3u7u/9YUrWkPTuVf8zd78vMj9uU7qx+P/NZbpE0ycyGdDGvAKBs0TEE0J16SSO6uQbv7U4/b5VUs135ZZ1+HidpSccv7p7KvD++U5nVnX5uCn4fKElmVmFm3zezhWa2SdLiTJkRSnfYKrebduefw8zMLjKz1zKnpjdKGpypL1fb1mU6xB2/q6N9ANDb0DEE0J2nJLUofcr13fJOP6+UtGvHL2ZmkiZKWvEu6j1H6Rtgjle6Azepo1pJayW1S5rQqfzErtqWuZ7wXyV9XNJQdx8iqSFTHwD0eXQMAXTJ3RskXSrpSjM73cxqzazKzE42syveRZW3SjrFzD5gZlWSvqp0x/PJd1FXXeZv6yXVKn3XdEe7k5Jul/TtTJv3kvSZPOprV7pTWWlml0oa9C7aBQC9Eh1DAN3KXGv3FUnfVLrTtEzShZLufBd1vSHpU5J+IWmd0jdzfMTdW99F065T+rT0CknzJD293fsXKn0k8W2lr328WemOZC73SbpX0vxMvc2KTz8DQJ9k7t59KQDoA8zsB5LGuHt0dzIA7PQ4YgigzzKzvczsAEs7VNL56nrYHQDYqdExBNCX1Sl9neEWSb+V9GNJfyhpiwCgB5jZNZmB9f+a430zs5+b2YLMAwMOyateTiUDAAD0LmZ2lKRGSde5+37B+x+S9P8kfUjpQfz/y91ndFcvRwwBAAB6GXd/VNL6LoqcpnSn0d39aUlDzGxsd/WW3UPj+1cM8MGV2z40YHRtfBNhe7IizlPxkGMD+reFeWNTvzAfNLk6zP/62qYwr1J2PVtS9WHZykT/MM81XFqV4rZsTa7LUU88bwZWDA/zVjXHebIxu+ZEPL9S74zxu63hlfE01yc3hLnl+P9K1TZPJfubpOKbWWtzjDHcHNyUOjAR1z2mf/ywj+Vb4vk7vDpefvUt8ZH5uqr4s25uS4X5plT2PsAsrsNyrEuJHONUJ3KsM20erxv9LR7FpSLH8hveL27P0EFd3SS8rYbN8bq3Nblj/8etq4znb92IeB1WrhMrNVU5JhCsexsa4rJDB8f5mnh/P39lvJxac2wHg3Lsa3KsetrSHs+bhGUvv02p+DPl2larg32kJI2uiWfwwLp4n72+Id4fVlpcT3Mq/+2spiIuW12RYxuujPcRiURcvrU9Xn5Jj7eP1hz5yD3ieax+wTq5aXNYtG1NPH8ra+O2r1wfz/f4E0kN7fG8aVM83f12jdeP5MbselY2xm1pScbrryQ1JN9e5+4jcxYoghNPPNTr63PsC7bz/PPz50rbfDnPcvdZOzC58dp2VIXlmWxVV39Udh3DwZVD9Klx/7BNdvG0JWHZNRvqwnx9c7zCTN9/ZZg/+fKEMP/AdXuE+T7vvz/Mxyaz63ly6w1h2RED9gnzCou/aEb7bmH+fMPVcT0V8Zf2tAGfDPMliTfCfNmmp7KyuprxQUmppT3e+Xx0xHlhfvPG28K8KscX2ejKvcJ8g8ejiUyzmWH+pi3KymZWx3V/bf+1Yf7VZ+Mv83N3i79Url8U72iPGRvvCB9cFXeWHth6U1ZWWRHPr8ocHfgBlfF+sSYRf6bVzfPC/ICqE8N8sMXt+eTkeN3+u2Ozl0eOvq7ufWTXMH9hQ7zN53LUyK1hfuzn4nVYOb5sbM94W0gdlb3uJe6YE5c94+S47itvDvMT/z1+2t6SirfC/ITarDNMkqRxtWGsp9fG615tRfbX/wMt8WcaVTk1zKfa5DD/l73ijsJhR8XfXzfPifeHw6vjel7fHK8fj6/O7kzvOTjebnYfGG/DR4+J/3NeWxN31Bevj7ezzW3x9rE8x4GLv589Jcx93LisLHH/w2HZlb+Kx5QfdVDc9v+4If5OrKuM5809b28M8zUV8XJ99opdwnzjHW9nZd9+bPew7JLG3KNe/WH95XFnoojq6xv0zF/ye5x6ZcWxze4+vcBNyp5usScIAACwU3JJqdxHNXvYCm37tKcJyuMJU1xjCAAAUBQutbfn93rv7pL0mczdye+X1ODuXZ5GljhiCAAAUBwuqYdGgzGzmyUdI2mEmS2X9C1JVZLk7ldJmq30HckLJG2V9Nl86qVjCAAAUBTeY6eS3f3sbt53SV/c0XoL2jE0s6SkV5W+1TYp6UJ3f7KQ0wQAAChbxbvG8F0p9BHDJnc/SJLM7ERJ35N0dIGnCQAAUH6Ke/PJu1LMU8mDJMUD1wEAAPR5PXcquVAK3THsb2YvSaqRNFbScQWeHgAAQHlylyV75I7jginmqeTDJF1nZvv5dg9oNrMLJF0gSXUVOZ4CAAAA0NuV+RHDoo1j6O5PSRohKeuxC+4+y92nu/v02ooBxWoSAABA8biklOf3KpGiXWNoZnsp/UjF+OHBAAAAfRrXGHZcYyilh6w5191zPKUeAACgD9vZ70p29+ynrgMAAOysfCfuGAIAACDDXWov7xOndAwBAACKgmsMAQAA0IGOIQAAAOSS0THcMYOrUvrQ2K3bZJVV8Uzc98bDw9wWvhXmqcNODvNjvvObML/2Y9VhvnTL02HeWntAkMZDRW5qWR7mQ2smh3kiRz0VFYPCPJlsDPNliUVhvmtqzzBfU/VaVmYWtyWZag7zeY0NYT66ep8wX9c6P8zPHD41zO+u7x/ml+zbFuZ/WLF3VvaZ3eJRlKZcGA+4/s0ftYT5vnuuDvOETQrzg0euCvMvnBQvv8/fcG5W1pyMt4+H2+4L8/+eemiY37qkKsyfClNphOIxR48e0y/Mz75oY5inPv7ZrCzxZLyN1T2xKcz3GxyvkxNrt4b5wTPi5dTychhrw/KaMB/TujTM2+5+PSur/uxhYdnETXeG+ZZX47bXJrKGgk1P0+Pys7f8JcxbG5vielJxPQ2Ni7Oy9uTmsKwPiK+haq6M1+ujLo/XydRR8T77U5eHcU5b3vdgmM9T9r5mSFO8XxpeHa9j8zbE+4hDxqwL88fXDQzzvepaw/zoHPVs/ma8L09UZI9/V/s/nwvLjhvwTJinZrwvzP/jq/E6o6tuD+O3r9w1zKcNrYune8axYT5kVPb+oP2ReH61psr7+j3J09cZlrGy6xgCAAD0WRwxBAAAAHclAwAAIG1nH+AaAAAAndAxBAAAADefAAAAII1TyQAAAHhHiiOGAAAAcJfa20vdii7RMQQAACgWjhgCAAAgffMJ1xgCAADAxRFDAAAAZHBXMgAAADhiCAAAgAyelQwAAAApfcSQm08AAAAgOaeSAQAAkEHHcMcMqG3VjINWbJNtra8Ky/qoUTuUPzjzvjA/7onzwvyGn8wO88qK/mHez2qzsvbk5rDswJrRYT5YY8J8uOrCvCJRE+bt7RvDfHRqlzBvVf4jsddUDgnzhq0LwvxNPRfmqWR8OD1X/bvUxtdlVKyLV+MNrfF6MyqYZaOHx8vJZxwR5ode8kqYp+bH68b4N5rCfPmmeLlWL4qXx9gBiaysuT07k6SJjQeG+clPnRDm9Yc8HObLFk8O872H9wvzF+vj5XrXz7K3D0lK/fTBrGyPQY1h2Z+/MTDMdx0YrwO7DRwU5k/cE+dvxZNVQ0v8mY5/Ky5/xNg12W1piNexX/1gaJivahoe5ou0MMwTitf3fbRPmG/wrWE+X0/H9Seyl7elqsOym5tWhHlzZUOY33HxAWE+pOreMN918KYw3+2nB4X5HnXxgh2XnJiV7T043p4WbY6/zD8yMf5ME/5wVpg37/lYmJ/2ufowr3+4JcwfnZfddkla1Zy9HnzwI78Ny048Psd+/7AZcT5gQJx/9dNxW374UJhPGB9/JrW2hrG/sjgru3RGW1j2rPuHxXWXC56VDAAAgHdwxBAAAADpZyVzVzIAAACkdOewjNExBAAAKIZeMMB1fJUtAAAAelhmuJp8Xnkws5PM7A0zW2BmlwTv72JmD5nZi2b2ipl9qLs6C94xNLPTzczNbK9CTwsAAKCspVL5vbphZhWSrpR0sqR9JJ1tZtsPQ/BNSbe6+8GSzpL0q+7qLcYRw7MlPZ75FwAAYOfUcSq5Z44YHippgbsvcvdWSbdIOi2YYsf4XIMlreyu0oJ2DM1soKQjJJ2vdE8VAABg5+Qutafye0kjzOy5Tq8LtqttvKRlnX5fnsk6+7akT5nZckmzJf2/7ppY6JtPTpN0r7vPN7N6M5vm7s9vXyjzYS+QpIkDcwygCQAA0Mt5/jefrHP36e9xcmdL+o27/9jMDpN0vZnt5577gc2FPpV8ttKHNpX5Nzyd7O6z3H26u08f0T8eSR8AAKDXc8/v1b0Vkjo/AmdCJuvsfEm3pifrT0mqkTSiq0oLdsTQzIZJOk7S/mbmkiokuZld7F7mg/gAAAD0tJ4druZZSVPMbLLSHcKzJJ2zXZmlkj4g6TdmtrfSHcO1XVVayCOGZ0q63t13dfdJ7j5R0luSjizgNAEAAMpXD9184u7tki6UdJ+k15S++3iumV1mZqdmin1V0ufN7GVJN0s6r7uDc4W8xvBsST/YLvt9Jn+0gNMFAAAoP57/GIX5Veezlb6ppHN2aaef50mauSN1Fqxj6O7HBtnPCzU9AACAcuft3Y9RWEo8Eg8AAKAYesEj8egYAgAAFAsdQwAAAPT0NYaFQMcQAACgWMp8xL6y6xgmhvVX7dn7bpPVzHx/WPbZY/4Y5tOOiYfoGV4zPMztf24N8xFV8RiQVamaMG/1rVlZZUX8JJeEVYV5m7WE+cocww5VVw0O82SqOcyHJwaGeYVZmA+q3v7pOtLefmg8zbr2ML96z3iEopcb4sHMjx65Ocx3GRY/4vHPq8aG+bUL443vrWR2PS3JSWHZ6ae+GOZ7Dt8Q5m9vHhTm1yysC/ODh4WxvjRvY5hX+NtZWUNiXVg2qbYwT/zpoTB/YUNFmE8ZEK8z//n5xWFeMS7+rI/8un+YH/3N7G2h4eZ4/p64ZmiYD6hMhnlbKl6vVzXHI3Uta2wN8yXKnu+S9OUh8fKuSATr3oDasOweAxrCfENrvJ9pVLy8N7YujdtSvW+YD7Z4eZji9aB/Vfa8b2tvjOuweP6mPN5H3Lsy/ipKWJzPy7Gcxh21/di+afsMGRLmNdaUlVXmGMRt8sB4XZr62+PD3L/7mzD/ysx4v2SHxQ+4SD3wUpiPq83+vpGklxuyP+szb48My/7x6nhZX1B/TZg3x7NXdZ+eEuYrW+N9+R9Xxstj6WFPhPknz8nel722Kv5MaxI5GlkmXFLuZ46Uh7LrGAIAAPRJro7nIJctOoYAAABFwhFDAAAAMFwNAAAAOuGIIQAAAOQu54ghAAAAJHHEEAAAAJJc8iRHDAEAACBxxBAAAABpDFcDAACAzHA1pW5E1+gYAgAAFAGPxAMAAEAaRwwBAADQIZUsdQu6RscQAACgGDhiCAAAgA5cYwgAAABJkpf3+NZ0DAEAAIrCJaWs1K3oUvl1DAcOUGrm+7eJ/IrrwqKDawaE+bdu3CPMz9tjdZi//rt4NvRLJMI8YVVhnlRbVlbTb3hYtrYizhMeT3N4aliYv9q0OMz7VcX1X3XUmjCfNXeXMD/YTsjK/u1zb4Vl7/zdsWF+4lMfiPMw3XFHH/hImN+6rDHMV9rrWdlf1o4Iy17y3Mwwf/ykR8P87pXxOpneG2SryLF/mNZ/fJivacpex/btNyos29Qen6/43GfjdWxVc1OYL0y8GeYtS/uHec2/nhHmz3w3nmdHnnhUVlaXY+U44oQ7w/y+VfH2cdK4+jD/7eJ4eddUxPNmSNvgMF/VVBPmL24YlJWd8YMXwrL1rfG2d//qhjBf3744zBOK2/6qzw3zVsXbh3t8ZXxlInt5u7fEZSuGhPngmolhvqY5e73uyhHDh4b5GRM2hvm/vhzPm2ZrzsqeXBN/Hxw8PN7vW0s8DxKT4nnws+viffO5a+eFeUNTXZjfuyquf2Nrdvax2+PvxF+dEu/LtyyOd0wXPbhbmPd/IN6/fXB03MY969rDfFFjPO8Tu2Vvr39tiPc/U2zXMJekeA4XF8PVAAAAIM1NqSRHDAEAACCOGAIAAECZU8nOEUMAAAC45GV+80l8RW4PMbPRZnaTmS0ys+fN7Ckzi69MBwAA6OPc83uVSsE6hmZmku6U9Ki77+bu0ySdJWlCoaYJAABQztwtr1epFPJU8nGSWt39qo7A3ZdI+kUBpwkAAFCW3LVT35W8r6R44C4AAICdTmmPBuajaDefmNmVko5Q+iji+7Z77wJJF0jSLruMLlaTAAAAiiq1E998MlfSIR2/uPsXJX1A0sjtC7r7LHef7u7TR46MnzAAAADQq+V540mfvPlE0oOSaszsC52y2gJODwAAoGx1jGNYzjefdNsxNLMrzGyQmVWZ2QNmttbMPtXd37m7Szpd0tFm9paZ/UXStZK+9t6bDQAA0Pv0+o6hpA+6+yZJH5a0WNIeki7Op3J3X+XuZ7n7ZHc/1N2PdfffvvvmAgAA9FJuSiYTeb3yYWYnmdkbZrbAzC7JUebjZjbPzOaa2U3d1ZnPzScdZU6RdJu7N6SHKAQAAEC+0qeSe6YuM6uQdKWkEyQtl/Ssmd3l7vM6lZki6euSZrr7BjMb1V29+XQM/2hmr0tqkvQFMxspqfndfAgAAICdWarnThMfKmmBuy+SJDO7RdJpkuZ1KvN5SVe6+wZJcvc13VXa7bFKd79E0uGSprt7m6QtmQkDAABgB+zANYYjzOy5Tq8LtqtqvKRlnX5fnsk6myppqpk9YWZPm9lJ3bWv2yOGZvYxSfe6e9LMvqn0EDTfkfR2d38LAACANNcOHTFc5+7T3+MkKyVNkXSM0o8kftTM9nf3jV39QXf+3d1vM7MjJB0v6YeS/lvSjPfY2JCvWKfkpddskz310Niw7KIt/cN8dVMqzJ9bOzzMz76iIsxPu6QtzEfXnxLml3/8zazsngenhWUXNFaF+cjquO2TB8Rn77/Q8i9h3pzjwtUxtx8b5hd9/powr/2fzwXpkWHZsffMDvMt58d11506Lsw3/C7+P0dbS7ycmlO7hvmSxPy4fFv29jCiLl4eGjAgjO9ZFef/cuCyMK8b2hLmAw+vC/Pzz/9YmCcefSKoPG6LXl4Yxtf8V7wdTB4QX/iS9Clh/ssH4+1v8a6PhvmSxqYwn3HkvVnZyxvjka3WtcRtf2jd+jDftTYeF/XNTe1h/tfkW2G+f7/dwvzVhnid3NKeveO/7pXJYdmKHN8RE6vjdWNT4vAwr/J+Yd6krWFeZ/G83FSxNsyXNDyQlU0afEJYdmxqUpi/6c+G+UJfHubnjonne/+KeD+5eEu8LexVF3/VvRCsky+mXgnLDtl8SJg//bH4MyUsnubkAfG6N2R4vJx+9HD8/Tc23vz08oYtWZk98VJY9pUN8fax9Mndw3xeU7xuPP717GlKki76TBg3fPo3YW5vTAzz1BknZ2WnXXtzWPaLP8gaKvkdiQ/nfKt4XD15x/EKSZ1n2oRM1tlySc9kzvi+ZWbzle4oxiuu8rsrOZn59xRJs9z9HknxHggAAAA5mJKe3ysPz0qaYmaTzayfpLMk3bVdmTuVPlooMxuh9KnlRV1Vmk/HcIWZ/VrSJyTNNrPqPP8OAAAAGR2nkvN5dVuXe7ukCyXdJ+k1Sbe6+1wzu8zMTs0Uu09SvZnNk/SQpIvdvb6revM5lfxxSSdJ+pG7bzSzscpzHEMAAAD8TU8OXu3usyXN3i67tNPPLukrmVdeuu0YuvtWSbeb2Sgz2yUTv57vBAAAAJCWKuFzkPORzyPxTjWzNyW9JemRzL9zCt0wAACAvsS9bzwS7z8lvV/SfHefrPSdyU8XtFUAAAB9UEqW16tU8ukYtmUuVEyYWcLdH5L0XsfVAQAA2Km4TMlUfq9Syefmk41mNlDSo5JuNLM1Sj/9BAAAADuglEcD85HPEcPTlH5O8r9IulfSQkkfKWSjAAAA+qL0dYbdv0oln7uSOx8dvLaAbQEAAOizdvCReCWRs2NoZpuV/gyW+fedt5QeGmdQgdsGAADQp3iZn0rO2TF09/hBnQAAANhx3jfGMXy/mdV1+r3OzGYUtlkAAAB9i8uU9ERer1LJZ8r/Lamx0+9bMhkAAAB2QMrze5VKPsPVWOZZe5Ikd0+ZWT5/BwAAgE7K/RrDfI4YLjKzfzazqszrS5IWFbphAAAAfUn6ruTyPmKYT8fwHyUdLmmFpOWSZki6oJCNAgAA6ItSbnm9SiWfcQzXSDqrCG0BAADo08r8puS8rjEsqs2b++mh+ydskx173PKw7CFr49l75LJ4iMW2VHyAdOGPUmF+xl8+FudhKil1RFZ0WqJ0dxZF/nr8nWG+3/2fC/PEs89nZXdeuCEs+/VFy8L8lGV7h/mx81rC/MdvjAzzPQYMDPO5W+rDvMUaw3xM1b5Z2d5D4v+d2f/cGubfW/DxMN9RuXYQ0XyXpNRRM/Ov/OCDwviMB34T5oOvPy//uiU9OOXRMD9lXLxcVzdXh3lNRfZyemVDPGdSOR4H8PlJg8N8ZHVzmI+qqQ3zn0+YEObHfnpNmH/9h2PD/D/OmJ+VNa6Id7fDjozny7/9ZFyYf3//eB7c8PqoMP+nmQvCvGpIXE/1tN3DPLU6u36rrQrL2uB4v9c+L647MTSuZ9PzC8N85uz2MD+i/9QwP3BY/FmXNmZ/V/RP7RGWPX5MMszfbq4J81PPiL+3lvw+Xvfq9qsI80NfjadbmeOAUuvI7P3kpj/F60DC4u1maWM8zeOHx+tYYly8D/YV8TzIta/5u6tvC/P152Q/W2PcneeGZeNv8w6Xd/luMbhL7b11gGsAAAD0LKdjCAAAAFd3RzVLr9uOoZlVS/qopEmdy7v7ZYVrFgAAQN9T7k8+yeeI4R8kNUh6XlJ88RAAAAC6YWU/jmE+HcMJ7n7Su52AmSUlvdopOt3dF7/b+gAAAHqjjnEMy1k+HcMnzWx/d3+1+6KhJnePb48EAADYiSR7680nZvaq0p3bSkmfNbNFSp9KNknu7gcUp4kAAAC9n5f4qSb56OqI4Yd7aBr9zeylzM9vuXvWMIBmdoEyT1MZ1S8egxAAAKC367XXGLr7Ekkys/dLmuvumzO/D5K0t6QleU6j21PJ7j5L0ixJmjpwbJn3pQEAAN6dcj9imM9jOf5bUudHEzRmMgAAAOSpYxzDfF6lks/NJ+b+t+dQuXvKzBgYGwAAYAeV+5NP8jliuMjM/tnMqjKvL0laVOiGAQAA9CUuqd3ze5VKPh3Df5R0uKQVkpZLmiHp8/lOwN2zn+gNAACwE/I8X6WSzynhKe5+VufAzGZKWluYJgEAAPQ96QGue/+p5F/kmQEAAKALvfaIoZkdpvQp5JFm9pVObw2SVFHohgEAAPQpvXyA636SBmbK1HXKN0k6s5CNAgAA6Gs6hqspZ10NcP2IpEfM7Dcdg10XpUGW0sj+zdtk9z8wMSw7prYpzJ9ZNzjM96yLy9e39gvzyQ0NYZ5YsSLMU/vsk132jjlhWT9gapjbK/PDvP62NWFeO6I9zCvr4msYXm+YFOYLZzwQ5jcE95+PqKkKyy7Z8mSY3xOm0ua2vcK8PrEqzHfLcR/TqsTCMB+pyWG+QSuzsiNHbgrL2i7D4/zBR8P8xcs3hnm/imSYz5of13/wsHjXsbIpnm7kgMGtYX7rkklhft3d94W5H7R3mH/nzaPCPJFj3mhNPG823LM+K7vq+HVh2cpJ8batfeNlnTo4Hlv/6MfjdTV1xIlhnus/+N89ryXM7X+z9xHDZ9QFJaW2F+Nt+xvHvRnmtb8+P8wvClNJipdTLjvypZVrvuTK87l2qbO7Dnk4zI8bGNfU0BK3fo8B8bawTtnb/RFDRoVll24NY02pawvzB+4aG+aDq+J9dmKXoWF+8tRlYT7sN58I87uOeiwru/a53cOyEwaEsY4dHS/BefFuUv/wtbjt9S1vhPnVH/pzmD/yevxdv3jL6Kzsny77v7Dsd27cI8zLh/XqZyX/zN2/LOmXZpa1lrj7qQVtGQAAQB/jvfhU8vWZf39UjIYAAAD0Zb3hVHJXR/bnmtmXJX1M0l6SnnD3RzpexWkeAABA32r5+lcAABIxSURBVJHy/F75MLOTzOwNM1tgZpd0Ue6jZuZmNr27OrvqGF4rabqkVyWdLOnH+TUTAAAAkZ4arsbMKiRdqXQfbR9JZ5tZ1s0OZlYn6UuSnsmnfV2dSt7H3ffPVHq1pL/kU2HQoKTSncsOt7j7999NXQAAAL1VeoDrHqvuUEkL3H2RJJnZLZJOkzRvu3L/KekHki7Op9KuOobv3Grl7u1m7/oumiZ3j28NBAAA2Fm4lMy/YzjCzJ7r9Pssd5/V6ffxkjrftt7x2OJ3mNkhkia6+z1m9p47hgeaWcfN6Sapf+Z3k+TuPiifCQAAAGCHjxiuc/durwnMxcwSkn4i6bwd+buuxjHsqaeb9Dezlzr9/j13/20P1Q0AANBr9OBoNSskdR78cUIm61AnaT9JD2fO+o6RdJeZnerunY9EbqOrI4Y9pdtTyWZ2gaQLJGlMdTwILAAAQG/Xg9cYPitpiplNVrpDeJakczredPcGSSM6fjezhyVd1FWnUNrxgegLwt1nuft0d58+tKp/qZsDAADQ49J3HFter27rcm+XdKGk+yS9JulWd59rZpeZ2bt+CEkxjhgCAABAPXrEUO4+W9Ls7bJLc5Q9Jp86i9Ex3P4aw3vdPecgjAAAAH2Ra4fuSi6JgncMe/AmFgAAgN5rB55qUiqcSgYAACgS78n7kguAjiEAAEAR9PCTTwqCjiEAAECRlHm/kI4hAABAsXDEcAdVJFyDa5q3yeqbq8OyyVQ8zk9FjuF/Vjf3C/ONbfH9MbOOeinMb17WEOZzznkqK/vMrZPCspMGrgjzacNqwvzME5NhPvRnT4b5gH6jwvyeg0aG+e+WDQ7zhmRjVjauoiose9GEvw/z6op4KxjaL843rZ4Y5nsMihfs9OHvC/M1LfEwnVMGjs/K6vptCMs++d24jhH9N4b5FXOHhfkLydfC/IzBw8P8+fXxdO9qzF7e7vG60bQqbmNjy9thvvzcs8L8M5MWhfnZH/hzmC/9a7wu7T7nzDAfetDr2eELQSYpdc7pYZ5LYt72z5LP1HPE4XH5398T5i/+Kp7HB845Kcxt3wlZWfMd88Oy1VNrw7zi8rPDPHHLXWHuM+PnCNz+0TfDfGi/9jD/48p4LNlNrdnb65B+8TbZlgpjTR0Ub/MXnL4wzPcfOiLM17YOCfNBQ+L2HDh+TZhfpuz95LHHLwjL3nLP5LjuEfVhPvmX8X7JNsbfHz4ofrjDN766PMyvqo6/F6sS2fN4+tDNYdkVTfGy/rtPxN9PB3yvNcyblf09IUlrmuPt7+S7PxjmZ4yJv6Oj9em7N+0Rlr1p/cthXi64KxkAAADv4IghAAAAJJecjiEAAABcUo4rLcoGHUMAAIAi4YghAAAAJHHEEAAAAEo/9SRZ5nef0DEEAAAokvLuFtIxBAAAKAoeiQcAAIA0p2MIAACADC/zk8l0DAEAAIqAU8kAAAB4R7LMBzKkYwgAAFAkZd4vpGMIAABQDDwSDwAAAO/wMj9kSMcQAACgGBiuBgAAAFLHqeTy7hmWXceweo+hmnT3mdtkk35yQ1j2/27cJcyrEvFMX9MSf9xXNsTlH2p+IcwbUsvC/LxbP5yVPdI6Jyw7wT8S5g+vToT5+rt3D/N+FS+FedJbwvyGxYPDfFi1hfnE/rVZ2ccmbgnLHnnTvmH+2Dlzw3yPkevD/M5lw8L8M3vF5cfc/skwb/rHq8O8/1XnZ2U3TnsoLFtbEV8NsnRLTZjPGFkR5q+sqgrzQXGshzesCfPqxMCsrCG5Iizb0r4pzM3ideyV1ANhftnScWH+1C3vC/PlW5vD/IoP3hHmW9qyZ8LPX4/X05/dfV2YP7ZofJhPG7M2zG9ftC7Mf7+iPcxHV2bPd0n63V1/CvPNc7KXX2tTvP9paYjn133vezDMH3g7njdn7TovLr86exuWpMe3xPuxt5qfCPO9qj+Qlc1dH+/fRvSfGua7bzogzE97pX+YP7k2/qz9K+J99q618X5v5G2fCvPjwzT2ycviPHXZ/8Vv/Okvcfl/+MQOTFWadf3K+I37Hw7jGZNWZWUr1sTz8ZWN8Q7oo5/9UJgPvfzxMB9sE8Pca+L9Z20qXt4zRzSGeXsq+/tp0Za4jklr4u9KSVqQ853icTl3JQMAACCtzPuFdAwBAACKhVPJAAAAyDz5hI4hAAAAxLOSAQAAkLHTDnBtZklJr0qqktQu6TpJP3X3cp8nAAAAPS59V3J5d4MKecSwyd0PkiQzGyXpJkmDJH2rgNMEAAAoW+V+80k8oFkPc/c1ki6QdKGZxQPmAQAA9GEdA1zn8yqVonQMJcndF0mqkDRq+/fM7AIze87Mnlu7bmOxmgQAAFBU+XULS3e6uWgdw664+yx3n+7u00eOGFLq5gAAABRAvt3C0h0xLNpdyWa2m6SkpPhZXwAAAH1Yb3hWclGOGJrZSElXSfqle5mP7AgAAFAQrqS15/XKh5mdZGZvmNkCM7skeP8rZjbPzF4xswfMbNfu6izkEcP+ZvaS/jZczfWSflLA6QEAAJStnjxiaGYVkq6UdIKk5ZKeNbO73H1ep2IvSpru7lvN7AuSrpD0ia7qLVjH0N0rClU3AABAb5TquRtLDpW0IHNzr8zsFkmnSXqnY+juD3Uq/7SkT3VXKU8+AQAAKArfkTuOR5jZc51+n+Xuszr9Pl7Ssk6/L5c0o4v6zpc0p7uJ0jEEAAAoApeUsrw7huvcfXpPTNfMPiVpuqSjuytLxxAAAKBIevBU8gpJEzv9PiGTbcPMjpf0b5KOdveW7iqlYwgAAFAELldS+d1xnIdnJU0xs8lKdwjPknRO5wJmdrCkX0s6KfMUum6VX8dw0yYl5jywTdS8oi0sevbMhWF+8xO7h/mgqriX/unJzWFesfiQML+zMZ63r6eWZmUDKkeGZRdsiqc5rrY6zPceFJdvatsQ1zMgbvuJY+L/LNy4OF4VBvXLvoeoORnfV+QTdwnz6956K8y/P2FtmB8+cmCYL1ufDPOxa+Ll0f+q88M8cUf2JRaffP7ksOxVBz4S5tWJ+K6ytc3xEx9btDXMR9fE62SjNYT5kk2PZWUDa8aGZXOxHKNUNWxdEuaJAVVh/vLWePktSD4Z5lfNPz3M/2nPdVnZotZ4Pb3pr5PCfO7GeHn8bsnoMB83IIx19sR4gP2lW+Llmjr1yDD/xb8/nZV941vxU51+8p14mm9uij/T/C2bwvynr9eEeWNqc5i3VMTrZHsy3tfU2/KgbFxHY9vqMJ9b8XiYn/T41DDf5HH59M2Y2c4dNjPMP3jDHWG+9u7GrGzkP8SjeaSOeH+YJy79bJjbPX+O89/fE+YaNyKe7jHxOtZ28dVhPmxa9vb60h3xuvH1k+aH+dxP1If5B0cNC/MFm+L92IDWoWHeYPE6PKhf3M62VPY+a+HqHN8TFdnfw+XFlVL8XbbDNbm3m9mFku5T+sly17j7XDO7TNJz7n6XpB9KGijptswTiZe6+6ld1Vt+HUMAAIA+qicfd+fusyXN3i67tNPPx+9onXQMAQAAisDlO3LzSUnQMQQAACiSnjqVXCh0DAEAAIpih8YxLAk6hgAAAEXgciU9vqG2XNAxBAAAKBLnVDIAAADS9yRzKhkAAGCn5+rZ4WoKgY4hAABAUbjcOZUMAAAA9eizkguCjiEAAEARuFwp7koGAAAAN58AAAAgzcU1hgAAAJB48gkAAAAkZYarcTqGAAAASD8Ur9SN6BIdQwAAgCJJpdpL3YQulV/HsDUpX7lhm6i5PhEWtTjWnBVxb/y4sfHHnb7vijA/+nt7h/nCU48M84nVtVnZ7Zv+EpbdUDUlzJc2rQzzEat3C/NkqjnMN7XH9dy7alqYP9X2fJiPaZmUla2cXxeWfWTKY3HdzfPD/O//PDnMU4o/0+3168L80ENeC/MDhr4e5pcuuScr++NPPSyb9IFhfuOSxjAfmOgX5gmrCPOrlr4d5m2JljBvb9+QlTW11YRlx9TuH+brmt8M81y2tKwJ87eq43WmLRkvv7sbnw7zr1btmpXtVTMmLDt3Y7ycnmp6K8wPCOqWpEFVYawTx8eftaG5Ov6DfvHy/sbvR2RlqX2OCst++cTsZSpJt57wYpjP2Rqv79Ms3rZPHR9vryu27hfmT6RGh/nkymFZ2V1Vb4RlLcfOeXhlvB9rVVOYj1S8j2jWljCft7E1zO+5Mt6Or1kwKCv7Vv36sOxuu18f5gO+cXyYr7thVZjXjYuHKqkcvjbMn/pafZhPHROve6M+vk9WdmxqbljW/vX8MF95+J/C/I2G+Lv1tbb4szYq3mev8Y1hnky9L8xnrxySlW1oifcFCc/RMSgTzl3JAAAA6MA1hgAAAJCcR+IBAAAgg+FqAAAAIMk5lWxmYyT9TNL7JG2UtFrSl909viMBAACgD3JJKd+J70o2M5N0h6Rr3f2sTHagpNGS6BgCAICdCEcMj5XU5u5XdQTu/nKBpwkAAFCWdvaO4X6S4sHOAAAAdioucfNJ98zsAkkXSNIuw+KBSAEAAHo1L/8jhoUeInyupHg4/k7cfZa7T3f36SMH9i9wkwAAAIrPlR6uJp9XqRS6Y/igpOrMEUFJkpkdYGbxM+UAAAD6LJd7W16vUilox9DdXdIZko43s4VmNlfS9yTFD4gFAADos9J3JefzKpWCX2Po7islfbzQ0wEAACh/5X2NYVncfAIAAND3uVTmN5/QMQQAACgSl5e6CV2iYwgAAFA0HDEEAABAL3gknqVvHC4fZrZZ0hulbgdKYoSkdaVuBIqO5b7zYtnvvEqx7Hd195FFnuY2zOxepT97Pta5+0mFbE+kHDuGz7n79FK3A8XHst85sdx3Xiz7nRfLvnwVeoBrAAAA9BJ0DAEAACCpPDuGs0rdAJQMy37nxHLfebHsd14s+zJVdtcYAgAAoDTK8YghAAAASoCOIQAAACSVUcfQzE4yszfMbIGZXVLq9qAwulvOZnaema01s5cyr78vRTtReGZ2jZmtMbO/lrotKJzulrOZHWNmDZ22+UuL3UYUj5lNNLOHzGyemc01sy+Vuk3YVllcY2hmFZLmSzpB0nJJz0o6293nlbRh6FH5LGczO0/SdHe/sCSNRNGY2VGSGiVd5+77lbo9KIzulrOZHSPpInf/cLHbhuIzs7GSxrr7C2ZWJ+l5SafzfV8+yuWI4aGSFrj7IndvlXSLpNNK3Cb0PJYz3uHuj0paX+p2oLBYzujM3Ve5+wuZnzdLek3S+NK2Cp2VS8dwvKRlnX5fLlaUvijf5fxRM3vFzH5nZhOL0zQAJXSYmb1sZnPMbN9SNwbFYWaTJB0s6ZnStgSdlUvHEOhwt6RJ7n6ApD9LurbE7QFQWC8o/QzbAyX9QtKdJW4PisDMBkr6vaQvu/umUrcHf1MuHcMVkjofGZqQydC3dLuc3b3e3Vsyv/6vpGlFahuAEnD3Te7emPl5tqQqMxtR4mahgMysSulO4Y3ufnup24NtlUvH8FlJU8xsspn1k3SWpLtK3Cb0vG6Xc+bC5A6nKn39CYA+yszGmJllfj5U6e+l+tK2CoWSWdZXS3rN3X9S6vYgW2WpGyBJ7t5uZhdKuk9ShaRr3H1uiZuFHpZrOZvZZZKec/e7JP2zmZ0qqV3pC9bPK1mDUVBmdrOkYySNMLPlkr7l7leXtlXoadFyllQlSe5+laQzJX3BzNolNUk6y8thuAwUykxJn5b0qpm9lMm+kTlajDJQFsPVAAAAoPTK5VQyAAAASoyOIQAAACTRMQQAAEAGHUMAAABIomMIAACADDqGAMqGmQ03s5cyr7fNbEXm50Yz+1Wp2wcAfR3D1QAoS2b2bUmN7v6jUrcFAHYWHDEEUPbM7Bgz+2Pm52+b2bVm9piZLTGzvzOzK8zsVTO7N/O4LZnZNDN7xMyeN7P7tnuqDgAgQMcQQG+0u6TjlH5s4g2SHnL3/ZV+csYpmc7hLySd6e7TJF0j6fJSNRYAeouyeCQeAOygOe7eZmavKv14xXsz+auSJknaU9J+kv6ceQxvhaRVJWgnAPQqdAwB9EYtkuTuKTNr6/Rs3ZTS+zWTNNfdDytVAwGgN+JUMoC+6A1JI83sMEkysyoz27fEbQKAskfHEECf4+6tks6U9AMze1nSS5IOL22rAKD8MVwNAAAAJHHEEAAAABl0DAEAACCJjiEAAAAy6BgCAABAEh1DAAAAZNAxBAAAgCQ6hgAAAMj4//rA3LG6I8GCAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# for plotting features\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure(figsize=(10, 4))\n",
    "librosa.display.specshow(features, y_axis='chroma', x_axis='time')\n",
    "plt.colorbar()\n",
    "plt.title('Chromagram')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extracting mel features\n",
    "\n",
    "import os \n",
    "import librosa\n",
    "import math\n",
    "import json\n",
    "import librosa.display\n",
    "import numpy as np\n",
    "from mel_features import log_mel_spectrogram as lms\n",
    "\n",
    "path = \"./breath/pos_new/0pos-0426-126-breathe-m-46.wav\"\n",
    "sr = 22050\n",
    "\n",
    "\n",
    "# Zero crossing rate\n",
    "y, sr = librosa.load(path, sr = sr)\n",
    "# zc = librosa.feature.chroma_stft(y=y, sr=sr,n_chroma=12, n_fft=4096)\n",
    "\n",
    "features = lms(y,\n",
    "            audio_sample_rate=16000,\n",
    "            log_offset=0.01,\n",
    "            window_length_secs=0.025,\n",
    "            hop_length_secs=0.010,\n",
    "            num_mel_bins = 64,\n",
    "            lower_edge_hertz = 125,\n",
    "            upper_edge_hertz = 7500\n",
    "            )\n",
    "       \n",
    "# print(features.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAoUAAAEYCAYAAAAqOYzUAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/d3fzzAAAACXBIWXMAAAsTAAALEwEAmpwYAADMtUlEQVR4nOz9eZRlWXbWCX7nzaPZezabz0N4zJERmRE5KSUkoQFRGqtoaArR1VXdNIsGFk0xVYMoSouiulAB3YzNIlfDKijU0EWDWoAmJIGGJCXlGJmRMYfPbm6z2Xtmb55u/+HhkZ4pt/3bHve5hVnE22tpKT3Os3vucIbvfHvvb4coijSxiU1sYhOb2MQmNrEPtiXe6xuY2MQmNrGJTWxiE5vYe28TUDixiU1sYhOb2MQmNrEJKJzYxCY2sYlNbGITm9gEFE5sYhOb2MQmNrGJTUwTUDixiU1sYhOb2MQmNjFNQOHEJjaxiU1sYhOb2MQ0AYUTm9jEYloI4cdDCP/0vb6PiU1sYhObWDybgMKJTWxiLgsh/IEQwhdCCI0QwmoI4edCCN/6Xt/XxCY2sYlNbDw2AYUTm9jE0EIIf0rS35T0f5O0KOmMpP+npB9+wOukxn5zY7KjfG8Tm9jEJnYYNgGFE5vYxEwLIUxL+suS/lgURf8qiqJmFEX9KIr+TRRFf/btn2VCCP8khLAfQng5hPDCPX9/LYTw34QQviqpGUJIhRB+6O3f1UIIvxJCeOKbfv9nQwhfDSE0Qwj/MISw+DYzuR9C+KUQQvWe3/+LEMJaCKEeQvi1EMJT97TNhhD+TQhhL4Tw+RDCXwkhfOae9iiE8MdCCG9KevPt//a3Qgg33/6bL4YQvu2e3//42/3907fv5aUQwqMhhD8fQth4++++9yF8holNbGITe+g2AYUTm9jEyD4pKSfpp4zf/JCkfy6pIulfS/q739T+n0v6/rfbL0j6Z5L+pKR5ST8r6d+EEDL3/P73SPoeSY9K+kFJPyfpL7z9+4SkP3HPb39O0iVJC5K+JOkn72n7e5KakpYk/e/f/r9vth+R9HFJT779789Lek7SjKT/t6R/EULI3fP7H5T0v0iqSvqypF94+55O6g54/gf36WNiE5vYxI68TUDhxCY2MbJZSVtRFA2M33wmiqKfjaJoqDuA6dlvav/bURTdjKKoLel/K+lnoij6xSiK+pL+uqS8pG+55/d/J4qi9SiKViT9uqTfiqLoy1EUdXQHnH747g+jKPpHURTtR1HUlfTjkp4NIUyHEJK6Ay7/uyiKWlEUvSLpH9/n3v/HKIp23r43RVH0T6Mo2o6iaBBF0d+QlJX02D2///Uoin7h7ffxL3QHqP7Vt5/ln0s6F0KoGO9qYhOb2MSOpE1A4cQmNjGybUlzEHO3ds//bknKfdPvb97zv09Iun73H1EUjd5uP3nPb9bv+d/t+/y7JEkhhGQI4a+GEC6HEPYkXXv7N3O6A9ZS39T3vf/7vv8thPBnQgivvu2Orkmafvt6B93b1ttg+O6/dff+JjaxiU3sONkEFE5sYhMj+w1JXd1xs75bi+7537clnb37jxBCkHRa0sq7uO4f0J1kl+/WHfB27u5lJW1KGkg6dc/vT1v39nb84J+T9PskVaMoqkiqv329iU1sYhN7X9sEFE5sYhMzLYqiuqS/JOnvhRB+JIRQCCGkQwi/O4TwP72LS/6vkr4/hPBdIYS0pD+tO6Dzs+/iWuW3/3ZbUkF3sqPv3vdQ0r+S9ONv3/Pjkv4Lx/UGugMoUyGEvyRp6l3c18QmNrGJHTubgMKJTWxiaG/H1v0pSX9RdwDTTUl/XNL/711c63VJf1DS35G0pTuJGz8YRVHvXdzaP9EdV/SKpFck/eY3tf9x3WEQ13Qn1vGf6Q6IPMh+QdLPS3rj7et2dH+X88QmNrGJve8sRFHEv5rYxCY2sfeBhRB+QtJSFEX3y0Ke2MQmNrEPtE2YwolNbGLvWwshPB5C+FC4Yx+T9H+ULa0zsYlNbGIfWJso+E9sYhN7P1tZd1zGJ3Qna/hvSPrp9/SOJjaxiU3siNrEfTyxiU1sYhOb2MQmNrGJ+3hiE5vYxCY2sYlNbGJH0H08V5mKziwvHNg+SOUObJOk4GA+k40ds33U7ZvtiTS/tpDPm+1Rz060DAnG61E6Y7Z302W8Rn+UNNt7fVuezXGbok8SQAEu4VCIK6Q69jXe0RY+4B5GdrskjRL2d+8pi9eIQO4uGez76A157I2ieJJ6ieCYQwn7N71B/PPmcGS307iReOyNoI9MmvvoW3VeHH1Ivnlk3kPf8c2S8eayY9nD951K8n0223Z7t2O/8FTaXtMkKcCiMuzbHy2V4Q82GtrPOoQBHjnGzQAG33DI65pGMb2F41DwpFvw9OF4jEb99a0oiuY9t/Sw7PlEMdqD/eiuvaXuL0RR9H0P+Za+wY4cKDxTKekzf+YPHNieKBTtC0xNYx9Rfdds71y3FSiSWRuMSVLmwgX7BzBZW49+FPvYKZw029tRAa+RDjYA3h/YhRlSCdgRJa3u299kptCy29N17IPA1ontr5rtqWYN+xgWbLm6vcpZs12SdpP2evTGzsEHIkna2uMNr1Kyd5Mk7Gdnpvl9zyc3zPbVwRJeY69rH5waXXt5ItAoSTfW7HFRyNntW7s8vpMAMGar/M0K9llXm7v2jnfzZhP7SKfs+2i17IPqpUss1/jkGfsaM3m+z3/9m/Zh9o2vrZrtlXm+z3LZPsDt71uqRdLIMfhyeceJwrBej4HD9VftvarfZZWnbgtQOFjkAJUEwsdhnvv4zL/+HdfxRw/Z9qKh/maK9wpJ+oHBG3P8q/HakQOFiiKNjIHcXds0/3zUt0GOJO3fsq+xc3XbbC8v8aIztW5fY++W3T597ivYx6lv/3azvV9dxGv0c/azbCQ+bLbPZG2GTpJKWXthKqdsUDjXuoF9pNs2kElf/prZ3rnBUnTEEM888xxeozh7ymw/oy+Z7Z0TfMitZ+3fJGVvNtOtNbNdklJ9eyMpp+05JkkRoNNhyT58XQ6PYh9Ube7Vy/a7SKd5MyuX7OfodHmzunLNnkfZrA3oqlVAlZI6HftZaVNttRkI3dy2wdZL+3ygbjTs9aDbsdeT2uYe9pHNzZjthGH293nd63ftA8VgAGMvw9tzGgiKXtsGtx6jcTEOwOcBdO8bC1JwrCuS7sjoH7IdPVCYLyjxzPMHN+/aDMXeb/wWdjH7zCNmO4HC21+6jX1svm5viumcvcjPPnG/alzfZH17wncKs3iJzazdz6tXbGb2sRPsRukM7Gftj6r232efxD6WM7fM9pk5+3vkZm2GThL6K7sA+CRpq2IzyGsd+z5afWYfInD5LxRsAD0ocB+ZyN4U1x1MYS5pj19ypVfCPvcxawOIt27YY6/XYyB0FQBdscjvM5ezl+Ji0Z5DnQ7fZ75gX6NYtBm6x87xRpaAb7a+hZfQ7qbNJlbnbc9DgqhwsSt9p24fehZPskeq07IJihtv2oxnoezw9EB8Q9IBLMX41rTDYAoTSWbbI4+//QhYSAQl8/w8kr5eSf0Q7ciBwuv1ov7Qz3zywPZyxZ4oqdkfxT4++mzFbM991B7ka46F7Qu/ZU/4Ytk+2X/8Y/ZJVuIYnxd/lV016ys2QBj2bZbuRYer5uJF+zfEQDSbtptRkoZDG+hnMo+Z7bOzzGCcBNzY2OWF78Vfgfc9pKMhHx1n5+05Uq3YHokhxEJJUiFvP+t+gxfoQt4eF82WfY3qNC+sJdhXNzYaZnsP4tckKQOALgcHQEna3LRX/5tXbfZs+TSDFPqup0/Za1LNflWSpGbb7iMBsaiSlEjZoK62YTOBs8sV7GNrzT5QZHI22NpcZTaytWd/0xGEEO3vcB/EBLpiCsEI0HlA4QeKCSQLUiJ1dEupHzlQGEJCufzBG/TVl+2QgNllZse+/BV7kW427IlGbgGJwevmLTvZ5bc+h13gZrR23WY8JSlbsN09s0v2ZuM5AV67Zi/A6Yz9PfbrfJStzNjvu922T+1f+Ow69jH8uM0EzlT4XZw7Z7MxxYK9IZ5f4hihvbY9Ltbtoae+I4Pj8VP2fQxGfI1b2/Z3nyrZ7XsN3mhee8u+T0oGePQxBlvdnn0NYqUkaWHBPvgUgL0tFh3Jb3AbtboNIGbPMANHiWl7ezx+UxD7+Nizdiz1wjwf8K5csREurUkeKbddeI5+z16TiAWUpCQwA+09JgaGjpAryw6DxTsuLKDLHsR9/B7YkQOFZJUF291DIEaSFhZtADGcsxfobpdPXxQDRIHOXQfwzGbtz7d0lgEyuTgIWNKpXpIC7EanLtjM1ZNPO54D3GcpOJl95EN27JkkzZbgXdX4XVSm7d+Qt6cHrniJs49LBbv9Y6dtlluSdns2y3djh11fHs+WZTPTvLBOlex5Ng9s41TWZugk6aUbdojF1jZvuuSmTsE8yzg2mQSA0+qU3Z5K8sZcAcGDloOR78O7IFZ/3t4iJEmJR+z5nob1oj9gUEjfrN20D7sE+CSp27EJjAQAU4lBHbF8RwWwhXA8FPZCCBOm8EEsl0/p0hMHg4Bez57xC3N8ulqcsQdxGmQTko7B1+rbv7m+artqvvJFDvbPGoyqJJWmWCLlxCl7c6f2jEOaYX/f3hSffty+z8ghsVIGDzN5L2B9liQVs/ZzPL7MFyHpnN7Q/qY36iwzNJ23DxTnZmygQ7F8knQaYjgrjsPZILI3oxdX7ISZjEPeZAcSqecg+mE44vF9cg4SOCJeZp85a7sa68D+1pp8nzt1+31Rouq5OU5aGEb2fTTavCYVztuTuVSwn8PjqTw1b+8B/aG95mw6DoBkZx6xx3ez4ckctr9JD0CjxG7scYAtusY4gOVRAadoE6bwwWw4jLS/f/CmRozPwBEP1ezYA5QkJgawYEhSC/DBzZu2+2LhBG+qU1M2gJie4lMixW3NVOxrkJSGJH15y97wZiCI/twUZ7LW+jZYur1nszmQCChJavXsA0fOcforwIxbSNlu7NPTV7GPlWDLHby5VTHbLzIxq5C251mjz3GgtY79m4UpG4SPHOJlzy3ZiWmZYG+aN1uOhJm0PXhOOVTRTubs715MVcz2Ryockb5SsT/sWt0GbFtNBnQzRRvIXFzg+2zAPNtvw5qUZXBQLdjffTpju10rBV6fU0l7fJ+es8f3TpMX16+Bx+nq6wzo2nv2XjQOsIXA8xAka46MjTGmMIRwWtI/kbSoO0qNn46i6G/FueaRA4WV0kg/+C0HT9hrO/bmv7PHL7uUswdotWAvbLRoSdLqpj0ZybWwvMybaqNhM0KNJk/mE4v2AkvswVKFT7Pf/zvs90WLjoe5OpG2XZ7zc/biuduvYB8zaTsYLzfk+B0SWC3v2wzx7tQZ7KIYbCbwkTl7jpwdvoV9tCN7Hg5GnCg1DXJG+ZS9cc+IM756wd5Ydwe252GpAAGYkoojOyEgOFR1h8AmUqZ2Hr65JC3B4auQtg9OmSSHtHSH9lw/n7qC17gBGm4liLXLpdhdT7qmjb4d/lDKMANXBJ3CNLjjZ4qO2PUy9JHlvepQ3K4J+1kPg0k8KhaClHR42Jw2kPSnoyj6UgihLOmLIYRfjKLolXd7wSMHCoMic/FZmrI3ku06n65KWQBTAPo2avzavu0pO7mCXKJNR/DvjS2bKfQknmWB8dmFBLg3VnjR2du3J/R3P2ufVElAW5KSKfub0iYwneZMvyFMl2yfQWFqYI/fbq5itr/WOId9nC/brFMpZd9nN8HxgMmRPT4f1at4jXTPvo8RAIxOhl3pI9lA6JGOrQfay/DY6yftA0cnCYL7khZ3XzfbhzO2JmN9yMzVCFy71Yw9B4piCaB2yn7WTuB3MZe0iwv00/a46EYct9jq2/vEbLZmtqfFh+Hysg3UydV+a5+DI6fK9jVI5UJyaB12YmrWHJJ5El6OhoWxMaNRFK1KWn37f++HEF6VdFLS+wcU9kdJ3d4/ONDnZ/+9DSBOneYFYWva/s25GXvxq+YY5W+3oMwdgMJ8mk+Jnzq3YrYXItaQ2FPFbO/0bAHsQo5ZkO981HbhEcvxVoM1GxN5OHUn7ISZ3YiZrWKC2RiyCE7EqaG9AH84+xL2kQKZimzDZthSdXbXqw+bYp43/1HKnoejnH2N/K49/iVptu7QjzIsM8MC8PVZWw6p3GO2Mdm215wKCIoPHNUzhhDD2Yvsayx2Wc1gpmOHN3RzDF538yfM9vmeHc+6n3XEP8DrGgBzO9PnmO+y7O++mrJZ/3NTPHbbPVsn6woIwEvxXbdHJcHj2MjeBCk4tDQf+LIhnJP0YUks1mzYkQOFoyio3Tv4hRVACLYNqv2SdBtcu1s1m4Hw1DK9dBLU7AEUbjcdyv9dO97pjRs82ffBBV2t2BMtx7eptbZ94h2O7EW8kGbWtN612a31gZ1R4AHh+TxkCw6ZPSDmKeNgG8lu5y6a7ZVsxWxPVxiEEztW3Xa4oKfs8dvMVMz2QWAgtJS2QfQwY4+bRolBYTdpX6PYq+E1+kV7jmQ6wOJluDRhrmcDz7XcebO9kWOwlU/YC0IQL55DYHepwHK54wBTWfub0UF1kOSFL9+1v0k+Y68n5N2QpGLG3u8eu8RM95WX7d9QGbzDcO16gOdxcTEHsRLAPTYXQvjCPf/+dBRFn/5t1wyhJOlfSvqTURSx28uwIwcK251IL7958EAnGZYbl/k0m3kcyoCV7UWp7age8OU3oIQXJMRsbzHLV9u2AcTcErvXZmdt98JLL9osX+sx3iiGI3DVgNTLHmTkStLtXRsgnJyx+1jMM5uTDjbo++XWp/Aaub49dnIpe5EfdnkxmUvY4+Kl5iW8Btkj03YMZ6rCsY+7KZvlIDcgxU5KDDxzDZsV7UGygCSlRzYzm+7zfRKD3ARwSn8vSS04DCx2bQ1Yqn4kSeWh/T6JpZakCGSbAmz+PYiNlJgVTcOBY7bh0DVN2Qenma49hxpZ9l6kIAN/p+aQTwOt2qMQ73dcAJ/LwgOxs1tRFL1gXi6EtO4Awp+Mouhfxb29IwcKC7mgDz9x8CBc3bYn6/Q0n+wrU/YgL6PkAS/AjZb90VfXbZBy8hQvbI8+aoO+m7c4FqTVtheNxRM2w0aaYpK0vmX/prZvD8Ozi9zHpSV74yV3/Zu7nGUagj0uSllegPtDe+wQQ5xL87u41mUXnWVnpjl2bG9gj70bEHYgScspG4hXg92+0ec+dhL2u6jOLpvt3ZFDQgVA+KjAJRQ3hvazpBOwXvQ4gYOY7H7aZs/mu1wfnMBpbdZmsSVmNPfy9vtsRxwTuzywn6W4a5cy7ed5jiWgOlGiax/8ExmHrBMoYSzMcZxdMqYr0xPLR9nHZO8nplAKD8IU2le6IwL8DyW9GkXR/30c1zxyoDCSXQ2htge1NddZ8iCbsQFXtQyF4bs8QKnc0/S0/eoduqUiXdLpaXav1evA0tVtFmT5BLsnSLB5umhPZk9prJmsvZGQRErBkQ1G2ZfXtxnIE0A+e8J+VmIGJKkALqVLJbt04fW2XTFCkhZydjLAdJE9GFM9mzXazQBQSnJYwdLQjj/LtEGzccAhAUOIjRyAq12STgPrmW3Z7zMxjFeVQpKSA2A8G8ymU9mUMM2HL3qWuZadlKMEg5R+FsSrt+x41XSHwyOGSzazGsANXoB9SpKeK9oJXc3e03iNEcTiIaBz4L24pfIiRyfHJdEkBCmRHtu9fkrS/07SSyGEF9/+b38hiqKffbcXPHKgsNOT3rh28ADpg/uNtPskFjdNAD7IZRwgBaotXFq02QWPDMtvvWmzNW1gASVpYcHesKgeKgmBS1Imad/HVA4YDGDXJKk3sgFwb2QPdU8fUxkbFF5a4LiuM7Nwn1CxpNZmoP/IPAfBW3Yux4zQvmwWozTid9FO2+N3oWknLWTB9StJjaotb1LYuGy2hx5Lj0RZ+8BBCTOSVJ+5YLbnN0Cfcp/ft4pwgKPNP81ra3fWLgXZyNvViyQp37MBcGbfPkwk1uxDjySlM45gaMMI8EnSKG2vnVESShfW7AONJF1f+oTZvlhmkoRK9h2GhuBxAXRjsQdzH5sWRdFn7lxxfHbkQGEmJZ1aOniAhGAPHmJiJGl72wYh+/s2QPAUuF+et79Tqw9yGyCZIEmbW/ZzeGqAkhg4iVv3yjweOz2Qg8nDouQY8ld3K2Y7uUTPl+x4Kknakb2h1Xu8+e937W/yuZfs9/19H+NElCScqtcgY5GyVCWp3Qd5nhKDqaWazXKMgGFrT9muX0maWn/DbA+QRT3KMxNOjM/OnC0nI0nNhB2mkT75pNme/w//X+yj/oY9xitP2lnU+tBHsY/taTtZ5WaHv1khZX+T9DIA6JMcNnP61n802xMde54lLrPiRzhjv8/XT3y32f7FXQbQcztUTQcvgTaO7GJy7VL7UclwHo+Nz338MOwIgsKRTs8cfLp5/bZ9Kv/8Z5nl6HXsRWfprJ2IsrDIm/8qkBhXb9qDYuiIwbh51XbnvPhZTrrJQkmSfNF+3191nHha+yCm/KTtrmw7SjW19m12IQV0/cnTnHzRgcz2leu2S1WSZhZskDEY2Ivjr0OmoCTt7NjTutO2Gc/gcNd3OzYDMV3hOLrTp+xkFHJrUX1aSXpk+SNm+2rNBp7bNUeFJDg4zfUYZINUnEog06IXzFh0SdLWI/bGulCNlwQlSV/+vN1H2lHeazCwv0mxYPdBnh5Jand+n9lOYCrrqLu9+roNTq/+aztZZYrzTNSsQ5Z0n7/Z5k3bs3AUYvU893BcgGMYI1P4MOzIgcL+MKGV2sFAhRI0Lj3NJ9ERZP6u37YBRh3i7CRppmoDx5mKzRTeus0xQpmcfY3Tlzg2rN8D6RxYVMj1IEmVWZsFyeXtTXPk6KPXsZ8jBzpul1/nrMj6lj0uZpdZbDYP90FxoJQkJUmnl2ygX87b39Sz+V9es8Fpq+MIsZiyfzOEfcAD2L74hv0+TwB2neNPqnOw5Fxb5fs8MWu/c6p+kYQkKEmKYGxlUhCaA2EeknT+lJ3k0ezwZkjx1LNle67vNnhba0HMdw/ClEpF7mNhwZ6He/WK2Z5K81wfwSFyd5PDCgKg6EDxfg7ARjGD4wBJRwG8eo3e+XtpRw4UZlIjnZs9+PSThGoLtx26uxsbUFWiYwOyC/N8hJsq2YOcEr4unuXYMUFcV8pxKp8q2YCsDok9rRbr+9V2bFYpCVR6tcqB+gnwMZ8/ZzOeew1mf2/esK9RgjqkkvS9H7cXrj0AU6uOWP8dyPG4MGe7xgpQUk2ShosVs323yeP3/Kx9ozste66fBpkhicESVa7Y6DAqJFf6qUWeh+en7UNJd2RTiW9u8X3Ol0HUHGJ/d9s8vikb9sIiu3a3GzHj/RzYoNu1f0R14/P8KpTP2u/ikx+rmO1vXGUQTuVSO22+xs5tOvi/92DLwwIeZfbtG2zCFD6YDUdBe52DF4UN2BRf/gqjwqVTNpj6vu+zj/7LFd40e3CC29q3N825Mm94ifP251vbZPagVrdB3fnTtLnzENrds1fQ1TUoTg/Zy5J0+rR9KofKWCoVuY/HHrMZz7V1XoBv7tj3ubFjfzPw5kuSlmbta1B92pk0swuljD0+Z/OszbfWsNnGk2X7PnIJB3gFIeTr+3aoyKkSo/AeDK5pR/3ZbCDAZrc/MsubDCVbpRMgT+Woxbtcste9JriGJenktD12CmmomQ0JTJK0CGtKF9bvLLCqktSD5LUMsL/7iwyOX9mKX2WJbBxuWUgFiB1zKElRPNWbQ7MQgpIOFvi9siMHCpsd6fNGDO9Tl+wNr/odHMt09aa9ob1xxW5PXOSFjVx01G7J8ty1Ug5igBzl+JqwprTgYH92weHmRn0d+31ubjv6AEmZ5Tn7XS2UOEsvB5I0PQDpknRz117oL56MfyqnQyhpNl5rsGzIyaIdr5oI/ByPVWymsA8Cw8vNN7GPQdpG0cmSPQ9Jm1KSliJb4L0N8ieStLD9mtm+P2Vn9b7ZhJhDSY9O29ms3Qjiix1i4fWRfeAm6ShJSkDVk412xWxfLjiAPIwtqo2cT/GBpJKsme1f3Tlntk+x80IffaFitr/0Ch9IVq/Y8jvDfny5o8Ow4xJTKE3cxw9k7WZPL33+4GSRC2fszLMZiDeRpBUAEM9dolML93F51X61+Zy9MZMGoSTNTdn3cWqeN+bFGbujMsQRrQLIkaR81t5YL5ywN+YkSDdIUhKQ0FNzthjt9ICTcjope5VuRbyKZ2btb9Yf2d/jzQ0GGEsV+5udzNsgZrdfwT46IOocxGAqA4LMM0P7Pj3afIOsDVII9J3a5VrTEYQuTPUZQHRLdqZpGmpifzT7Jewj1bCvka3biQ+dCsdrD6AMaR3KTUpSNWUnbC0B6MuJD3jlyB47JwAA7yT54JQb2GEaZ6ZrZnu7yORDe2C/741Fdi3kCvZvGrv2enJUXKFxBbIPzSbu4wezYimr5z91sLZYvUG1eBmBP33RBkvdgX2NUpZB4bPn7EWl1rZPoqs7DlkQo0a0JHWgXZJ26vb7fOqs3f6x03apJkm62bBL4VH82Yk5BhgUBN8e2u97d2RLaUjSyqYN+p6Ys0GMJG1BnBzVYD41wwBjCmqqNkf2PZCbUZIGA/swsJznxJ2lui1CnN0GnbYWl4LMlmwQUqjaDFuyzcxWIIFrB4NRBLFkgcj2aJF18yLQGRxl7DmS2+fxHYp2RvlJsYZgqmu/z42MzZqmIg7jKHZsYDmE2sYe2SaqDz4f2SB8O22HNkjSbgcYTSAfpPgAhZJIJNYhHIckDbmoj46FCSh8EEsmper0wV+XsvROTnE8FDEUe6A3R8Hpkq8Unvn3VdYpJHviBIsYTy/bi+O27IXJwwhRfNl+x964B44D4ImCzfQVZW/uV7pcq/ejc3YVg3yfAURx2gYyqZG9ob3aAi05SYsp+7vn+vY9rKcZYDQGNrtQHPI8TIzgcNWD2IWOI54qFW+JixyiugFix4YO8erRlM0UZrZt0NiesYGSxELdJKbcmOexl4nsb0aMpySl+jbTd6pnu9qpwozEoI+E1U817HuQpNy2DYDXztrC040el+ubK9hsZLTEY69UtZ+1sesQRgeLC/reT5I00oQpfCBLp2zg98YN+2UmzrF7otO3B89e025/jD0HWsrZp+ozWXtjLmc4w/mtbfs3V2p80rxQsdtTIITcHPLCtd60XZ5FKMtWyvLJ/1bT3lRPQCLJQr6GfbzRPGe2U/k4j83d/KLZ/m2Jr+A1bpY+ZbYHCJI/v/t57COCjTezwWLgpEoe7doHljDD4r5Rwd7wsluga7rrkDOAzah/mesSp6fhYASbSO4al13r1e1DS3reZvQ5cEGaqtkMcVTmer7RTbt6C8VjRSAaLUn9sj12ci2ISxzxSXV/6TGzvbJvM+GtMr9xivPsgmyZJE3N2P2sX33vxavfT3ZHp/DoAtgjBwoTipQxddLsW/6pn2EB4d/z/RWz/cwJe/HMJBmkJGCQb4/sRcnDwM0VbVfily4z2zgY2Yk5Zys1s91Tjo9qRRey9rvaafFz5NL2fXxpxUbyZ+eYdTpRtDeK5MghkQLxqBtnPma29wOzINu9itl+Ptis6vqsXT1Dklojm4HoFOznkKRiClyiF+1xs9PlLNPZrJ3MUgjAtDjK6YzAlTh6njeAza59wKNKOEslZnNSMFdpzZkdsOehccqWxqkNGBRmT0Nxgb59+Bo5ah8PEjZY2kvY3yMdeK5TolQnbbuXWz1e9zoDG8i/tcpxiStv2geK4wLYjst9apJ9PF575DSJYDJ7cH3DXuiLIMOynIT4H0m5jg0s+zl7ka/1efEk3bBbt7gkWnXKPiUWMna7RzSXtMtIm5qkGyTWWLu0ABVPEtxHrQesU5bj/eYa18z2CDa0Zo4Z5BNpO84zGgDbDtmfkrQwtOdAcsgHpyRkNfbSNvDsp9nlT5v3PGT9bs4+jn1M9Wx2bDeziNeoZuzxmQz2u6CMckkqBxs4RsB4ZnscHtFL2WEFpyPODO4G2/tAoK/Q5NjHQdruY1Cw1+eO2ENCLF5bIG5tSLPdtVbffhdUKUeSkqAOcRhl7j5oNmEKH8BSyaGWigcvXhTgW8pyDMVeB3TFMjaYaieY1q9l7RNcRlBzFXuQlsu2C/qHvp1dB3nQpKPYyEyCk24SUDaNpF4GjvjMkzk7aLs+tEG2h/GsdewN7/UOV5DZL9ljZzplf49exCf/0si+RhJi+RIOwS8Koi/ucwJSpwi1pNMQEgCMkSTVk/Y1alU7wWi6bY8rSYqgHr0nxKIN+n3ZpA1uLza+jH0kAKivzzxhtteLnH28P7Ld4KkUs7vlqGa2J8B1u15hIJ+UPQfSI/uAV4xAIV7SRmR7J8537frJj4jjL1+dtutRf2Gbw6ky+Xhi4RPA92A2KXP3gNYdpPTWzsGAam3bfpkfu8CT9ZGCHe9EUgKpLk9WcsV00zY42BpxhYIrW/Y1LsxxduZC2mY5hjBEggO+5pP2+8ol7MDy2qCCfSxu2wvsyZY9Loihk6THYDPqQ7KAJA0gWzDdA3em4z5rZTvpIN23+2hAdqckAQ5SdcgxbsVtex4mp+2NObfLjH2l8Vtme3/xnH0BR+yY4JsMHfpSlLxW7YM8z4BZ6lHKBp7TXTt+Mtvm0JxCwZ4DOxkOyCbQV6zbsXj5FDOFtco5u482SFQ5wgoujOz3mW7bh7e96jnsowhr6xmotS5JSUcylWUTJvHBbQIKH8CiSOr1D35hv+sxmx1IR7w4Xm7bbidS9s+lmB0rpeyNd7ZvMymn0xAALym5YMtp7HVZo+oLdZsp2dmzJ3ylzO7jb5+xkyOKLXvxLOcYIEdJeyjvL1wy22+nD5ZBeqcPQEKU+CNJM2mbrSkU7XYam5I0m7A3tG7WZg+qsJlJUjNpM69vzNjJLpK0IHsOlMANSN9UkrZP2uzWQscGph4gJNjQHul9FS/Rlc2gZQDIezQb+zmoQQ7JFX2HCPf07ZfN9ozjoJrds787SQD1qxXso7JrJ7MkO/C+HVJF/Vnbc9ApgbKDAyi1B/YhswehO5LUbtiHcgJsJDcj+WRr4tph9DEeCxP38YNYJjnSqerBsRj7Q3thChB7I0n74D7OQRDoCcgsltjN97XBU2Z7ZADju3Z5ze7j4hID5Asz9mk1BHvzf3YetOQkdSPbfZZN2QtbpscbyRCu0YP2SuDNf2dkhwScArkZSXpsYAOELGRv0sYuSamufR8EoIlFkaTZtv3dTzRZDDy1bh/woj1wgz/2HPZRHNmi5clmzWwPbY7L7Zy03ZU9iMuVpOoKA0fLdk49h7/J9Wy2fKtqZ+12xIfMk5X4THeCWH3QW0wMed0jI6HuxBRXzsrAHMgBuK3NP4N9FCM7brGS52SV6Tl7jd/frpnthyEa7WIjHXHhR8Im7uMHs94wodv1g0HE+rYNMJ46ywHuj1btOCGql+qxguyNOZe0N/cvX2NwSzWDqXycJH3svH3iPV2x2wtDdtenYZEegltLjqSFXM1mnbLQ7mGdChl77O12+ZtR2EDhti0HE2oMXsNFOzaMNub5vS9gH3KApbgWFuyNOfUaV/EIWdgUq7a785VLvxf7mAv2IbHUZiHvEUj8JOo2wCjv2+BXkkYJe7lffvUXzfb6Rc4oTzXs8RmVeW3tLF402xvgom4kK9jHXNv2xAzgEFneZcmlRMteO0dT9iEzO+LKLA0QCmpDIooUH9R5AFtc9/D7S6cwKMR02T9MO3KgsN2J9NVXDwYBrZbtJun3Oag7XGI3n2W7Ldb/m87Z97las4HQ2gYDoZ1te9GYneUAYnJ5ru/YJ5obVZYvyaSglNiUDSwzaXaN/dq+7a4cgMd/4y1+38tL9vt85izHmr7Utd/X00+Du36dRXNDH+4DFthhgdlI5e3NqJ+v4CVyb9nJEVEdAMYFG/xK0trSs2b7wrrN0J3t2VVXJGk9d85sL4lB4bBYMdt7wEwRqJSkJIyLtce/y2y/1eNEqiQcrjw1g8kojGMw4G0tl7MB2b5s9mzKMS5IVoFYf0/CFxEYeZDqkqROy/4m43Afa3RcANvDt0miyQPacDDS7s7BYIfe5WuvsuTBSy9C5lnWfi3laabk83nbRb1+247bmqqyq2bQtyf8L/8bO75HkkoVULOvwWl3wCe48qy9+J27ZEt29PqcRX3rLTtGqN203SypNPdRv2jf55WrPJ02VwHogIt6ZvH7sI9i0QYIp07aB5LdNd5Irr4BZcKGPC6WTn3SbO/17PuImCjUfs0+OF189IfN9uUF3vBurdnryVSJs3a3d+yDT6tl9+F53314nzsbFCfHLHUqbb+vRJLBQRtqNA9h3StM8dpZnbO/SatpHxLPXWAGeWvLHns7n7G9SZ5vurNqJ1v1oWSgJPXa8YC6h2mMy+K9v5jCiSTNA1mkSCNjMtR27Im0s86gkCbB1GwFrmBv3JJULNkb89yizbS8+TXOrEyNQV9qb8uO20plbbA0NcOsUrZgg5DqDGiCdXhBWAcg39q3T+2tPY4HvH3Nfp+DPicgZXP2u+h26NTOwdTrcB+XX7GvQRu7JG2v2u7MQpld6fmi/S52NuyxubvGcYtpj1CbYasO8d8OeC+s9ezr17A372bddtf3e7z50/gsV+y57GE3dtdA4B3WLIlBBjFTzTrP5fUbdghRAtbOay9fwz4o8SEBWelD0PGUpBH0MXLUCI3t2vUkeBxCvN+xyWAOk9rHD2TdVldvvXiwnEWuZLuHcwU+JdIgptNVv8ubf71mn3b7XXuyZkHcWpLqAOg8GzOBkMau7doddHnhOvWo7Xbq9ezJvLfHJ1lygRA48CxsHcjSy5V47PV79vsawsa9veqIT4ONgDbdoWMjIWbVo322t2sDndaeze7SpioxgLjxWvzShDR2PBsAjc8hfDPP+I1G9jzbr0GCh6MPfFbHwYmAI82RkQMc0D5BjGaHcSca7TME+CQGjofBnrmyjw8DeB4jmzCFD2DpbFqL52ypFctyjs2o07bB0t62vTjWtzi5otexJzyBmP1tLluVydsshufETAtTqWqzB2Uopi5Je8DuvrFvA2iPe6IH4NbD4sW11n785Ata/Ghjl+KzY90WB7hTHzu3GbwSqKNDjWdctPftseeKhwLLT9n3SSBmHPfRh0xWScoCECK2MQ1eA4nDScjz4DECniM4cEvsGUAWr8fflOYIPUdyDAmPHqYwdh9jyD6md+E5WB0n4DhhCh/AEsmEypWD2cDdTRuQ3VphjTVyu9Kp3DMJEin7JED3UJ7lMncEhDwbDcX7kYaVh7lKZ+zF0eP6imu0YHhcNUlgxzzXoJM7LRZ0EJCkbtMG2QRiyjMV7INiND33mQFXOoGQbosTe+ib0aaZzPAS2WnEZzTp0ELjwsUUAlszjnlK65rnPhGcwn164oMHVGIRQoziHrwkBum0vksOBnkMLtVxMOGH0cc4DniHYSEEBcea8F7ZkQOFw8FIta2DA55zEJ926hHOkOt27AWB3LLEPkjxJ4pnYSPzgFeajLQxJxw0OLkjeZHnYUoAYQQxLfkya8nhPThYPPom6bSdxEQARGLgSdfwuDYI0HnGHn1XcuF53IQE+oqQaOVxpXcH9sHJw9YQyCCXqudY1YdQD1qzKM5O4o25vcdsOoFoAo0eFo8OLbT+etZ3YtxD136fHkBHngMPUIobwzmWJJAxxBwem5hC3QGGY7rOP5L0A5I2oih6ehzXPHKgMJVOam7pYJZs9YbNTBXKLElDgfTFaZtJyTtix+jkTyCmUOY+BGKy+7sMXgn0tUH2xuNqpEUnBRuiBwhlizaYGvVBhsXh4iMQ4gHyCdiMaAEmhkPiTXUAIQPJMcS7+OLo7OWHEpR6HY6ZpRAKWpw9BxKNgfHBpAT4Jp77JKWB1j5l6MffLjwbN5TmRvbW00cb1hQCjQmHxAqx1GQUdy5JSXJzj8F9TO/TxVLLvo9xMOHHxsJYYwr/Z0l/V9I/GdcFjxwolOwBsnDSzvztO06JrX0byFA8oGfDowU07YjFI+vCfXo2I8oWJCDkcY3FPZW3Bp7YSMgApeQLj3uCFmAHO0YxnJRIVVlgjc3ahv1NCXgSCyhJhSn7PilJROJA+rgstiRlC/ZhgeYpAdM79wEsHngmJGn9GlRegfv0JJXt79ihNzQu6LAsyQz9kQ4n7tYDxnIwdsh9PHAI6tN9UM3h4AjBoPv0JSDFA1yHER93lGPwHtzGl30cRdGvhRDOjeVib9vRA4WRNBwePEg7bXsyFks8kfKg40Z6XT24B0nqIBNoL54ejSq6D8+mOXfCrg5AWoibt7jkHxmdZovTDKDjnjQ9iSjjcE9k4TBA35TAryTNLNvflCRS6FAkSdOz9nPkgbmVeGzNL9vxrmu3+Ht0IaGrumjH7rYgCUqSlk7a1+g61Ap4LttLted950v2bx57yhbIttblu5ZO2/NwZ53rmNM3IyDvAZ7kfcjkoPTmFIeb0LpGbnAPM0uHXU963TiYwLhG7uXj5BpGC5L8TOFcCOHeElOfjqLo0+O/qa/bkQOFo9FI7ebBi8Kt1+3yRGefPIt9pMH9QLFMHrkNAilZuEZqDMHpHlmbJDxrs86MDxktwJgM4GDgiCkZgE/KEydKLN6Zx07hNZJJe1zUtu0NbbrK4REEQgjoeA4TmQxkDsPBS5LK0/YBLp+jWCZOxkrAPCyWQTfScQA8eQLc3H0+qF5/w970ZhZsgPzMMxXswyFxadpegzfmqZK9nsyfqOA1uh17/FL4T7fFgI3IBeqjAABbkgYA2EiVoQRsvCRtr9mC4k0oPiAxsCTXrwc0xk0CGYsc0hGyB7jXrSiKXniY9/LNduRAoRSUMOJ8TkAiCQE+iV3M2yt23GLJ4folYEmxTLubLElDrFE6w5s7Jd3QqdvzLqgCwThcIJTtPQMZtdNzLMJNDBoBJYmrdDz2tM3WZNJ8wrx5gzcCyzzvu9mwv1k2x2PvicftsUMgptnkuT6zAIkkwJo+9axdxUaS5oH8and5A1g8bbO7cwv2oceTDJuGcpPNNhxkM/wcFD9Ph1BJWliyQR2Ni7pjv51ZsPsoFuMn+pG+Ku0BiyccayuMX09MYdwM5nEAtsPIcD4qFsKk9vEDWb6Q0lPPHlxbmHBOo+mofrFuxxRWFipmO2UnS8xckTuoOs8sSA30FEvTfNKkSgo0WRu7DEDmTtq1os88ZpecKgObI0n7+/YCvH7TjrOjrHZJWjpjx/P1gOGQpFze/u7tFsQ+ZnkBfuQRG+BevRpfefeTH7U31QszrOW5smc/a6NjL5wvPMMbd6cHoSJwm5CbIUmaztvfPeNg/bM5e1zMz9nt5+c54YvAVKMAMXCBx14xax8yK44s/02opjddAp3CyCGHBACZwG0DALQkdeGbrW3YY3Nx3qELObRPJBSiIXHsOQpkOwAbeXtQqsuR+X6cXMwT8eoHsFRSmq8ePMi+/BKIjjoGaAEWP4rP8WT1Ukxhu2EvXJ4Yt5nFitnuYU23oRYvMYGeWqYkm0AM2wAyhyXO1Fs4aS+enngpYjk+/OEKXuPmir3Anj4JWoiOZMLrN+1FfmbWHt9LC7wZnZi2GeRKmg9OU7P2gWK1ZSeV1dpMj23s2OvBo6fseTZX4PCJdNL+pj/3JY6jW1y0M5hPzsXf8JYK9sFoO1kx27ea7DKdT9mHzEfnGLwuQrz1fN4eW60BKzdc3gZR/rw90S7OctziMLLH3nU4cZyp8j7zuaENsnM5Zrr34GTUadjPOg7ZGwJ0HlB4vGofj02S5p9J+g7diT28Jem/i6LoH8a55pEDhSHYbo6PPmczcC2Hq6YOdXCHQ3uyej4oga0kuDtzBWb5KDZs38HiUfwYSeOUpnijoOSf6Yq9uVPw+p0+7KF85Q27Tq6npnBzz97QPO5jYiCuXAO3bNajFQfVAeAm+gOHux6eIzfiTbOftMdFMW0DjDdWefM/MW9vNtM5EIAPjpJpCfsaFEcqSZub9lzemrHXg0cczGxjAJWcOvY8HIz4OVoDqPme5cPCQj6emH0qwQfq2ZIN5Jtdey7Xu7zulbP2N10s2+1Pdr+IfdyY+3azPesgBjwxxJZ55LzeT+7f2BaCNCYAG0XRfz6WC91jRw4UdnuR3rx28Kni9i0b6NS3GQiVKvbi6Il7IZs7absaaWP2BLhTOT4PnZ4I9sZMJf1qjthHitdLJG0mpVzmRQvHBTwHVeiQWKplzzH2iAGm0IVzl+zYM0m6fb1mts8t2oeeBXB7SdLNXRukvNV/Aq/RaEFGOOw1m1ssuVR4xP5mv/xle3O/cY2zj2fn7bmeyzG9u7kKcjGwXrz4YnzVhQaUftvZcJTehOS2Fz5uh4pI0sy0/az5rL2uvXGNDzUnl+LFdF1f4z2i1bbn0YlF+x6uRd+BfdxasyfJVz9nJ2ZK0vYYFCTi2jhYvuPlPj66IPnogcLuUNevHMyyUVYYLUoSA8fahs3y9dq8UWSLNosxf8pOKMgV+DlIsqC2YbNjEiereGrtch82wG0BA+fT/7Ofg+QfPMHSLDjOOm60cJFczJXX4i/gL3/hstn+0m/yNyftSc88nAE5GMrOJL1RSdrZtt1nW7ftuU6aj5J0E2IGPdIiLMlhj09PSAvNEbpPT5m7+pYNUn4NBLIl6cylJbO9WrXH3isv2pqPknQF4q0p1npqhuc6eXJqu/b4Jw+KJO3XoVyfQ14qLpg6DIAzlqopR8kmMYV+y2aTOvfIwczRa19dN//eMwlQygUEbysLHCNEbCNlZ3okabCguuMaJEdAgs30riSOJ6H4S8+CQPV+x7Fw0XN4wGu+yGEB9t8zI0S2uxYf6Hegks3IcZjotOyDUx+kdZJwQJSkhUV7854CWZzXHbqQxIR7BPUpNpfWAw/wpBhkSkrwyFPRNyGtROnOHmBZJgNrq0MybPm0Dch6Hftd5Av8vm+BssMQ5sjIEdKSAjUCTzm1w9AIPFaA7SHbJPv4Aa3bGejy6wczXBQU66nFS0b1OVt1dhNWF203X6IC9Wkd7mMq4UXl4ySuO0zm2Yw6VG0ksjdNT83V/JS9+RPL5wGNlEVNIEZiAEzA0pOARL9ZPGezZ8TGS9LKmzYbk3VURaEM/Apo83kqhZB1IUHp1CMcqE/vq93kuUzVXSicxHMg2a/Z6xpl4JPskySVq/Y8nJ1nho1A4VTZbqd3KUmba/baefK0PfYW53ndq1bsPaDdiQ+2SJLGEyt9GIbZx7D+vv+yjyfuY7eFEMxFdmbRZukodkziFPzyrH2K9NQ+Jrfr2nXbDegpHE9iyh6RbdIQbEGc0ThqBhPooxJfkjRwMDqW5aB+rcQLcM7BgsQFfZ5SY7NLFbOdXNTXXr2BfdA388RonpqyQyjIpU8uPkkawjUef9TO3tytM9hqNOxv9shF/mY3btiHhbUVqHji0CRdPmtnc5OtXuf3XZ6258DUFK9JgwHUGIczi0fgfbpqz/dnLtGmzePi6iowmpA0VsgxcOj37e/uWS/qm3aIRBS/fHJsEHScAB/aGBNNHoYdOVA4GAy1s147sJ2CYmchVk+S5k/ZjA9t/s06AzYCniNY+AjwSXz68mgI4gkNmNfiNOuOxV0QPPVnq/P2faSA5bjy8gr2Qd+sNMXAkgAZ1aL2sHinTtuJJDev2wenCqkxS2pBbNj8CQYgJBBcrdjtpbIDYICc0fauPYfqNUdFk5M2EJphXXTdhD3i9Hk7maXZ5EMRuYfXrtsxyCVgASXWQvS4RE8s2d+9XACg/wRrvFbhmxAJnc8wSKEy5RSuulNjNLa8aG/hjX2eh7ff4mQUy8YhSTMOF/axclFPmMLx2eknzpntnk2T3HzE1njqzzLoswEE1UaWGJxStqzEgIvqkHoAG8nv0Ps8e5EXticu2QBhvwnxl8nT2AdlgJYcWdJFADKzS+QyZWZ2d9f+ZhSaQMBV4vHpiffbABdeqVSx78ER10VVOGarwHiCgLYkUdGIayv8PiuQVEBspAcgX35lzWwnRt9jCwv2upZKOUSfezbo29qx77PfZ+BZnbK/K1V/WSgxE766DTqES/ZztBxjLwfgNJf3xKZDXCKw7YdRG9ljx4lNPMoA9siBwtFwpE7j4CB2D/tFlgcWbu6EfcTzBEtTZiRlZxITc+c3NigkJXpJ6vfsd0GD11PdhSbr3Ak79mZrg5nZV6GdYsd2Njh7kxjPeo0X4CZkJHYaVJeYp+zuJodQWOZhdjuQ2LNPpULEhxYSos9AFRBJOnHCZreabXtDW7/tqB0LAsLT03xYGDi0Ic2/dwi8kxRX66a9ZjV2eR6ugnbk1BQfIms1u73Vih9LOj9LpfTAgyL+Xosz9jdZ37HX1q1dR4ISHFRvXavhNcjGUeYO+wB3/FGOwXtgC5owhQ9iyVRSZcPfQsHQHWC2JGbYGpBlR+yZxPINqXTFbC85YkF6wPj0HQMvbt1hir+U2JV+6w2OYSN7C9rjFmSX+ETtAeGFaZs9oASNnXWWSImbUe5JPopbGkvisUeHHo9c0uvwzUguyRNrunbdBn2eTZOUAgZUaszBPNCaNIQsaUrmkqSbb9hspCchhsSUs3CgbjjCey5/zW4vQyUnT1ZvdZ6KJIDMUI2JAVLb2Li+itfAaiOHUJf4MDKgj45Nso8fyFLppOZPVA5sp0w+j+uLkiNqUHwz6figWPsYlObbwMRIUqdhLxqeyZpK2wswSY8QOJB4wuehHiotWpKUhkxrAiAeaR1Knsg6AARlrtM39QAMAhBpwbsa8Tcl4OkxAmShP4bNBkAhZc97Mm55HvI1iDWl7z4aOepuw/gc5cAV6ZgjdJ+LZ1h8nbKHc1CmlGLCJZYMO3fJjjvfWGMGmUKZSmV77Hmy60narL7Fa1K3CTqxjrEV1z5QsjdBE53CB7Fep69rrx4c9F+Yst2dHrFZYkoos5IAiMRxifs1213pWYCn5itmOzGJktQF0DcO8Wrqg4yEkiUpB/Wq2V3P7AIZgRyPecYWGR0GaPy7mC041HikiiiW1CMST4YuaBhbFBss+Q4DZDTf+z3QvHOAdGK36B48YTN02J2uxH9XlNjjcaVT0g0RgTOzzJpurNkhFARMSbxdYn1LYn89dlzkYo6PiznwAHsP7ciBwnQmpaVzB2cQk4CqZ4EmVw2xY54FuFy12S9yZ3piComxdGVswYmFQErDETt2GJO1Dxsabe6ehY2ATncMCzC51+hUL/E3o2tQNR6JD07EzErMUhem7DnkSaRq7NrjE1lsR21vMk8oyPoNW5Sf3reH0SStTnIve0BhF9itvToD/Ucfr5jtna59aPGIhWchNrcDCV3k+pWkpZN2aA0dWHYdJURJqzYLh+U7F7Gb+5Ro8n7LDD4E83gP3is7cqAwiuwNHDOlHACEQB3JrNBElKS1a3YsBwE6YmIk3ijIjegxct3mHFnSxDwRw0bMlsTfhE7MHobOs/GSkaA4gSnPYkK/yZdskOLJQqVv5hm/dA0CyONgNGm98LDtFI6yv+2oGQyMJQFo0hOV+MBMhx5Pbe/HnrOz+G9dZU/Or/68LTtGh3YqISrx2Ln22i2z3XPooXhrqjAzDk+Px8hdPw6Wj64xltrHRyQLGi1oolP4YBaZMSGbN+0TNVWdkFhGZWvF1uvyxBQWIaGA4pA8RiAlnWCgQxszLRge9wSB00zePs16Eh9oY44AWHoEsmkB9gAhehYC+p5DD90nxQN6+iAQ7XmfZJmkPU89GzMdKEgk3vMu0AUNJdMkLtMY5SHOzqFrGtc8guTrt21mtu1guklwmQ71Hh3ZbssenwQwSiR0KPZOrN/YtP/eEUuNgM4R/oOJJLAmee6T5pHnGnH7ODoWJtnHD2K9bl+rVw9m2ShWz5N5RqXyaOMmgCFx7Vc6RVIGnsTsmCfGjRgIet8eNpLc9bRwuRZHOIlS4o8rYSZm0oIU/2TvYgpjnkI9rCkCz5aDxYPDVQIqPniMxhaBRk9CTQuAJR0QJUccKHx3ipmVpLYh9SVJA0go8LyLzVs20PHU/qY4OarU5GHYPAd7yzxatbQeEBjzHHrIPOxZ7BJ048g+HsM1jgtTGIIm2ccPYtFwpLbhCpmCaguegTG7bIshD0l4uuBgriADbnvVdqOMI6uXYrI8xi48PonG3Xg9rjFyidKi45HKmJqxgbyn4gNpvVHCS9vBMNM3oYOAB7jS2BvHmkdA3yOHZGmeSuyK9AD9TMZmCulQ5LkPmiPtBgNoBOHgefAkfNH666ndXduome00lz3vmxJ36BDpSWgk9hZdqg6glAA9RY/jdxxyXXFtHFqIx4opnLiP/ZbKpDV3eunAdpKc8ehHbd60T7PNOjCJnvgzmGgVqIFUmWf3RHPP3kioZJrEG+84MkAF+JYWcQ+4JVa0A7iS3LYSA7b6li1lJMUPMPYwHJVlO4SiCwcOSkjwmCcJhMAnqwAwm06bSQqu4dmMmiAz5AGWOP7gEp7xi3Ggkd2Jp6QleS88oJDWg2bNft+eMA76JuTp8cQ5E9DxZLbHNU884FHQIcS/d9zDcWEKJU2yjx/EUpmkFk4dzORtr9bMv+84NjQ68YaEzUB4WCU62ZOo6O6aHdcoMTj1nOzJ/TuCPoh1kvh9EQPn+abkAhnBIu4RKaZYpuk5rhlM8aykkdlxuOsJ9JF5DgJxXU4Sg31iiOtbNezDA8gs82w0VCGJ3MsSz2XS0fRIWBG7Ra7duC5Xycfi0ZpEDJxn7A3A/TuWKh1wDWKHPe+b1iQXGEvEA6eudxWzD48dH6ZQE53CB7HRcGQquRNA8LAHxPjQRpLLOiQ7IC5mUIJYPUfMCoE+j3QOxcYQu+BhTTNpcK+ByGse6kRLLKdBcaKe+EuK8aFkF4nL1BHIri6y+C+5xuhZKfFH4g1rHBnMZOOovEKHt2KF4wFpXBQclUCINaKxRZUtJE6OIC8MrWkSfxNP7CN5e+ibBnCpSnxQTad4HyEjVpTWTk8MJ3keolH8BA6+h2MExo6ChYn7+IEskUxqyojNomDpndu2a1iKnx7vYeBoAa4uVMz2lONEvXFry2xv7zti8aCaCDEUnixqcrX0e/biOAShWYmzoDug++hxB9Hm7nGZ7u/a32QcFTYGTRsUelx4ZJgYMfLo5sVLuvG4xuh9JmF8zxtei7u2tWIzcB5hdHqfUd9mYzxxYRSPSgfq1j6/7+k5m/X36K+OI3yBjNY1Yr/GER4xFrftw8d8h+JeHgcze6zA6RG+1yMHCvvdvlauHFw/kxiK6hIzKbTo0CnSM/hoI0DRUU9JKdgUPULe5AZsg1i4ywCE0MZdcrA1FAMk0NgehxaXR7+STv+e0AQyEsXF8nKe+DSqn+xgkOlAQuyvJwOUNm9ina6/cp37gEMiJS1IDNSpkpPHlU5hAZSg5JG92YXa3J5EKXqfWPKvy3MoG3NTdslkobLDw3dhu2LxDgFYkh0rQDcOS7z3yT0H2ZEDhal0UrPLBydhbNywhU09myoGGQNAoBg4id055DL12GhgX8PDCMXVtBtHRiLZ1i1bm9JjmZwNsnsdZie4frIjLpG+CQA2T4UCckFTWMFh1JqWGJDFlZOR+H2PQ9Scxvc4mMLGrv1NPPFncWOQPX2QdM5YDG4j7QCvcV3UHs1GT/JPXMMDoMPLclyYwPeNhTCJKXwQiyKp3z14oR6HKG7cqhGNXVb2ry7Z2cVkHq0t+nr9Jl8jbt1WTzYtxbVQPBUJtErxgX6pykCfgLyHuaJDC2VOejIW2117w6J35XGN0YHDA7LJCMh7Nqu4cYue9YZ08zzrBW2aUQAReQeQj1vzmkoGevrwxHzTPMKQFkdYAgG2uNV2JAZsSTi8jYa8nnjWxveDve+YxDFmH4cQvk/S39Kd49L/K4qivxrnekcOFIZwhy08yCjpwONGocWP4mJoE5AY1O3t2KWvxpHp56k2ggkDlNXrYCNpcSRzZVETwwbgl7KCJd7cPQsX6iXCN+s6so+pdjGNf5JYkRzZxw6WBEvMAbAcRx95iBP1sKbtPRuEU3ayFJ+56jUYhBODnMw//NKbHrcrgUJifz0JGp02xVfa9+ljkO0DHK69juegoELPHIngGuNgElEAG+5zHOE9R8rGxCKHEJKS/p6k75F0S9LnQwj/OoqiV97tNY8cKEwkEypXDt7USDg673CvkWuXKp7srtsJHpIjwxNYEM9GQnF042A54gZkS8wq0cbbBT1Gj9Gi4okHpGf1vO9RzCoFBPgk3piJ5fCwayQt4tko4pa2con7wn1SfLFHnJ0OHKSrJ/F90rN6pKHimgekRIn4bCM9a9z1ROL3jZWaHIczDLGAQ4+HBaTxOQ4whUlQ42ArD0Gy5sjYeN3HH5P0VhRFV+5cOvxzST8s6f0DCkMIShrSB+sQyOxxT9S3bJYOF+AxiFfTouSRbqDJuF/jBZhiCmnDc1VrwKxH+1k94tVxS0K59BZD/Fg8crfnSjYId5X8AxezJ4mJbBylCeMK2rreBfymACXoPDJDxOJ5mG46UNB9EMsn8bpG89STaMKhOR5WiWRW6DDMczkD6zMdjMYhCk3m8bAMBvHqFktSgKEzjsorGB4xBjbyWMUtjs99fFLSzXv+fUvSx+Nc8MiBwm67p8tfu3lgO7l2PUbZgMQEeuLwiK2hSgoew+oBDsaHnoWyHj3iwJTYcBiugXHUpyWBYI82H22a9D7TDs1GAoUkYjyOBdhzcKLxOY54KZIRIkA3jjAOT8Yt9UMHJw+7S/OMvllcEO+1uDWDPfF+mIFP7k7KdpEn3jpeTXhJiqL4SgJx19/DAJ6u+zg2cYfhQbKP50IIX7jn35+OoujTD+Gm3rEjBwrTmZSWzy8e2E6adaPIk5RAZdVOm+1dRxJIs26DkHFkLNLi59lUCaRsQFUJj9BxXPeDB3hSbBjGIXlixyDj0FO5gsICqFavpyY2xW0R40NSMBIDOk88K4EQugZ9c4m/qyfGjSyCd+EZv3Q4o/Hr2VTJe0HvwsOaYua7Y56Re5jmkMddHxdAxE1gktjL4ik+gMkqrrjEePa+i/d72Bb0IO7jrSiKXjDaVyTdC1hOvf3f3rUdOVA4HI60t3MwEFm7Yj9vebaCfdAgJqBUnmW2kjYCiq0ZR2blOEpK4ancEVtDixvFyXmyYSM4DPRgM/KUuSPGx+MmpI2517fHnmczoo0ZkxYcZe48gfZk3Wa8WNG2A4STjWNzp7nqmYdYpQP6oBhlyTHXkUlk7wZ9E8+4oW9C5Q89fWDMIJgnttdzMLIs6didj4IXZhx2GLI4R8UiSdH43Mefl3QphHBed8Dg75f0B+Jc8MiBwmgUmYkgs6cWzL9PZ+K7ZSnloA7uN4lZEFqUPFIvFGvnqXKAQd2QCegBQiQRQTVXPewCxUOxzJAdZyrx+0ylx8A6gcvI877jmie+chzxO3F18zxxpHSfpRm78tA4+vAYxXlSrLRH9obmAGX9ZnK8JlGstKciTzKKty15JGnIXU+Z1h4NwgjexWGweJ49IG7Cl8fIxXwY7uWjY+MrcxdF0SCE8Mcl/YLuxCP8oyiKXo5zzaMHChWZiywtXCQPIUk5iMsikOJhrjyadZaNQ/h0HKcrzgTk5yxOQyk9SKqhxCCJAQQydJ34cXQeAIGxSrBxu9xvAJCJofOIQo9D6DhulqmH5aP3iYcJxzyk9+XZ/CkLmto9fdB90jfzhC6QubQ8YU0hdmwctbtpbI3GkVwBQMfjPh6HNBTZOORiDiNu8VgBxzHG50ZR9LOSfnZc1ztyoFCRvRnQwkQAROLNiLJ2PWWryAh4UoKHJO0DOzCO+DMCIXOL89hHvwdAHjTWPO4gcilRzFZlvop97INUUcYxLvZ3amY76T56gs8JCKG70wHoyDU2Sjz8zGCXRArVsCVw60mkolg7xzXo0IIxhY5x4VFmiGvjKBFK74IAguewMKI+4D7Hwa4hewbyPpIkuI9xAKWxXCMmm05aitLxcjGP0X08djtyoDCdSWnp3MEu4vUbm+bfuxI0YEMrQq1dT+A4bawklpzN8QI+6NtAx1NyikRzaSOhpB1JatQgBmgMLv/50wcnJ0l86q5t7mIfxAR6wCsleYzDFUkyQ6VZu9qOJ1CfNNY874KC5OldeKSKSMuTmEIP++spPRjXKJ6PvB8SJzFhaIJjmtLhbAQSKhK/z3HMEdQ6hPViHMLpqHvqKisIbLsjrpFA32GIV+Pfv5+YwvBA2ceHbkcOFA6HI+3vHgzsKMats8/uY9pMxgFScEGAxIj9GoNbAsAePbr5EzZAoAxmj5YWlYcrTNkb2sBRv3NrxT4s0PeYnqtgHxS3RRuixKAwkbYXPw9gI6P79CzgeZhDniB7YrdQm8/h5qZDTTSyx56Hbe82bYA8NV/Ba9A8XL9hy2R53LLEehKA9hyGaW31hJvQGKeY13GImscN85AcwDJm+IR0NMSrx2GHJXd0ZGxS+9hvyWRC5erBbjgCIZ4MOYzDwAUhfjwJMXSesmsJQ+Tba50WVXexF+hyldka1PeDd9UGhkPib0LZxZ6KJmQedqxQJjkYe9P0yLC0HGXqLPM8B/2mPeLDGbqxx1DnnL4rlmVzrCcBYozPXFrCa2SzNkhZu06sKYeb0DxCMfxtju2lBA0P+0XAchwVNjyAK64R8KTYSQ8IJ/M8J4HXuEyiy8ZQ0eT4AMswcR8/iHVaXb315TcPbF88d8L8e4+4NZW5213bNtuXL9j3IElZKB/XACZw1PfoLdqfj7QSJb7P2WU71m52nkEKHVZbAEytCjd3jcBpC0rlFcrsfiPrwriS2E1N8aqUcCBxLB6xXy4hZNgIPPF+xJo+9uGzZnuvx31c+doNsx1jDh1JZWQXL8SPQb72hn2w8rCmcUFGvjSLvyE23QMgaPxhvWpH/BkZHXpcZRxBRJ7/3qPpaAPLQ8ks9riPYyarHB/A57CgsSaajNuOHCjMFrK6+OwjB7Y36jYDUSx7Ms8gULlqy1R4XDXE4i2ett1FZ87wRtIH4HjtKpe5q23av2kAsCyWeNMcDu0J3+3Yp9mC45t2AVievGAnxLSaDIRItNzj5ibQR0yKJ5Eqbua7xzVGjLyHsScWbnbW/u7ra8xGkquR5Es87mNKjNjaZrYmB3IvlvdE8oGtDMQpN4ywHUnqOA4kU3Aop+IDErNbmATiGL8UWkPg1sN4Uh9NYPQ9jP04WLy4njOPxXVje/7+OAHH6AjfK4LCEML/JOmvSGpL+nlJH5L0X0dR9E8fxg0lkwmVpg6eTJVZm13IZhnnUoJGvmgDnRZky0oMPAn0XTrFi2e9RVl6NriVpNSj9iJ+9aq9cHXaDEDIRZ2Gb3b7yjr2MQcxWQRM203e/GsbNbsPh6umVOVyepYNAEBLvICiHMeAATKJFHtkbaoLFbO9XLLnUD3vqRRiz+W4IsYSv8+LZxkgTxXs+X7lLXtjPvtIBfvYrxOrFL/GbWnaXp/rWxzaQBnMU9O8rpHtrttas8QQe+Y6xXwT6HO5fmPK3kiHkwQSl7F0MbOHwIqOx8I4ax+P3TxM4fdGUfTnQgj/qaRrkv4zSb8m6aGAwnQ6oRMnDwZMnY794ff2HCXo9u3f7GzYsTMeoVhaVOpnK2Z7e4E31XrDHljr68ykfPQ5exGvlO0FeGuXF50rb9nv88brt8x2T0jAo0/Yrq1nLtjjZrvBfXz2t+x35WEKiW3c27bHVn4MtY9ps6HsZYnF2T2MJrHptMbPzjJLfQP6oPv0uABPX7IF9Z9c4sz2RLD7WT5pHyYqU7xe9Hukixc/ros8OeMwAggeRQRan6mkpcc8c8CyoUNJYxy1uT1VqeLaoSSrHCdJmmOefXz3N98v6V9EUVSnJIk41m739fJXDmaG0hDIPDvPE/GJJytme/SE3X7tOk/WNrgjp6ft5yhkGWA8ecpmt54+40iIiWzW88Ut2wXy5uu84RWKNoB47lOPmu1pyMiVpA9dhDJ2KRsInZnlhbH1bMVs/9or7K7PZOzFgIClS48OsucTPft9nnr0DPaxPwaZoZNnKmb7o6fsb9YZ8ML61S/Z86zTsucyJQZJUqUCnoUBz8M06Dp+8kP2NbYdeVJXrtprEiY5lTiM4/ylObP95lUHQE7ZB9HtVfsaOUeSHjHI9KxTM7zPbN622cjqvH0Q9TzH9qod/+5hCkcx4/0OQ8fwOAE+tDC+iiYPwzyg8N+GEF7THffx/zmEMC+J/afv0kqltD75qeUD2/ca9gB841V7kkhSDzbe+XmbjSkUeMNLJu1BvLJin0S3t/nTfOcL9sC6kL2G10gPbWA584SdOfmJR3nhurJlL6AkEVirsxulO7DfxWze3hDXGrzIE8753m9zCI63bSCzuWtvFF/6PLvSiSkh15nH8lD7df4EM6/DIdTVjiCjPMUHpwwcIvtQTYekoySp3bbvYxTxBtDs2/dZzdvrRTLBDPLCAuma2nG3SUes3sK8PUm2t3i9uPaqXd++AmEHJZC4kqQGJJ5RCFHN4QYnoENhNRTyIjFT6EkuGoFSwLB/TPT/jomNufbx2A1HTBRF/9e34wrrURQNQwhNST/8sG4oSEoZd9VqxY8b2N20j9W3r9maYHmHWG2uYG82i8s2CJmf82gl2otKL/B9Vpt2dmYo2AvCdorlNuYhUaTVtRfgfp8ZoXIOMpiBidned5RlOwLzuDLLCUhpYCPrW5C96XEjQgx8xuPCy9ggo5q3DyyZBB8WTp+3s+f7fRu8UsyWJD37pD1+H0t9Ba9RS9uAbLdns2etHr/vpXnQC83bh5rbq8wF7O3HX59JwooOPTNzfDij3wwACHkUEbbXamZ7txW/bKCnljReA1i4w4jUG0syyzikcQ7LjjNTGEL4vZJ+/m1A+BclfUR3Ek/WHsYNJZNSpXTwhBws2GApk7ETDiTpi5+9CvdANUA9NVftQb66Yp80H7/IzzGdBd0x8UTbnj5vtm/27c2q2eYMuZNFm709UbTvszNidqGYjBsDxO87m7YXncUihxU0OvbmPlWy30XewVJ3u1AzGNyElLQjyUwGk6RMjkFKqQTsWMYeN0UxW3PxrB2asLppb/4Ls7x4L09DLemIt9V8ZI+d9aH9TZ6YuY19bBVtgPyG7IPqTs5Rro/0QvM8fpfP2y7oftd+nwnw0khSvWYDXPIGlaZ4TWo3bcaSWGiS0ZL4AIdVasTZ8+P4+7h1oF2ZxWPQOjws8+zN75V53Mf/bRRF/yKE8K2SvlvSX5P09yV9/KHcUGKkueLBJ6g6TDSPUPjv+oGDJW8kieaRo2iKtnfsjTmdtgfFmRnuJJccQ4F6oHwWUzb2n2nYAFuSmmm7BF22a8fidbLsimwm7d90IxvEPLrIQVnENuZTzKQMhvZ91vftxZNEjiWpXIYqHk+eNNv7AColKQeZvxfP8mGBLB/sOVCAcSNJnzhlM+FfzZ0y2ykBROLYxsjB5uzLHhe9IcjedLl29x4w8pQwXi7x2Fu5bX8zhzdey8s2k317xZ6ru1uOtRNA3wDAVtrBhJOqwtyCDcLfeIkPmcQU7m3X8BooXg2Abjji9SK24PgxAnxs4XhL0ujr7PH3S/p0FEU/E0L4Kw/rhrKJni7mrh3Ynl62g+A3IDFCkrZ3QfKgbH8wR0U0zUA2IBUH+OJldoF862MACh3jLsiejAuNK2Z7slnDPtpVG4SnB/aJuJ1mGZe1rp0BWkzbfZzIMvHdjuxvstNj8JpN24vb4iyUvnIkV2xt20z2C8/Zm26nx6fYQtYeN6BYI0mCsES9vmfP9YtTzDptdm2GrZCO7xxbKNiMZbbN4HUqZY+tUIRyk+JNcwrCCnab9hzK5XhckCTYlOObUTz26TP2euAp9ETgtd202fTKDMct0rvYuG2PC4rblThm0MOw7W3FjzGOa+OoUhOX8Tw0C8c/+3glhPAPJH2PpJ8IIWTlghsPx1IQR0SATpJ+7Zevm+2Xnjo40UWSuj1mQW5ch8LwcGR+9hnW4mr0bQB8NmvHRkrSwsqXzPZByWYgEms2EyNJy/Cb5lPfarbfaHMFGUqqKXbtbJZ2xqHpGGywlc0xU7gAG8VWz37fzY5H5xBK0HXtBbhaZoBBjM+2Q6rozLz9m4+lP2+2px11oHeKnzDbswAKK1lmnfZ6NqBLDnhczK9/zWy/VvndZvtyjuc6nP9E++5unQF0EUICPEzhGoiSE9iqVnl9pozxESRBeRKQiI2k2Mh8n59j5cqG2T4O8XWqEOMBbHGzi48N4HNY9D5gCn+fpO+T9NejKKqFEJYl/dmHdUO9UUbXegczBNe3bZbDkwzwXb/rnNlOQvKemNjTi/Z9ZlL2JKnkOZ6kmolfrzfRsSmdzLqtIagSg6n9cx822xs5W2Pwid4r2Ed+z94UEyDInN9mcHv7lB0xsdG1Y6EkqZK2gcyJ9KrZ3pzjjWIwsDe8p07a99BybEYk8VN6jK9xfsrOpM6v2GMvsWtviJK0+PRFs/3Nrs1G5hK8qRYhMzi1wzWDw8Du59neb9p97PNasDb/jNm+XLHvIZXk+LRba+B5cMRoSjZDRq5dj1HG+AxU06F4WElqNEAPtGs/h6eMIwl9e0pWkh0FuZhxyN4cKTsKWYsHmCf7uCXpX4UQFkIId1fQ1x7WDe13EvrM1w6ekIOBPdFWb/PiSFpxzz8HQrFFnqwt0IKjk+ZMllkQMnINS9Le6WfN9vZF+10sXvks9lG+/arZnp4/Z7bfKj2OfSzm7HFRfstmnaK9GvZRWHrSbJ+DcSVJC3tvme2Zhp1c8VLhAvZB4RE3ynYs02LZERs5grJsWd6M5tuQ+T6EChxzNqMvSemh/SyJALIhjiSnSqpmtg8LfHBKb9kAmNjGQY77KPVrZvtuy641/cqbDi3Plj0P9xvsPi4W7Xk0N2Nfw8N0P3rGvsaV27B+O9ixHoiFU0JYfZeJgSyIcPcciSa9tn0YOC4agccp+/hYM4UhhB+S9DcknZC0IemM7oDCpx7GDQXZMSWFvP0yUxBvIkm1XXvDohKfo4g3f5pHPZBZuRK4+PyHqzbAKHRreI3Shn2N8i1IJCmwRErr0Y+a7emOHVtzsvk69rFRssFS+8nvMttLbXa/1VP2N1lscdLN2pSdDVvL2+O3ucNjrzJtD75i1t6MEgleXHeaICM0dMSfzdggu3LGToiZ3+f3Xd22x/e5BXv8vrVv34MkrY5sQDabYhY6DSLanTM2C72aPod9LAxt/b8PzdnAdHDpNPbxxlV7faYaz5J0wc79US5tj98eaJZK0oyRzChJ+1U7JMATM9uFLGmq+d7rMAgnEe5+j+MSMZHkECqefLAsHPvs4/9e0ick/VIURR8OIXynpD/4sG6okIv0kUsHD8Ir6/YkeOkrHDQ7v2RvvN2ePUlKnAOi/ab90amP+Sk+7W6NbLmY2SwPvHITguChzF33jL2xS9J6wZa9Od2y4xozTRYkXxzZC3ByCMwV/L0k5Qv2TjAKjlJjkc1Q5JL2fVLYgSTNAGl05bZ9n9UpxwAHq5Y4I3E2BVJFr/2i2T64accGS1J46iNmexLipaay7D7ebtkbbz/NG3Nm0UZChZoN2JYr2IU6aZshprFJZTUl6elLNiDr9vkarY79mynwCjw1fQ372BrYIHs4sucAqDpJkpaX7e9+6xbJsHgqUtnrgae296GUuYtZNcXVxzFhNKVjzhRK6kdRtB1CSIQQElEU/YcQwt98WDc0GgU1ugffVqdrTwJPmTsKIiYh+ak8D+AKZAt2+vagSCe5D5LL8IhXRwX7fQ1fe9lszwDDIUlzz9kopVa1QeMgwfFpxV7NbN/N2lmo0007lk/ipJxEl909pxdskPFazo6/LGQYbJXAYzQLU2SnyS6+q7ChXZhnMDVXs1k8stR0BX9Ds6gn+2V5JGnOT9ssc+mWncEvScOczVhSPGAq4o291LFBeDdvg5jHT3LSzWhkb8wLMzW8Rmto30ezb3+zzJDnYTphv68SrPGzZQeb3rAPX+trdnsmx/OwvmUf6sk1LElDkKQZh2GpvDHELbq0DI+ChaDIQSC8V+YBhbUQQknSr0n6yRDChiQHef7uLIRIaYMNWQSvairFp/LFGXuAVgv2RMoleWOmEl1ru/YmsLnLsSC1qr1oPDfHcYkjcFcOP/Hd9gX6vAAXd2/a7RAsvQOxfJK0kraB5cLABn25OkvSbJ/4kNneSbAr/XbHZncX0lDXFZjEcVh3wJp3n3jC3lSXspwEspW1pYpKeXuyF/7D/4p9pLJQHeNkfJZksWe7hweemMLrdpj2iZ4d07J28nnso7RiJ2xl5myAMazwdkESQPsDZqF7Q7ufDKy/G4FjTa/X7DF+vmonBwWIRZWkJmhHnj9n71X7+wzoMnn7wNzv8T5CySgTlm+8duzL3OlOSbuOpP9a0o9Kmpb0lx/WDYUgZQyWrD+wb/n6dQYpVCf0pddtFH/+DDNwTbiNF79sn9r/4A8zuH0sacfa7US2aLQkXZ23JTvObX3ObE+88kXsI1G0wdLwjB1nlx0wQ3FmYG+qCUhaaFc5diwJIq2bA64EkgYB7P2+TeORgLYk9Uf2+KUM6FKGgdJywR6/i9ucMd4p2i68AJtN6vGnsY9u1ZYzKoFUUTFw5jCFJqRe/zJeo1+z++k+8S1me6XJFU2iDFSh2bQPb/NpBnQRVCe61bQPRZJUa9uH3TPT9vidj/iAl5+xQXatbx+We7APSdK1FXv8djpQjcSRuEYMm6tkJfUxDtAXE1h6WMDjlKF8rN3HUfQN9Zf+8UO8F0nSKArqGEkYN1ftTbE6w67GuYo9eIpQiumLX+EMZ6oKUZ2xF9gXrzDwPP846BRu2oBOkpK3Lpvto0U7uLz9sd+FfeTrNksX+vam6nEfX848ZrZfGL1ptpd2OT5tY8kG0E/XbdkQSdov2bWiKS5xLzCLF0CncBTZC9JCvuboA8IjAPBJXMkm8bl/b7a3HdUa8h+23fFbjwMT7rBC02ZFRy0+qA4atvMl92s/bbbv/87fj33kUva4CG37HlJ9Ppz1RnYfVJpTktJJe42/tWcDtlSFmcKy7DW8nLKfdXvI7O/ZE/Y8uw44fmPFPrBIzLB5YgoPI2t3BN6gDxJTKB3TMnchhH3dYTqDvlH2NEiKoijiWfEubDCUNuoHb4zb2/bC5QnOnanYgGweQOPTT3Lc4mtv2ItKOmMvGM9d4MWzD7ph3SJnMBeKIH46BZVCbjEjNCralT7eWP5Os30qwRUhzgxsUNdL2QD6yuKnsI+Zgf2uEj3+ZqWGrc23V7aZrfaQDwutgb0xD6GeWTbBLuqFns0qYWKPuO720qN22EDimiMmsWmzSglgF2qjCnaxCBqYu1+zDySSVP2QzZYPP/JtZvsgyQenETCFIW8z+r0sL/mdob0mUViNJF3fstdn8sI8NsNjrxfZ7yuXsDupte11UZI2IOexXLL3gJNQA1qSVm/YwDFbYI8TStKMg9WKWabOwwIem5jCQxKvDiH8Xkk/LukJSR+LougLnr87EBRGUeQpnTB2y6eHevZU7cD2Z0GuYAVOkZK0Csmst7fshcsTDlCt2ovjzo7tvljZZVfNYs5e2DItPmn237Rd0OEVO9EkWmR3UMhCAPubP2W2j0oV7IPAK8VwTHeZKdyrQInFeVZpSo3sDYuYwN0Ojwsan/WOPTbPTXO2dxfKst0MrC1Z3weB4EX7PsotZuwHc3ZYwCIkD83l+QBIrCgBPkmKgElB1+4tBp6CpLJR1v6mno1sACx05ACFM2Xby3Jm1ma/soF1NvdH9j6x17UBsifh69nzNpD5jVfs9XtzlQ/DzTpIGTWY3U3AIZGyk8dRgu6D5z4+FKbwa5L+M0n/4EH+yKNT+AlJL0dRtP/2v8uSnoyi6LfezV1if7JdU7WuvXDd2uTB8+UvbprtzzxrL/IphyxIs2kvGhUAjZUCLzrNof0uNuc5QWP5FLiPazawpM1MkpS03fEhBe0QDyhJKyXbfdwDuY1Hm7+GffSSNrjtRHwq78pmTZsOXTEy0mmbztnMQKPP91DM2L+pJh31ftP2hpboQHktGJuSlBqAAPaM7c5PNbgPwYaVPGtXVZEkQSKJhrAeODbEUcEGQv28PTa7aU6kGsHY22kx090B2Zpa075GIcWxve0hrDkQHtEf8j6z07H76Pfju1SpjF3OoZ/WrMcrlOC5z7ju4+ME+MiiEFzyZbH7iaJXJZ/39F7zJJr8fUn3in017/PfxmYhjJQ1MixDsDejnEOb7+Kj9qLxpc/bgcqnz/Oi02rZmxGVMOqBG0bi2DBPgsbglJ0Bmmx9xb6HDp/KKcljVLTdUr0Cx9EtdO0M0BRUhOgX+Jt2ZW9GW117U5WkQspmCjcbkC2b5AMJJYoQUzgEWRFJmqragG4qYjBF2dprOdu9fOEUs7uD175qtqdg7A2hXZK6Ofu7pzv2wUuSoj3IdiWR+CIzmomWvflnuvZ6sT/9BPaxVbfHbxsqPUnSzp49/s4s2AC56SjTOIKYrmbXBnTzRRbgSCXseRYAGHRbnH2cTNtbOIExj40j5vCDFjNI9gAxhXMhhHvdvp+OoujTD+GW3jEPKAzRPQqZURSNQgiev3tX1h+ltN6qHNj+8nV7ws9WeACfO2kvTM2mDRBuXmWB7OqcvUg/dsk+wZVzDLZe37A3o5UcuzM/MW0vGpVZO0kk2eE4uhEEuHdLNjO7X+D4nVzfXqT3ijYj5AHQM137sLAy4higV9ZsgJtJ2+N3Ku8RU+YDhXkPDuBJLrqrXbtkmiTV2jaAeLxqV+AIfcemOWWDuigNGzeU1ZSk0jq4bh3C6NEZ+3DWK9lrUmbP9n7cuQ/7WcK2HTNbOcF95NJ2OImn0s3yrD3+ml0bTBETLknbDWC6C/bhrZphJrw3sr/Z/Kw99q6XmbEfANvY2o+vHkeAbhyuXXIve8Dt8YkpfKDs460oil44qDGE8EuS7re5/VgURXZ22gHmAXdXQgh/QnfYQUn6o5JYjfVdWrMT9LlXD76tl75ox9Y8+jRnnu3t2YvGlZftzWh2iZmr6Yo94csF2PwzDApPnrBFcz2nkWt9uzzcOSi1W/nqL2Mf4Zodt5hfsheu/kl2W5W27CFZTNvANLlpf3NJalyIT45/7JRdmWJ/YB8mSile5NfrdrLK4wv2oaaU5D52+vaG5yEGKnl7jBeHIAcD2bKSNASpl1TWBvq987ZotMQLfG6DhdE7v2bPo/ynfofZ3ptm+ansTShZn7CfY2qHl/2n50AyLHMOr0GyS+QhmcswS72YswdoWjYobEfslm337XexsWUz+jlQwZCknb4N9D2SNEkAZIdR8WQcjOZxsnFlH0dRFF8+4ZvMAwr/iKS/Lekv6k4W8i9L+sPjvpF3bigpzVQOHqQ/9MO2RIrHvVZvQsD10N5Umw0+iW6s2RvWKmhYtT/JmcNP27ep5WADEEmaDnYw/27KZunKp22GQ5IGWft997J2rNNGBrKLJN1aAvFq2RvzDMV0SUpE9sK16JBy2QP9s3rXZs9WepxIVS3aG0UhabO72YjZ37ORzRp1MuzOJHY2Rd8ENjNJShTtzbtz0ZasSQwdbGTPfo7BDifuJHOQGQyxkQHGpiSpBK7wLRsgRwneLkjLM5tkgJGBaiN7ffuQOIj4PqdG9sGok7T7KIrj8AZQhjSTsYHB5moN+yBLpJiRSgzteXQY4tXjsKNyH2TRIWUfv1vz6BRuSGIRrDFZMiGVDBatCXUxPdltxNJ99CP2htbu8sZMtY3pYHTSIatQ79ruhbMpBjq5tr1h5dI20xIleGNOt+1rpBv2Ar0wx30MYcMiQJfoMOtU2LCZksrpCl4jnbZd/utNezPyxGRBOVS9Vbdd6U9U7PhMiXUdczmOxSNJmsWdq2Z773VgviR1QMtwqgRzmYCUpAhYaA8HMiI2BrQ8W0UOschs28J4g12bYUvNcNhMf/aS2Z53ZAZXh/aBI5u1PTUj8Ry5PbLJhQCTaCbN7yILeotpSLCbWeAY5Z0Ne23tdz06hZD5S+5j8fpMTOAHqsydDkenMITwn0r6O5LmJf1MCOHFKIpQWPihxQa+Wyume/r4iYM3pY2uzaBREL0kbezZkzGXsQdo0jH2HjtpL+KnSrbrtzxkFwhprAVHTcsBVCmgBI3Uqr1xS1LjkY+a7ZRQsNnmxTEXoOxaxo6X6lz8Vuxj4ZYtX1LoOMRmc/bYeqFog/RhmafsWz0725UyJ3f7HB4xC4k5G2WIO5A01beflYD6wBEA31iBusQX7G8Wugxidl74AbO9ClIvkpS/fc1s75ywZW2K++yipgzlJADk3gyH5hS6NbO9A1qIkrSRsPtJwQFvdsAVTdJpKu0Gh3oHG5lP26zphVP2PvS1rzIxkC8Cw+zIPO2CuDpmDo8BjMWVrJGOlwvac3CJa1EU/ZQkW+/tPnbkQOEdDH3wAJjKwEYB8SaSdLJqD7Dtpn3yX9vkwXfztn0f3/O8zUAUUw6NqqQNlqoOAeHCFbsEV1SxXSC907YUjCRlW/ap+mzb3pjnyrwZTW3bLF6AzMphsYJ90DU8upCUNBPBifirNZY36Q7sazw+ZzMxqcDJFWtlO2xgum+DMUnaSNoaglNpW5w6+4xdi1qS5iE7M4KDUyLHgC7bs/USE2vMvNa/8KLZXgL5nc4n/xPsIx3ZTGHI2gfqzCrHFIZFO8GonuYM5jS4jymmMN+pYR+kF7qdtWNz+iBxJUkF8NRQhZnSlEO+pw2u9h12c5NO4TgsLnD0/P0hqLyMyYKiQwCF79aOHCjsjTK63mIQcJARZS9JqZjq6kvzPPqabfujv7Vhn8rTS/wOZhM205IaMsvROWvXjyW5DXKzeGzqqi20XgCdQ0lqVm13EIkte0rp1WbsMne1LsfRnQn2xrw3sg8Ll6rMgnSh1Nhq02YCnyhwpZDMwGYXMgCUJCldsg8c6NLfq2Ef/e/8EbM927Tn0O0FTjSZ37ETqTwu6Olvs2sbj7ZtIO85kIwK9n3gNrXvqAPdtr/7YoUZzXaC55H597kK/qaXskNvSGmgWGOgf23hk2Z7vW1vv8USr0mNPV7jyYY9PgRadlxi+Y6KRTqmZe7uWgghK+n3SDp37++jKPrLD+OG+sOg9frBJ9YuCJuemeNJMhzZy18P+tjc4UkwGNhgqVK2gWUxyRIp5a69oa3l2YWXBrcrDd4TW7aOoSSlbgHIgCD6AZSok6QEBLiXoD5tvWyzVpJ0pmNv/qcdp9mvdZ4z2wcwNi/meTNqgYj2XN5mD3J9BnT0voNDhmWpYY+LIVSyGb70IvYxePEls73w7d9utp9Y/bfYRzRjs7+9l+17kKREFupVf+K7zPbalF1tR5LmbsNchdhIlTmMYwTVi7oJZl5vNu33Wcnaa2MO6hZLUhoSiJrZitkeVXlTp4QYKlDgcf0mIZbJozGIwtGHoFN4nOIBx2HHGhRK+mlJdUlflMSpeDGtkO7pheWDZWd++U178csl+dSTy9qugyGUQEo4svB6fXBxZG1gudmtYB+7wT75z4YaXqMwtN3UfRBg9SSaqAjB/MD4pDvsSieB7GTD7sNT13Uvb29WM3WOr5ydtp+lMbA31Yb4PtPg/t3sgGBzhsuyzUOM5koCUuMlLWZt9muxZzOFmUc5dCHTAraRwKtDCzFK2utBeoGTQCKIXUyt24k9M477TDTtsTfatNmxRJUVEQagJGCFBt21S4VrZvueKmY7sdiSlIbf5CE+2JM9eqJgs6InU/a7yH3iHPbxM5+xgWe5wlqHr38RQm8OQadwHGzjcQKWxx0Unoqi6Pse+p28bUGRqRH1fZdsdqEZsevhpXVb0+sayMUszfMHbXXs09X8tN3H+QxXayh1IFB/3wGQb79h/wA2zchRl5hKdEVz9vfoOqqNjBK2i7kA4DXTZeBZrNsSP1+p2GyOJH3pDXt8zgIZc7LCU/Z0zt6MznfthJluYEZoENns7UyWXY2ZkQ2EyBUZOdzHfBMgXj3NYy8M7ENmRCXqJCVOAtMHbLqnIHvrtF32Mg110AfbnHHbeNxmXrMjBmy9pD22ulCxhJQI7twHSOfs2Yee1ZPPYx95YNwHsGbVO+w+PgWFGNbW+bAwOAQdwriA7f3log7HHhR+NoTwTBRF7AMZgwVFZhDwtuw4pK02MylFYOmq0zaAWKrwJDp5yj5pEiN0tccVIUpZG0xdbL6I1yCXUb9qy5ekL9tlxCRpdMLOLq7P2zIWFA8oMUuXbNmgL3LELY7gNx45pN26DbK7PXvsPbHArjES1t2p2N/DU92FYgYLjm+WGkFCAbgiUxVmrtpfedFsT66tm+2ZR+yxKXGG8qjHCV+7v/jvzfbpJ6EcpaMOdA7A6Shtj+9EgVmn2RXbRT1wJHTtzNjPSjqbweHu7IGO5nDWXn9n69ewj50pSLoZgTzVriN2vQVC30NHdSL4rr32Q3cQjocpPCal9CJxstR7aQeCwhDCS7pz/ylJ/1UI4YruuI+DpCiKIk79excWKaibOHiQzg9tF0ca4jQk6VdW7CSOKVBNODdlnyIlaaZn32chA6X0WjYYkxiEeFyiOahikOyC+82RnZmog/wOlBrLQBk8SVqvPG62n679qtmeXuHkivY5O+ng0ehlvEb5BRuQrTTsJJB8wqE9GdmgrhXsDXG67ajuArp4TbGW51LPdh8TUO8DsyVxZmVqxn7fw1U7MUiSEmX7fUZQdUKSSmfsNWmwb4PwzDxXNAkQxpHYsb/HCOozS9Iwb/eRdOiB0qEkHWyQUll7BfsY5exFfpi2gRIdECUp37djd7MJ+zmTSV73Wi17bKXSDD4KU/YavrfFDHFcG4dO4XGSpDmuTKEtvvUQLWnIve4mbKbwq2uOck/Ays9P2Sf7/ogXBMpuI1twVMfog1bWlYhZjudGXzPbE6CfpgLrjgnqx44y9rtqQdC3JJ1+/RfsW7hqg77UWRusSSzCnUqxayyZtd2ExYw99i6D8LQkPVO2a/EuN+z2YZoTe4bkrpcjWQWYq/SO7QaPANBJ0o1/++tm+/IneS6TEYOWOsHxlRHFPtKG59BTJBts24e3EejZSVLii58x28OTz+I1ArBGnbR92B2UeFwke/azUCJVYsDsWQliYneqtrwU6eVKDJZSDjDVa5Nm48OvaEI6he83O5agMIqi65IUQviEpJejKNp/+99Tkp6QxEFv78KGUUrbg4PdQlQXc3maF8ftFtQlho15x1HRhESKp6Ka2V4bVrCPCz2bmUr1eREfFu0FNpqyXXQpYAElKQLB2syuvfnPNtg1RnVbR8DWRA73W5i12ZxEs4bXKFZs9qCastmaVo7H3iiCBZaK00PSjiT1g32yWl639S8lKUCVDkry6N5meZ7qBfuQmHncjrPrzdtSR5I0pCSPF38Dr0GAK71kP8dg0x43kjRs2iBl2I0HDiQpNW2vJ40FLotJ1YdIY5AE+SVJ8CzkXu46yjgmIB57ugGxv7OsiLC1a6+t9T1mqTH7+BDi+T5YiSbBFWr0XpknpvDvS/rIPf9u3Oe/jc2yo6YebXzuwPYIFCo3p1ncd7dtx3qQmv1TCQ6vzEMdUWLHth0u09dTtgf/TJZxe6Vjg5T9aXtTzBb4VJ7bs+O2Qsu+B9fJH2IjM8s2wzaCDVOShll7Ae6XOcs0N7D7eblvi/teSHE963LbBgiUXBTVGSBPP2eP3/YU62yW3vq8/YOMzVimwG3rsdG67R7uf+7gteiukQB2qsheA7pGEsZnqspzhOpAh92a3ccUH0jIyjd57Vy99B32NWh873N4zwiEowttOwY562DTd6u2JFgRYmoLaSY4klCjueyogFSu2vNoe8Vevz0Wl208PoCPLZI0Oo5M4T0WoujrCsVRFI1CCA9N9Lqhsj4TfueB7amkDdiKcNqVpFbX/iCvNezF78SZCvaRS9lxGMOkvSgNRvyKZzK2O7N6m2vD6rIdfzP1GExmx2TF+BuI72nnOQN0WLSZlBlgNAcrHDuW/PyvmO2px1nouLhvA65PjT5rto+mObmiX7YPFCPIGA9Fzj4u7h4sGyVJiU2OSySX6KhsA53EM5wBmn7lRfsaUNqt+CyHTo9m7LE3fJGBZf0N+wCXWbC/e+syVxtJUCIJxdV47JIthh/1HbWP9+2DzwiUBAjwSZx4RjGHKajXLkkLEP5ANlVhguMj521wutNm8Ppbv/7eu4+pD09M4Tj0FA/LjqX7+B67EkL4E7rDDkrSH5XEK9C7tFJo6JPJgzdGAlO30uye6A/sk9GJGXuSvNY4h31USvZGMZWyT/7LEW+q5T2bjeyVGEBkaGP+nB2TlXbES42adnxZgMLwZUfVlM0TdqzS3pnnzPYSABBJ0lu2uz5x2xFRMW33s/rE95jtSWAXJCnfY3kdyyieSpLSW/bGvfPEd+A1qrdt1mj0si2dQ7V6JWn3FTuWNDdrs07Fj30M+6CDVfMGg4PstA1CRh3bRb1/g9mcTNlmCgtL9nrReJN1OMs5qMXrqO6SAHWHDlRZyrRr2AeJq1NoQ7/Csb0BwjDoHlojjtfebtks9M1NjtWrbYIm4xhYvLjX8CSRHBs2MTqm2cf32B+R9Lcl/UXdYT5/WdL/6WHd0DCk1MwfvDj1QcOq1maXUrVkD7Ddpg1SSNJGYtBHBbEprkZi4WgC0JKUW4Q4uXOwMDUdtTUhGSUCZqo7zclD5ZbtMsrv2CBm8JINQCQpNWszlq2n7FJlkpTftiuSLKzbEj8YhyfOAE1ThZkULwv9pXNm+8z1L+I1eq/YoDADB47uDZutlKR0wY4fRiawzoxQ4oSdPFTYYXc8xfNFoCU3+xGuKdy8as+B9oZ9n9kKg/CoYzOB9BySlMnZ95mAuPHUnq3fKkn9ih3qkQDtyZQjfpjKNJLcVz6wNFQHqm95LAX1wccBtsYB+rCPY5OrcvxjCi9FUfT77/0PIYRPSeLI5ndhyVFfU0YAbidHjA4H516o2ItGPmEHfQ8cr22/b4PTubTtzqyus+s3AoaNEgokSSP7N/0p2xWZJlFdSerY75MqQrySZjfh+aRNXmfftLOTk474tDBvA+jCrVfxGiQgvFOwx+/Slp0tLknpDSiFByCdauS6zDEu6t//h8z2UsteYjz5/RlI0Nj5lf9otlee4eouiaqtiJAsMuOTnrfn2f5rNpDPg2tYkornT5nt21+yx+/AoVeXO2OP38ETL+A1RgDIevmK2Z6qcUwhzZHGWdvz4BG7z0CGM7mwexX7QCNJXaicNV1iL0u/C6VOD4MpPCYag+OwY1/7WNLf0W9PKrnffxuLDZMZ1coHL1412S6ODMifSNKVmn2NT5RttqbYYFfNftlmOb5Ys8FBYplPTgu3XzTbRxCoL0mjir2hpWmBJWAqqXMWYmNgwXiq9wXso1mwv2kSGNHBbU7gCGv2b4JDnqdw03Y1ZmZttmaQ4T4CxLiR/l9qld2EnQsfNttHS7yh5bt2TGz2C79sttff4PvMTNlgPzdrA+AAFU/u/Mhe4D2izxEczpLE5kC7x8i9nMxwHxQKQmEHEh9KUmmbQds9y4fI6U1blqm4aR8yI8faur9sa6duZ+094vqevTZL0u0Ne9zs7DCQH0GS02HEFJIdG9ew044lUxhC+KSkb5E0H0L4U/c0TUl6aERtqt/R7ObBJ9YKZICuVmywJUlf3bHdIC9l7Gt8OMfB0sWmDaYegWoMG32H3uKCzWIUmwxeE117gSVg6VkccyvAeoJ49WCaM7Fnanbc1nAeGORb7IoUaWk5Kmwg87pnM8ipPn/TEbjjk2v2s44WbUZJYjd45Aj2J93HvsPtSkZu2cIF2/UbysyaRpv22Os7ysN1Nu3f7N8E1rTO2fPTj9vZsGSZCr+L0QmoxHSVBcd1we5nd/qc2T5/266qIvG61pi3Y9M9xABJgu1AOckra3zgrtdt0Le9yePiMOoSxzXPPRwn4Pjev9GDzTr6ZSSV3v7NvShqT9L/5mHdUCM5rc8Wv//A9pm8PchzEZ+MTs2DaG7S/mS3s5wV1gWJlG7XnvCnMpxosiWb/dqc4iSQQsV+n3N7NhuT2wJXpcTuY9Ax7Od4M3pr5hNm+yM7v2m2B4d4apgFyRlHvF9602ZKhhXIHJ5mDbYevK8MxBxSgLwkDUCKKNVl8epkywaFmZP2+M0++hj2QSw0iUYPbvH4Ts2A9wK0ECVp0LDleWaesMFWGhhRSQqQXVx9wXaZRo6QgOga1FL3GDCv6SGs8Y460ORi7s3awv/DKQ5TSg3t9WAuYx96ziwww7yxCRWpkgyU8hA60+/Gr2hyGGzjUQCvXjuWTGEURb8q6VdDCP/zXSHrw7BMoq+zpYNPYa/t2pPx0Spn+j06bcuPUBJIP+IT3M26vTF3evag2Mkx8KzkbcZyIefQm+vaDITlypekrEMuZj9t/2YIxPPJGsfRPXbz5832sAuSNJ4FBUDfqMo6hdsLdkIAbST0LiWpG9nM6+tN2y31qWmuZ129YYtT751ieZ7mvA2W5ko2qxS+ZFfPkBwyLCV7QyR3qCT1L9oyLMlXOPyhccueh+1dG7wufIhZwP4tUCto2OtJrsqJJoWz9vqcOM2Vg7pF+8AxgGo6VEtdksoZmy2nij39wGEFrYS9HnQHNkhPJTgeMJOx96pCyZFsWLBZU4qePIzs4+ME+MgiheOZfRxC+JtRFP1JSX833EfNOYqiH3ooNxQNVOkefIp7oVwz/z70efB8pmZnHF6cgSy8BDNCJ6bsrNwbu/YC2+zyoHmyap92SYRbkt4KdtxLSTbLt7DJrpppqDZCotD1KXZnFlMgdFyywVSS3MuS+hDrtFPhDW+mZjOvEbAcW9MsCj2EBefbpmxAl2vxYaI3a7N46T67reb37MNZsm3PoTDPIHzvC/azTn2rnTGenGPpEVpxEg4X9PQjUDnlLRvEJPMMUupX7fe98C3Pme2e5xDUV94/9RReopmpmO21yAaNt1v230vSfOWc2Z4YAogRr62doQ3IekPK+uU+RiDX1dhjz1l9y57v4wBshyFefZySVY5rosn/8vb//+uHcSN3bRQSamcOXlioxNG1ETNsj8zadHg+YZ+YV1oc43ZjywYpxOr3OF9Gr9bsjeRkmQvYP9WyS3CNkhDv58hUTe/b7zsFDNygwiyIJWMkSXNbNtnd+4pDkgZq7SY+ZcenSdJK1WaVZrv2xj075FimrYQNZF4Z2AzdmUo80V1JWtjiTOxBFlg6qj/b5jKOxXP2HBmVgcV2jO9O3h4X6TJnw2bBVT7/gh0eERo17GMBStCJ9ECnKtiHoBpO0lEz+PLQnke3d+21tVpylHYDUHe9VjHbaw0GKZT78/hizWzvDDhUZGHW7uT2LQ6LiSv6PI5YPqp97LnHccjaHIpF0lHW2bZG1MshhD8p6RFJL0n6h1EUOaBKPBspqVY4eLO43bU3/5/6RV50Pv6CDeqeWLQXtuGIJ8GNFTv+5uJZezJ/5DTHcVDVk+6QXV9NqARCeonTr7OrkcqVbV76VrOdDgKS1EtC2cDTdrZsJc8xWRsLNqCrQCUGSUrn7fFZy9rfY7tfwT6ubdqb/6mKnVy0N2QgRJtq0RHPmh7YoC7TAHHf574D+yiu21Iuoys2eB196FPYRwCWozvH9ZOzIGSsWzbDPGyzpl0AV3rI2MzW6BZHEXU+aouvbxUhEUXSl79i30cCqlo125wlXXSUqbNsr+Fh8ez2xxZstmihwBqwG1l7P2w3eT8kO4yKJeOoRnJcEk2OsyTNP5bUl/Trkn63pCcl/V8e9g3V22n9wssHbyjTZftlfuKj7EbJpED+IWEv0OdLzKQsfwLcmcHG16URs3w3I3uz+dJVBjrdM3aw/hACYp98nE+iKQgM34hsZquaZHdmpQll6sAtO8rwqXzxli3IfPvMx/EamaHNQt/u2PF+N7b5Pk9U7T7om261OHbs2qY9z54/43D3ZO2N4NGUzbC1Mwxek7M26xTN2SDFU8axn7IPJOu5c3iNcx07ciuxY8ccDj/ybdhHPwsl/a7YrvbReRbI3ivYh5ovrTNAfvMNkGXq2+vzqbMV7KMJ5d86HXuP2NpiEL4AiSL1rn0PpQwDuo1t+z5312t4jeEhMGyHEVN4rMrcHcdEE0lPRlH0jCSFEP6hJC7gOQabzTX1o48cnIm3m7UBhGfRqRagEDmIV6cdGc4ZgWwNjN/LXY5PK6bt+zi7yJOdAMJivma2Jx1+7m7ajhk807c1w3I1O0lEkroF2w14I2PL97SgUo4kFabsb7q17wjETzvEvg3zxBkliMVL2eNms8G1j0/P2eztMOLDQiVpH3xqs3YoyGZwlBor2O9ikLBZqetdjmf9ja/Zm/90mYHl7unfZba3nvoBs73T5/d944Z9HyeWbBe1Z8+9Bo6D6zeY/RpBR7T5k0yLJG2s2zGvyaS9LubyDi9My15/txr22Nt3CJJTH4UyZzC39mylANIx9Fhc2RsXC5g4PskojsqtsS2E8Nck/aCknqTLkv6rKIpq9HcWKHxn94qiaBAcaf7jsMRwoFzzYBCwAFIXH1vkTfcnP2ezA+un7I3gwhzLbVTS9uLXGtqTtZRmLUSioD3Za9WsvTje3LfdE28NWGD10aodB3eqazOvCUccUiNn3+crq5y1S5ZMMEtH9u++aoPo3/lxeyN4bIEZ5Ft1G5xe27LHXrPN46aQAxcfSzYqytrjd1cVs/3WHoPwy5H93asFG9y+tcqHhd1de652u+zOvJKzD04vv24fVE8sM4C4es2e66ugi/fYRfbCpNP22Mnn+D7LZbuf3R37XdS2OckplbZBdHdgA4yTp9gLk83aQObmmv2uMmnec1dX7L2otc9xtwSyDyNWj5JEXLWPj0miSaSAyYBjsl+U9Offxm8/IenPS/pv6I+s1erZEMJdn0aQlH/730FSFEXRGGph/XYbJtNqlg/OsMz07EmQG/CC8CMftcvcrTbtwPGtJoODUcH+6JWMDRr3+9zHZ1+3N8VFx8Zczdsb1rkp24XnYYSugTJ/mLYTH6Ipnuy3W/bD9gf2NTZ3+JT5oYvx9C0l6Xu/xX7fn3/VHjczlQr2AXJ0ysDGXcjx+56bshfpTJIX8RLUB+9G9oPkUtzH6SkQhYZ5lnWIs09Pg+yNY7PaqdvfJJWyx8XGJsfddlrgISnYz0H3KEk7u3Yf588xc3X5qu2azebsOTS/wGsnAfkBqFgU8vE39VrDXk+KRT5MEKM5cNSaPgz3cVw7LoDPa4fBFEZR9O/u+edvyqkvbekUviflpZONmqb+408d2J6o2ICt+9R3Yh/ZYDNPDTjZt7o8QKfz9m86Q/s03BnwifoHn7FlKnIjBsgjqCJOlVU87kyyr23aMisLZWZNF/LgiizYyUWrG/wc1zdskDI7zaDwuTk7WH/heZuBaA0YpDR79n2+douE1T1uGHvc5NMO7TJgumehvPqbXWZ/X1qzZWsqBXtjbjLRoloNqqYUeHPvwppSKtnXIDeiJE1N22MnDd+s1+c5cv4MjC3GrhoASzc/bwPLy6/bh35JSgDIJs/Y+gY/SCoF11i1CY6Tpx0JXwCWPG7XcbiHyd5voC6uvQeJJv8HSf8fzw/jF8wct41GGraMlRhS1ytrLIVRgjJg1ZIt8lqb5RJ0r9Vs3bt0wZ6I5wpc0YQyg3fE0jn5AIK1SRtAV8QL8CBvv4vpnL1wLeVY0qMHgs1t0H1sOrL05mftsVdI8+L6GyscK2pZCjIvJXZLzc8AGOOQQs2X7feVcIQu1HpQbrJmxwxemCFZXY6ZJVvL8uGMQB+BLUnqdOyxMxzaz9EGFlDixIfNTXstOHuK3wVFGbU6PC5OnIgXpnH6vE0cSOzarddt0LdXd4T3gFt2CFqIBBolab9un1r6PQavCNhgWfMAPnoXHyjQ+GCSNHMhhHvV7z8dRdGn7/4jhPBLku63UP5YFEU//fZvfkzSQNJPejo8eqCwWFL46MGZdIlmzfzz7pSjiPi0nUVHmUG9kWOjyEB2ccqezPUR78zTCZsdq8pRnggGZwSTdQ+EZCUpDdnc80k7pnCly/Imb6zbAKNStO/h2z7Ci1IhbTOvKzXezBqQtLjXgFimRb7P86DDvQHJ3KU8r1iUGbnV4nfxlVs2kD+zBAkFkL0pSd2hDeRrTbvds3gvzNnrwdoGA7ZKxb4Gad5FDn9UDmI4Z2bs7+EJsZgqQdgMqEdI0s3b9trZH8C4qDFgI/dwHxLo0hneOs9DDeeVW/Z6knIcJggUpkFmSJIGQN8O+/GV6D5QoA8s0gNlH29FUfTCgdeKou+2/jiE8F9K+gFJ3xV5FgkdQVAYeh2lrr92YHsEAzQxxezYfNuuZ5rp2gwESTtI0ik4MjeD7frqBo69GYFrwKPvRzT2atuO1ZvNcTZhKmF/MypbdTZ5DftYOG27Xde6thtxs8EA47Ub9n3u1njxXF60r9GDzeqNK7wxnz5pbwSLM/Y1ZguOKggOQEb2/U/buo4zHVtm6Oe2WAIolwGQXbEBxPVtnockRP/IOT5EUrxeGlyRHvBa34vnJuw7qkVlM/Z9bmzxHCFX+Usv2p6D0hSPTQJs3Z79rJ02v8tm0/5Np20fFlJ9B8PcsueqB9DF1ffzSMEQKDxOcjLjsEPKPv4+SX9O0rdHUcQaSm/bkQOFw05XzdcPFpxNQhR9ao9dSlkIrE3P2cAyU3VkcHTsb1CiiZh2nPBmbPdaoudwcUA//UWbPehDMoAk7Q9swJYEzUaKe5Sk5Q273N4ClNJ7s/I89lFr2kD+xLwnDBdcRpv2arG4wO97p2b3cX7B3oxyKWa2iIGbLXAw3s2mDdQbkJFbyjNIuQJ64lMX7OfwuOtrcC7yxPGXiyCBkrHv4+wpHhcbW/aNEKC7tcL7Sj4PsabQLkkDYAJzeftZiQWUpNu37WchN3iSTgLiJBCS3un1eOD02vbBn2InJam/Z19jHIDuMEDfcWIj44a1OO3vSspK+sW3Y2R/M4qiP0J/dORAYaJUVuHbvv3A9ta8HZOV3uf4s9RNqHKwZAvedipcf5aqHOQ37AoFanE8SQrKx+2dtCtwSFIS2MT5XVtDsDbFpd1O73zBbCeB4PCV38I+EgWb0QlnbZ3CPGgQStKHTtrv+9V1dqXfhoSW0ydsVmmpwoCt0bU33strNtDfbzCzVZmGetZQ91WSCkDo9EHfzyO5NFOx38XrN+1nLRZ48T4xZ8/1G2t8jVur9sGIYuA2NhiwnTxpu/QJjJXK/E0JWHqs07XvgxJmdnf4XfS79vuertrrycp1FtTPFaD2cceeyyTSLUkZAMgej2EG6mZ3IOZlHGDsg8QURgqHIl4dRdEj7+bvjhwoDMOBwn7twPbiFtTqbTkybukHAOj6aY6XKjSgRu2+HQ8YAdMoSSELivhbV/Aa9KyDss2KFru8OA6BpUt87t/bF/AUQwf2dpi1F3mSzZGk27s2gFissKumDkkJhZy9OF5Z4ylLMWwZiFV68hF+3w0gAlMO0nRxGoL5O/azFjO8abY69rP0IaP2RJW/KX2TWyvMmnYBpFSrlDnML7xWs8fFPMRG1h2SNNs7dh8nlxlYNlv2mjQcxRcpJgaNMoeXT1ewj8Y+JGOBYL6nRF1r394nPJI0R6Fm8HFi+WLbMa59/J7YqNNR++WXD2zPnbaFpYc1FvcN4IK+cfJ3mO3nb/4H7qNvb3hR1XZRhw3OPh6t2b8JZzkeygLgknTj5HeY7dWRLRsiScm+zcKlKBg64ZBV2Lbvo3Xho2b7cprHjWQn/7QcVSV2IO6wVoeYrHU+LGSz9rT+0GN2+2Vb6UiSVAG1DBAJkCS9sULxlfbfz4A+oMRxdM2m/T1eSzKIIU9iJsMvY2qKZILs5+g7WCVyZ+ZBn7IKyTAeazuyjxug35eE9YCEqSV2/1Lm7+Iyi1fPztnkAVVe8TCFSZhonnjBASTVjMN9/IECfQ47jJjCd2tHDhRGw5EGhjhYd8UOPk+Qcq+kACejM9t2jdvLp78L+3jkrX9rto/esqVzhk3e/BOQkhh97UW+BgCy8/O/abYHR9ziMG8n5iTy9uLpYX8TZy6Y7XtQOH4qcCxqadq+j37Em+ZswXYxf+WqzR4sLDJLff6UvRF0+/YCfXKRV6wzM/a7KKQ4yemNzYrZvgZqR40W3+fmps3S5UAI2RE6hgs8iVtL7HbtgWwTgUZJymTsa+w17AfxaCH2IEFjZobX5/NnbXcmEVsbW/y+Nzft9ZVYunqN+5iZtecygSlyL0vSiMrDOcBYAmj9IcRoTiRpHtzeA51Ctx05UCjZcRC5xx83/3awApHlkoZNe0PLtKEWpKNEzf7pZ8z23JzNeGbWr2Efg5u2EHLCwVCMLE1ISelNYCwzXPqKXNAq2qfuRIqH6WgFRKFn3jDb98sse1OH2m0pSJiRpGwKkmoAYLRBz06SqkV7Ec/BPZDcjMSSCh5R85NVe+ztg6yNR0z5kYt26MKrr9gM8cICj++tLRsAz8462EaYqlu37W/iYSPJ60qudJKCkaSlRcp852uAfJ8abWI0PcksNmCj6i7T0/xNSXuSzJNFvbNuh+94so9JvPowAN0HCTRGmriPH8iiKNKwc/Aiu/ebnzP/Pj3FtH4CCo2P3jzYfS1Jj9bYZdo/cdFsH6Zt127v7Iewj/albzXbcz2HXEwfxE9T9uKX7jDDlt47uJa1JA1Xbfa3t8Nxi+lp25+ZqdkxnrVpft+5hP2uCj1+FzcHdsbthRP2jvjsBY6nevO2vZkszQDrBJnFknS2bM+BxsAGY5LU7MZzR5YcSSD1fXv1rYKLz7MXUbbr+hqz6WfO2N/s6Segisd1ZpUKBfu7TkNiT88hkUJaiDt7/EIzMCyKEHcL51xJ0uyM3UmzZb+rlkeSpmEfFijMo9Vktj2bg6SxDh/wsCoKHAAPw338fnNRjyEs9qHZkQOFIRGULh68AOZP25m/4ckPcx9XbNdt/0k7/iy0ePMfZG1wGkYgi9NlQJdp1+w+hnxKTG3ZTOD+BVuqpTXN2cfVvs3SJWdsqZeUQ5Wfso9JIigReJZmhvbmvpXirPRe295sVrft9ld56GmmYrePYJGnMo+SdC2ywW0xw99soWgz8rd3bVe7R0yZ9on12/Y8W7fPK5KkqYoN6BYXObaXwFQb9vZOh+d6uWSPrfq+/T495Q/fumIjMoqdlKQcuMpnpkFb1QHCl0BqywP6yMqQrd1qQcY5hDZI7D6mmMPDsqPANh4ViyJeg99LO3KgMJFOK3/yYP29QIP8LZvlkyQl7cdODOwN7fUlrq98tm0Dz/zq6/YFGo7df6piNvfBRS1Jgzm7/EUvZW9onSQzQpQ9PNq3nzVyZNBRoFFUsOMab7dYe7IK2d4emYHdpj32SB+t0+GNmcqW31y3O/GUQj27bPcxX4gvnH52zkZCzVZ8iZRzFypm+xSwZ5KUSdt9PHaCqasbIJI9hNtIOfTosgA8ib3Y23NUTQEgQ4k9d35jt3cgK3005PukDOeTSyAn4whd6Pbs32CoiIMpJOs7ik0fRvZx3GSV48QCemySaPIgFoKCkUARztt6c8MCFxFP7dkR7N2iDRAe2/hV7CNRt/uIdmyXapizmRhJ2rrwCbM9OeIFuLryVbM9V7KZlHLPrhMtSf18xe5j1paDyUAiiiQN6jWzPQkn6lSCwdavvGozmvU6v+8+ZKWTK/L0aY4zIp03CtH0aPPdAsWllY0KXoPOd7SfeZirjc14JboW5jimsAkJL7/+ErvJq9P2NWgToSogkrS2br/QhXkbCHn6oMzhmSq/i6V5e/y98gaUZXP45uo1+xq9nv2sKTgISFICTngkDeXJos4bXjVJau9xwmIv2IevCNbOcdj7DfSRTUDhA1g0HGqwdzAQCV/9kvn3SYgtkyTl7YlUhC82mGJWKUFSAZA5PMpzbGS5YQOy9L4NPCVJXdvVku7Zx/adqbPYxcKK/c16l20x8f6eQ8gbFsdEEuJIR7woUS3evTJveF960U5soI1gnvWx0a7ftp9jt87MwQuP29coZJjdbfdhDsA38VQbef2WDXRurdjj3xP7Q67fETC3nmsQQK5Ocx+7O1D9AshGEtCWpAZM1c0tjnGr7dnP0obycJUKA/nFeXuubu7Y4LbtyMQ+ecK+j719CCFyJA+RpbLxZYTG4ZaNe41xZDgfJTvKt3rkQGEoFJV+9uA4NorFG74FbllJg469EUQX7aSDvSnb5SpJszW7skoEcXL0nJIUIvs3t09xbdjsyHZt7crWU1wYsJ5ie9aOOyyesjPGMw73BjGFStmL41ye3fXNvn0YWGnGX7guXrDd8dugYyixjMppuzqi1iCuUZIGMD49rvTBKF7N1StrvOE1oP7stTfsedrYs7UpJenMOfsgujzPz7mxbaNPcld6mIcUMFONpn0PHlA4P2eD8LV1BoV9KO82GADr73Cl08GnCEk5l856sr3tj5JM2NfodBxSRjn7N0clpjCu+/j9lGgSybc+vld25EChel3p1sEl4IYA6Poe8WoYPJmGne3arj6GfYzKNqWToIolA2ZaBhkbQBT7/C5GsDAlwK2aHPF95vZsRpNiCj1KyJRRTu782Tnwh0r6Wtt26ZPbVuITIm3MpEEoScWsveHtQ7ILZUBLUi5l9+Fxxw+GxLDZ7YtVPixUICZw5VbFbC9AqTJJ2t21D3jFAjNXJxftZ72xav/90BFHlwDx6gFco7nLc/3xi/Y8DIHfxfqG/T5PnLDXvSfsSqiSGESv79rvaq7E4HarAe74gn0TxSKve5g57ABK+BuYZkdFp/DYMIURyy69l3b0QOFoZGrnhac/Yv557tZl7GKwZbtVe1N2jNuJlc9jH6OMHfuFW7uj9vHt/LsqbfgNdqrxmtm+V7bj6DwUBYmFJ5bthJhom+tZh/lF+xp5eyP5YuNJ7OPaCkiPrHNCwey8HR+ZhvJa5PqVpBSI0RKTWN/jFasLe2Klwhm3pwAIkcj22iaDwixgkFzeBjHrt/lg9dhTXCKRbBUiPSjOczDgTXV11Q4FmZqCjNwWg8ItKAXpiVeNm2zVgEQUSSpm7WuQLE6tzSweLY3tLsz1qzz2Oi0bQHeoHqXig6mjAsaOF1P4Xt/FwXb0QGEiYcqLBAJ9OU5KSJ224+CSdZs18iSzJDr2AhxN2Uxid+Y09jHft123xTq7dpNtO5GkmrPdZ40cx1cWkvZ9jFZBcNwhXi0AfTsn7ZCAV19iBu7KWzaDvLfDQL5YtsFSrWxvzGfO2lnUEstpJGHxJNeZJO3u2qjw2pus5Xl9zo6brczY76peY+mRYslmayyhfEk6fQEOReIFfmfXEf4ALlGSLylATW2Js1mfeoLWNWb5Wo4ydmTPP20DrtVte65OFzi5iNhykjsq8ZlH5TzNQ/s5OhA76TGq8XznN/a7OIxEk0n28dGxowcKcwUNHj+YDbxZfc78c0/5mGyAWryRPRn3ogr2sTC0gVB6YJ/ggmMittM2QGgsPI3XIO2939iyK8h0erzoXJi39fsufYsN0ncyEAQnKQMZdG/u2WwkVXOQGEDkHG7C5r793UfQRyIwKJwCPbrNbXt8b27Gl8IoTvGuSfVjF+ft5emqw2VKYIvivrahHJoknTptf5Nmkzd3EjqmBKTVFdY1LZbs8fnWFftZlxY5851kb67f4rG1A1nOe3v2XJ9xaCEOIYlpYdZuv7nOY29hxn6OLgyL6hzLfe3X7XFR22Th/wiyqTjej/eqAAD4g1Zf+YiQq/e1IwcKd0fT+pfN/+TA9vXr9gC8yASbnpmzAVthaC+wuRQzFJ/ZedZsp5MCxVtJ0rVX7Xdx7Qq7H3Y37Xi+E+fsjSKbZVbp53/OXphae/amWqryu3jmw+fMdtr8v/L5m9jH/rb9rrIF3jSn52w2pgxu17fe5G9KYIvWzs1VTrrptuyNubXP9aqptmu3a7PpHsHm5j7dpz2XPRvNLrCRs7N8WCgWbXbs5nX7u/d7jncBiw5V2Lh5M76e3dIye3JmKjaAqMMU2KrxfdxasQ9n67ftOVCZZcD28lfstbM4ba8Xezt8IBlCcFoKYq2l+O5fYhrH0cdxAnxo0YQpfCDb2+vp3/38we7E/V0bsO2+wFHGs5+wM2qXHXEvZKWsvUh/6XUqo8SL/OVX7ASOqRmWtSlN2Yv0yhXbDTg9x8xVAoLYiIGjDDuJNes21m2QEiiQSdIQYiP7jsor7aYNQvZrtgs6BVJGd/qwgdDWbTuAbQTsmsSVFMoVDrGgDe3m1R2zfW+b2TEaWyNgOTzv+1rD/qarDgZ5bsl+XxSa0KwzCC/AXKc50DPKj961EXzT66/zPHv0Q/bJvg/q6l/+Ah+c6H12oTzc1oqduCY53K5wDh1HrF7XUfMvrnv4qLB4RyW2kSzSpMzdA9lwOFLLENw894QtB7O9yYvjv/plG2ScPXPCbPeMve1te1FZv22zZ8TESFIbNiNq9xgBnbrjWE4LQjpjMy31Ld78L/ftjSIHWaRph54XAQTSppSkPviMMjn7Pvd3HNI5dXvDo3vw2Ag2ZtpUJWlvx9688yUbxCQd4r4kyUFAx/OukhCDWd9ikEL30XYkDJDtwGEgR+/bwQh5KmiQvf6VG2Z7GrJAOi1e9wgs0bto73P8MK0HxLANHezvAKo9eaqVHAaYoj7oXR1GXONh2gQUPoBFo5HazYNB4Wuff9P8+1KVGYpB355sX/usvfGmsxyzQiCjB8VMPZO5VLWTQGjBkKQuMFfEjnmYlNEAxMBhI2nWGRTu12w2ZmbRThjoOliQdsN253g0wQggL1+wDyS5U5zpunnLZnfpHgpldo3Rd69t2CyfxBtFC0TLy7OsIdgAz0IPJK4yOQ4JIFe5hyWhaxB7m3TMw6n5itneBTDV78Z3H3vuk4zWJA94pTU8B6EgnvWZ1ngCfZ59ht4Flf+UHLI2VKZ0DGXyDqPU3lGxO7WP3+u7ONiOHChMJJLmpkRgizYBiTebfMneFD2xY+gSdVyDbH/bZiAouPfObwAgFB21jakPWFRoEScXn8Tjghg4ape4ZFQmz25CApZr1+yQgEyO+6CNlzarnoPlI0CXNRQE7hqxNXR4o/HvMWIo8uByHZcRu0vMrOcASBtvqWqHgvTafHAikJJwrEk036ndw3zR2EOw5TBef+3n6Dhcv0NPXfiYNg7ANnEff6MRPngv7eiBwmRQ3sj3p/g0z+DL5G1ARqWBPEH0dB+0yHsAxvScHYhPcV8SL/SZPIMlsr3Nmtmen7KBp2eyE2Abgnu5PMOxkTQuPIs4ZerRApxysHgEplCCwvG+qQ8PsKRnJaZkHFmPdJjw1I4ll6mH8aH1gL5JMsNLOX0zOlB7gBI9hyfxgdZOeg7PuKDEM/JeeIwOwzQPEx4gBe/T9S7gux6FJA96l5Ikh2D+UbEjjAmPHigkW71sa9oVpnlzTwPeogV47iS78IawOFIMkcctO47YMAuAS1Jx2mZK8kUGr3SftCiRi1viTZEYOmqXGEB0mwwKC1N28g8xgfs7NeyDxh5tzK02v2+6hse1S8CRXKYe/oLeJ7HUnT5/UzrAeeLsaA4QiMkkeR6GtL25d2AOeDZmjJNzAMuk7GuMw8sypEMNsOkuoA+HcgLQnhjlEADcHpHgtbgxhePo4yjZEfks97WHCgpDCENJL0kKurOG//Eoij4Lf2RqcpVnKuafezJAacLTwlbbYO0n6gOZRE8sCCzSMyfsLGuJs493N20X3WjIGc4Ua0ebkccS5KLO2UPd874xSQQYaIk3RYrrcp2YXXApnlHgt6eSAoKpkT2XR44YNwKeHHbAYIvGLzFbd+7DBhlZOCR6gCclT5DL31UyzTU+baM1nNYTj2AzgcJxsGPENibhYDWW8nGefQSuMQ4mMW4iietAckTqPJNFH3BJmrakH5f0U5L+kKT/UdK3W38wHAxV3zo40aNZs10cHrcruZQI0HlcC7QAF4HR9JyoS9O2K5EkPyRp87YtrUAgZXcN6nOJF67SjJ0cNAXtktQCUWiKI/VobVHMlYe59bhVLfMwFLTAUrvHxYeB+o6SD7SZ9NogmusIj4hApJie1QO2iP31hBXQoYaSf5oOloTAq+e7k6VAt5TCfyQOWaH4S48RS0fj28V4xgR943BhezxOdGTB9cQx9iI4qI6DJDkuoFD6YNc+DroDCCNJ/0CSQgh/OIqiT5t/ZOhljSN+J+41PMCTFnk6DXsGOAEhzwJMk41iUioLXOaOnpU2Iw+4pUB7YmvyjnFD7FdlwY7xlKT6lv2+x8GaEuND78IDhBIwbjxxt8T40MbryQwmo6QFz7ugDc2zMdN8p/fpiveD3xA48Kx7HgBMdhhCx7zuAXsGoPLOfUDWruMaZAPIYE45QsLjVjQZB4OMTKGHNT1GsjWH4eoOIfz3kn5YdzKaNiT9l1EU3aa/e2igMNxBdnfpsr6kjKQrkr5f0qe/6bd/WNIflqRccdlkU+hU7nEd5GDTpISCgYsRshk2YhI9unnETHkWaDqNjsOlRPdJIIVYPomZPvrmnk2VruHJzqT4SAIHnoxyel906PHIOtGmSmy7xJsiPes4ABvZOKo1eGRYaE1BGZaxxGTZmyrpiUqeJBAHqwT3QX3QPJWkFMwBigf0PEcmTyw0xFo7wg4ICI1DN3IcJegOIwlkHKELh2GHKEnz16Io+m8lKYTwJyT9JUl/hP7oYTKF947o1yQtvP2/f8T6o2QqqRMXD66Vu7Vi65+Ru1PiCY8LLB+YkbkiRsizAJOWlge8kouDzMMI0SIfEvb79rC/KG8CJ2rP5k+Mj0fWhgAXsaoUPiExmEIQc0hB37TZkHg1JdRIHDtGm6ZnXHS79tiLO8fGZfRdM/C+PUZzxFMth75qAdQKPH2QJ4fCHzz7DII+zDh3ZNePAQiNwz0c+x7GwEYeK6bwEEBhFEX3xuEVdcdji/awQeHdXfJpSXdH7y+GEP5oFEVv3O+PRsORNm8dHOdGDERxmhMfCBSSm9DDglA2LGVnehadxq4tsu2Ke6GKD7RwOSYiufnGwcyyKzEeQ+cxj0YmjT06DBBTLvGGRt98Z9UWv/YYgXDJkakKY8+XHBQP9Hk2ZlI88LiP6Zv0e/Yc8BwW6HBFDLNn7NE88gA2ehfEFHrcsjR2aA6NQ56Hxp4nSYS0Dn3MbDxANg7Q+MGraOJ+Z3MhhC/c8+9PU/jdvRZC+B8k/ReS6pK+0/M3hyVJk9AdlBokPSZpUdK9oPDbJD0hScNB02RcaOHyZPpRIPPcSTtrd+hY2KgWKZ0iPUwh/cbzLshIssPD1tCEzjoyPMmIEcqX7Q3NUyOUMlFdiSZQHquzH18gmyQ7GnsMIMiI/aLwCInfF8eiOmL1YPxSmIYHxNDhzHOfcRNePJsmHUjom43DXe8R2SbwmhiB0gAku0j8LOQB8XzT+GDLk0gVPxkrbh/jsPcb6LMs0gMxhVtRFL1wUGMI4ZckLd2n6ceiKPrpKIp+TNKPhRD+vKQ/Lum/ow4fJii8d8RHunOkSUoqR1H069/021/X26AwlS49byVIUNauRymcFvoBCB17jDY0mmjkGpYY9HmATlzgWFmwy8d5jOrkelhTsiTpZDnchJT16NkoSFKJYmI9wJPeF/XhcUmhCLcDTMU1Tx8E5GkzGgeI8RgBNna1sz4rHZzaezYQ8hxI6KB6GMyVS98PQixQTmYMJS3JPBJXPdAUjRxbWdz7HMc3/UBZFGk4pvcRRdF3O3/6k5J+Vu8xKPzmWXd3Ft2PsnmHKRz0GxoZ2aaeRBIycsVQQXUSfJYcGbXAsHk2f2I5PO4eSkbBk6hjcNOpmxY2X61p+zfU7qloQoLj9a0aXoMAG7mlPMK9cRd5z9/TgSPK8bjIFu1nIYDhAeF04AgglOxhPAlMjYUdG0MVGnKJ0gHRc4AkBs4je4OVmLDcLyMhGjt06PH0cSgZpgBuSYNQis/SeUB43D7GEaN8lOwwiNEQwqUoit58+58/rDu5HWgPExTeO1L6ugMSU5J+9D6/fYcpzBWmnr/wxP3Y0DuWzoAQcpsXrtUbtvj0zqqt3ecZoLNLtjwJDeCt23ZCjSRT5FuSeg6CjZ6FAJvHKLYmW7Q3K48rvQ1uWQIxnjhROpB42AN638hsjeFUjnFfjg0vDwcOz/uMG0Ixjvq09K48KgDEoPlKb8YLofD0Qd4HYl4LjjrQFDbjqRxERnqhntjeCCqBjCO7PuUoPWiZJzSHwnsiD1ACdzyuB2PILB5LoskxYSPvuI8P5V7/agjhMd3x0l6XI/NYOryYwpS+zhz+RAjhywclmgwGI926ejAgopfp0ebb27FjgFBI0wHzqRLIzGLFbD9xYRH7qG+DILPjXWBGIrgwPEHdzILY9+nR7qNg/9ll282ddizga9c3zPZenzcKBIUADlwbBbIgBEyZjaQ54GHx4pYro1g+yZOJbT+HJwSD2C9PaAJWyxnDYYHGDq2t+w6wResJZZRLnDTjAX1kcV3QuTIfVGkOELB8gGSEA80jYRWN3vvax+Po49gwhdHhlLmLouj3vJu/O8xEk7uWk/Q9OiDRpN9tqGUE23MVBA6GJvfv4mkCELzI7+3Gc8t6jBZ5TyweAYjyrC2h4tqMYjI6Hlc6ufBWqdyZg42kmCyPxV24XC5T+O4e3TwykvQYjvhdUVZ5L4pX/UVi0EfuYc/4JiaFEiMkqQOyNkNwQZOeqMdIUsk1D5uQKOVIKqODEWt5xhevJptd5hKiVhEGaTweKVpbxyGQTfZ+c+0ehh0SU/iu7GGCwr6k+x2hZyX9u2/6b++4j9OZ4vPWxofaT47JTrExte14mcOStLdlM4W0MZNbQOITtS/uJV5JtMo8V/GgWLy9HftdebLwiI2kuK5Rkt8VbVYeWRC6DwJbC2cWzHZJajftd7G/HY8pl3gOedixuALB4yh9RUB/HIc3z/uk8csamQy2dlbtkpT0zajUnsTvy3NQpWcll74n9pGSqehdtPbYe4EEBhQ48MTqJSLIYHaMvbh1oA+joonHjpP7+Cjf6mElmtyVo5GkcE/w42+3EMxJv7tmn65IH1BiUDeOExxp79HiOI6NJOdghLj+Zrx3JXGAO202tQ2OrySAQBUMPAA6AafuYoWTVdIZe1wQQN7bYeBJjA99c8+7OAztMrwHR5Y0SrnAfVIyjMSxeJ6KPOSCJr1QDxDyAHXLPO56Mg+jSQBhHPdBRutJYww6nBSm4ZEA8uiBksWdy4fiXvaIdB9C1ZSxWCSNhkcXFR6VRJNvdB8bp7C4GYsSb3oJyEikZomDusmF7WEjSSLFU+bOw8JZRkkiEi/iyDo5Fp24VSOGjsWVXDGezErKck5C8tDuOgPkTgOEkEluw3Gw8rixyWiu0sHJpRVHYuHIMPO4oPvwzBEaf7Rm5aHKh8QHPJQqcqyteODwlG5DEBIvTlRiwEag0FUYAOYReQ3GkUjlMQJcAfY7l0Ym7bkT8eojYw87pvAuQ9iRdDc47X6JJu+4jxPJ/PP1rYOzg6mu5SA4NveYA9SzGRFgoyy8cQjeBkdGY1zVfXJ3StIINCRIe9LjJowbW+M5lcfdSCSpDmEFcV05HovLGElcw5lcY5KUK9lABkW4HYwRzSP67h45mQQ8hyeui75racqO7U04EgrixpJaNenvGn2zcQjqc81sR0whXQPWtbRDIJvWxkQ6XjlKiWNNXZqjh1DmDtnIMYC+41P7OPrAxhRKX3cZ37ui3S/R5B1LJBOaXZ4/8ILkSvRkBhOwpI2i7+kDmEBq39+2wYPEQMhzYkYwBPuARwuRYvEoftKzGZERYPNoIWJZNgeAiKtp52EjezGf1QNi6F14mFsae3QfHgmguIcaV1gBMGyp+4ZWf6PRukUsn2fdY22+h18T25NxzuLrMPZc4yKeSLxnjtD7psx2D5uEDNthADpPMlZM8sElxYVVs4+OHWXi86gkmrxjIQRlCweDiKULJ8xOPYOn27JBBm1WHokUlBsYg2tgGDNeSmLARtmZHncQLbANkAjy6H2h9h4lLQwcJbz69uLoAZaYWdml2rHxk4dIe9KT5DSC7+55F/QsdI1xCAgPYSMZR7USD0NHbxwPTuPQhYQ4aJc23xjqmMdNJPEwhbSGj8O9R++L1CM8cjKHYcjAOWL5QiLegWIcySxHyUYfUKbw3i/U1B2QWJE0uk+iyTfEFLabB5+gSHLmMGIPfGXAIC4mZvabxMkTHgDhWegtcwWOA3gdh0uUNgo6tSccrnZ6Vx4QTtdAFsQBkGkOHE4mIF+DniVu6TeJy5XRu/C8b/qmHmHquIdIT6ZqXPkdz7sgsOVhuj3x1JYNHHJIcd3DHq8Asf5cLSo+cTAO8oFA3+FUbnGIVx9l+u2b7IPqPu7r6wfge4Nu/uV9fvv12seZ4vOWS4cqV3iEd+O6XcexUXDyBC/ynuQIMsqYpQnviX1MV+14qE7DdqN4mBaqNU0A2fMuWeibN39a3ChW7zBq8VLpN0nKlexn9cSOxY3zHAeLRxtJ1zigem0c90mHL8/hbhyJO2QYd+vYuGn9JQbNEx8c91k9xEASuqDMYQ/IOQxAdhRc0O8nLcQokoYf0Ozje9FZR1JWd8qt3LD+KJ1O6uJTywe2r07Zi+Pqldt4Y7So0MJGMYkS6/cRKOx3eVOlZBbPAkxxLeOg5CkLehxuwrjB5x6gT8BxHEAobrkzzzXGwXi2IDbS8xzEYmC5PgebTnMdA/VdFSHiV0WhGMxxyILQbzDG01NTeAwAmL5rxyhuIPlAOCbpkVQX9iAN+1D1B9acccTlehhNsnGAwuOiIXhYdpTfx2FVNLkLEJOS/lgI4StRFP2z+/0wiqRe7+DFp7VnL7AedyZNeHKBeE7lpEfX6dmT1aNtFtdlKkn5sp0oQu/CV3PV/iYEXj0sHm0kWNbKA35JYsLBUJDhidkB9Ombedh0sgQABFdSAmzedCBxzcOYfXhiTWlz9xys6LBAwNPjliVLF6CiCbDxEs9VD5AnIDOO7Hk8aEKzKxkrZo1xD6AbB+gjOwwG7iiDpHFbFEUf2JjCjr6RLbxr96MQ3okp7HUa2q8f7ObogAvEk6lK6v8lcHd6Jklr32ZS6Bok0yJJ+zs1s90l/wAsBulCknCvxPIkBOQHEW9Gvba9OFKgvivsADY8D0tCwLHdoHHDAAPrJ49hM6JAfI/bNW6SU7POQt50MKIZkop4iSQ3oEenkO4TmSvHXCddyHHEn7Gb0BFrSrWk4Rqe56Cww3FIqAy68ZKYXDJZMYGnx45KzOD7yY4yCH6YoPDerxzpjus4Kal1H5bwG3QKt1cP1imsb9gVNPIOiRSMIwLg6XEH0abJZa345E+L58ATD0WyCRDjRhu3JCX69sJFmo0edz1JTNBBwBPcTvF85NaSOIaTANs4ZEF6vfgSP+O4DwK4BPo8ouZxZVY8GzOBWw+r1AegMw4ZFlr3CFjGFbqXfOsaeg768TQGpfjST76wmngxs54D4GHUNj4ugO0oA61vtqN8rw8TFN577aCvZ07cb8a/wxSOhm3NLh8cj3fi3MEahpK0u8Vu1/pWzWwn0EcuV49RH55qJLQRjCO2hlml+K7dKQDyVFheYtY0btav5zce6Zy4i8E44hbjZidLvFGkUzz2aPwm4TBwGBUfxiGt4zECAOxe5nEVF8h7NEnpgOf5ZqinCGuWJ2yGgDy9T0/CF9k4EpDoMBE5hmbceNWjkIjiucaRseiDW/v43hXoq5KWdUejcPY+v32HKUxni8+njM23A25CkqyR4svFeNxWxGKUqtNwDzzAqaKDZwGm+sikO0ZizB4b5eyFzaMLGVeexxNT6KnSQUaxolgv1ZHAEVeawZe0EA/ESPGZJw+TQuOXwk08z0HzbGq2gteIW6OZnlPiuUqhIqSrJznK3DnigwnoYJ1oR6wpPSsdvsbjah/HHHrvs4/HAcaOQn3lw7JIkUbDoyufc1iJJk/r67qFV0MI3/tNZe7esdEo0n7tYBBQ2zzYtSxJ+SK7M6maCJ3g6DQsxWcPPKddckV6XF8EIJq79kbieU4CMuSu9yzAyOLFzDaUmCEmuSTJkfgA9+EqmQYAtzxrH0hIIkhiJtvj8o/L9Hm+GTHZnFHOjBCGBIyBSaQDSX6K171hzMx2T6IJxS16xgWBunEczuKCKR+bToL58Rh9jx0GwzaOUnofKIs+2LWP79q9o+Z+Ze6+Qby6UTt4cx1Hlh2xQrvbdtyiB7DRSTSZtkFM15Ews79dM9s97gk6EdMi7mGuaGOlYOkuY3BM4KBJ6IlPo/jJccj3EAD2LCaUpEQg3APWyBXpCX+gdz4OYWna8LAEo6MPSrYaRcyOobg6vG9PWcy4LtPiNLuPszl7zaHa3xKznhRz6GH9kQmMWZZN4rHHTCF2wYDuMNzHYwB8WK7vfQYqP8ji1d4yd++4j7P58vOnLi4eeNGtNXtR2V7ZxBsjpsSqvSz5Nk1adBq7tgvaAzzJBe3JxKZ+sA70GNjIcdTvpKSbzBgSCgrlotlOLIkk1bdspptcqp5MVopXpbHpOUzgN3WAbPpmVO7Mk/nOgI1clTwP6aDqAch0n+MoN0lGIvGedzEEt5hHiJ6+ewLAlMeTQ4eaccSJ0uGs3Ywff0mHyHG4XQ+jAtJhZL4fFYv0wU00uXckRfp6NnIwy9z1mkpnD74tyjyrLt0vZPEbzaqtLPFp1hNHRxvrwpmDga8kZXPMiNa3bGDpqWFLjCW5EschzEuAzKP/R8CSguTHEfTtSQIpTNnAchzK/iOgB4j99WxG49A6JEA2Dh1OAlvEKnmy0imzfRyVQuIKfUt8qMmVQH7KEQdFkmG+2vRUeSXeYcJzH+PQQqSYbwJTnpjbcUjS4Ddx1DaOa4cRl3hkLDra93pYiSZ93QGJKUk/ep/ffj3RJFN83qrmUV202bHWPsebdBrxFq7ZkzaTKElTMzYIISal24mf3ZZIshuFFmDamAsOPUUyAoUeF7Wn9qtlpSo/B7G7JIsj8diiRBQPU0i6jxQb5mFNe2MQU8b4Scp8dyQtxJX48YADYuSzBQbQlCBH78JXCtJeUzJD+2DkOfTkYXx6MqAJiNN9EEiXDkcFgNYtmmce3dNxlGEkQ4H3Q1ABGEfFnqNjH1zxaunrDGFH0l1F6J8IIXz5oESTRCKoPH3wIvr6V27GvylYEOZO2mxjucIbMzEUlEXdBuAq8anc41KK6zIqTdvsgyQNgN2l7zGORYfAAckUSQ6tQ0cgPtdPtgGEix2DZ8X4tTHovLlY06I9dsahLYnAEto9IQEooeJhd+E+6NDjmetFmKv0zfIQJy1JSTiIJh0HJyrfSeMiA3GNEsd00WHZA0DispGeLGoyV0JMTFJ0HO/igyRJE8nHur9X9rBB4d2vdG+JkPslmrxjyWTQwsLBk75xYcHs8M2vXMWbInfk4pk5s31ujkFhf2AP4suvrJnt+zu260HiDW0ccho00fZ3OeOWwBIBJc+CgMkspJU4YzPQEocdtBzi1bTZkIvPE6A8u2TX3d5es+Ma67Ud7IOYQA8jRMLouPnnefP3SFTFtXG4uUkRgcJmylVOAqEkpZkFew6sXLHXLInj/QiYStLMsn0oJ08PMaISz0MyTxY1MYGJkf2uPOOG44P5GgQ+xxG3GDvD+ZgAPpd9gLOPHyTR5J2Ywub+nl78wsGLz/aqnRl8+tFTD3ST97OddTum8PbVdbxGFjY0dGt53EExtc0kaTSM56YeR2zENIQEeFzDBMio7OB+jUH4oG8H+3sWcYpdpHHhed+bt+05Qou8x0VNbOI42AP6Zp5DD7nSkaV2pG9Oz4HmqENPkauJ2Pe5dGYG+9jZsA9wDagrPzVjl/+UmAlMO7K5200bpLjKd4IRgzbo0z3ED++hceEZN+OoyEN2XADZONQfDss+qDGF936h1yRd1B2QOLpPosk7MYWpdPH5rnG6p4XJE1NIrkJyS3lihMgogWMcWb2eSULaexQv5blPysIj93Jvn+Nm8LQLAMKTaEKMp4e5yiTixW15Fmh0fcG48LitKAjel+QUTzvSs2lSIhRdI6n4CQceIXpak4hB3lhhqZdcAfRXgbn1uI/7EOdZnnYwbHCNAmgy7m3zAS/uuPAks9A8ojnkiWfFPcAxR2ibiFsn+rDsKAOtb7ToUCVpQgh/WtJflzQfRdEW/f5hM4V3R/Vj+ror+V9af1QoJPXDP3LmwPZ6wx6Ab13hzOAcbN4dcDn5Aq7tBbRUtRf5BohGS9Luuv19PVmkNKErC7Yr0rPhNev2s9D79IAUeta48hCS1AUmhdyhkrR4bhl/Y9neDm/+xB4Mhvb49lTgGNGG5tiM4vbhiaNDJhA21cpCBfuYnrXH1u4mg5TpObsfitUjwCcxS7d42mYb+11eT6ZmWESbrDpnH1SvvHLLbM+DK17iQwutOZ5YUzKPpBIZysWMmNFExQOYI6P+wy9z936yKBpPHXGPhRBOS/peSTe8f/MwQeG9qOheSZqPhxAePSjRpN0Z6d//6sHur1zePqGtXbddZ5JUWbDZxnzRdv3ubvLmv7djZ6oWyvbiWZzmxbXfs91WHhavBSX7iCn0ZPrRqTpufI/EQIj68ADPyoId6+TJPt6B8IdxCPPGjTUlMCZx6UEC4RIDcbpPDwtC4zPnYHzIttdqcA+eZBUSA4fa3Q6lAdp4UzB+m3VH8hscqIslXi/qdfvwRWEalIDnsbjlESVef2n8etZW0sD0sGdxPU6eRBWW34GkHEcI0WFkQY/LDjGm8P8h6c9J+mnvH7wXFU2CjIom0bClZ587ePNduQ2Lo2NjJhdzuWIDspnFCvaxecve/Klc3ywEW0tSqWKfqD3VA8ozFbN9HDQ3xYbR4udZ2Dz1kS3LgziwxJsqAX2JwwbSSRtAeIRiR/14LiXPhleq2gcrT2YwyWlQJrYnjKMKTB8lSnlieykByRNHR9aCEIpCkYFnsWT/ptmIH3+2sGSPi1Sa1+d2074POmR6mELyTrT27PXEsyaRJA0xhZ55SNJmkSe29/hgKdOODdsYPVDt47kQwhfu+fenoyj6tOcPQwg/LGkliqKveDx6d+0wE02GuuNOPi2jokmuUHo+lz34AapVe2HzBPtT/A69QE9G12PPHuwCl6SlZSjIDtnLkvTKS3b1ltllDj4nRrMLmX7EbElSuWJvFMSK1rfY/UZVamiB9ZxEc7D5k5tcii+SPXTE2FP9bwL6HoaZZEM82cfT8zw+LfMwmmT0TbvA0ElSCxiKqRlmTQn0EQihuFyPVaqwJjn6aLUgNMGQG7trVBWFMso9zGwubX/3NhzePLWm6ZuNYw4NIeHFxRSS+xiSrcYBxj5IZe4esKLJVhRFLxzUGEL4JUlL92n6MUl/QXdcxw9kh1nR5O7/H9wn0eQd63SG+tX/cPvAi2bg1E3xgpJUOH+/d/h1ow9GLmxJKpZIC87++26PJwEt0tsrGFOKzCpl2Z24eBL7KJTtjYACyysglSHxpkobhUfGorZRM9s95cxIyoJAtmejSMAcoE3VU6WGzLNReN6XZQR+JVYroHq+JEIvSVNV+5v2Oozk6XBGh91mw8HuTtlAqFKx17W64zC8X4MEuh4Dy8qsHW9N8ZV7OyyTReOT1CM8LmqKWyTG05MomOjbfYSEI648iq8RGNfiupelY8QUykemeCyKou++338PITwj6bykuyzhKUlfCiF8LIoiU1vqYYLCjqS7szvo60kn9xMSvMd93NaFRw/WCVxftSe8x1dPkge0+Xs0wfYhLmZ20WYPZmb5RE1u7p3V+JOEygZ6WNO9HZtBI0BGp3YpfhkwAgce89T7JZBNjGc6wwcSYrfoXXlK2MWt1iDFr0zh0myEMAzyCtCBRZLWb9XM9oyjZCWNT0pcq86x/l86YwOIeh1UF7J8IMnMQznJjEdmhao9UXayQ1IJgGVtww698bh2KZlgHGAL44ePSU3huOLWnmscGTuEMndRFL0k6R1R5xDCNUkvvNfZx/euUl+VtKw7GoVP3ee33+A+nqkevIjeumYvCCSrIDHDloNYJasM310rVexFmgSwB47MtP1a/ILqPWAYODM4/uZPC9c4lP2LFcg+dgBPep+kiScxWKKYw16S3zfFVBGYauxyLOo4ShOSzMo4EpDImnV7DpFKgCRNAwihQ6gknblkey8ae1Ca07HHbG/az1ougzg7uIYl6cRJGxRm0ry5r9y277MHJUA9MZy0KZfg4O9iwmEuE1vkSTShRMHI4VmIm33ssbjJKu8v93F0pAHsYSWaPKOvZx/fCiH8mSiK/tn9fphMSqcWD35hs99pVzRZ3eTB8+brdpLH9Vft7G1PUkKjZrNjX/uS3b5w0paCkaQsMBCe+EqK9xtHNiy5qCuz9j143EEkz0PxZ57g9HGA1ywwPgSUPOCVEntogfbEV7Lwroc1tQ8LlOTh0YrDrF0Ym75qOiCH5Mi4pVqo6O50sHjLy6Dvt2eDrak0M8i1XagWtcTXyFN4zqz9HB7NRhK7p/ftSaSisZWBuMbKPIfN0MFp6BB4R8mZMagAKPHwQd1xch8fpk7h2/2d8/72sEDhvV/rftHl77iPu+19LZQO3viutO1N8/KbNbwZcvNhvV/I+pU8Ljo77otcJJK0D8CTSpFJrHtHMWzzp+exjxTE1iydshe/6jyzNWRciswR1B2zjq7EgItAn2fhm56xDxTkrvecyuk5PLGPTWA5KF51fpkrbOxu2XOkQOzYPgP9Ilyjts0JSAQ+ex1gmB0sXjSCcQGbP4I1SZmc/d2bTZ4j9d14c6AKLmyJQ282VuxYVM88zGTJ42TPw9omg1vaq7rNcWgIxtccHYd7+H1jkQ+sv1d2WNnHkaSR7sQVdu/DEr7jPs7mys+v1A6esG9dsxe/pROc6ZeBuJbKjM0ard6sYR8ECin5ogr3IElpYAduvL6K1yhVwa0K0gwNR8ZtFgDw5VftTcBTHYM070hslhJVJD4xT88xu0uWAdkQH3MFDBzEUy1DIpYkba3Y9ZF9MVf2s9y+vGK2k+tX4pjXXdh4PTGc5K4kRl/i2EZiFkjwWZKmK/bYmqnY82yn5gB0NXt9zmYdVZZAXoeyk/cAVEp8GKAQIo+HJOk4aFpG7mfJU/vYUamJyqU6Sj0+bHs/JZpEio60O/ywso+lryea3G/1eocp7LQbSiXf/enm1vUa/oYYNioMTy4piWPtZkBAmxY+ibMaPRsaMYUEhPKKX8EgX7IX4DbI4niMYoQ8FSFuX7a/KZXBkzihpVC2DwOeSjfDEWSlb9iAbvMmdoExmp7NiN4FZXh6MkBDwn6fxGgW4XtI0iOX7PVic8sha9OyWaP+tt3ukYtpNu31IpeDsAIHYVQq22tO36Gq0Gnbz0qsvof178D7pvXbk8xCGpjk+vWEcaQBCA3GAJRIjcBzn2QE6Dw1yI+NHUKiSRx7mKDw3pHS1x2QmJL0o/f57TtMYTpben5t++ABUq9BwLXjZc8uVcx2WmBp45ak+WW7j1zBXjxvvGVrEEoMLHMAtiR2eZJ7wvO+iU2kxdOTQUfxZ6Qh6NGjy0HcoSfphlxG28AOlKYdWaZUQaZjX8MTq0fP4XEfUwgFu/x5+Zqes8HrNMjJFIv8Lgj0DYc8R3bWbVc6fRNPYhqJU6dS9sZsJQDetcfP2fdxY52vQe9rY81eLzxC3hSeQ1JEnoRGilOm8e0pFUm6pQOPCoAjw96ysbBeo/gu6uNkH1RQKH29vF1H0l0E8xMhhC8fVOYumZSW5w5+YdMlGwitbbEbZeWWDRDocJUBaQfJV1nFstMXOFZvH+KdPMHQZBTjlnHoQhKQIZDtqmgCgIxcfB7ZEEoe2t/mGCB6X7SRuArcgyuSKq9QEL7EOm6e+MpGzd7cien2uIuostDWin2NKahrLHHM7DQkRkjS2UcOluGSpBtX7OfwxBRmcvYcuHHVTsC7dY3f960Fe/3dAvArsWbj7JIdptGCiigSV7UahysSlRsgxMJz6KE+PFqH9KyHoWM4Dhf1cXEfS9HYdAofhj1sUHj3K92L5HL67WXu3rEokhrtgwfyiy/ZG8nCArN4pNx/9U07k3XLIQq9dHbRbM9A8gRlI0osdeGJxSMQQuV4PHEvtLkvnbSBPumWSdLKDRuQ7azXzHZP9nEXgv09LlMqbdWHAvYeNjJuZpunXF8ZpFrWrm3gNZbO2UoC9K48RgcOiq+cW+RDJrF0jT1moek+pqv2Nzl1hsHr9jaAENA6zDo0SWdm7DkwO8uZ2O22DfrqNZLRcngWINGEwnf6PV5b6T5IfN0DCklpYBDFL104jiQRDzg178EBouL2cVgWfYDdx99c5u6uzeq3l7n7ekxhC1wD4Ha9fpXZmu3VmtlOg5wAn8QZtWR7juLzdKL2DDxiE+k069HSoszJ9VX7OcqO0ljEBJIOXM9RaYHKI3pAIcU2Umb80AGUyBVOSTmjIT8HAYS5k1y7m/TmugCAh2Mo7Ubmie2lChye8Utuasqivn6VS0HOzIEkDYhX7zvWpFTaPuB5XNDNFoAp2ANGQ97WEnDQ3N2254gneYiM2HSKtZakIST/eIBSNAK2cQwMHIG6cQC6o5y88Q0Wcaz+e2mHlWjymqSLugMSR/cpc/dOTGEmV3q+vn8wmCGX6fwSn+xJzf7yV+9XdOXr5il9tbtjM2jbazWz3cOSUNxWf8DuY6rO0u/aCxPVLZY4/obcnZ5sQgKF2xs28KQqNpLUadgbhSfgmrQMKRvck1zR3rMBBG1G5JKSpBPnbHenZ9Mkd2anZYNTT+wYeQV2YJ565iEdvmZm+ODUaFDSmP2ufGyN3T6/YM/lvqMe+6ll+5vcXvOI3dvvvAvFA9IODwkBXKr+4inX5xHRtsyzntDYSzjeRdR578HU/7+9O421qyrDOP5/vBRQhNBwacBS2mJaaxlUKBpQCRI1ILFqaQhOgcRIiIDxA0ZNjCCECMQpURCNaXCkRtCGqLQRGRUwFFooQ6i0Ml3AYqVCSelAXz/sjWyv7Vn79J6zJ55fctKzz9T3vXs471l7r7WqGCC7OV67vY+LW+Ocwv1re71p27Zg7dqdtxamDtKPr+l97Q2kWyAOnHlg8jNSUqd2U9fqvVziiznVSlfmyz01N2xqwvX1T6e/KJLFa6K1cnKJQVz/8UTvU/qpnn57JIagANiUuByqzN87dZBOHeRThSnAlpd65zqSGL9y84vpInz1it4/nFLFLaRPfaXG2Swz5Mfjib9narsYndq7+C0jNRsJTHzu7qnT08MhPTPWuzUxdbqzzLh5KxMtINNmvyn5GakfzHu+ofdxr0wxljo+p6aoK/Mjct8p+/Z8PtWRZN3jPaeoBdKd8CbaiWRQqjh9PIgp/aoQlJuOty5VdTR5nlevK3yXpNnjOpr89/Txpo0buPuGe4YcllUp1V750sbexcGzjz01uGBqtnVT7wLhxcQQQYMwiFMXLyR6Nb6wvnenhbZ4/tnew/c0xZoVdUdQzoZn0qMqdEVX9oHBaEfBVolI/+io07DbW4sdTaLw2AfGve424CHgoZHdJj7unZmZmVnzZHMfl7nVoaqOJhuAvfLlafx/RxMzMzOzznutXlNYbIV8jKyFcA6wZQcdTczMzMw6LSIa3ftYEx3TbKcfLG0ELgEuAh4lu65wLvCpiPjVuNeeCZyZLx4G3D+UoOo1CqQHOGwf59UuzqtdnFf7dDW3LuQ1PSLSM0MMkaSlZH/LMv4ZEScOM57xhl0UfgS4AZgC/AK4IyLOT7xveUTMG0pQNXJe7eK82sV5tUtX84Lu5tbVvOx/DfP08euBb+X3x4BngSmSDgKuiojbhvh/m5mZmVkfhtb7OCJGIuLtwFRgCdn8x5OAA4D0AExmZmZmVplhj1NIRDwFnNrHW340rFhq5rzaxXm1i/Nql67mBd3Nrat5WcHQrik0MzMzs/Zoy2SBZmZmZjZELgrNzMzMrDlFoaQTJT0s6RFJX647nn6kYpd0nKR7JG2TtHDccy9LWpnfrqsu6v6UyPEsSavyPP4saW4dcZZRdluTdIqkkDQvX54haVNhfV1ZXdT9KZOjpFMlPSjpAUm/rDrGskpse98prJPVkjYUnuvK/jVd0p8k3Sfp5nwUh8aTtEjSOkk7HHtW0hxJd0jaLOm8quPbVSXy+mS+rlZJul3S26qOcVek8iq87ugdfZ9ZB0RE7TdgBFgDHALsDtwLzK07rkHFDswAjgB+Ciwc99zGunMYUI77FO7PB5bWHfeu5pK/bm/gVuBOYF5hPd5fdw4DWl+zgBXA5Hx5St1xT2R9FV5/LrCosNyV/evXwOn5/ROAn9Udd8ncjgOO3Nl+QzaG7dHAxcB5dcc7wLyOLexbJwF/rTvmQeSVv2YEuBH4w/jvM9/af2tKS+E7gUciYm1EbAEWkw183QbJ2CPi0Yi4D2juhIe9lcnx+cLiXkBTezCV3dYuAi4lG0qpbcrk+Fng8oh4DiAi1lUcY1n9Hhs+DlxdSWSDUybHuWRfxAA37eD5RoqIW4F/9Xh+XUTcBWytLqqJK5HX7a/sW2Q/LFvRspvKK3cucC3Q1GOGTUBTisKpwBOF5Sfzx9pgorHvKWm5pDslfXSgkQ1OqRwlnS1pDXAZ8PmKYutXMhdJRwLTIuL3O3j/TEkrJN0i6b1DjHMiyqyv2cBsSX/Jt71Kp1LqQ+n9S9J0YCavFk/Qnf3rXmBBfv9jwN6S9qsgNpu4zwDX1x3EIEiaSrb9/aDuWGw4hj5OoSVNj4gxSYcAN0paFRFr6g5qV0TE5cDlkj4BfBU4veaQ+ibpdcC3gTN28PTTwMERsV7SUcASSYeOayVti93ITiEfT9aKcaukwyNiQ51BTdBpwDURUZxtviv713nA9yWdQXZZwxjwcs93WO0kvY+sKHxP3bEMyHeBL0XEdkl1x2JD0JSicAyYVlg+KH+sDSYUe0SM5f+ulXQz8A6y64uapN8cF9PcX5KpXPYGDgNuzg96BwDXSZofEcuBzQARcXfeKjobWF5F4H0os76eJLvOaSvwd0mryYrEu6oJsbR+tr3TgLOLD3Rl/4psEoAFAJLeCJzS8gK+8yQdAfwYOCki1tcdz4DMAxbnx8ZR4EOStkXEklqjsoFpyunju4BZkmZK2p3s4N7YnoLj7HLskiZL2iO/Pwq8G3hwaJHuumSOkmYVFk8G/lZhfP3omUtE/DsiRiNiRkTMILseaH5ELJe0v6QRgLzlaRawtvoUkspsk0vIWglf2fZm095ckDQHmAzcUXisS/vXaN6KDfAVYFHFMVofJB0M/Ab4dESsrjueQYmImYVj4zXA51wQdksjWgojYpukc4BlZD2bFkXEAzWHVcrOYpd0IbA8Iq6TdDTwW7IvrQ9L+npEHAq8FfihpO1kBfolEdG4L60yOQLnSHo/2QXjz9HQU8clc9mZ44ALJW0l6zR0VkSkLsquXMkclwEflPQg2WnILzaxNaOP9XUasDgiih2curR/HQ98Q1KQnT4+e6cf2CCSriaLfVTSk8D5wCSAiLhS0gFkLe37ANslfYGs53WjL8lI5QV8DdgPuCJvVdsWEfPqiba8EnlZx3maOzMzMzNrzOljMzMzM6uRi0IzMzMzc1FoZmZmZi4KzczMzAwXhWZmZmaGi0IzaxBJ+0lamd+ekTSW398o6Yq64zMz6zIPSWNmjSTpAmBjRHyz7ljMzF4L3FJoZo0n6XhJv8vvXyDpJ5Juk/SYpAWSLpO0StJSSZPy1x0l6RZJd0taJunAerMwM2s2F4Vm1kZvBk4A5gM/B26KiMOBTcDJeWH4PWBhRBxFNi3cxXUFa2bWBo2Y5s7MrE/XR8RWSavIpoZbmj++CpgBvAU4DPhjPs3YCPB0DXGambWGi0Iza6PNABGxXdLWwpzH28mOawIeiIhj6grQzKxtfPrYzLroYWB/SccASJok6dCaYzIzazQXhWbWORGxBVgIXCrpXmAlcGytQZmZNZyHpDEzMzMztxSamZmZmYtCMzMzM8NFoZmZmZnhotDMzMzMcFFoZmZmZrgoNDMzMzNcFJqZmZkZ8B9chTgmzCtqlwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# for plotting features\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure(figsize=(10, 4))\n",
    "librosa.display.specshow(features, y_axis='chroma', x_axis='time')\n",
    "plt.colorbar()\n",
    "plt.title('Chromagram')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extracting only Log mel Spectrogram"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Processing breath2/neg\n",
      "breath2/neg/1neg-0421-078-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/1neg-0421-091-breathe-m-47.wav, segment : 0\n",
      "breath2/neg/1neg-0421-089-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/0neg-0423-142-breathe-m-24.wav, segment : 0\n",
      "breath2/neg/1neg-0420-058-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0423-146-breathe-m-35.wav, segment : 0\n",
      "breath2/neg/0neg-0422-114-breathe-m-26.wav, segment : 0\n",
      "breath2/neg/0neg-0423-140-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0422-106-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0420-064-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0421-082-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/1neg-0422-099-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-068-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0420-045-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/0neg-0423-147-breathe-m-34.wav, segment : 0\n",
      "breath2/neg/1neg-0423-122-breathe-m-36.wav, segment : 0\n",
      "breath2/neg/0neg-0420-075-breathe-f-17.wav, segment : 0\n",
      "breath2/neg/0neg-0423-136-breathe-m-40.wav, segment : 0\n",
      "breath2/neg/0neg-0420-052-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0420-046-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/1neg-0420-063-breathe-f-26.wav, segment : 0\n",
      "breath2/neg/1neg-0422-116-breathe-m-44.wav, segment : 0\n",
      "breath2/neg/1neg-0421-076-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/0neg-0423-123-breathe-m-48.wav, segment : 0\n",
      "breath2/neg/0neg-0422-095-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/1neg-0420-073-breathe-m-49.wav, segment : 0\n",
      "breath2/neg/1neg-0420-051-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/1neg-0420-047-breathe-f-22.wav, segment : 0\n",
      "breath2/neg/1neg-0422-114-breathe-m-26.wav, segment : 0\n",
      "breath2/neg/0neg-0422-117-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0423-134-breathe-m-41.wav, segment : 0\n",
      "breath2/neg/1neg-0421-081-breathe-f-54.wav, segment : 0\n",
      "breath2/neg/1neg-0422-107-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/1neg-0420-064-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0420-051-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/0neg-0423-149-breathe-m-21.wav, segment : 0\n",
      "breath2/neg/0neg-0420-057-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0420-069-breathe-m-31.wav, segment : 0\n",
      "breath2/neg/1neg-0420-075-breathe-f-17.wav, segment : 0\n",
      "breath2/neg/0neg-0421-081-breathe-f-54.wav, segment : 0\n",
      "breath2/neg/0neg-0422-101-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0420-062-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0421-083-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0422-097-breathe-m-37.wav, segment : 0\n",
      "breath2/neg/0neg-0420-062-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0420-074-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/0neg-0421-076-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/1neg-0420-065-breathe-m-59.wav, segment : 0\n",
      "breath2/neg/1neg-0422-101-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0422-097-breathe-m-37.wav, segment : 0\n",
      "breath2/neg/0neg-0422-102-breathe-m-64.wav, segment : 0\n",
      "breath2/neg/1neg-0422-115-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-066-breathe-f-52.wav, segment : 0\n",
      "breath2/neg/1neg-0420-044-breathe-f-65.wav, segment : 0\n",
      "breath2/neg/0neg-0420-058-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0423-148-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/0neg-0420-045-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/0neg-0420-046-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0420-044-breathe-f-65.wav, segment : 0\n",
      "breath2/neg/1neg-0420-052-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0422-102-breathe-m-64.wav, segment : 0\n",
      "breath2/neg/0neg-0423-132-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0422-106-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0423-138-breathe-m-24.wav, segment : 0\n",
      "breath2/neg/1neg-0420-068-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0421-080-breathe-f-50.wav, segment : 0\n",
      "breath2/neg/1neg-0421-082-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0420-074-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0420-057-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/0neg-0420-063-breathe-f-26.wav, segment : 0\n",
      "breath2/neg/1neg-0420-066-breathe-f-52.wav, segment : 0\n",
      "breath2/neg/0neg-0420-069-breathe-m-31.wav, segment : 0\n",
      "breath2/neg/0neg-0421-083-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0420-065-breathe-m-59.wav, segment : 0\n",
      "breath2/neg/0neg-0422-107-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/0neg-0423-141-breathe-m-20.wav, segment : 0\n",
      "breath2/neg/1neg-0422-095-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0423-144-breathe-m-42.wav, segment : 0\n",
      "breath2/neg/0neg-0423-139-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0422-117-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0423-150-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0421-089-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0422-112-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-072-breathe-f-33.wav, segment : 0\n",
      "breath2/neg/0neg-0423-151-breathe-m-20.wav, segment : 0\n",
      "\n",
      " Processing breath2/pos\n",
      "breath2/pos/1pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/1pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/2pos-0420-048-breathe-f-30.wav, segment : 0\n",
      "breath2/pos/0pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/1pos-0426-129-breathe-f-26.wav, segment : 0\n",
      "breath2/pos/1pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/2pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/2pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0421-084-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/1pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/0pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/2pos-0420-055-breathe-m-59.wav, segment : 0\n",
      "breath2/pos/1pos-0422-096-breathe-m-31.wav, segment : 0\n",
      "breath2/pos/1pos-0420-048-breathe-f-30.wav, segment : 0\n",
      "breath2/pos/0pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/0pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/1pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/1pos-0421-077-breathe-m-60.wav, segment : 0\n",
      "breath2/pos/2pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/2pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/2pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/2pos-0422-109-breathe-m-40.wav, segment : 0\n",
      "breath2/pos/1pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/0pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/0pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/1pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/2pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/0pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/2pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/2pos-0421-086-breathe-m-65.wav, segment : 0\n",
      "breath2/pos/1pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/1pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/2pos-0420-049-breathe-m-48.wav, segment : 0\n",
      "breath2/pos/1pos-0421-093-breathe-f-24.wav, segment : 0\n",
      "breath2/pos/0pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/baat2.wav, segment : 0\n",
      "breath2/pos/1pos-0426-127-breathe-m-34.wav, segment : 0\n",
      "breath2/pos/1pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/2pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/0pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0421-086-breathe-m-65.wav, segment : 0\n",
      "breath2/pos/1pos-0422-109-breathe-m-40.wav, segment : 0\n",
      "breath2/pos/0pos-0426-124-breathe-m-38.wav, segment : 0\n",
      "breath2/pos/0pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/0pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0426-127-breathe-m-34.wav, segment : 0\n",
      "breath2/pos/1pos-0422-103-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/1pos-0426-124-breathe-m-38.wav, segment : 0\n",
      "breath2/pos/0pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0420-049-breathe-m-48.wav, segment : 0\n",
      "breath2/pos/1pos-0420-055-breathe-m-59.wav, segment : 0\n",
      "breath2/pos/1pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/0pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/1pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/0pos-0426-129-breathe-f-26.wav, segment : 0\n",
      "breath2/pos/2pos-0426-129-breathe-f-26.wav, segment : 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "breath2/pos/2pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0426-124-breathe-m-38.wav, segment : 0\n"
     ]
    }
   ],
   "source": [
    "import os \n",
    "import librosa\n",
    "import math\n",
    "import json\n",
    "\n",
    "\n",
    "DATASET_PATH = \"breath2\"\n",
    "JSON_PATH = \"FYP_test1.json\"\n",
    "SAMPLE_RATE = 22050\n",
    "DURATION = 4  # measured in seconds.\n",
    "SAMPLES_PER_TRACK = SAMPLE_RATE * DURATION0pos-0426-126-breathe-m-46\n",
    "\n",
    "def save_mfcc(dataset_path, json_path, n_mfcc = 13, n_fft = 2048, hop_length= 512, num_segments = 5):\n",
    "   \n",
    "    # dictionary to store data\n",
    "    data = {\n",
    "        \"mapping\": [],\n",
    "        \"mfcc\": [],\n",
    "        \"labels\": []\n",
    "    }\n",
    "    \n",
    "    num_samples_per_segment =  int(SAMPLES_PER_TRACK / num_segments)\n",
    "    \n",
    "    expected_no_of_mfcc_vectors_per_segment = math.ceil(num_samples_per_segment / hop_length) \n",
    "    \n",
    "    for i, (dirpath, dirname, filenames) in enumerate(os.walk(dataset_path)):\n",
    "        \n",
    "        # Ensure that we r not at the root level\n",
    "        if dirpath is not dataset_path:\n",
    "            dirpath_components = dirpath.split(\"\\\\\")  # \"genre/blues\" => [\"genre\", \"blues\"]\n",
    "            semantic_label = dirpath_components[-1]\n",
    "            data[\"mapping\"].append(semantic_label)\n",
    "            \n",
    "            print(\"\\n Processing {}\".format(semantic_label))\n",
    "            \n",
    "            # Process files for specific genre. \n",
    "            for f in filenames:\n",
    "                \n",
    "                # laoding audio file\n",
    "                file_path = os.path.join(dirpath, f)\n",
    "                signal, sr = librosa.load(file_path, sr = SAMPLE_RATE)\n",
    "                \n",
    "                # divide signals into segments,process segments, extract mfcc and at last store mfcc vectors.\n",
    "                for s in range(num_segments):\n",
    "                    start_sample = num_samples_per_segment * s  # if s= 0 -> then start_sample = 0 \n",
    "                    finish_sample = start_sample + num_samples_per_segment \n",
    "                    \n",
    "                    \n",
    "                    features = lms(signal[start_sample: finish_sample],\n",
    "                            audio_sample_rate=16000,\n",
    "                            log_offset=0.01,\n",
    "                            window_length_secs=0.025,\n",
    "                            hop_length_secs=0.010,\n",
    "                            num_mel_bins = 64,\n",
    "                            lower_edge_hertz = 125,\n",
    "                            upper_edge_hertz = 7500\n",
    "                        )\n",
    "                    data[\"mfcc\"].append(features.tolist()) # we can not append mfcc directly blc its a numpy array so we have\n",
    "                    # to first convert it into list.\n",
    "\n",
    "                    data[\"labels\"].append(i-1)\n",
    "                    # i mean in each iteration we are in different genre folder. why r we doing i -1 that's blc in first \n",
    "                    # iteration we were not inside the dir containing all the genres folder. \n",
    "\n",
    "                    print(\"{}, segment : {}\".format(file_path, s))\n",
    "\n",
    "    # final step: saving everything as a json file            \n",
    "    with open(json_path, \"w\") as fp:\n",
    "        json.dump(data, fp, indent = 4)  # indent mean spaces while writing. fp mean file_path  \n",
    "        \n",
    "        \n",
    "if __name__ == \"__main__\":\n",
    "    save_mfcc(DATASET_PATH, JSON_PATH, num_segments = 1)\n",
    "                "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Combining 4 features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Processing breath2/neg\n",
      "breath2/neg/1neg-0421-078-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/1neg-0421-091-breathe-m-47.wav, segment : 0\n",
      "breath2/neg/1neg-0421-089-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/0neg-0423-142-breathe-m-24.wav, segment : 0\n",
      "breath2/neg/1neg-0420-058-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0423-146-breathe-m-35.wav, segment : 0\n",
      "breath2/neg/0neg-0422-114-breathe-m-26.wav, segment : 0\n",
      "breath2/neg/0neg-0423-140-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0422-106-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0420-064-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0421-082-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/1neg-0422-099-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-068-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0420-045-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/0neg-0423-147-breathe-m-34.wav, segment : 0\n",
      "breath2/neg/1neg-0423-122-breathe-m-36.wav, segment : 0\n",
      "breath2/neg/0neg-0420-075-breathe-f-17.wav, segment : 0\n",
      "breath2/neg/0neg-0423-136-breathe-m-40.wav, segment : 0\n",
      "breath2/neg/0neg-0420-052-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0420-046-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/1neg-0420-063-breathe-f-26.wav, segment : 0\n",
      "breath2/neg/1neg-0422-116-breathe-m-44.wav, segment : 0\n",
      "breath2/neg/1neg-0421-076-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/0neg-0423-123-breathe-m-48.wav, segment : 0\n",
      "breath2/neg/0neg-0422-095-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/1neg-0420-073-breathe-m-49.wav, segment : 0\n",
      "breath2/neg/1neg-0420-051-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/1neg-0420-047-breathe-f-22.wav, segment : 0\n",
      "breath2/neg/1neg-0422-114-breathe-m-26.wav, segment : 0\n",
      "breath2/neg/0neg-0422-117-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0423-134-breathe-m-41.wav, segment : 0\n",
      "breath2/neg/1neg-0421-081-breathe-f-54.wav, segment : 0\n",
      "breath2/neg/1neg-0422-107-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/1neg-0420-064-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0420-051-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/0neg-0423-149-breathe-m-21.wav, segment : 0\n",
      "breath2/neg/0neg-0420-057-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0420-069-breathe-m-31.wav, segment : 0\n",
      "breath2/neg/1neg-0420-075-breathe-f-17.wav, segment : 0\n",
      "breath2/neg/0neg-0421-081-breathe-f-54.wav, segment : 0\n",
      "breath2/neg/0neg-0422-101-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0420-062-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0421-083-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0422-097-breathe-m-37.wav, segment : 0\n",
      "breath2/neg/0neg-0420-062-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0420-074-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/0neg-0421-076-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/1neg-0420-065-breathe-m-59.wav, segment : 0\n",
      "breath2/neg/1neg-0422-101-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0422-097-breathe-m-37.wav, segment : 0\n",
      "breath2/neg/0neg-0422-102-breathe-m-64.wav, segment : 0\n",
      "breath2/neg/1neg-0422-115-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-066-breathe-f-52.wav, segment : 0\n",
      "breath2/neg/1neg-0420-044-breathe-f-65.wav, segment : 0\n",
      "breath2/neg/0neg-0420-058-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0423-148-breathe-m-38.wav, segment : 0\n",
      "breath2/neg/0neg-0420-045-breathe-f-29.wav, segment : 0\n",
      "breath2/neg/0neg-0420-046-breathe-m-22.wav, segment : 0\n",
      "breath2/neg/0neg-0420-044-breathe-f-65.wav, segment : 0\n",
      "breath2/neg/1neg-0420-052-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/1neg-0422-102-breathe-m-64.wav, segment : 0\n",
      "breath2/neg/0neg-0423-132-breathe-m-45.wav, segment : 0\n",
      "breath2/neg/1neg-0422-106-breathe-m-19.wav, segment : 0\n",
      "breath2/neg/0neg-0423-138-breathe-m-24.wav, segment : 0\n",
      "breath2/neg/1neg-0420-068-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0421-080-breathe-f-50.wav, segment : 0\n",
      "breath2/neg/1neg-0421-082-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0420-074-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0420-057-breathe-m-39.wav, segment : 0\n",
      "breath2/neg/0neg-0420-063-breathe-f-26.wav, segment : 0\n",
      "breath2/neg/1neg-0420-066-breathe-f-52.wav, segment : 0\n",
      "breath2/neg/0neg-0420-069-breathe-m-31.wav, segment : 0\n",
      "breath2/neg/0neg-0421-083-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0420-065-breathe-m-59.wav, segment : 0\n",
      "breath2/neg/0neg-0422-107-breathe-m-46.wav, segment : 0\n",
      "breath2/neg/0neg-0423-141-breathe-m-20.wav, segment : 0\n",
      "breath2/neg/1neg-0422-095-breathe-m-53.wav, segment : 0\n",
      "breath2/neg/0neg-0423-144-breathe-m-42.wav, segment : 0\n",
      "breath2/neg/0neg-0423-139-breathe-m-32.wav, segment : 0\n",
      "breath2/neg/1neg-0422-117-breathe-m-28.wav, segment : 0\n",
      "breath2/neg/0neg-0423-150-breathe-m-23.wav, segment : 0\n",
      "breath2/neg/0neg-0421-089-breathe-f-20.wav, segment : 0\n",
      "breath2/neg/1neg-0422-112-breathe-m-50.wav, segment : 0\n",
      "breath2/neg/0neg-0420-072-breathe-f-33.wav, segment : 0\n",
      "breath2/neg/0neg-0423-151-breathe-m-20.wav, segment : 0\n",
      "\n",
      " Processing breath2/pos\n",
      "breath2/pos/1pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/1pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/2pos-0420-048-breathe-f-30.wav, segment : 0\n",
      "breath2/pos/0pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/1pos-0426-129-breathe-f-26.wav, segment : 0\n",
      "breath2/pos/1pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/2pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/2pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0421-084-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/1pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/0pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/2pos-0420-055-breathe-m-59.wav, segment : 0\n",
      "breath2/pos/1pos-0422-096-breathe-m-31.wav, segment : 0\n",
      "breath2/pos/1pos-0420-048-breathe-f-30.wav, segment : 0\n",
      "breath2/pos/0pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/0pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/1pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/1pos-0421-077-breathe-m-60.wav, segment : 0\n",
      "breath2/pos/2pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/2pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/2pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/2pos-0422-109-breathe-m-40.wav, segment : 0\n",
      "breath2/pos/1pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/0pos-0418-042-breathe-m-33.wav, segment : 0\n",
      "breath2/pos/0pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/1pos-0420-054-breathe-m-45.wav, segment : 0\n",
      "breath2/pos/2pos-0420-061-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/0pos-0426-128-breathe-m-37.wav, segment : 0\n",
      "breath2/pos/2pos-0423-121-breathe-m-36.wav, segment : 0\n",
      "breath2/pos/2pos-0421-086-breathe-m-65.wav, segment : 0\n",
      "breath2/pos/1pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/1pos-0421-094-breathe-m-51.wav, segment : 0\n",
      "breath2/pos/2pos-0420-049-breathe-m-48.wav, segment : 0\n",
      "breath2/pos/1pos-0421-093-breathe-f-24.wav, segment : 0\n",
      "breath2/pos/0pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/baat2.wav, segment : 0\n",
      "breath2/pos/1pos-0426-127-breathe-m-34.wav, segment : 0\n",
      "breath2/pos/1pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/2pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/0pos-0418-043-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0421-086-breathe-m-65.wav, segment : 0\n",
      "breath2/pos/1pos-0422-109-breathe-m-40.wav, segment : 0\n",
      "breath2/pos/0pos-0426-124-breathe-m-38.wav, segment : 0\n",
      "breath2/pos/0pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/0pos-0421-087-breathe-f-40.wav, segment : 0\n",
      "breath2/pos/2pos-0426-127-breathe-m-34.wav, segment : 0\n",
      "breath2/pos/1pos-0422-103-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/1pos-0426-124-breathe-m-38.wav, segment : 0\n",
      "breath2/pos/0pos-0420-053-breathe-m-29.wav, segment : 0\n",
      "breath2/pos/1pos-0420-049-breathe-m-48.wav, segment : 0\n",
      "breath2/pos/1pos-0420-055-breathe-m-59.wav, segment : 0\n",
      "breath2/pos/1pos-0421-092-breathe-m-53.wav, segment : 0\n",
      "breath2/pos/0pos-0426-126-breathe-m-46.wav, segment : 0\n",
      "breath2/pos/1pos-0422-113-breathe-m-50.wav, segment : 0\n",
      "breath2/pos/0pos-0426-129-breathe-f-26.wav, segment : 0\n",
      "breath2/pos/2pos-0426-129-breathe-f-26.wav, segment : 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "breath2/pos/2pos-0426-125-breathe-m-30.wav, segment : 0\n",
      "breath2/pos/2pos-0426-124-breathe-m-38.wav, segment : 0\n"
     ]
    }
   ],
   "source": [
    "import os \n",
    "import librosa\n",
    "import math\n",
    "import json\n",
    "\n",
    "\n",
    "DATASET_PATH = \"breath2\"\n",
    "JSON_PATH = \"FYP_test1.json\"\n",
    "SAMPLE_RATE = 22050\n",
    "DURATION = 4  # measured in seconds.\n",
    "SAMPLES_PER_TRACK = SAMPLE_RATE * DURATION\n",
    "\n",
    "def save_mfcc(dataset_path, json_path, n_mfcc = 13, n_fft = 2048, hop_length= 512, num_segments = 5):\n",
    "   \n",
    "    # dictionary to store data\n",
    "    data = {\n",
    "        \"mapping\": [],\n",
    "        \"mfcc\": [],\n",
    "        \"labels\": []\n",
    "    }\n",
    "    \n",
    "    num_samples_per_segment =  int(SAMPLES_PER_TRACK / num_segments)\n",
    "    \n",
    "    expected_no_of_mfcc_vectors_per_segment = math.ceil(num_samples_per_segment / hop_length) \n",
    "    \n",
    "    for i, (dirpath, dirname, filenames) in enumerate(os.walk(dataset_path)):\n",
    "        \n",
    "        # Ensure that we r not at the root level\n",
    "        if dirpath is not dataset_path:\n",
    "            dirpath_components = dirpath.split(\"\\\\\")  # \"genre/blues\" => [\"genre\", \"blues\"]\n",
    "            semantic_label = dirpath_components[-1]\n",
    "            data[\"mapping\"].append(semantic_label)\n",
    "            \n",
    "            print(\"\\n Processing {}\".format(semantic_label))\n",
    "            \n",
    "            # Process files for specific genre. \n",
    "            for f in filenames:\n",
    "                \n",
    "                # laoding audio file\n",
    "                file_path = os.path.join(dirpath, f)\n",
    "                signal, sr = librosa.load(file_path, sr = SAMPLE_RATE)\n",
    "                \n",
    "                # divide signals into segments,process segments, extract mfcc and at last store mfcc vectors.\n",
    "                for s in range(num_segments):\n",
    "                    start_sample = num_samples_per_segment * s  # if s= 0 -> then start_sample = 0 \n",
    "                    finish_sample = start_sample + num_samples_per_segment \n",
    "                    \n",
    "                    \n",
    "                    zc = librosa.feature.chroma_stft(y=signal[start_sample: finish_sample],\n",
    "                                                     sr=sr,n_chroma=12, n_fft=4096)\n",
    "\n",
    "                    mfcc = librosa.feature.mfcc(signal[start_sample: finish_sample],\n",
    "                                                sr =sr,\n",
    "                                                n_fft = 2048,\n",
    "                                                n_mfcc = 13,\n",
    "                                                hop_length = 512\n",
    "                                                )\n",
    "\n",
    "                    chroma_cq = librosa.feature.chroma_cqt(y=signal[start_sample: finish_sample], sr=sr)\n",
    "\n",
    "                    pitches, magnitudes = librosa.piptrack(y=signal[start_sample: finish_sample], sr=sr)\n",
    "                    rolloff = librosa.feature.spectral_rolloff(y=signal[start_sample: finish_sample], sr=sr, roll_percent=0.1)\n",
    "\n",
    "                    features = np.concatenate((zc, mfcc, chroma_cq, pitches, rolloff), axis = 0)\n",
    "                    \n",
    "                    data[\"mfcc\"].append(features.tolist()) # we can not append mfcc directly blc its a numpy array so we have\n",
    "                    # to first convert it into list.\n",
    "\n",
    "                    data[\"labels\"].append(i-1)\n",
    "                    # i mean in each iteration we are in different genre folder. why r we doing i -1 that's blc in first \n",
    "                    # iteration we were not inside the dir containing all the genres folder. \n",
    "\n",
    "                    print(\"{}, segment : {}\".format(file_path, s))\n",
    "\n",
    "    # final step: saving everything as a json file            \n",
    "    with open(json_path, \"w\") as fp:\n",
    "        json.dump(data, fp, indent = 4)  # indent mean spaces while writing. fp mean file_path  \n",
    "        \n",
    "        \n",
    "if __name__ == \"__main__\":\n",
    "    save_mfcc(DATASET_PATH, JSON_PATH, num_segments = 1)\n",
    "                "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(128, 183899)\n",
      "Model is running .....\n",
      "Testing Accuracy: 0.7\n",
      "\n",
      "Printing Classification report               precision    recall  f1-score   support\n",
      "\n",
      "           0       0.83      0.71      0.77        14\n",
      "           1       0.50      0.67      0.57         6\n",
      "\n",
      "    accuracy                           0.70        20\n",
      "   macro avg       0.67      0.69      0.67        20\n",
      "weighted avg       0.73      0.70      0.71        20\n",
      "\n",
      "\n",
      "ROC curve value is:  0.8095238095238096\n",
      "\n",
      "Confusion matrix, without normalization\n",
      "[[10  4]\n",
      " [ 2  4]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# training the cough model\n",
    "\n",
    "import breath_model as cm\n",
    "\n",
    "\n",
    "# Calling random forest class here:\n",
    "DATASET_PATH = \"mel_spec.json\"\n",
    "save_model_in = \"FYP_test1.pkl\"\n",
    "test_size = 20   # 20\n",
    "ran_state = 10\n",
    "rf_maxdepth = 15    #15\n",
    "rf_ran_state = 15    # 15\n",
    "rf_estimators = 90     #90\n",
    "# 66\n",
    "\n",
    "ff = cm.RF_Model(DATASET_PATH, save_model_in, test_size, ran_state, rf_maxdepth, rf_ran_state, rf_estimators)\n",
    "mod = ff.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Deep learning models testing\n",
    "# Output: No promising results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_33\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "max_pooling2d_93 (MaxPooling (None, 275, 32, 1)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_80 (Conv2D)           (None, 275, 32, 32)       832       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_94 (MaxPooling (None, 69, 8, 32)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_81 (Conv2D)           (None, 69, 8, 32)         25632     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_95 (MaxPooling (None, 18, 2, 32)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_82 (Conv2D)           (None, 18, 2, 32)         9248      \n",
      "_________________________________________________________________\n",
      "flatten_30 (Flatten)         (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "dense_89 (Dense)             (None, 128)               147584    \n",
      "_________________________________________________________________\n",
      "dropout_58 (Dropout)         (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_90 (Dense)             (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dropout_59 (Dropout)         (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_91 (Dense)             (None, 2)                 258       \n",
      "=================================================================\n",
      "Total params: 200,066\n",
      "Trainable params: 200,066\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/45\n",
      "3/3 [==============================] - 0s 141ms/step - loss: 0.6711 - accuracy: 0.6023 - val_loss: 0.6830 - val_accuracy: 0.6087\n",
      "Epoch 2/45\n",
      "3/3 [==============================] - 0s 113ms/step - loss: 0.7185 - accuracy: 0.5227 - val_loss: 0.6844 - val_accuracy: 0.6087\n",
      "Epoch 3/45\n",
      "3/3 [==============================] - 0s 119ms/step - loss: 0.7344 - accuracy: 0.4886 - val_loss: 0.6835 - val_accuracy: 0.6087\n",
      "Epoch 4/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.7238 - accuracy: 0.5341 - val_loss: 0.6774 - val_accuracy: 0.6087\n",
      "Epoch 5/45\n",
      "3/3 [==============================] - 0s 108ms/step - loss: 0.6834 - accuracy: 0.5795 - val_loss: 0.6752 - val_accuracy: 0.6087\n",
      "Epoch 6/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.7162 - accuracy: 0.4659 - val_loss: 0.6740 - val_accuracy: 0.6087\n",
      "Epoch 7/45\n",
      "3/3 [==============================] - 0s 107ms/step - loss: 0.6964 - accuracy: 0.5682 - val_loss: 0.6739 - val_accuracy: 0.6087\n",
      "Epoch 8/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.6744 - accuracy: 0.5682 - val_loss: 0.6728 - val_accuracy: 0.6087\n",
      "Epoch 9/45\n",
      "3/3 [==============================] - 0s 131ms/step - loss: 0.6824 - accuracy: 0.5455 - val_loss: 0.6711 - val_accuracy: 0.6087\n",
      "Epoch 10/45\n",
      "3/3 [==============================] - 0s 108ms/step - loss: 0.6921 - accuracy: 0.5455 - val_loss: 0.6718 - val_accuracy: 0.6087\n",
      "Epoch 11/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.7055 - accuracy: 0.4886 - val_loss: 0.6728 - val_accuracy: 0.6087\n",
      "Epoch 12/45\n",
      "3/3 [==============================] - 0s 104ms/step - loss: 0.6682 - accuracy: 0.6591 - val_loss: 0.6726 - val_accuracy: 0.6087\n",
      "Epoch 13/45\n",
      "3/3 [==============================] - 0s 133ms/step - loss: 0.6722 - accuracy: 0.5909 - val_loss: 0.6704 - val_accuracy: 0.6087\n",
      "Epoch 14/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.6663 - accuracy: 0.6136 - val_loss: 0.6681 - val_accuracy: 0.6087\n",
      "Epoch 15/45\n",
      "3/3 [==============================] - 0s 104ms/step - loss: 0.6654 - accuracy: 0.6023 - val_loss: 0.6686 - val_accuracy: 0.6087\n",
      "Epoch 16/45\n",
      "3/3 [==============================] - 0s 104ms/step - loss: 0.6582 - accuracy: 0.6364 - val_loss: 0.6689 - val_accuracy: 0.6087\n",
      "Epoch 17/45\n",
      "3/3 [==============================] - 0s 109ms/step - loss: 0.6805 - accuracy: 0.5682 - val_loss: 0.6696 - val_accuracy: 0.6522\n",
      "Epoch 18/45\n",
      "3/3 [==============================] - 0s 103ms/step - loss: 0.6437 - accuracy: 0.6818 - val_loss: 0.6692 - val_accuracy: 0.6522\n",
      "Epoch 19/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.6747 - accuracy: 0.5682 - val_loss: 0.6718 - val_accuracy: 0.6522\n",
      "Epoch 20/45\n",
      "3/3 [==============================] - 0s 133ms/step - loss: 0.6636 - accuracy: 0.5795 - val_loss: 0.6753 - val_accuracy: 0.6522\n",
      "Epoch 21/45\n",
      "3/3 [==============================] - 0s 109ms/step - loss: 0.6870 - accuracy: 0.5227 - val_loss: 0.6743 - val_accuracy: 0.6522\n",
      "Epoch 22/45\n",
      "3/3 [==============================] - 0s 103ms/step - loss: 0.6523 - accuracy: 0.6364 - val_loss: 0.6710 - val_accuracy: 0.6522\n",
      "Epoch 23/45\n",
      "3/3 [==============================] - 0s 137ms/step - loss: 0.6504 - accuracy: 0.6477 - val_loss: 0.6718 - val_accuracy: 0.6522\n",
      "Epoch 24/45\n",
      "3/3 [==============================] - 0s 119ms/step - loss: 0.6320 - accuracy: 0.6477 - val_loss: 0.6696 - val_accuracy: 0.6522\n",
      "Epoch 25/45\n",
      "3/3 [==============================] - 0s 118ms/step - loss: 0.6411 - accuracy: 0.6591 - val_loss: 0.6694 - val_accuracy: 0.6522\n",
      "Epoch 26/45\n",
      "3/3 [==============================] - 0s 101ms/step - loss: 0.6221 - accuracy: 0.6591 - val_loss: 0.6716 - val_accuracy: 0.6522\n",
      "Epoch 27/45\n",
      "3/3 [==============================] - 0s 134ms/step - loss: 0.6472 - accuracy: 0.5682 - val_loss: 0.6723 - val_accuracy: 0.6522\n",
      "Epoch 28/45\n",
      "3/3 [==============================] - 0s 150ms/step - loss: 0.5974 - accuracy: 0.6932 - val_loss: 0.6748 - val_accuracy: 0.5652\n",
      "Epoch 29/45\n",
      "3/3 [==============================] - 0s 130ms/step - loss: 0.6529 - accuracy: 0.5909 - val_loss: 0.6766 - val_accuracy: 0.5652\n",
      "Epoch 30/45\n",
      "3/3 [==============================] - 0s 124ms/step - loss: 0.6441 - accuracy: 0.6136 - val_loss: 0.6780 - val_accuracy: 0.5652\n",
      "Epoch 31/45\n",
      "3/3 [==============================] - 0s 112ms/step - loss: 0.6398 - accuracy: 0.6250 - val_loss: 0.6779 - val_accuracy: 0.5652\n",
      "Epoch 32/45\n",
      "3/3 [==============================] - 0s 107ms/step - loss: 0.5904 - accuracy: 0.6818 - val_loss: 0.6814 - val_accuracy: 0.5652\n",
      "Epoch 33/45\n",
      "3/3 [==============================] - 0s 106ms/step - loss: 0.6004 - accuracy: 0.7273 - val_loss: 0.6888 - val_accuracy: 0.5217\n",
      "Epoch 34/45\n",
      "3/3 [==============================] - 0s 117ms/step - loss: 0.6137 - accuracy: 0.6250 - val_loss: 0.6913 - val_accuracy: 0.5217\n",
      "Epoch 35/45\n",
      "3/3 [==============================] - 0s 105ms/step - loss: 0.5836 - accuracy: 0.6818 - val_loss: 0.6930 - val_accuracy: 0.5217\n",
      "Epoch 36/45\n",
      "3/3 [==============================] - 0s 112ms/step - loss: 0.5904 - accuracy: 0.6477 - val_loss: 0.6895 - val_accuracy: 0.5217\n",
      "Epoch 37/45\n",
      "3/3 [==============================] - 0s 102ms/step - loss: 0.5736 - accuracy: 0.6705 - val_loss: 0.6857 - val_accuracy: 0.5217\n",
      "Epoch 38/45\n",
      "3/3 [==============================] - 0s 104ms/step - loss: 0.5691 - accuracy: 0.7273 - val_loss: 0.6886 - val_accuracy: 0.5217\n",
      "Epoch 39/45\n",
      "3/3 [==============================] - 0s 117ms/step - loss: 0.5811 - accuracy: 0.7159 - val_loss: 0.6937 - val_accuracy: 0.5652\n",
      "Epoch 40/45\n",
      "3/3 [==============================] - 0s 104ms/step - loss: 0.5545 - accuracy: 0.7159 - val_loss: 0.7028 - val_accuracy: 0.5217\n",
      "Epoch 41/45\n",
      "3/3 [==============================] - 0s 103ms/step - loss: 0.5357 - accuracy: 0.7386 - val_loss: 0.7242 - val_accuracy: 0.3913\n",
      "Epoch 42/45\n",
      "3/3 [==============================] - 0s 113ms/step - loss: 0.5467 - accuracy: 0.7159 - val_loss: 0.7170 - val_accuracy: 0.5217\n",
      "Epoch 43/45\n",
      "3/3 [==============================] - 0s 107ms/step - loss: 0.5617 - accuracy: 0.6136 - val_loss: 0.7143 - val_accuracy: 0.5217\n",
      "Epoch 44/45\n",
      "3/3 [==============================] - 0s 111ms/step - loss: 0.5381 - accuracy: 0.6932 - val_loss: 0.7116 - val_accuracy: 0.5217\n",
      "Epoch 45/45\n",
      "3/3 [==============================] - 0s 105ms/step - loss: 0.4897 - accuracy: 0.8182 - val_loss: 0.7166 - val_accuracy: 0.5217\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 - 0s - loss: 0.8574 - accuracy: 0.4865\n",
      "\n",
      "Test accuracy: 0.4864864945411682\n",
      "Target: [0], Predicted label: [1]\n"
     ]
    }
   ],
   "source": [
    "# training the cough model\n",
    "\n",
    "import breath_model as cm\n",
    "\n",
    "\n",
    "# Calling random forest class here:\n",
    "# DATASET_PATH = \"mel_spec.json\"\n",
    "# save_model_in = \"breath_model2.pkl\"\n",
    "\n",
    "\n",
    "import json\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "import tensorflow.keras as keras\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "DATA_PATH = \"mel_spec.json\"\n",
    "\n",
    "\n",
    "def load_data(data_path):\n",
    "    \n",
    "    with open(data_path, \"r\") as fp:\n",
    "        data = json.load(fp)\n",
    "\n",
    "    X = np.array(data[\"mfcc\"])\n",
    "    y = np.array(data[\"labels\"])\n",
    "    y = y[..., np.newaxis]\n",
    "    return X, y\n",
    "\n",
    "def plot_history(history):\n",
    "    \n",
    "    fig, axs = plt.subplots(2)\n",
    "\n",
    "    # create accuracy sublpot\n",
    "    axs[0].plot(history.history[\"accuracy\"], label=\"train accuracy\")\n",
    "    axs[0].plot(history.history[\"val_accuracy\"], label=\"test accuracy\")\n",
    "    axs[0].set_ylabel(\"Accuracy\")\n",
    "    axs[0].legend(loc=\"lower right\")\n",
    "    axs[0].set_title(\"Accuracy eval\")\n",
    "\n",
    "    # create error sublpot\n",
    "    axs[1].plot(history.history[\"loss\"], label=\"train error\")\n",
    "    axs[1].plot(history.history[\"val_loss\"], label=\"test error\")\n",
    "    axs[1].set_ylabel(\"Error\")\n",
    "    axs[1].set_xlabel(\"Epoch\")\n",
    "    axs[1].legend(loc=\"upper right\")\n",
    "    axs[1].set_title(\"Error eval\")\n",
    "\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "def prepare_datasets(test_size, validation_size):\n",
    "\n",
    "    # load data\n",
    "    X, y = load_data(DATA_PATH)\n",
    "\n",
    "    # create train, validation and test split\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size)\n",
    "    X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train, test_size=validation_size)\n",
    "\n",
    "    # add an axis to input sets\n",
    "    X_train = X_train[..., np.newaxis]\n",
    "    X_validation = X_validation[..., np.newaxis]\n",
    "    X_test = X_test[..., np.newaxis]\n",
    "\n",
    "    return X_train, X_validation, X_test, y_train, y_validation, y_test\n",
    "\n",
    "\n",
    "def build_model(input_shape):\n",
    "    # model always take input shape in the form of 3d tuple. \n",
    "    \n",
    "    # build network topology\n",
    "    model = keras.Sequential()\n",
    "\n",
    "    # 1st conv layer\n",
    "    model.add(keras.layers.MaxPooling2D((2, 2), padding='same', input_shape=input_shape))\n",
    "    model.add(keras.layers.Conv2D(32, (5, 5), padding='same',  activation='relu'))\n",
    "    \n",
    "    model.add(keras.layers.MaxPooling2D((4, 4), padding='same'))\n",
    "    model.add(keras.layers.Conv2D(32, (5, 5), padding='same', activation='relu'))\n",
    "    \n",
    "    model.add(keras.layers.MaxPooling2D((4, 4), padding='same'))\n",
    "    model.add(keras.layers.Conv2D(32, (3, 3), padding='same', activation='relu'))\n",
    "    \n",
    "#     model.add(keras.layers.MaxPooling2D((2, 2), padding='same'))\n",
    "    \n",
    "    model.add(keras.layers.Flatten())\n",
    "    \n",
    "    model.add(keras.layers.Dense(128, activation='relu'))\n",
    "    model.add(keras.layers.Dropout(0.5))\n",
    "\n",
    "    model.add(keras.layers.Dense(128, activation='relu'))\n",
    "    model.add(keras.layers.Dropout(0.5))\n",
    "    \n",
    "    # output layer\n",
    "    model.add(keras.layers.Dense(2, activation='softmax'))\n",
    "   \n",
    "    return model\n",
    "\n",
    "\n",
    "def predict(model, X, y):\n",
    "    \n",
    "    # add a dimension to input data for sample - model.predict() expects a 4d array in this case\n",
    "    X = X[np.newaxis, ...] # array shape (1, 130, 13, 1)\n",
    "\n",
    "    # perform prediction\n",
    "    prediction = model.predict(X)\n",
    "\n",
    "    # get index with max value\n",
    "    predicted_index = np.argmax(prediction, axis=1)\n",
    "\n",
    "    print(\"Target: {}, Predicted label: {}\".format(y, predicted_index))\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "\n",
    "    # get train, validation, test splits\n",
    "    X_train, X_validation, X_test, y_train, y_validation, y_test = prepare_datasets(0.25, 0.2)\n",
    "\n",
    "    # create network\n",
    "    input_shape = (X_train.shape[1], X_train.shape[2], 1)\n",
    "    model = build_model(input_shape)\n",
    "\n",
    "    # compile model\n",
    "    optimizer = keras.optimizers.Adam(learning_rate=0.0002)  #0.00008\n",
    "    model.compile(optimizer=optimizer,\n",
    "                  loss=keras.losses.SparseCategoricalCrossentropy(),\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "\n",
    "    model.summary()\n",
    "\n",
    "    # train model\n",
    "    history = model.fit(X_train, y_train, validation_data=(X_validation, y_validation), batch_size=32, epochs=45)\n",
    "\n",
    "    # plot accuracy/error for training and validation\n",
    "    plot_history(history)\n",
    "\n",
    "    # evaluate model on test set\n",
    "    test_loss, test_acc = model.evaluate(X_test, y_test, verbose=2)\n",
    "    print('\\nTest accuracy:', test_acc)\n",
    "\n",
    "    # pick a sample to predict from the test set\n",
    "    X_to_predict = X_test[0]\n",
    "    y_to_predict = y_test[0]\n",
    "\n",
    "    # predict sample\n",
    "    predict(model, X_to_predict, y_to_predict)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(390, 1)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X, y = load_data(DATA_PATH)\n",
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_34\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_83 (Conv2D)           (None, 547, 62, 32)       320       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_96 (MaxPooling (None, 274, 31, 32)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_35 (Batc (None, 274, 31, 32)       128       \n",
      "_________________________________________________________________\n",
      "conv2d_84 (Conv2D)           (None, 272, 29, 32)       9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_97 (MaxPooling (None, 136, 15, 32)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_36 (Batc (None, 136, 15, 32)       128       \n",
      "_________________________________________________________________\n",
      "conv2d_85 (Conv2D)           (None, 135, 14, 32)       4128      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_98 (MaxPooling (None, 68, 7, 32)         0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_37 (Batc (None, 68, 7, 32)         128       \n",
      "_________________________________________________________________\n",
      "flatten_31 (Flatten)         (None, 15232)             0         \n",
      "_________________________________________________________________\n",
      "dense_92 (Dense)             (None, 64)                974912    \n",
      "_________________________________________________________________\n",
      "dropout_60 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_93 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "dropout_61 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_94 (Dense)             (None, 2)                 130       \n",
      "=================================================================\n",
      "Total params: 993,282\n",
      "Trainable params: 993,090\n",
      "Non-trainable params: 192\n",
      "_________________________________________________________________\n",
      "Epoch 1/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 1.6260 - accuracy: 0.4545\n",
      "Epoch 00001: val_loss improved from inf to 0.98495, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 2s 553ms/step - loss: 1.6260 - accuracy: 0.4545 - val_loss: 0.9849 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 2/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 1.2859 - accuracy: 0.5455\n",
      "Epoch 00002: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 457ms/step - loss: 1.2859 - accuracy: 0.5455 - val_loss: 1.1102 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 3/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.9745 - accuracy: 0.6136\n",
      "Epoch 00003: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 446ms/step - loss: 0.9745 - accuracy: 0.6136 - val_loss: 1.2050 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 4/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6746 - accuracy: 0.6818\n",
      "Epoch 00004: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 440ms/step - loss: 0.6746 - accuracy: 0.6818 - val_loss: 1.2527 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 5/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.9156 - accuracy: 0.6023\n",
      "Epoch 00005: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 2s 588ms/step - loss: 0.9156 - accuracy: 0.6023 - val_loss: 1.2670 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 6/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6422 - accuracy: 0.7159\n",
      "Epoch 00006: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 463ms/step - loss: 0.6422 - accuracy: 0.7159 - val_loss: 1.2542 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 7/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6923 - accuracy: 0.7045\n",
      "Epoch 00007: ReduceLROnPlateau reducing learning rate to 9.999999747378752e-06.\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 442ms/step - loss: 0.6923 - accuracy: 0.7045 - val_loss: 1.2141 - val_accuracy: 0.5217 - lr: 1.0000e-04\n",
      "Epoch 8/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6580 - accuracy: 0.7159\n",
      "Epoch 00008: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 456ms/step - loss: 0.6580 - accuracy: 0.7159 - val_loss: 1.1520 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 9/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6937 - accuracy: 0.6818\n",
      "Epoch 00009: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 497ms/step - loss: 0.6937 - accuracy: 0.6818 - val_loss: 1.0916 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 10/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6304 - accuracy: 0.6818\n",
      "Epoch 00010: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 480ms/step - loss: 0.6304 - accuracy: 0.6818 - val_loss: 1.0361 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 11/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6116 - accuracy: 0.7273\n",
      "Epoch 00011: val_loss did not improve from 0.98495\n",
      "3/3 [==============================] - 1s 446ms/step - loss: 0.6116 - accuracy: 0.7273 - val_loss: 0.9862 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 12/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5464 - accuracy: 0.7500\n",
      "Epoch 00012: val_loss improved from 0.98495 to 0.94125, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 462ms/step - loss: 0.5464 - accuracy: 0.7500 - val_loss: 0.9412 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 13/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6535 - accuracy: 0.6932\n",
      "Epoch 00013: val_loss improved from 0.94125 to 0.90168, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 480ms/step - loss: 0.6535 - accuracy: 0.6932 - val_loss: 0.9017 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 14/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5814 - accuracy: 0.7727\n",
      "Epoch 00014: val_loss improved from 0.90168 to 0.87036, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 2s 515ms/step - loss: 0.5814 - accuracy: 0.7727 - val_loss: 0.8704 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 15/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5112 - accuracy: 0.7273\n",
      "Epoch 00015: val_loss improved from 0.87036 to 0.84359, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 465ms/step - loss: 0.5112 - accuracy: 0.7273 - val_loss: 0.8436 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 16/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5106 - accuracy: 0.7727\n",
      "Epoch 00016: val_loss improved from 0.84359 to 0.82072, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 464ms/step - loss: 0.5106 - accuracy: 0.7727 - val_loss: 0.8207 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 17/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6126 - accuracy: 0.7159\n",
      "Epoch 00017: val_loss improved from 0.82072 to 0.80090, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 470ms/step - loss: 0.6126 - accuracy: 0.7159 - val_loss: 0.8009 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 18/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6961 - accuracy: 0.7386\n",
      "Epoch 00018: val_loss improved from 0.80090 to 0.78338, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 465ms/step - loss: 0.6961 - accuracy: 0.7386 - val_loss: 0.7834 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 19/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6369 - accuracy: 0.7273\n",
      "Epoch 00019: val_loss improved from 0.78338 to 0.76866, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 2s 514ms/step - loss: 0.6369 - accuracy: 0.7273 - val_loss: 0.7687 - val_accuracy: 0.5217 - lr: 1.0000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 20/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5160 - accuracy: 0.8068\n",
      "Epoch 00020: val_loss improved from 0.76866 to 0.75809, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 465ms/step - loss: 0.5160 - accuracy: 0.8068 - val_loss: 0.7581 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 21/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.4226 - accuracy: 0.8409\n",
      "Epoch 00021: val_loss improved from 0.75809 to 0.75008, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 458ms/step - loss: 0.4226 - accuracy: 0.8409 - val_loss: 0.7501 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 22/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6055 - accuracy: 0.6818\n",
      "Epoch 00022: val_loss improved from 0.75008 to 0.74373, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 463ms/step - loss: 0.6055 - accuracy: 0.6818 - val_loss: 0.7437 - val_accuracy: 0.5217 - lr: 1.0000e-05\n",
      "Epoch 23/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5748 - accuracy: 0.6932\n",
      "Epoch 00023: val_loss improved from 0.74373 to 0.73744, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 484ms/step - loss: 0.5748 - accuracy: 0.6932 - val_loss: 0.7374 - val_accuracy: 0.4783 - lr: 1.0000e-05\n",
      "Epoch 24/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.6084 - accuracy: 0.6932\n",
      "Epoch 00024: val_loss improved from 0.73744 to 0.73114, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 2s 500ms/step - loss: 0.6084 - accuracy: 0.6932 - val_loss: 0.7311 - val_accuracy: 0.4348 - lr: 1.0000e-05\n",
      "Epoch 25/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5193 - accuracy: 0.7273\n",
      "Epoch 00025: val_loss improved from 0.73114 to 0.72592, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 477ms/step - loss: 0.5193 - accuracy: 0.7273 - val_loss: 0.7259 - val_accuracy: 0.4348 - lr: 1.0000e-05\n",
      "Epoch 26/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5118 - accuracy: 0.7159\n",
      "Epoch 00026: val_loss improved from 0.72592 to 0.72147, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 454ms/step - loss: 0.5118 - accuracy: 0.7159 - val_loss: 0.7215 - val_accuracy: 0.4348 - lr: 1.0000e-05\n",
      "Epoch 27/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5533 - accuracy: 0.7159\n",
      "Epoch 00027: val_loss improved from 0.72147 to 0.71694, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 451ms/step - loss: 0.5533 - accuracy: 0.7159 - val_loss: 0.7169 - val_accuracy: 0.4348 - lr: 1.0000e-05\n",
      "Epoch 28/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.4461 - accuracy: 0.7614\n",
      "Epoch 00028: val_loss improved from 0.71694 to 0.71325, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 456ms/step - loss: 0.4461 - accuracy: 0.7614 - val_loss: 0.7133 - val_accuracy: 0.3913 - lr: 1.0000e-05\n",
      "Epoch 29/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.5523 - accuracy: 0.7386\n",
      "Epoch 00029: val_loss improved from 0.71325 to 0.70973, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 455ms/step - loss: 0.5523 - accuracy: 0.7386 - val_loss: 0.7097 - val_accuracy: 0.4348 - lr: 1.0000e-05\n",
      "Epoch 30/30\n",
      "3/3 [==============================] - ETA: 0s - loss: 0.4420 - accuracy: 0.7841\n",
      "Epoch 00030: val_loss improved from 0.70973 to 0.70717, saving model to sickweight4.h5\n",
      "3/3 [==============================] - 1s 471ms/step - loss: 0.4420 - accuracy: 0.7841 - val_loss: 0.7072 - val_accuracy: 0.4783 - lr: 1.0000e-05\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 - 0s - loss: 0.7385 - accuracy: 0.4595\n",
      "\n",
      "Test accuracy: 0.45945945382118225\n",
      "Target: 0, Predicted label: [1]\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "import tensorflow.keras as keras\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "DATA_PATH = \"mel_spec.json\"\n",
    "\n",
    "\n",
    "def load_data(data_path):\n",
    "    \"\"\"Loads training dataset from json file.\n",
    "\n",
    "        :param data_path (str): Path to json file containing data\n",
    "        :return X (ndarray): Inputs\n",
    "        :return y (ndarray): Targets\n",
    "    \"\"\"\n",
    "\n",
    "    with open(data_path, \"r\") as fp:\n",
    "        data = json.load(fp)\n",
    "\n",
    "    X = np.array(data[\"mfcc\"])\n",
    "    y = np.array(data[\"labels\"])\n",
    "    return X, y\n",
    "\n",
    "\n",
    "def plot_history(history):\n",
    "    \"\"\"Plots accuracy/loss for training/validation set as a function of the epochs\n",
    "\n",
    "        :param history: Training history of model\n",
    "        :return:\n",
    "    \"\"\"\n",
    "\n",
    "    fig, axs = plt.subplots(2)\n",
    "\n",
    "    # create accuracy sublpot\n",
    "    axs[0].plot(history.history[\"accuracy\"], label=\"train accuracy\")\n",
    "    axs[0].plot(history.history[\"val_accuracy\"], label=\"test accuracy\")\n",
    "    axs[0].set_ylabel(\"Accuracy\")\n",
    "    axs[0].legend(loc=\"lower right\")\n",
    "    axs[0].set_title(\"Accuracy eval\")\n",
    "\n",
    "    # create error sublpot\n",
    "    axs[1].plot(history.history[\"loss\"], label=\"train error\")\n",
    "    axs[1].plot(history.history[\"val_loss\"], label=\"test error\")\n",
    "    axs[1].set_ylabel(\"Error\")\n",
    "    axs[1].set_xlabel(\"Epoch\")\n",
    "    axs[1].legend(loc=\"upper right\")\n",
    "    axs[1].set_title(\"Error eval\")\n",
    "\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "def prepare_datasets(test_size, validation_size):\n",
    "    \"\"\"Loads data and splits it into train, validation and test sets.\n",
    "\n",
    "    :param test_size (float): Value in [0, 1] indicating percentage of data set to allocate to test split\n",
    "    :param validation_size (float): Value in [0, 1] indicating percentage of train set to allocate to validation split\n",
    "\n",
    "    :return X_train (ndarray): Input training set\n",
    "    :return X_validation (ndarray): Input validation set\n",
    "    :return X_test (ndarray): Input test set\n",
    "    :return y_train (ndarray): Target training set\n",
    "    :return y_validation (ndarray): Target validation set\n",
    "    :return y_test (ndarray): Target test set\n",
    "    \"\"\"\n",
    "\n",
    "    # load data\n",
    "    X, y = load_data(DATA_PATH)\n",
    "\n",
    "    # create train, validation and test split\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size)\n",
    "    X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train, test_size=validation_size)\n",
    "\n",
    "    # add an axis to input sets\n",
    "    X_train = X_train[..., np.newaxis]\n",
    "    X_validation = X_validation[..., np.newaxis]\n",
    "    X_test = X_test[..., np.newaxis]\n",
    "\n",
    "    return X_train, X_validation, X_test, y_train, y_validation, y_test\n",
    "\n",
    "\n",
    "def build_model(input_shape):\n",
    "    \"\"\"Generates CNN model\n",
    "\n",
    "    :param input_shape (tuple): Shape of input set\n",
    "    :return model: CNN model\n",
    "    \"\"\"\n",
    "\n",
    "    # build network topology\n",
    "    model = keras.Sequential()\n",
    "\n",
    "    # 1st conv layer\n",
    "    model.add(keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=input_shape))\n",
    "    model.add(keras.layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same'))\n",
    "    model.add(keras.layers.BatchNormalization())\n",
    "\n",
    "    # 2nd conv layer\n",
    "    model.add(keras.layers.Conv2D(32, (3, 3), activation='relu'))\n",
    "    model.add(keras.layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same'))\n",
    "    model.add(keras.layers.BatchNormalization())\n",
    "\n",
    "    # 3rd conv layer\n",
    "    model.add(keras.layers.Conv2D(32, (2, 2), activation='relu'))\n",
    "    model.add(keras.layers.MaxPooling2D((2, 2), strides=(2, 2), padding='same'))\n",
    "    model.add(keras.layers.BatchNormalization())\n",
    "\n",
    "    # flatten output and feed it into dense layer\n",
    "    model.add(keras.layers.Flatten())\n",
    "    model.add(keras.layers.Dense(64, activation='relu'))\n",
    "    model.add(keras.layers.Dropout(0.5))\n",
    "    \n",
    "    model.add(keras.layers.Dense(64, activation='relu'))\n",
    "    model.add(keras.layers.Dropout(0.5))\n",
    "    \n",
    "#     model.add(keras.layers.Dense(64, activation='relu'))\n",
    "#     model.add(keras.layers.Dropout(0.5))\n",
    "\n",
    "    # output layer\n",
    "    model.add(keras.layers.Dense(2, activation='softmax'))\n",
    "\n",
    "    return model\n",
    "\n",
    "\n",
    "def predict(model, X, y):\n",
    "    \"\"\"Predict a single sample using the trained model\n",
    "\n",
    "    :param model: Trained classifier\n",
    "    :param X: Input data\n",
    "    :param y (int): Target\n",
    "    \"\"\"\n",
    "\n",
    "    # add a dimension to input data for sample - model.predict() expects a 4d array in this case\n",
    "    X = X[np.newaxis, ...] # array shape (1, 130, 13, 1)\n",
    "\n",
    "    # perform prediction\n",
    "    prediction = model.predict(X)\n",
    "\n",
    "    # get index with max value\n",
    "    predicted_index = np.argmax(prediction, axis=1)\n",
    "\n",
    "    print(\"Target: {}, Predicted label: {}\".format(y, predicted_index))\n",
    "\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "\n",
    "    # get train, validation, test splits\n",
    "    X_train, X_validation, X_test, y_train, y_validation, y_test = prepare_datasets(0.25, 0.2)\n",
    "\n",
    "    # create network\n",
    "    input_shape = (X_train.shape[1], X_train.shape[2], 1)\n",
    "    model = build_model(input_shape)\n",
    "\n",
    "    reduce_lr = keras.callbacks.ReduceLROnPlateau(\n",
    "                    monitor='val_loss', factor=0.1, patience=6, verbose=1, mode='auto',\n",
    "                    min_delta=0.0001, cooldown=0 #, min_lr=0.001\n",
    "                )\n",
    "    \n",
    "    earlystop = keras.callbacks.EarlyStopping(monitor='val_loss', patience=15, restore_best_weights = True, verbose=1)\n",
    "    checkpoint = keras.callbacks.ModelCheckpoint(\n",
    "            \"sickweight4.h5\", monitor='val_loss', verbose=1, save_best_only=True,\n",
    "            save_weights_only=True, mode='min', save_freq='epoch' \n",
    "        )\n",
    "\n",
    "    callbacks = [reduce_lr, earlystop, checkpoint]\n",
    "    \n",
    "    # compile model\n",
    "    optimizer = keras.optimizers.Adam(learning_rate=0.0001)\n",
    "    model.compile(optimizer=optimizer,\n",
    "                  loss='sparse_categorical_crossentropy',\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "\n",
    "    model.summary()\n",
    "    \n",
    "    # train model\n",
    "    history = model.fit(X_train, y_train,\n",
    "                        validation_data=(X_validation, y_validation),\n",
    "                        batch_size=32, epochs=30, callbacks = callbacks)\n",
    "\n",
    "    # plot accuracy/error for training and validation\n",
    "    plot_history(history)\n",
    "\n",
    "    # evaluate model on test set\n",
    "    test_loss, test_acc = model.evaluate(X_test, y_test, verbose=2)\n",
    "    print('\\nTest accuracy:', test_acc)\n",
    "\n",
    "    # pick a sample to predict from the test set\n",
    "    X_to_predict = X_test[2]\n",
    "    y_to_predict = y_test[2]\n",
    "\n",
    "    # predict sample\n",
    "    predict(model, X_to_predict, y_to_predict)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}