trainSirenDetector - 4Layer - 16-32-64-128.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Last updated: GHS\n",
    "# email: gabrielsarch@gmail.com\n",
    "#\n",
    "# Last edited: 4/25/2020"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This file trains the Convolutional Neural Network\n",
    "## Uses siren, car, and background noises (split into 3-second chunks)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "\n",
    "RATE = 22050 #Hz - sample rate of training files\n",
    "\n",
    "# paths to numpy array files\n",
    "# These files contain numpy arrays in format (nSoundBytes x nSamples) \n",
    "# nSamples is usually = RATE*seconds\n",
    "# see generateTrainingData\n",
    "sirenFile = r'C:\\Users\\Gabe\\Documents\\SeniorDesign\\Data\\Senior Design data training\\sirenNoise.npy'\n",
    "backFile = r'C:\\Users\\Gabe\\Documents\\SeniorDesign\\Data\\Senior Design data training\\backNoise.npy'\n",
    "carFile = r'C:\\Users\\Gabe\\Documents\\SeniorDesign\\Data\\Senior Design data training\\carNoise.npy'\n",
    "\n",
    "# Where to save CNN Model WEIGHTS (not full architecture)\n",
    "saveModel = 'C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5'\n",
    "\n",
    "# Where to save model architecture + weights\n",
    "modelSave = 'C:/Users/Gabe/Documents/Driver-Alert/Models/siren_detector_V2.h5'\n",
    "\n",
    "# Parameters for generateData\n",
    "num_samples = 999 # number of samples used for training from each array\n",
    "num_samp_noise = math.floor(num_samples/3)*4 # need more noise samples than siren samples\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Array shapes before:\n",
      "(1839, 66150)\n",
      "(6566, 66150)\n",
      "(613, 66150)\n",
      " \n",
      "Array shape after (valid, train):\n",
      "(184, 66150)\n",
      "(1655, 66150)\n",
      "(657, 66150)\n",
      "(5909, 66150)\n",
      "(61, 66150)\n",
      "(552, 66150)\n"
     ]
    }
   ],
   "source": [
    "# Load in numpy arrays containing data & splits data up into training and validation sets (use validation for ROC curves)\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "sirenNoise = np.load(sirenFile)\n",
    "backNoise = np.load(backFile)\n",
    "carNoise = np.load(carFile)\n",
    "\n",
    "print('Array shapes before:')\n",
    "print(sirenNoise.shape)\n",
    "print(backNoise.shape)\n",
    "print(carNoise.shape)\n",
    "print(' ')\n",
    "\n",
    "percentHeldOut = 0.1\n",
    "\n",
    "perm = np.random.permutation(len(sirenNoise))[0:round(percentHeldOut*sirenNoise.shape[0])]\n",
    "sirenNoiseVal = sirenNoise[perm]\n",
    "sirenNoise = np.delete(sirenNoise, perm, axis=0)\n",
    "\n",
    "perm = np.random.permutation(len(backNoise))[0:round(percentHeldOut*backNoise.shape[0])]\n",
    "backNoiseVal = backNoise[perm]\n",
    "backNoise = np.delete(backNoise, perm, axis=0)\n",
    "\n",
    "perm = np.random.permutation(len(carNoise))[0:round(percentHeldOut*carNoise.shape[0])]\n",
    "carNoiseVal = carNoise[perm]\n",
    "carNoise = np.delete(carNoise, perm, axis=0)\n",
    "\n",
    "print('Array shape after (valid, train):')\n",
    "print(sirenNoiseVal.shape)\n",
    "print(sirenNoise.shape)\n",
    "print(backNoiseVal.shape)\n",
    "print(backNoise.shape)\n",
    "print(carNoiseVal.shape)\n",
    "print(carNoise.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "bool"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x = True\n",
    "isinstance(x,int) #or x==False\n",
    "if not isinstance(x, bool):\n",
    "    print('Hello')\n",
    "type(x)\n",
    "#raise TypeError('HI')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy import signal\n",
    "import pandas as pd\n",
    "import os\n",
    "import librosa\n",
    "\n",
    "sos = signal.butter(5, [50, 5000], 'bandpass', fs=RATE, output='sos')\n",
    "\n",
    "def get_mfccs(audio):\n",
    "   \n",
    "    try:\n",
    "        audio = 2*((audio-min(audio))/(max(audio)-min(audio)))-1\n",
    "\n",
    "        #Filter\n",
    "        audio = signal.sosfilt(sos, audio)\n",
    "        \n",
    "        mfccs = librosa.feature.mfcc(y=audio, sr=RATE, n_mfcc=40)\n",
    "\n",
    "    except Exception as e:\n",
    "            print(\"Error encountered for a file\")\n",
    "            return None\n",
    "\n",
    "    return mfccs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Fit #1\n",
    "\n",
    "from keras.callbacks import ModelCheckpoint \n",
    "from datetime import datetime \n",
    "import math\n",
    "import random\n",
    "from sklearn.model_selection import train_test_split \n",
    "\n",
    "\n",
    "\n",
    "def generateData(carNoise, backNoise, sirenNoise, uniformSNR=False):\n",
    "# This function \n",
    "# 1) combines siren and noise data to create 2 groups: siren present group, siren not present group\n",
    "# siren group: siren + car, siren + environment, siren + car + environmental (4 signal to noise ratios for each group)\n",
    "# no siren group: car, environment, car + environmental \n",
    "# Then\n",
    "# 2) Normalizes the audio and extracts\n",
    "\n",
    "# Inputs:\n",
    "# 1-3) car noises, background noises, siren noises (np array, nSoundBytes x nSamples)\n",
    "# 4) uniformSNR - if input = false (default) training data will consist of multiple SNRs (-5, 0, 5, 10 dB SNR)\n",
    "#               - if input = True will deafult to SNR of 0 dB\n",
    "#               - if input = Db (integer) will do SNR of that dB\n",
    "    Db5gain = 1.778279 # gain for 5dB increase\n",
    "\n",
    "    if isinstance(uniformSNR,int):\n",
    "       # if uniformSNR==True:\n",
    "        #    uniformSNR = 0.0001 # Set SNR to 0 if True\n",
    "        if not isinstance(uniformSNR, bool):\n",
    "            if uniformSNR > 0:\n",
    "                DbMult = uniformSNR/5 # Excludes if uniformSNR = False\n",
    "            elif uniformSNR < 0:\n",
    "                DbMult = 1/(-uniformSNR/5)\n",
    "                Db5gain = 1/Db5gain\n",
    "            elif uniformSNR == 0:\n",
    "                uniformSNR = True\n",
    "            else:\n",
    "                raise TypeError('Wrong uniformSNR input')\n",
    "    else:\n",
    "        raise TypeError('Wrong uniformSNR input')\n",
    "\n",
    "    start = datetime.now()\n",
    "\n",
    "    # get desired number of samples from each array\n",
    "    if len(carNoise) < num_samp_noise:\n",
    "        extra = num_samp_noise - len(carNoise)\n",
    "        perm = np.random.permutation(len(carNoise))\n",
    "        carX = np.concatenate((carNoise[perm], carNoise[perm[0:extra]]))\n",
    "        if len(carX) < num_samp_noise:\n",
    "            extra = num_samp_noise - len(carX)\n",
    "            perm = np.random.permutation(len(carX))\n",
    "            carX = np.concatenate((carX[perm], carX[perm[0:extra]]))\n",
    "    else:\n",
    "        perm = np.random.permutation(len(carNoise))[0:num_samp_noise]\n",
    "        carX = carNoise[perm]\n",
    "\n",
    "    if len(backNoise) < num_samp_noise:\n",
    "        extra = num_samp_noise - len(backNoise)\n",
    "        perm = np.random.permutation(len(backNoise))\n",
    "        envX = np.concatenate((backNoise[perm], backNoise[perm[0:extra]]))\n",
    "    else:\n",
    "        perm = np.random.permutation(len(backNoise))[0:num_samp_noise]\n",
    "        envX = backNoise[perm]\n",
    "\n",
    "    if len(sirenNoise) < num_samples:\n",
    "        extra = num_samples - len(sirenNoise)\n",
    "        perm = np.random.permutation(len(sirenNoise))\n",
    "        sirenX = np.concatenate((sirenNoise[perm], sirenNoise[perm[0:extra]]))\n",
    "    else:\n",
    "        perm = np.random.permutation(len(sirenNoise))[0:num_samples]\n",
    "        sirenX = sirenNoise[perm]\n",
    "\n",
    "\n",
    "    # Add gain to background noise\n",
    "    splitback1 = math.floor(num_samples/3)\n",
    "    \n",
    "    if uniformSNR == False:\n",
    "        \n",
    "        print(\"Mixing various SNRs\")\n",
    "\n",
    "        carX[0:splitback1] = carX[0:splitback1]*(1/(2*Db5gain))\n",
    "        carX[splitback1:splitback1*2] = carX[splitback1:splitback1*2]*(1/Db5gain)\n",
    "        carX[splitback1*2:splitback1*3] = carX[splitback1*2:splitback1*3]\n",
    "        carX[splitback1*3:] = carX[splitback1*3:]*Db5gain\n",
    "\n",
    "        envX[0:splitback1] = envX[0:splitback1]*(1/(2*Db5gain))\n",
    "        envX[splitback1:splitback1*2] = envX[splitback1:splitback1*2]*(1/Db5gain)\n",
    "        envX[splitback1*2:splitback1*3] = envX[splitback1*2:splitback1*3]\n",
    "        envX[splitback1*3:] = envX[splitback1*3:]*Db5gain\n",
    "\n",
    "        #sirenX[0:splitback1] = sirenX[0:splitback1]\n",
    "        #sirenX[splitback1:splitback1*2] = envX[splitback1:splitback1*2]*Db5gain\n",
    "        #sirenX[splitback1*2:splitback1*3] = envX[splitback1*2:splitback1*3]*Db5gain*2\n",
    "        #sirenX[splitback1*3:] = sirenX[splitback1*3:]*Db5gain*3\n",
    "        \n",
    "    elif uniformSNR == True: # SNR of 0 dB\n",
    "        print(\"SNR of 0 dB\")\n",
    "        sirenX = sirenX*1\n",
    "    \n",
    "    else:    \n",
    "        print(\"Uniform SNR with SNR = \", uniformSNR)\n",
    "        sirenX = sirenX*Db5gain*DbMult\n",
    "\n",
    "    sirencar = sirenX[0:splitback1]+carX[0:splitback1]\n",
    "    sirenenv = sirenX[splitback1:splitback1*2]+envX[0:splitback1]\n",
    "    sirencarenv = sirenX[splitback1*2:splitback1*3]+carX[splitback1:splitback1*2]*0.5+envX[splitback1:splitback1*2]*0.5\n",
    "\n",
    "    sirenSet = np.concatenate((sirencar, sirenenv, sirencarenv))\n",
    "    noiseSet = np.concatenate((carX[splitback1*2:splitback1*3], envX[splitback1*2:splitback1*3], carX[splitback1*3:splitback1*4]+envX[splitback1*3:splitback1*4]))            \n",
    "\n",
    "    print('Finished scrambling data set')\n",
    "    print('Extracting MFCCs...')\n",
    "\n",
    "    mfccSiren = []\n",
    "    for i in range(0,len(sirenSet)):\n",
    "        cur_audio = sirenSet[i]\n",
    "        #cur_audio = 2*((cur_audio-min(cur_audio))/(max(cur_audio)-min(cur_audio)))-1\n",
    "        mfcc = get_mfccs(cur_audio)\n",
    "        mfccSiren.append(mfcc)\n",
    "        sirenSet[i] = cur_audio\n",
    "\n",
    "    mfccNoise = []\n",
    "    for i in range(0,len(noiseSet)):\n",
    "        cur_audio = noiseSet[i]\n",
    "        #cur_audio = 2*((cur_audio-min(cur_audio))/(max(cur_audio)-min(cur_audio)))-1\n",
    "        mfcc = get_mfccs(cur_audio)\n",
    "        mfccNoise.append(mfcc)\n",
    "        noiseSet[i] = cur_audio\n",
    "    \n",
    "    # Create labels for the data\n",
    "    sirenLabels = ['siren']*len(mfccSiren) \n",
    "    noiseLabels = ['noise']*len(mfccNoise)\n",
    "\n",
    "    # Comibine MFCCs and labels \n",
    "    featureSiren = list(zip(mfccSiren, sirenLabels))\n",
    "    featureNoise = list(zip(mfccNoise, noiseLabels))\n",
    "    features = featureSiren + featureNoise\n",
    "    random.shuffle(features)\n",
    "    \n",
    "    # Convert to Panda dataframe\n",
    "    featuresdf = pd.DataFrame(features, columns=['feature','class_label'])\n",
    "\n",
    "    from sklearn.preprocessing import LabelEncoder\n",
    "    from keras.utils import to_categorical\n",
    "\n",
    "    # Convert features and corresponding classification labels into numpy arrays\n",
    "    X = np.array(featuresdf.feature.tolist())\n",
    "    y = np.array(featuresdf.class_label.tolist())\n",
    "\n",
    "    # Encode the classification labels\n",
    "    le = LabelEncoder()\n",
    "    yy = to_categorical(le.fit_transform(y)) \n",
    "\n",
    "    duration = datetime.now() - start\n",
    "    print(\"Generating training and test data took: \", duration)\n",
    "    \n",
    "    return X, yy\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:02:01.592300\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_5 (Conv2D)            (None, 39, 129, 16)       80        \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 19, 64, 16)        0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 19, 64, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 18, 63, 32)        2080      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 9, 31, 32)         0         \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 9, 31, 32)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 8, 30, 64)         8256      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 4, 15, 64)         0         \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 4, 15, 64)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 3, 14, 128)        32896     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2 (None, 1, 7, 128)         0         \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 1, 7, 128)         0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_2 ( (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 2)                 258       \n",
      "=================================================================\n",
      "Total params: 43,570\n",
      "Trainable params: 43,570\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "400/400 [==============================] - 1s 1ms/step\n",
      "Pre-training accuracy: 52.0000%\n"
     ]
    }
   ],
   "source": [
    "# Create model architecture\n",
    "\n",
    "# Format of input data\n",
    "num_rows = 40\n",
    "num_columns = 130 #x_train.shape[2]\n",
    "num_channels = 1\n",
    "\n",
    "# generate data for initial model evaluation\n",
    "X, yy = generateData(carNoise, backNoise, sirenNoise) \n",
    "\n",
    "# split the dataset \n",
    "x_train, x_test, y_train, y_test = train_test_split(X, yy, test_size=0.2)#, random_state = 42)\n",
    "x_train = x_train.reshape(x_train.shape[0], num_rows, num_columns, num_channels)\n",
    "x_test = x_test.reshape(x_test.shape[0], num_rows, num_columns, num_channels)\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Convolution2D, Conv2D, MaxPooling2D, GlobalAveragePooling2D\n",
    "from keras.optimizers import Adam\n",
    "from keras.utils import np_utils\n",
    "from sklearn import metrics \n",
    "\n",
    "num_labels = 2 #yy.shape[1]\n",
    "filter_size = 2\n",
    "\n",
    "# Construct model \n",
    "model = Sequential()\n",
    "model.add(Conv2D(filters=16, kernel_size=filter_size, input_shape=(num_rows, num_columns, num_channels), activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=32, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=64, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=128, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(GlobalAveragePooling2D())\n",
    "\n",
    "model.add(Dense(num_labels, activation='softmax')) \n",
    "\n",
    "# Compile the model\n",
    "model.compile(loss='binary_crossentropy', metrics=['accuracy'], optimizer='adam') \n",
    "\n",
    "\n",
    "# Summary of model\n",
    "\n",
    "# Display model architecture summary \n",
    "model.summary()\n",
    "\n",
    "# Calculate pre-training accuracy \n",
    "score = model.evaluate(x_test, y_test, verbose=1)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "print(\"Pre-training accuracy: %.4f%%\" % accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generating data for fit # 1\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:01:53.178506\n",
      "Train on 1598 samples, validate on 400 samples\n",
      "Epoch 1/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 1.1787 - accuracy: 0.7303 - val_loss: 0.3493 - val_accuracy: 0.8300\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.34925, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 2/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.3701 - accuracy: 0.8511 - val_loss: 0.2911 - val_accuracy: 0.8950\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.34925 to 0.29114, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 3/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.3207 - accuracy: 0.8698 - val_loss: 0.3366 - val_accuracy: 0.8675\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.29114\n",
      "Epoch 4/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2872 - accuracy: 0.8849 - val_loss: 0.3561 - val_accuracy: 0.8525\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.29114\n",
      "Epoch 5/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2477 - accuracy: 0.8911 - val_loss: 0.3802 - val_accuracy: 0.8275\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.29114\n",
      "Epoch 6/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2851 - accuracy: 0.8880 - val_loss: 0.3965 - val_accuracy: 0.8050\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.29114\n",
      "Epoch 7/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2569 - accuracy: 0.8992 - val_loss: 0.3726 - val_accuracy: 0.8175\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.29114\n",
      "Epoch 8/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2368 - accuracy: 0.9080 - val_loss: 0.3319 - val_accuracy: 0.8575\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.29114\n",
      "Epoch 9/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2293 - accuracy: 0.9036 - val_loss: 0.3456 - val_accuracy: 0.8425\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.29114\n",
      "Epoch 10/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2313 - accuracy: 0.9043 - val_loss: 0.3043 - val_accuracy: 0.8800\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.29114\n",
      "Epoch 11/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2131 - accuracy: 0.9143 - val_loss: 0.3332 - val_accuracy: 0.8550\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.29114\n",
      "Epoch 12/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2063 - accuracy: 0.9199 - val_loss: 0.3565 - val_accuracy: 0.8250\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.29114\n",
      "Epoch 13/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2075 - accuracy: 0.9118 - val_loss: 0.3359 - val_accuracy: 0.8425\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.29114\n",
      "Epoch 14/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2121 - accuracy: 0.9149 - val_loss: 0.2981 - val_accuracy: 0.8800\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.29114\n",
      "Epoch 15/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1981 - accuracy: 0.9168 - val_loss: 0.3398 - val_accuracy: 0.8425\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.29114\n",
      "Epoch 16/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2178 - accuracy: 0.9161 - val_loss: 0.3424 - val_accuracy: 0.8325\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.29114\n",
      "Epoch 17/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2083 - accuracy: 0.9199 - val_loss: 0.3263 - val_accuracy: 0.8475\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.29114\n",
      "Epoch 18/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1822 - accuracy: 0.9299 - val_loss: 0.3134 - val_accuracy: 0.8575\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.29114\n",
      "Epoch 19/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2173 - accuracy: 0.9143 - val_loss: 0.3126 - val_accuracy: 0.8600\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.29114\n",
      "Epoch 20/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1843 - accuracy: 0.9305 - val_loss: 0.3353 - val_accuracy: 0.8350\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.29114\n",
      "Epoch 21/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1787 - accuracy: 0.9324 - val_loss: 0.2872 - val_accuracy: 0.8725\n",
      "\n",
      "Epoch 00021: val_loss improved from 0.29114 to 0.28720, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 22/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1757 - accuracy: 0.9343 - val_loss: 0.3894 - val_accuracy: 0.7975\n",
      "\n",
      "Epoch 00022: val_loss did not improve from 0.28720\n",
      "Epoch 23/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1815 - accuracy: 0.9330 - val_loss: 0.3695 - val_accuracy: 0.8150\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.28720\n",
      "Epoch 24/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1749 - accuracy: 0.9280 - val_loss: 0.2648 - val_accuracy: 0.9025\n",
      "\n",
      "Epoch 00024: val_loss improved from 0.28720 to 0.26478, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 25/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1692 - accuracy: 0.9393 - val_loss: 0.2828 - val_accuracy: 0.8725\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.26478\n",
      "Training completed in time:  0:01:57.539897\n",
      "Results from fit # 1\n",
      "Training Accuracy:  0.9161452054977417\n",
      "Testing Accuracy:  0.8725000023841858\n",
      "Generating data for fit # 2\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:01:51.770469\n",
      "Train on 1598 samples, validate on 400 samples\n",
      "Epoch 1/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2282 - accuracy: 0.9168 - val_loss: 0.3475 - val_accuracy: 0.8425\n",
      "\n",
      "Epoch 00001: val_loss did not improve from 0.26478\n",
      "Epoch 2/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2196 - accuracy: 0.9105 - val_loss: 0.3351 - val_accuracy: 0.8600\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 0.26478\n",
      "Epoch 3/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1999 - accuracy: 0.9224 - val_loss: 0.2901 - val_accuracy: 0.8700\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.26478\n",
      "Epoch 4/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.2003 - accuracy: 0.9193 - val_loss: 0.3181 - val_accuracy: 0.8575\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.26478\n",
      "Epoch 5/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1863 - accuracy: 0.9299 - val_loss: 0.2650 - val_accuracy: 0.8900\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.26478\n",
      "Epoch 6/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1889 - accuracy: 0.9274 - val_loss: 0.2280 - val_accuracy: 0.9250\n",
      "\n",
      "Epoch 00006: val_loss improved from 0.26478 to 0.22804, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 7/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1935 - accuracy: 0.9174 - val_loss: 0.2507 - val_accuracy: 0.9050\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.22804\n",
      "Epoch 8/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1802 - accuracy: 0.9362 - val_loss: 0.2949 - val_accuracy: 0.8675\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.22804\n",
      "Epoch 9/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1796 - accuracy: 0.9280 - val_loss: 0.2963 - val_accuracy: 0.8575\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.22804\n",
      "Epoch 10/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1768 - accuracy: 0.9368 - val_loss: 0.3128 - val_accuracy: 0.8500\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.22804\n",
      "Epoch 11/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1748 - accuracy: 0.9362 - val_loss: 0.2455 - val_accuracy: 0.9075\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.22804\n",
      "Epoch 12/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1712 - accuracy: 0.9355 - val_loss: 0.1857 - val_accuracy: 0.9425\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00012: val_loss improved from 0.22804 to 0.18568, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 13/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1570 - accuracy: 0.9406 - val_loss: 0.2340 - val_accuracy: 0.9225\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.18568\n",
      "Epoch 14/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1649 - accuracy: 0.9355 - val_loss: 0.2311 - val_accuracy: 0.9225\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.18568\n",
      "Epoch 15/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1500 - accuracy: 0.9418 - val_loss: 0.2239 - val_accuracy: 0.9125\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.18568\n",
      "Epoch 16/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1563 - accuracy: 0.9368 - val_loss: 0.2027 - val_accuracy: 0.9425\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.18568\n",
      "Epoch 17/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1615 - accuracy: 0.9393 - val_loss: 0.2410 - val_accuracy: 0.9025\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.18568\n",
      "Epoch 18/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1397 - accuracy: 0.9418 - val_loss: 0.2083 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.18568\n",
      "Epoch 19/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1502 - accuracy: 0.9418 - val_loss: 0.2734 - val_accuracy: 0.8900\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.18568\n",
      "Epoch 20/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1436 - accuracy: 0.9412 - val_loss: 0.2162 - val_accuracy: 0.9250\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.18568\n",
      "Epoch 21/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1366 - accuracy: 0.9487 - val_loss: 0.2039 - val_accuracy: 0.9350\n",
      "\n",
      "Epoch 00021: val_loss did not improve from 0.18568\n",
      "Epoch 22/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1490 - accuracy: 0.9374 - val_loss: 0.2323 - val_accuracy: 0.9125\n",
      "\n",
      "Epoch 00022: val_loss did not improve from 0.18568\n",
      "Epoch 23/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1444 - accuracy: 0.9424 - val_loss: 0.2971 - val_accuracy: 0.8600\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.18568\n",
      "Epoch 24/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1412 - accuracy: 0.9462 - val_loss: 0.2802 - val_accuracy: 0.8750\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 0.18568\n",
      "Epoch 25/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1239 - accuracy: 0.9493 - val_loss: 0.1890 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.18568\n",
      "Training completed in time:  0:01:44.326599\n",
      "Results from fit # 2\n",
      "Training Accuracy:  0.9605756998062134\n",
      "Testing Accuracy:  0.9375\n",
      "Generating data for fit # 3\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:01:48.351531\n",
      "Train on 1598 samples, validate on 400 samples\n",
      "Epoch 1/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2150 - accuracy: 0.9212 - val_loss: 0.2199 - val_accuracy: 0.9500\n",
      "\n",
      "Epoch 00001: val_loss did not improve from 0.18568\n",
      "Epoch 2/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1703 - accuracy: 0.9299 - val_loss: 0.2111 - val_accuracy: 0.9250\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 0.18568\n",
      "Epoch 3/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1833 - accuracy: 0.9268 - val_loss: 0.2174 - val_accuracy: 0.9325\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.18568\n",
      "Epoch 4/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1521 - accuracy: 0.9380 - val_loss: 0.2687 - val_accuracy: 0.8900\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.18568\n",
      "Epoch 5/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1642 - accuracy: 0.9374 - val_loss: 0.2644 - val_accuracy: 0.9000\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.18568\n",
      "Epoch 6/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1454 - accuracy: 0.9412 - val_loss: 0.2061 - val_accuracy: 0.9275\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.18568\n",
      "Epoch 7/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1501 - accuracy: 0.9431 - val_loss: 0.2163 - val_accuracy: 0.9225\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.18568\n",
      "Epoch 8/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1468 - accuracy: 0.9431 - val_loss: 0.1671 - val_accuracy: 0.9650\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.18568 to 0.16707, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 9/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1473 - accuracy: 0.9431 - val_loss: 0.2202 - val_accuracy: 0.9200\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.16707\n",
      "Epoch 10/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1385 - accuracy: 0.9487 - val_loss: 0.1772 - val_accuracy: 0.9525\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.16707\n",
      "Epoch 11/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1191 - accuracy: 0.9568 - val_loss: 0.1693 - val_accuracy: 0.9475\n",
      "\n",
      "Epoch 00011: val_loss did not improve from 0.16707\n",
      "Epoch 12/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1229 - accuracy: 0.9524 - val_loss: 0.1729 - val_accuracy: 0.9425\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.16707\n",
      "Epoch 13/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1204 - accuracy: 0.9524 - val_loss: 0.1769 - val_accuracy: 0.9525\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.16707\n",
      "Epoch 14/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1117 - accuracy: 0.9543 - val_loss: 0.1735 - val_accuracy: 0.9400\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.16707\n",
      "Epoch 15/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1121 - accuracy: 0.9543 - val_loss: 0.1684 - val_accuracy: 0.9400\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.16707\n",
      "Epoch 16/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1138 - accuracy: 0.9562 - val_loss: 0.1781 - val_accuracy: 0.9500\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.16707\n",
      "Epoch 17/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1003 - accuracy: 0.9643 - val_loss: 0.1573 - val_accuracy: 0.9475\n",
      "\n",
      "Epoch 00017: val_loss improved from 0.16707 to 0.15735, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 18/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1153 - accuracy: 0.9537 - val_loss: 0.2747 - val_accuracy: 0.9100\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.15735\n",
      "Epoch 19/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1227 - accuracy: 0.9518 - val_loss: 0.1975 - val_accuracy: 0.9350\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.15735\n",
      "Epoch 20/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1123 - accuracy: 0.9599 - val_loss: 0.2271 - val_accuracy: 0.9175\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.15735\n",
      "Epoch 21/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1043 - accuracy: 0.9606 - val_loss: 0.1579 - val_accuracy: 0.9450\n",
      "\n",
      "Epoch 00021: val_loss did not improve from 0.15735\n",
      "Epoch 22/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1019 - accuracy: 0.9618 - val_loss: 0.1395 - val_accuracy: 0.9525\n",
      "\n",
      "Epoch 00022: val_loss improved from 0.15735 to 0.13947, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 23/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1149 - accuracy: 0.9543 - val_loss: 0.2321 - val_accuracy: 0.9175\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.13947\n",
      "Epoch 24/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.0887 - accuracy: 0.9656 - val_loss: 0.1625 - val_accuracy: 0.9500\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 0.13947\n",
      "Epoch 25/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.0769 - accuracy: 0.9737 - val_loss: 0.2152 - val_accuracy: 0.9350\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.13947\n",
      "Training completed in time:  0:01:42.072419\n",
      "Results from fit # 3\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy:  0.9693366885185242\n",
      "Testing Accuracy:  0.9350000023841858\n",
      "Generating data for fit # 4\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:01:48.504546\n",
      "Train on 1598 samples, validate on 400 samples\n",
      "Epoch 1/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.2174 - accuracy: 0.9274 - val_loss: 0.2507 - val_accuracy: 0.8975\n",
      "\n",
      "Epoch 00001: val_loss did not improve from 0.13947\n",
      "Epoch 2/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1656 - accuracy: 0.9393 - val_loss: 0.1564 - val_accuracy: 0.9550\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 0.13947\n",
      "Epoch 3/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1461 - accuracy: 0.9487 - val_loss: 0.1932 - val_accuracy: 0.9425\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.13947\n",
      "Epoch 4/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1377 - accuracy: 0.9531 - val_loss: 0.1667 - val_accuracy: 0.9475\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.13947\n",
      "Epoch 5/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1254 - accuracy: 0.9556 - val_loss: 0.1883 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.13947\n",
      "Epoch 6/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1206 - accuracy: 0.9587 - val_loss: 0.1718 - val_accuracy: 0.9275\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.13947\n",
      "Epoch 7/25\n",
      "1598/1598 [==============================] - 4s 2ms/step - loss: 0.1279 - accuracy: 0.9556 - val_loss: 0.1700 - val_accuracy: 0.9500\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.13947\n",
      "Epoch 8/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1206 - accuracy: 0.9581 - val_loss: 0.1411 - val_accuracy: 0.9575\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.13947\n",
      "Epoch 9/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1218 - accuracy: 0.9593 - val_loss: 0.2373 - val_accuracy: 0.9075\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.13947\n",
      "Epoch 10/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1150 - accuracy: 0.9599 - val_loss: 0.1491 - val_accuracy: 0.9625\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.13947\n",
      "Epoch 11/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1120 - accuracy: 0.9643 - val_loss: 0.1320 - val_accuracy: 0.9550\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.13947 to 0.13197, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Epoch 12/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1094 - accuracy: 0.9643 - val_loss: 0.1904 - val_accuracy: 0.9425\n",
      "\n",
      "Epoch 00012: val_loss did not improve from 0.13197\n",
      "Epoch 13/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1075 - accuracy: 0.9599 - val_loss: 0.1752 - val_accuracy: 0.9450\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.13197\n",
      "Epoch 14/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1031 - accuracy: 0.9618 - val_loss: 0.1430 - val_accuracy: 0.9575\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.13197\n",
      "Epoch 15/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0959 - accuracy: 0.9668 - val_loss: 0.1428 - val_accuracy: 0.9550\n",
      "\n",
      "Epoch 00015: val_loss did not improve from 0.13197\n",
      "Epoch 16/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.0949 - accuracy: 0.9687 - val_loss: 0.1814 - val_accuracy: 0.9350\n",
      "\n",
      "Epoch 00016: val_loss did not improve from 0.13197\n",
      "Epoch 17/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0886 - accuracy: 0.9675 - val_loss: 0.1803 - val_accuracy: 0.9350\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.13197\n",
      "Epoch 18/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0969 - accuracy: 0.9668 - val_loss: 0.1601 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00018: val_loss did not improve from 0.13197\n",
      "Epoch 19/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.1208 - accuracy: 0.9593 - val_loss: 0.1492 - val_accuracy: 0.9550\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.13197\n",
      "Epoch 20/25\n",
      "1598/1598 [==============================] - 5s 3ms/step - loss: 0.0946 - accuracy: 0.9643 - val_loss: 0.1521 - val_accuracy: 0.9550\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.13197\n",
      "Epoch 21/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0897 - accuracy: 0.9712 - val_loss: 0.1606 - val_accuracy: 0.9400\n",
      "\n",
      "Epoch 00021: val_loss did not improve from 0.13197\n",
      "Epoch 22/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0841 - accuracy: 0.9750 - val_loss: 0.1953 - val_accuracy: 0.9425\n",
      "\n",
      "Epoch 00022: val_loss did not improve from 0.13197\n",
      "Epoch 23/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.1035 - accuracy: 0.9643 - val_loss: 0.1688 - val_accuracy: 0.9475\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.13197\n",
      "Epoch 24/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0808 - accuracy: 0.9725 - val_loss: 0.1776 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 0.13197\n",
      "Epoch 25/25\n",
      "1598/1598 [==============================] - 4s 3ms/step - loss: 0.0787 - accuracy: 0.9712 - val_loss: 0.1306 - val_accuracy: 0.9500\n",
      "\n",
      "Epoch 00025: val_loss improved from 0.13197 to 0.13057, saving model to C:/Users/Gabe/Documents/SeniorDesign/CNNModels/weights.best.basic_cnn.hdf5\n",
      "Training completed in time:  0:01:45.694032\n",
      "Results from fit # 4\n",
      "Training Accuracy:  0.9881101250648499\n",
      "Testing Accuracy:  0.949999988079071\n",
      "Finished model fitting\n"
     ]
    }
   ],
   "source": [
    "# Fit model by regenerating training data based on num_fits\n",
    "# Each iteration of num_fits, a new combination of siren and noise is generated from the array\n",
    "# adn the model weights are fit with epochs = num_epochs and batch size = num_batch_size\n",
    "\n",
    "from keras.callbacks import ModelCheckpoint \n",
    "from datetime import datetime \n",
    "\n",
    "#num_epochs = 12\n",
    "#num_batch_size = 128\n",
    "\n",
    "num_fits = 4 # number of times new training data is generated and model trained\n",
    "\n",
    "# For each fit (i.e. for iterations = num_fits):\n",
    "num_epochs = 25\n",
    "num_batch_size = 32\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=saveModel, monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)\n",
    "\n",
    "for i in range(0, num_fits):\n",
    "    \n",
    "    print('Generating data for fit #', i+1)\n",
    "\n",
    "    X, yy = generateData(carNoise, backNoise, sirenNoise)\n",
    "    \n",
    "    # split the dataset \n",
    "    x_train, x_test, y_train, y_test = train_test_split(X, yy, test_size=0.2)#, random_state = 42)\n",
    "    x_train = x_train.reshape(x_train.shape[0], num_rows, num_columns, num_channels)\n",
    "    x_test = x_test.reshape(x_test.shape[0], num_rows, num_columns, num_channels)\n",
    "    \n",
    "    start = datetime.now()\n",
    "\n",
    "    model.fit(x_train, y_train, batch_size=num_batch_size, epochs=num_epochs, validation_data=(x_test, y_test), callbacks=[checkpointer], verbose=1)\n",
    "\n",
    "    duration = datetime.now() - start\n",
    "    print(\"Training completed in time: \", duration)\n",
    "\n",
    "    # Evaluating the model on the training and testing set\n",
    "    \n",
    "    print('Results from fit #', i+1)\n",
    "    score = model.evaluate(x_train, y_train, verbose=0)\n",
    "    print(\"Training Accuracy: \", score[1])\n",
    "\n",
    "    score = model.evaluate(x_test, y_test, verbose=0)\n",
    "    print(\"Testing Accuracy: \", score[1])\n",
    "    \n",
    "print('Finished model fitting')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save model\n",
    "\n",
    "model.save(modelSave)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Anaconda3\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "\n",
    "from keras.models import load_model\n",
    "\n",
    "modelSave = 'C:/Users/Gabe/Documents/Driver-Alert/Models/siren_detector_V2.h5' # location of the trained model\n",
    "model = load_model(modelSave)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute ROC curve for validation data\n",
    "\n",
    "num_samples = sirenNoiseVal.shape[0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute ROC for various SNRs\n",
    "\n",
    "num_samples = sirenNoiseVal.shape[0]\n",
    "\n",
    "from sklearn.metrics import roc_curve\n",
    "from numpy import ravel\n",
    "from sklearn.metrics import auc\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "num_rows = 40\n",
    "num_columns = 130 #x_train.shape[2]\n",
    "num_channels = 1\n",
    "\n",
    "def plotROC(sirenSNR, Title):\n",
    "\n",
    "    num_samples = sirenNoiseVal.shape[0]\n",
    "    X, yy = generateData(carNoiseVal, backNoiseVal, sirenNoiseVal, sirenSNR)\n",
    "\n",
    "    X = X.reshape(X.shape[0], num_rows, num_columns, num_channels)\n",
    "\n",
    "    y_pred_keras = model.predict_proba(X)\n",
    "    predicted_proba = y_pred_keras[:,1]\n",
    "    desc_score_indices = np.argsort(predicted_proba, kind=\"mergesort\")[::-1]\n",
    "    y_pred = predicted_proba\n",
    "\n",
    "    y_test2 = yy[:,1]\n",
    "\n",
    "    y_pred_keras = model.predict_proba(X)\n",
    "\n",
    "    fpr_keras, tpr_keras, thresholds_keras = roc_curve(y_test2, y_pred)\n",
    "\n",
    "    auc_keras = auc(fpr_keras, tpr_keras)\n",
    "\n",
    "    # plot\n",
    "    plt.figure(1)\n",
    "    plt.plot([0, 1], [0, 1], 'k--')\n",
    "    plt.plot(fpr_keras, tpr_keras, label='ROC (area = {:.3f})'.format(auc_keras))\n",
    "    plt.xlabel('False positive rate')\n",
    "    plt.ylabel('True positive rate')\n",
    "    plt.title(Title)\n",
    "    plt.legend(loc='best')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mixing various SNRs\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:24.313402\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Mixed SNRs: (-5, 0, 5, 10 dB SNR)\n",
    "\n",
    "plotROC(False, 'ROC curve: Mixed SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  -10\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:21.725961\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of -10 dB\n",
    "\n",
    "plotROC(-10, 'ROC curve: -10 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  -5\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:21.698075\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of -5 dB\n",
    "\n",
    "plotROC(-5, 'ROC curve: -5 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SNR of 0 dB\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:21.919018\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmcTfX/wPHX24xQobJF1rHOjJiYCBGRpUK+JUpUv7Elpfq20CIk38gW2SlSipSolJZvfX2/yk6EMI19yZIlFDPj/fvjnrmNccdcY+6cuXPfz8fjPpxz7+ee8z4z477vZzmfj6gqxhhjDEAetwMwxhiTc1hSMMYY42VJwRhjjJclBWOMMV6WFIwxxnhZUjDGGONlScGYICIi20WkmdtxmNzLkoLJcs4H158ickJE9ovIdBG5Mk2Z+iLybxH5Q0SOicinIhKVpkwhERktIjudY8U7+0Wz94qylog86fxcjonIWyKSL5PHGSAiic7P5oSIbBKRuy9Q/jIRGSEiu53y20RkVKrXt4vIbyJyRarnuorI96n2VUROOu/fIyIjRSQsM/GbnMmSggmU1qp6JRAD3AD0S3lBROoBXwHzgVJABeAnYImIRDhlLgO+BaKBlkAhoD5wGKgTqKBFJDxQx3aO3wLoCzQFygMRwMBLOORsVb3S+Vk/AbwrIiXSKdsPiMXz8ysINAHWpCkTDvTJ4Jw1nfPdAnQA/i+zwZucx5KCCShV3Q8swpMcUgwD3lHVN1T1D1X9XVVfBJYCA5wyXYCyQDtV3aiqZ1X1gKq+oqoLfZ1LRKJF5GsR+d35xvu88/x0ERmcqlxjEdmdan+7iDwnIuuAkyLyoojMTXPsN0RkjLNdWESmicg+59vy4Iv4tvwgME1VN6jqEeAV4KH0CotIZxHZISKHReSFCx1YVRcBfwAV0ylyIzBPVfeqx3ZVfSdNmdeBp0XkqowuRFXjgSWc+7s1Qc6SggkoESkNtALinf3L8Xzj/9BH8TnAbc52M+BLVT3h53kKAt8AX+KpfVTCU9Pw133AHcBVwEzgdhEp5Bw7DLgXmOWUnQEkOee4AWgOdHXKlhWRoyJSNp3zROOpFaX4CSghIkV8XFMUMAHo7FxTEaC0r4OKxx3AZcDGdM69FHhKRHqJyPUiIj7KrAS+B55O5xipz1kNaIjzuzW5gyUFEyifiMgfwC7gAPCy8/w1eP7u9vl4zz4gpb+gSDpl0nMnsF9VR6jqX04NZNlFvH+Mqu5S1T9VdQewGrjLee1W4JSqLnWaZloBT6jqSVU9AIwCOgKo6k5VvUpVd6ZzniuBY6n2U7YL+ih7D/CZqi5W1dPAS8DZNGXuFZGjwElgATBEVY+mc+5/AUOBTng+/PeIyIM+yvUHHhORYukcZ7WInAQ24Ukg49MpZ4KQJQUTKHepakGgMVCNvz/sj+D5YCvp4z0lgUPO9uF0yqSnDPBrpiL12JVmfxae2gPA/fxdSygH5AX2OTWCo8AkoLif5zmBp38kRcr2Hz7Klkodl6qexPNzSW2Ok4Qux9Ns1EVEevg6saomq+o4VW2Ap0b0KvCWiESmKfcz8Bmevg9fauFJbh2AusAV6ZQzQciSggkoVf0PMB0Y7uyfBH4E2vsofi9/N/l8A7RIPRImA7tIvy39JHB5qv1rfYWaZv9DoLHT/NWOv5PCLuA0UNT5ML5KVQuparSfcW4Aaqbarwn8pqppP+zBU1Mqk7LjNL2d18zkvQDV7cAXQOuMgnBqROPwJOkoH0VeBroB16XzflXVOXh+l/0zOp8JHpYUTHYYDdwmIikdkn2BB0XkcREpKCJXOx3B9fh7JM5MPB/AH4lINRHJIyJFROR5Ebndxzk+A64VkSdEJJ9z3LrOa2vx9BFcIyLX4hmlc0GqehBP08jbwDZV3eQ8vw/PyKkRzpDZPCJSUURu8fNn8Q4QJyJRInI18CKepOnLXOBOEbnZGY01iAv8n3USWEs8icfX6084newFRCTcaToqyPkjkFI6kWcDj2dwPa8B3Z2fq8kFLCmYgHM+YN/B0yaOqv4PaAH8A8+34R14OmxvVtWtTpnTeDqbfwG+Bo4Dy/E0Q53XV6Cqf+DppG4N7Ae24hlyCZ4E8xOwHc8H+mw/Q5/lxDArzfNd+LtD9wieD++S4O1oPpFeR7Oqfoln9NV3znXv4O/+lrRlNwCPOuff55xrd5piHVLuUwBW4BkNlN4Q1z+BEXh+PoecY9+tqgnplB9EBk1Dqroe+A/wzIXKmeAhtsiOMcaYFFZTMMYY42VJwRhjjJclBWOMMV6WFIwxxngFdPKvQChatKiWL1/e7TCMMSaorFq16pCqpneXulfQJYXy5cuzcuVKt8MwxpigIiI7/ClnzUfGGGO8LCkYY4zxsqRgjDHGy5KCMcYYL0sKxhhjvAKWFMSzIPkBEfk5nddFRMaIZzH2dSJSK1CxGGOM8U8gawrT8Uzjm55WQGXn0R3PsoPGGGNcFLD7FFR1sYiUv0CRtngWb1dgqYhcJSIlnfnqjTEhatayncxfu8ftMHKUs2eTOXMmkVoRxXm5tb/rOWWOmzevXce5SyDudp47LymISHc8tQnKlk1vPXRjTFZy68N52bbfAahb4ZpsP3dOdPToUTZv3kx4eDg3VMjwhuRL5mZSEB/P+VzcQVUnA5MBYmNjbQEIY7KYrwTg1odz3QrX0DbmOu6vG9pfAI8ePcozzzzDnKlTqVSpElOnTuWWW6oH/LxuJoXdpFp/FigN7HUpFmOCSlZ/i/eVAOzD2T3JycnUr1+fzZs38+yzzzJgwAAKFCiQLed2MyksAHqLyAdAXeCY9ScY45/5a/ewcd9xokoWypLjWQLIGQ4fPsw111xDWFgYr776KmXKlCE2NjZbYwhYUhCR94HGQFER2Y1nHdq8AKo6EVgI3A7EA6eAhwMVizHBzFetICUhzO5Rz6WoTFZSVd577z369OnDa6+9Rrdu3WjXrp0rsQRy9NF9GbyueBYON8YvoToqxVfTTlTJQrSNuc6tkEwW2rVrFz179mThwoXcdNNNNGjQwNV4gm7qbBO6srrJJFhY007u9f7779OjRw+Sk5MZPXo0vXv3JiwszNWYLCmYHC117cCaTExuc/XVV1O3bl0mT55MhQoV3A4HsKRgstnFNgGlbjqxJhMT7JKSkhg1ahRnzpzhhRdeoGXLlrRo0QIRXyP03WFJwQRc6kRwsWPfrenE5BY//fQTcXFxrFq1invvvRdVRURyVEIASwrGT5fSyZs6EdiHvAk1p0+fZvDgwbz22mtcc801fPjhh9x99905LhmksKRg/HIpnbyWCEwo27p1K0OHDuX+++9n5MiRFClSxO2QLsiSQojztwZgnbzG+O/EiRPMnz+fTp06Ub16dX755RciIiLcDssvlhRCwIU++P1t47dOXmP88/XXX9O9e3d27NhBrVq1iIyMDJqEAJYUQsKFmn6saceYrHHkyBGefvpp3nrrLapUqcJ//vMfIiMj3Q7rollSCBHW9GNM4CQnJ9OgQQO2bNlCv3796N+/P/nz53c7rEyxpGCMMZl06NAh7wR2Q4YMoWzZstSqFdwrC1tSyIXS9iGE4tQQxgSSqjJz5kyeeOIJXnvtNbp3785dd93ldlhZwpJCELiUu4DBOomNyUo7duygR48eLFq0iPr169OoUSO3Q8pSlhSCwMXeI2Cdx8YExrvvvssjjzyCqjJ27Fh69epFnjx53A4rS1lSyMFSagh2j4AxOUOxYsVo0KABkyZNoly5cm6HExCWFALoUuf/T90MZM0/xmS/xMRERowYQWJiIi+99BItWrSgefPmOXaKiqxgSSELpU0Cl7rwuTUDGeOeNWvWEBcXx5o1a+jYsWOOncAuq1lSyCKzlu3k+Xnrgb+TgH2oGxN8/vrrLwYNGsSwYcMoWrQoH330Ef/4xz/cDivbWFLIIik1hCHtrrckYEwQi4+PZ/jw4XTp0oURI0Zw9dVXux1StrKkcIlSdwbXrXCNJQRjgtCJEyeYN28enTt3pnr16mzevDnHrISW3XLXWCoXpB4dZJ3BxgSfRYsWER0dzYMPPsimTZsAQjYhgCWFLJEyXNRqCcYEj8OHD/Pggw/SsmVLLr/8cv773/8G5QR2Wc2aj4wxISdlArv4+HheeOEFXnzxxaCdwC6rWVLIhNRDT21eIWOCx8GDBylSpAhhYWEMHTqUcuXKERMT43ZYOYo1H2VCSj8C2LxCxgQDVeXtt9+mSpUqTJkyBYC2bdtaQvDBagqZZNNOGBMctm/fTvfu3fn6669p2LAhTZo0cTukHM2SQjouNEWFNRkZExxmzpzJI488gogwfvx4evTokesmsMtq9tNJR+omorSsyciY4FCiRAkaNWrEhg0beOSRRywh+MFqCmnYzKTGBK/ExESGDRtGcnIy/fv3p3nz5jRv3tztsIKKpc007GY0Y4LT6tWrufHGG3nxxRfZvHkzqup2SEHJago+WA3BmODx559/MnDgQIYPH06xYsWYN29erlka0w0BrSmISEsR2Swi8SLS18frZUXkOxFZIyLrROT2QMZjjMl9EhISGDlyJA899BAbN260hHCJApYURCQMGAe0AqKA+0QkKk2xF4E5qnoD0BEYH6h4/DFr2U7vGgjGmJzr+PHjTJ8+HYDo6Gi2bt3K1KlTQ25G00AIZE2hDhCvqgmqegb4AGibpowCKWM7CwN7AxhPhlKGoFpfgjE518KFC6levTpxcXHeCexy69KYbghkUrgO2JVqf7fzXGoDgAdEZDewEHjM14FEpLuIrBSRlQcPHgxErF42/bUxOdOhQ4fo3Lkzd9xxBwULFmTJkiU2gV0ABDIp+FqzLu1wgPuA6apaGrgdmCki58WkqpNVNVZVY4sVKxaAUI0xOVnKBHYffPAB/fv3Z/Xq1dx0001uh5UrBXL00W6gTKr90pzfPBQHtARQ1R9FJD9QFDgQwLiMMUHit99+o1ixYoSFhTF8+HDKlStHjRo13A4rVwtkTWEFUFlEKojIZXg6khekKbMTaAogIpFAfiCw7UPGmBxPVZk2bRpVq1Zl8uTJALRu3doSQjYIWFJQ1SSgN7AI2IRnlNEGERkkIm2cYv8EuonIT8D7wENqd5wYE9ISEhJo1qwZXbt2JSYmhmbNmrkdUkgJ6M1rqroQTwdy6uf6p9reCDQIZAz+SDu1hTHGHTNmzKBXr16EhYUxceJEunXrZvMVZTO7oxmb2sKYnKJUqVLceuutTJgwgdKlS7sdTkgK6aRgk98Z464zZ87w2muvcfbsWQYMGMBtt93Gbbfd5nZYIS2k62VWQzDGPStWrKB27dq8/PLLJCQk2AR2OURI1xTAJr8zJrudOnWK/v37M2rUKEqWLMmCBQto3bq122EZR0jXFIwx2W/btm2MHTuWbt26sWHDBksIOUzI1xSMMYF37NgxPv74Yx5++GGio6OJj4+nTJkyGb/RZLuQrSnYjKjGZI/PP/+c6Ohounbtyi+//AJgCSEHC9mkYDOiGhNYBw8epFOnTtx5551cffXV/Pjjj1SrVs3tsEwGQrr5yGZENSYwkpOTufnmm9m2bRsDBw6kb9++XHbZZW6HZfwQ0knBGJO19u/fT/HixQkLC2PEiBGUL1+e6tWrux2WuQgh23xkjMk6Z8+eZdKkSVSpUoVJkyYBcOedd1pCCEIZJgURKSAi/URkorNfSURaBT40Y0wwiI+Pp2nTpvTs2ZMbb7yRFi1auB2SuQT+1BTewrNgzs3O/l5gSMAiMsYEjbfffpvrr7+e1atXM2XKFL755hsiIiLcDstcAn+SQmVVHQIkAqjqKXyvqmaMCTFly5alRYsWbNy4ka5duyJiHw3Bzp+O5jPOimgKICIVgDMBjcoYkyOdPn2af/3rX5w9e5ZBgwbRtGlTmjZt6nZYJgv5U1N4BfgSKC0iM4DvgOcDGpUxJsdZtmwZtWvXZuDAgezcudMmsMulMkwKqvoF0B7oBswD6qjqN4EOzBiTM5w8eZKnnnqKevXqcezYMT777DOmT59uTUW5lD+jj75S1YOqOl9VP1HVAyLyVXYEZ4xx344dOxg/fjw9e/Zkw4YN3HHHHW6HZAIo3T4FEbkMyA+UEJGC/N25XAiw24CNycWOHj3K3Llz6dq1K1FRUcTHx9tKaCHiQjWFR4ENQDXn35THImBi4EMzxrhh/vz5REVF0bNnT+8EdpYQQke6SUFVR6lqGeA5VS2rqmWcR7Sqjs7GGI0x2eDAgQN07NiRu+66i2LFirF06VKbwC4EZTgkVVVHi0g1IApPc1LK87MCGZgxJvskJyfToEEDdu7cyeDBg3n22WfJmzev22EZF2SYFETkRaA5nmakRUAL4H+AJQVjgtzevXu59tprCQsL44033qB8+fJERUW5HZZxkT/3KXQAmgD7VLUzUBObXdWYoHb27FkmTJhAtWrVmDjR00V4++23W0IwfiWFP1U1GUhyRiHtB2xyE2OC1JYtW2jSpAm9evWibt26tGpl81uav/mTFNaIyFV4JsZbCSwHVgc0KmNMQEybNo2aNWuybt063nrrLb766isqVKjgdlgmB7lgM5B4blkcoKpHgXEisggopKqWFIwJQuXLl6dVq1aMGzeOkiVLuh2OyYEumBRUVUXkM6C2sx+fLVEZY7LE6dOneeWVVwAYPHiwTWBnMuRP89FyEakV8EiMMVnqhx9+ICYmhldffZV9+/bZBHbGL/4khZvxJIbNIrJaRNaIiDUfGZNDnThxgj59+nDzzTdz6tQpvvzyS6ZNm2YT2Bm/+DO09K7MHlxEWgJvAGHAVFV9zUeZe4EBeNZr+ElV78/s+YwxsHPnTiZNmsSjjz7KkCFDKFiwoNshmSDizx3Nv2bmwCISBowDbgN2AytEZIGqbkxVpjLQD2igqkdEpHhmzmVMqDty5Agffvgh3bt3JyoqioSEBEqVKuV2WCYI+dN8lFl1gHhVTVDVM8AHQNs0ZboB41T1CICqHghgPMbkSvPmzSMqKopevXqxefNmAEsIJtMCmRSuA3al2t/tPJdaFaCKiCwRkaVOc9N5RKS7iKwUkZUHDx4MULjGBJf9+/fTvn17/vGPf3DttdeyfPlyqlat6nZYJsj5NV2FiJQGKqvqdyKSDwhX1ZMZvc3Hc2mHP4QDlYHGQGngvyJS3bkv4u83qU4GJgPExsbaEAoT8pKTk2nYsCG7du1iyJAhPP300zaBnckS/kyI939Ab6AwUBEoB4wHmmXw1t1AmVT7pYG9PsosVdVEYJuIbMaTJFb4Fb0xIWb37t2UKlWKsLAwxowZQ4UKFWx6a5Ol/Gk+ehy4CTgOoKpbAH86hFcAlUWkgrOKW0dgQZoyn+CZbA8RKYqnOSnBv9CNCR1nz55l7NixVKtWjQkTJgDQqlUrSwgmy/mTFP5yOooB76iiDAc8q2oSnhrGImATMEdVN4jIIBFp4xRbBBwWkY3Ad8Azqnr4Yi/CmNzsl19+oVGjRjz++OPcfPPN3HnnnW6HZHIxf/oUlojIs0B+EWmCZ5nOz/w5uKouBBamea5/qm0FnnIexpg0pk6dSu/evbn88suZMWMGnTt3tpvQTED5U1N4FvgD+AXoA3wLvBDIoIwxHhUrVqR169Zs2rSJLl26WEIwAedPTeF2PHcjTwh0MNlh1rKdzF+7h437jhNVspDb4Rhzjr/++otBgwYBMGTIEJo0aUKTJk1cjsqEEn9qCvcC8SLytoi0cPoUglbqhNA2Ju1tE8a4Z8mSJcTExPCvf/2LgwcP2gR2xhUZJgVnCc4qwKfA/wEJIjIx0IEFUlTJQszuUY/765Z1OxRj+OOPP3jsscdo2LAhp0+fZtGiRUyZMsWaiowr/LqjWVVPA/OB6XiGmt4bwJiMCSm7d+9m6tSpPPbYY6xfv57mzZu7HZIJYRkmBRFpJiJTgV+BB4B3gGsDHZgxudnhw4e99xtERkaSkJDAG2+8wZVXXulyZCbU+VNT6Al8CUSqaidVXZD6vgVjjP9Ulblz5xIVFcXjjz/uncDOlsY0OYU/fQr3qOpcVf0zOwIyJrfat28fd999N+3bt6dMmTKsXLnSJrAzOU66Q1JF5D+qeouIHOHciewEz31n1wQ8OmNyiZQJ7Pbs2cOwYcN48sknCQ/3az5KY7LVhf4qUwZHF82OQIzJjXbt2sV1111HWFgY48aNo0KFClSpUsXtsIxJV7rNR6p61tmcpqrJqR/AtOwJz5jglJyczJgxY86ZwK5FixaWEEyO50/9tUbqHefmtRsDE44xwW/Tpk3ExcXx448/0qpVK1q3bu12SMb4Ld2agog85/Qn1BCR353HEeAgaSa5M8Z4TJ48mZiYGLZs2cLMmTP5/PPPKVvWbpI0weNCo4+GAcWAUc6/xYCiqnqNqj6THcEZE2wqV65Mu3bt2LhxIw888IDdlWyCzoWajyqp6lYRmQlEpzyZ8keuqusCHJsxOd6ff/7JgAEDEBFee+01m8DOBL0LJYW+QBwwzsdrCjQKSETGBInFixfTtWtXtm7dSs+ePVFVqxmYoJduUlDVOOffhtkXjjE53/Hjx+nbty8TJkwgIiKCb7/9lltvvdXtsIzJEv7MffQPESnobPcVkTkiUjPwoRmTM+3du5fp06fz1FNPsW7dOksIJlfxZ+6jAar6h4jUB1oDs4FJgQ3LmJzl0KFDjB8/HoBq1aqxbds2RowYwRVXXOFyZMZkLX+SQrLz753AeFX9CMgXuJCMyTlUldmzZxMVFcUTTzzBli1bAChRooTLkRkTGP4khX0iMg7oCCwUkcv8fJ8xQW3v3r3cdddddOzYkXLlyrFq1Sq7I9nkev7c0XwvnnWax6rqEREphWdkkjG5VnJyMo0aNWLPnj0MHz6cPn362AR2JiRk+FeuqidEZCPQWEQaA/9V1S8CHlkWm7Vs5znrMxvjy44dOyhdujRhYWGMHz+eiIgIKlWq5HZYxmQbf0Yf9QbmAGWdxxwR6RXowLJa6oTQNuY6t8MxOUxycjIjR44kMjLSO4Fd8+bNLSGYkONPfbg7UEdVTwCIyBDgB2B8IAMLhKiShZjdo57bYZgc5ueffyYuLo7ly5dz5513ctddd7kdkjGu8afDWIDEVPuJznPGBL2JEydSq1YtEhISmDVrFgsWLKB06dJuh2WMa/ypKcwElorIR3iSwV3AjIBGZUyApUxJERkZSfv27Rk9ejTFihVzOyxjXOdPR/MwEfkOSJnuoqeqrghsWMYExqlTp+jfvz9hYWEMHTqUW265hVtuucXtsIzJMfy93+C08/jT+deYoPP9999To0YNRowYwYkTJ1DVjN9kTIjxZ/TRC8D7QEmgNDBLRPoFOjBjssqxY8fo0aOHd0rrf//734wbN85mNDXGB3/6FB4AaqvqKQAReRVYBfwrkIEZk1X27dvHu+++y9NPP83AgQO5/PLL3Q7JmBzLn+ajHZybPMKBBH8OLiItRWSziMSLSLp3QYvIPSKiIhLrz3GNycjBgwcZO3Ys4JnAbvv27bz++uuWEIzJgD9J4RSwQUSmisgUYD1wVERGisjI9N4kImF4FuhpBUQB94lIlI9yBYHHgWWZuQBjUlNVZs2aRWRkJP/85z+9E9jZyCJj/ONP89HnziPFUj+PXQeIV9UEABH5AGgLbExT7hU860E/7edxjfFp165dPPLII3z++efUrVuXadOm2QR2xlwkf4akTsvksa8DdqXa3w3UTV1ARG4AyqjqZyKSblIQke547qymbNmymQzH5GZJSUk0btyY/fv3M2rUKB577DHCwsLcDsuYoBPIaR99De3wjgEUkTzAKOChjA6kqpOByQCxsbE2jtB4bd++nTJlyhAeHs6kSZOIiIggIiLC7bCMCVqBXBdhN1Am1X5pYG+q/YJAdeB7EdkO3AQssM5m44+kpCSGDx9OZGSkd0W0Zs2aWUIw5hL5XVMQkXyqejE3rq0AKotIBWAPnkV67k95UVWPAUVTHf974GlVXXkR5zAhaN26dcTFxbFy5Uratm3L3Xff7XZIxuQa/ty8VkdE1gNbnf2aIjI2o/epahLQG1gEbALmqOoGERkkIm0uMW4TosaPH0/t2rXZsWMHs2fPZt68eZQqVcrtsIzJNfypKYzBsz7zJwCq+pOINPHn4Kq6EFiY5rn+6ZRt7M8xTWhKmcCuevXqdOzYkVGjRlG0aNGM32iMuSj+JIU8qrojzZQAyQGKx5hznDx5khdffJHw8HBef/11GjVqRKNGjdwOy5hcy5+O5l0iUgdQEQkTkSeALQGOyxi+/fZbrr/+ekaPHs3p06dtAjtjsoE/SeER4Ck8S3H+hmeU0COBDMqEtqNHj9K1a1eaNWtGeHg4ixcvZsyYMTaBnTHZwJ+b1w7gGTlkTLb47bff+OCDD3juued4+eWXKVCggNshGRMyMkwKznxH59XbVbV7QCIyISklEfTp04eqVauyfft260g2xgX+NB99A3zrPJYAxbGFdkwWUVXeffddoqKiePbZZ9m6dSuAJQRjXOJP89Hs1PsiMhP4OmARmZCxc+dOevbsyRdffEG9evWYNm0alStXdjssY0JaZuY+qgCUy+pATGhJmcDuwIEDjBkzhl69etkEdsbkAP70KRzh7z6FPMDvQLoL5hhzIQkJCZQrV47w8HCmTJlCxYoVKV++vNthGWMcF+xTEM8YwJpAMedxtapGqOqc7AjO5B5JSUkMHTqUqKgoxo0bB0DTpk0tIRiTw1ywpqCqKiLzVLV2dgVkcp+1a9cSFxfH6tWradeuHe3bt3c7JGNMOvwZfbRcRGoFPBKTK7355pvceOON7Nmzh7lz5/Lxxx9TsmRJt8MyxqQj3ZqCiIQ7M53eDHQTkV+Bk3gWz1FVtURh0pUygV2NGjXo1KkTI0eO5JprrnE7LGNMBi7UfLQcqAXclU2xmFzgxIkTvPDCC+TNm5fhw4fbBHbGBJkLNR8JgKr+6uuRTfGZIPLVV19RvXp1xo4dS2Jiok1gZ0wQulBNoZiIPJXei6o6MgDxmCB05MgRnnrqKaZPn07VqlVZvHgxN998s9thGWMy4UI1hTDgSjxrKft6GAPAgQMHmDt3Lv369WMMGzxnAAAXhUlEQVTt2rWWEIwJYheqKexT1UHZFokJKvv37+f999/nySef9E5gV6RIEbfDMsZcogz7FIxJTVWZMWMGUVFR9OvXzzuBnSUEY3KHCyWFptkWhQkK27dvp2XLljz00ENERUWxdu1am8DOmFwm3eYjVf09OwMxOVtSUhJNmjTh0KFDjBs3jp49e5Injz/3PhpjgklmZkk1ISQ+Pp4KFSoQHh7OW2+9RUREBOXK2SS5xuRW9lXP+JSYmMiQIUOIjo72TmDXpEkTSwjG5HJWUzDnWb16NXFxcaxdu5b27dvToUMHt0MyxmQTqymYc4wZM4Y6deqwf/9+Pv74Y+bMmUOJEiXcDssYk00sKRgA75QUN9xwA126dGHjxo20a9fO5aiMMdnNmo9C3B9//EG/fv3Ily8fI0aMoGHDhjRs2NDtsIwxLrGaQgj78ssvqV69OuPHj0dVbQI7Y4wlhVB0+PBhHnzwQVq1asUVV1zBkiVLGDlyJJ7VV40xocySQgg6fPgw8+bN46WXXmLNmjXUq1fP7ZCMMTlEQJOCiLQUkc0iEi8ifX28/pSIbBSRdSLyrYjYIPgA2bdvH8OHD0dVqVKlCjt27GDQoEHky5fP7dCMMTlIwJKCiIQB44BWQBRwn4hEpSm2BohV1RrAXGBYoOIJVarKW2+9RWRkJC+99BLx8fEAXH311S5HZozJiQJZU6gDxKtqgqqeAT4A2qYuoKrfqeopZ3cpUDqA8YScbdu20bx5c+Li4qhZsyY//fSTTWBnjLmgQA5JvQ7YlWp/N1D3AuXjgC98vSAi3YHuAGXLls2q+HK1pKQkbr31Vg4fPsyECRPo3r27TWBnjMlQIJOCr6EsPsc8isgDQCxwi6/XVXUyMBkgNjbWxk1ewNatW4mIiCA8PJy3336bihUrUqZMGbfDMsYEiUB+ddwNpP40Kg3sTVtIRJoBLwBtVPV0AOPJ1RITExk8eDDVq1fnzTffBKBx48aWEIwxFyWQNYUVQGURqQDsAToC96cuICI3AJOAlqp6IICx5GorV64kLi6OdevW0bFjR+677z63QzLGBKmA1RRUNQnoDSwCNgFzVHWDiAwSkTZOsdeBK4EPRWStiCwIVDy51RtvvEHdunU5dOgQ8+fP5/3336d48eJuh2WMCVIBnftIVRcCC9M81z/VdrNAnj83U1VEhNjYWOLi4hg2bBhXXXWV22EZY4KcTYgXZI4fP85zzz1H/vz5GTVqFA0aNKBBgwZuh2WMySVsjGIQWbhwIdHR0UyePJnw8HCbwM4Yk+UsKQSBQ4cO8cADD3DHHXdQuHBhfvjhB15//XWbwM4Yk+UsKQSBI0eO8Omnn/Lyyy+zevVq6ta90D2AxhiTedankEPt2bOH9957j2eeeYbKlSuzY8cO60g2xgSc1RRyGFVlypQpREVFMWDAAH799VcASwjGmGxhSSEH+fXXX2natCndu3enVq1arFu3jkqVKrkdljEmhFjzUQ6RlJRE06ZN+f3335k0aRJdu3a1CeyMMdnOkoLLNm/eTMWKFQkPD2fGjBlUrFiR0qVtBnFjjDvsq6hLzpw5w8CBA7n++usZN24cALfccoslBGOMq6ym4ILly5cTFxfHzz//zP3330+nTp3cDskYYwCrKWS70aNHU69ePe+9B++99x5FixZ1OyxjjAEsKWSblCkp6tSpQ7du3diwYQN33nmny1EZY8y5rPkowI4dO8azzz5LgQIFGD16NPXr16d+/fpuh2WMMT5ZTSGAPv30U6Kiopg6dSr58uWzCeyMMTmeJYUAOHjwIPfffz9t2rShSJEiLF26lKFDh9oEdsaYHM+SQgAcO3aMhQsXMnDgQFauXMmNN97odkjGGOMX61PIIrt27eLdd9+lb9++VKpUiR07dlC4cGG3wzLGmItiNYVLdPbsWSZOnEh0dDSDBw/2TmBnCcEYE4wsKVyCrVu3cuutt/LII49Qp04d1q9fbxPYGWOCmjUfZVJSUhK33XYbR48eZdq0aTz88MPWkWyMCXqWFC7Spk2bqFy5MuHh4cycOZOKFStSqlQpt8MyISoxMZHdu3fz119/uR2KySHy589P6dKlyZs3b6beb0nBT6dPn2bIkCEMGTKE119/nSeeeIKGDRu6HZYJcbt376ZgwYKUL1/eaqoGVeXw4cPs3r2bChUqZOoYlhT8sHTpUuLi4ti4cSOdO3emc+fObodkDAB//fWXJQTjJSIUKVKEgwcPZvoY1tGcgREjRlC/fn3++OMPFi5cyDvvvEORIkXcDssYL0sIJrVL/XuwpJCOs2fPAlCvXj169uzJzz//TKtWrVyOyhhjAsuSQhpHjx4lLi6OPn36AFC/fn3Gjx9PoUKFXI7MmJwpLCyMmJgYqlevTuvWrTl69Kj3tQ0bNnDrrbdSpUoVKleuzCuvvHLOHGBffPEFsbGxREZGUq1aNZ5++mmf5/jkk08YNGhQwK8ls1SVxx9/nEqVKlGjRg1Wr17ts9zs2bOpUaMG0dHRPPvss97np0+fTrFixYiJiSEmJoapU6cCsHbtWurVq0d0dDQ1atRg9uzZ3vd07NiRrVu3BuZigulRu3ZtzYx7J/6g90784YJl5s2bpyVLltSwsDDt16+fnj17NlPnMia7bNy40e0Q9IorrvBud+nSRQcPHqyqqqdOndKIiAhdtGiRqqqePHlSW7ZsqW+++aaqqq5fv14jIiJ006ZNqqqamJio48aN83mOevXq6cGDB/2OKTExMVPXklmff/65tmzZUs+ePas//vij1qlT57wyhw4d0jJlyuiBAwdU1fOz+uabb1RV9e2339ZHH330vPds3rxZt2zZoqqqe/bs0WuvvVaPHDmiqqrff/+9du3a1Wc8vv4ugJXqx2esdTQDBw4coHfv3nz44YfExMTw2WefUatWLbfDMuaiDPx0Axv3Hs/SY0aVKsTLraP9Ll+vXj3WrVsHwKxZs2jQoAHNmzcH4PLLL+fNN9+kcePGPProowwbNowXXniBatWqARAeHk6vXr3OO+aWLVvIly+fdzGqTz/9lMGDB3PmzBmKFCnCe++9R4kSJRgwYAB79+5l+/btFC1alJkzZ9K3b1++//57Tp8+zaOPPkqPHj04ceIEbdu25ciRIyQmJjJ48GDatm17ST+n+fPn06VLF0SEm266iaNHj7Jv3z5KlizpLZOQkECVKlUoVqwYAM2aNeOjjz6iadOm6R63SpUq3u1SpUpRvHhxDh48yFVXXUXDhg156KGHSEpKIjw86z7KrfkIOH78OF9//TWvvvoqy5cvt4RgTCYkJyfz7bff0qZNG8DTdFS7du1zylSsWJETJ05w/Phxfv755/Ne92XJkiXn/J+8+eabWbp0KWvWrKFjx44MGzbM+9qqVauYP38+s2bNYtq0aRQuXJgVK1awYsUKpkyZwrZt28ifPz/z5s1j9erVfPfdd/zzn//0Oa19hw4dvM05qR/vvPPOeWX37NlDmTJlvPulS5dmz54955SpVKkSv/zyC9u3bycpKYlPPvmEXbt2eV//6KOPqFGjBvfcc885z6dYvnw5Z86coWLFigDkyZOHSpUq8dNPP2X4M7wYIVtT2LlzJzNnzuT555+nUqVK7Ny5k4IFC7odljGZdjHf6LPSn3/+SUxMDNu3b6d27drcdtttgKdpOr2RMBczQmbfvn3eb9fguTejQ4cO7Nu3jzNnzpwzHr9NmzYUKFAAgK+++op169Yxd+5cwDN78datWyldujTPP/88ixcvJk+ePOzZs4fffvuNa6+99pzzpm6/z4ivpJL2Gq+++momTJhAhw4dyJMnD/Xr1ychIQGA1q1bc99995EvXz4mTpzIgw8+yL///e9zfgadO3dmxowZ5Mnz93f54sWLs3fvXr+Sq78CWlMQkZYisllE4kWkr4/X84nIbOf1ZSJSPpDxgGdU0fjx44mOjmbIkCHeCewsIRiTOQUKFGDt2rXs2LGDM2fOMG7cOACio6NZuXLlOWUTEhK48sorKViwINHR0axatcqv46e+Y/uxxx6jd+/erF+/nkmTJp3z2hVXXOHdVlXGjh3L2rVrWbt2Ldu2baN58+a89957HDx4kFWrVrF27VpKlCjh847wi6kplC5d+pxv97t37/Y500Hr1q1ZtmwZP/74I1WrVqVy5coAFClShHz58gHQrVu3c34ux48f54477mDw4MHcdNNN5xzvr7/+8ibBrBKwpCAiYcA4oBUQBdwnIlFpisUBR1S1EjAKGBqoeAD+/POUtz2zXr16bNiwwSawMyaLFC5cmDFjxjB8+HASExPp1KkT//vf//jmm28AT43i8ccf9466eeaZZxgyZAhbtmwBPF/YRo4ced5xIyMjiY+P9+4fO3aM6667DoAZM2akG0+LFi2YMGECiYmJgKdv4uTJkxw7dozixYuTN29evvvuO3bs2OHz/bNnz/YmlNSPLl26nFe2TZs2vPPOO6gqS5cupXDhwuf0J6Q4cOAAAEeOHGH8+PF07doV8NQEUixYsIDIyEgAzpw5Q7t27ejSpQvt27c/73hbtmwhOjpra4iBrCnUAeJVNUFVzwAfAGl7c9oCKb/VuUBTCdCdOKrKunXrWL9+PW+//TaLFi2ifPnygTiVMSHrhhtuoGbNmnzwwQcUKFCA+fPnM3jwYKpWrcr111/PjTfeSO/evQGoUaMGo0eP5r777iMyMpLq1auf8+GYolGjRqxZs8bbRDNgwADat29Pw4YNvZ3PvnTt2pWoqChq1apF9erV6dGjB0lJSXTq1ImVK1cSGxvLe++95+3ovhS33347ERERVKpUiW7dujF+/HjvazExMd7tPn36EBUVRYMGDejbt6+3I3nMmDFER0dTs2ZNxowZw/Tp0wGYM2cOixcvZvr06d6aytq1awH47bffKFCggM/kcynEV1tYlhxY5B6gpap2dfY7A3VVtXeqMj87ZXY7+786ZQ6lOVZ3oDtA2bJla6eX2S9k4Kcb2Lt3LwPaVM/yH6Ixbtm0aZP3W2Vu1qdPH1q3bk2zZs3cDiXHGDVqFIUKFSIuLu6813z9XYjIKlWNzei4gexo9vWNP20G8qcMqjoZmAwQGxubqSzm6YRzpyPOGHNpnn/+eZYtW+Z2GDnKVVddFZB52ALZfLQbKJNqvzSwN70yIhIOFAZ+D2BMxpggVKJECe9QV+Px8MMPZ+n9CSkCmRRWAJVFpIKIXAZ0BBakKbMAeNDZvgf4twaqPcuYXMr+y5jULvXvIWBJQVWTgN7AImATMEdVN4jIIBFJSfnTgCIiEg88BZw3bNUYk778+fNz+PBhSwwG+Hs9hfz582f6GAHraA6U2NhYTTv22ZhQZSuvmbTSW3ktJ3Q0G2MCLG/evJleYcsYX2zuI2OMMV6WFIwxxnhZUjDGGOMVdB3NInIQuPhbmj2KAocyLJW72DWHBrvm0HAp11xOVYtlVCjoksKlEJGV/vS+5yZ2zaHBrjk0ZMc1W/ORMcYYL0sKxhhjvEItKUx2OwAX2DWHBrvm0BDwaw6pPgVjjDEXFmo1BWOMMRdgScEYY4xXrkwKItJSRDaLSLyInDfzqojkE5HZzuvLRKR89keZtfy45qdEZKOIrBORb0WknBtxZqWMrjlVuXtEREUk6Icv+nPNInKv87veICKzsjvGrObH33ZZEflORNY4f9+3uxFnVhGRt0TkgLMypa/XRUTGOD+PdSJSK0sDUNVc9QDCgF+BCOAy4CcgKk2ZXsBEZ7sjMNvtuLPhmpsAlzvbj4TCNTvlCgKLgaVArNtxZ8PvuTKwBrja2S/udtzZcM2TgUec7Shgu9txX+I1NwJqAT+n8/rtwBd4Vq68CViWlefPjTWFOkC8qiao6hngA6BtmjJtgRnO9lygqYj4Who0WGR4zar6naqecnaX4lkJL5j583sGeAUYBuSGuaX9ueZuwDhVPQKgqgeyOcas5s81K1DI2S7M+Ss8BhVVXcyFV6BsC7yjHkuBq0Qkyxaez41J4TpgV6r93c5zPsuoZzGgY0CRbIkuMPy55tTi8HzTCGYZXrOI3ACUUdXPsjOwAPLn91wFqCIiS0RkqYi0zLboAsOfax4APCAiu4GFwGPZE5prLvb/+0XJjesp+PrGn3bcrT9lgonf1yMiDwCxwC0BjSjwLnjNIpIHGAU8lF0BZQN/fs/heJqQGuOpDf5XRKqr6tEAxxYo/lzzfcB0VR0hIvWAmc41nw18eK4I6OdXbqwp7AbKpNovzfnVSW8ZEQnHU+W8UHUtp/PnmhGRZsALQBtVPZ1NsQVKRtdcEKgOfC8i2/G0vS4I8s5mf/+256tqoqpuAzbjSRLByp9rjgPmAKjqj0B+PBPH5VZ+/X/PrNyYFFYAlUWkgohchqcjeUGaMguAB53te4B/q9ODE6QyvGanKWUSnoQQ7O3MkME1q+oxVS2qquVVtTyefpQ2qhrMa7n687f9CZ5BBYhIUTzNSQnZGmXW8ueadwJNAUQkEk9SOJitUWavBUAXZxTSTcAxVd2XVQfPdc1HqpokIr2BRXhGLrylqhtEZBCwUlUXANPwVDHj8dQQOroX8aXz85pfB64EPnT61HeqahvXgr5Efl5zruLnNS8CmovIRiAZeEZVD7sX9aXx85r/CUwRkSfxNKM8FMxf8kTkfTzNf0WdfpKXgbwAqjoRT7/J7UA8cAp4OEvPH8Q/O2OMMVksNzYfGWOMySRLCsYYY7wsKRhjjPGypGCMMcbLkoIxxhgvSwomxxKRZBFZm+pR/gJly6c3q2R2E5FYERnjbDcWkfqpXuspIl2yMZaYYJ811GSvXHefgslV/lTVGLeDuFjODXIpN8k1Bk4APzivTczq84lIuDOHly8xeKY1WZjV5zW5k9UUTFBxagT/FZHVzqO+jzLRIrLcqV2sE5HKzvMPpHp+koiE+XjvdhEZ6pRbLiKVnOfLiWcdipT1KMo6z7cXkZ9F5CcRWew811hEPnNqNj2BJ51zNhSRASLytIhEisjyNNe1ztmuLSL/EZFVIrLI1wyYIjJdREaKyHfAUBGpIyI/iGdNgR9EpKpzB/AgoINz/g4icoV45utf4ZT1NbOsCWVuzx1uD3uk98BzR+5a5zHPee5yIL+zXRnPXa0A5XHmnwfGAp2c7cuAAkAk8CmQ13l+PNDFxzm3Ay84212Az5ztT4EHne3/Az5xttcD1znbVzn/Nk71vgHA06mO7913rivC2X4OeBHPnas/AMWc5zvguYs3bZzTgc+AMGe/EBDubDcDPnK2HwLeTPW+IcADKfECW4Ar3P5d2yPnPKz5yORkvpqP8gJvikgMnqRRxcf7fgReEJHSwMequlVEmgK1gRXONB8FgPTmgHo/1b+jnO16wD+c7Zl41mgAWAJMF5E5wMcXc3F4JnG7F3gNz4d/B6Aqnon8vnbiDAPSm9fmQ1VNdrYLAzOcWpHiTIvgQ3OgjYg87eznB8oCmy4ydpNLWVIwweZJ4DegJp7mz/MWz1HVWSKyDLgDWCQiXfFMNzxDVfv5cQ5NZ/u8MqraU0TqOuda6yQrf83GMxfVx55D6VYRuR7YoKr1/Hj/yVTbrwDfqWo7p9nq+3TeI8Ddqrr5IuI0IcT6FEywKQzsU89c+Z3xfJM+h4hEAAmqOgbPjJI1gG+Be0SkuFPmGkl/neoOqf790dn+gb8nTuwE/M85TkVVXaaq/YFDnDulMcAfeKbxPo+q/oqntvMSngQBnqmui4lnXQBEJK+IRKcTZ2qFgT3O9kMXOP8i4DFxqiHimT3XGC9LCibYjAceFJGleJqOTvoo0wH4WUTWAtXwLF24EU+b/VdOh+7XQHpLGOZzahp98NRMAB4HHnbe29l5DeB1EVnvDIddjGcN4dQ+BdqldDT7ONds4AH+Xg/gDJ7p3IeKyE94+h3O60z3YRjwLxFZwrmJ8jsgKqWjGU+NIi+wzon5FT+ObUKIzZJqTCriWZAnVlUPuR2LMW6wmoIxxhgvqykYY4zxspqCMcYYL0sKxhhjvCwpGGOM8bKkYIwxxsuSgjHGGK//B/MbBln4wiWMAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of 0 dB\n",
    "\n",
    "plotROC(True, 'ROC curve: 0 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  5\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:22.489311\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of 5 dB\n",
    "\n",
    "plotROC(5, 'ROC curve: 5 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  10\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:23.264401\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of 10 dB\n",
    "\n",
    "plotROC(10, 'ROC curve: 10 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  15\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:22.761185\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SNR of 15 dB\n",
    "\n",
    "plotROC(15, 'ROC curve: 15 dB SNR')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "plotROC() missing 1 required positional argument: 'Title'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-94-849930b13ab7>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m# SNR of 20 dB\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      2\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 3\u001b[1;33m \u001b[0mplotROC\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m20\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m: plotROC() missing 1 required positional argument: 'Title'"
     ]
    }
   ],
   "source": [
    "# SNR of 20 dB\n",
    "\n",
    "plotROC(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmcjXX7wPHPZUaoJCGlsY11FrJMdiKyVMjTI0rU8xskKT1SkRZJimyRsZelSClRKS1PPT0VCkmWMI19yZIl++D6/XHuOY0xY44xZ+4551zv1+u83Pc533Of654Zc833+73v6yuqijHGGAOQx+0AjDHG5B6WFIwxxnhZUjDGGONlScEYY4yXJQVjjDFelhSMMcZ4WVIwxhjjZUnBBB0R2Swix0XkiIjsFpFpInJlmjb1ROQ/IvKXiBwSkY9EJDpNm6tEZLSIbHWOlejsF83ZMzIm51hSMMGqtapeCVQDqgP9U14QkbrA58B8oARQFvgF+F5EIp02lwFfATFAS+AqoB6wH6jlr6BFJNxfxzbGF5YUTFBT1d3AIjzJIcUwYIaqvqaqf6nqn6r6DLAEGOi06QKUAtqp6lpVPauqe1T1RVVdmN5niUiMiHwhIn+KyB8i8rTz/DQRGZyqXWMR2Z5qf7OIPCUiq4CjIvKMiMxNc+zXRGSMs11IRKaKyC4R2SEig0Uk7BK/VMYAlhRMkBORCKAVkOjsX47nL/730mn+LnCrs90M+ExVj/j4OQWBL4HP8PQ+yuPpafjqHuB24GpgJnCbiFzlHDsMuBuY5bSdDpx2PqM60BzoehGfZUyGLCmYYPWhiPwFbAP2AM87z1+D5+d+Vzrv2QWkzBcUyaBNRu4AdqvqCFU94fRAll7E+8eo6jZVPa6qW4AVwJ3Oa7cAx1R1iYgUx5PkHlPVo6q6BxgFdLyIzzImQ5YUTLC6U1ULAo2Byvz9y/4AcBa4Pp33XA/sc7b3Z9AmIyWB37MUqce2NPuz8PQeAO7l715CaSAvsEtEDorIQWAicO0lfLYxXpYUTFBT1f8C04Dhzv5RYDHQPp3md/P3kM+XQAsRucLHj9oGlMvgtaPA5an2r0sv1DT77wGNneGvdvydFLYBJ4Giqnq187hKVWN8jNOYC7KkYELBaOBWEUmZbO4H3C8ij4pIQREp7EwE1wVecNrMxPML+H0RqSwieUSkiIg8LSK3pfMZHwPXichjIpLPOW5t57WVeOYIrhGR64DHMgtYVfcC3wBvAptUdZ3z/C48V06NcC6ZzSMi5UTk5ix8XYw5jyUFE/ScX7AzgGed/e+AFsA/8MwbbMEzYdtAVTc6bU7imWz+DfgCOAz8iGcY6ry5AlX9C88kdWtgN7ARaOK8PBPPJa+b8fxCn+Nj6LOcGGaleb4LcBmwFs9w2FwubqjLmAyJLbJjjDEmhfUUjDHGeFlSMMYY42VJwRhjjJclBWOMMV4BV3yraNGiWqZMGbfDMMaYgLJ8+fJ9qloss3YBlxTKlCnDsmXL3A7DGGMCiohs8aWdDR8ZY4zxsqRgjDHGy5KCMcYYL0sKxhhjvCwpGGOM8fJbUhCRN0Rkj4iszuB1EZExzmLoq0Skhr9iMcYY4xt/9hSm4VnwPCOtgArOozsw3o+xGGOM8YHf7lNQ1W9FpMwFmrTFs3i6AktE5GoRud6pF29Mjpm1dCvzV+5wOwxjMnT27BlOnUqmRuS1PN/av+spuXnz2g2cuwThdue585KCiHTH05ugVKlSORKcCS4X+sW/dNOfANQue01OhmSMTw4ePMj69esJDw+netlMb0i+ZG4mBUnnuXQXd1DVScAkgLi4OFsAIhuE2l/HF/rFX7vsNbStdgP31rY/OEzucfDgQZ544gnenTKF8uXLM2XKFG6+Odbvn+tmUtiOZ7HzFBHATpdiCTnzV+5g7a7DRF9/lduh5Aj7xW8CyZkzZ6hXrx7r16/nySefZODAgRQoUCBHPtvNpLAA6CUi7wC1gUM2n5Czoq+/ijkP1nU7DGOMY//+/VxzzTWEhYXx0ksvUbJkSeLi4nI0Bn9ekjobWAxUEpHtIhIvIj1EpIfTZCGQBCQCk4Ge/orFnGvW0q3e4RRjjPtUlbfeeouKFSsyZcoUANq1a5fjCQH8e/XRPZm8rsDD/vr8YJLd4/8pCaFttRuy7ZjGmKzZtm0bPXr0YOHChdSpU4f69eu7Go/d0RwAUsb/s0vtstcwpF0VG183xmWzZ88mJiaGb775htGjR/Pdd98RHR3takwBt55CqEjdO0iZELbxf2OCS+HChalduzaTJk2ibNmybocDWFLIFdIbHkp9CWX09VfZUI8xQeD06dOMGjWKU6dOMWDAAFq2bEmLFi0QSe8KfXdYUnBRSjJI7xp6u4TSmODyyy+/EB8fz/Lly7n77rtRVUQkVyUEsKTgqpS5AksAxgSvkydPMnjwYF555RWuueYa3nvvPe66665clwxSWFJwmc0VGBPcNm7cyNChQ7n33nsZOXIkRYoUcTukC7Kk4GcXupw0lO4oNiaUHDlyhPnz59OpUydiY2P57bffiIyMdDssn1hSyEaZTRinZRPIxgSfL774gu7du7NlyxZq1KhBVFRUwCQEsKSQrdKrJ2TzBcaEhgMHDtC3b1/eeOMNKlasyH//+1+ioqLcDuuiWVLIgoyGhOx+AmNC05kzZ6hfvz4bNmygf//+PPfcc+TPn9/tsLLEkkIWZFRh1IaDjAkt+/bt8xawGzJkCKVKlaJGjcBeWdiSQhZZj8CY0KWqzJw5k8cee4xXXnmF7t27c+edd7odVraw2kcXYdbSrXSYuDhb6xAZYwLLli1baNWqFffffz9RUVE0atTI7ZCylfUUMpF6/iD1lUQ2TGRM6Hnrrbd46KGHUFXGjh1Lz549yZMnuP62tqSQidTzB3YlkTGhrVixYtSvX5+JEydSunRpt8PxC0sKPrD5A2NCU3JyMiNGjCA5OZlnn32WFi1a0Lx581xboiI7BFe/xxhjssnPP/9M7dq16d+/P2vXrsWzLhhBnRDAkoIxxpzjxIkTPP3009x0003s3LmT999/n9mzZwd9MkhhSSEDdqWRMaEpMTGR4cOH06VLF9atW8c//vEPt0PKUTankIHUE8x2pZExwe3IkSPMmzePzp07Exsby/r163PNSmg5zZJCKrYEpjGhZ9GiRXTv3p1t27YRFxdHVFRUyCYECPGkkLaGkS2BaUzo2L9/P3369GHGjBlUrlyZ//3vfwFZwC67hXRSSFvDyO5DMCY0pBSwS0xMZMCAATzzzDMBW8Auu4V0UgC7B8GYULJ3716KFClCWFgYQ4cOpXTp0lSrVs3tsHIVu/rIGBP0VJU333yTihUrMnnyZADatm1rCSEdIZsUZi3d6p1DMMYEr82bN9OiRQv+7//+jypVqtCkSRO3Q8rVQm74KGVyOSUh2GSyMcFr5syZPPTQQ4gICQkJPPjgg0FXwC67hVxSSJlctkllY4Jf8eLFadSoERMmTKBUKfu/7ouQSwpgk8vGBKvk5GSGDRvGmTNneO6552jevDnNmzd3O6yAYv0oY0xQWLFiBTfddBPPPPMM69ev9xawMxfHkoIxJqAdP36cfv36UatWLf744w/mzZvH22+/HTIF7LKbX5OCiLQUkfUikigi/dJ5vZSIfC0iP4vIKhG5zZ/xGGOCT1JSEiNHjuSBBx5g7dq1QbNWslv8lhREJAwYB7QCooF7RCQ6TbNngHdVtTrQEUjwVzxW9dSY4HH48GGmTZsGQExMDBs3bmTKlCkULlzY3cCCgD97CrWARFVNUtVTwDtA2zRtFLjK2S4E7PRXMFb11JjgsHDhQmJjY4mPj2fdunUAQbs0phv8mRRuALal2t/uPJfaQOA+EdkOLAQeSe9AItJdRJaJyLK9e/dmOaCUq47sMlRjAs++ffvo3Lkzt99+OwULFuT777+3AnZ+4M+kkN4sT9rLAe4BpqlqBHAbMFNEzotJVSepapyqxhUrVswPoRpjcrOUAnbvvPMOzz33HCtWrKBOnTpuhxWU/HmfwnagZKr9CM4fHooHWgKo6mIRyQ8UBfb4MS5jTID4448/KFasGGFhYQwfPpzSpUtTtWpVt8MKav7sKfwEVBCRsiJyGZ6J5AVp2mwFmgKISBSQH8j6+JAxJiioKlOnTqVSpUpMmjQJgNatW1tCyAF+SwqqehroBSwC1uG5ymiNiAwSkTZOs8eBbiLyCzAbeEDtjhNjQlpSUhLNmjWja9euVKtWjWbNmrkdUkjxa5kLVV2IZwI59XPPpdpeC9T3ZwzGmMAxffp0evbsSVhYGBMmTKBbt25WwC6HhWTtI2NM7lSiRAluueUWxo8fT0REhNvhhCRLCsYY15w6dYpXXnmFs2fPMnDgQG699VZuvfVWt8MKadYvM8a44qeffqJmzZo8//zzJCUlWQG7XMKSgjEmRx07doy+fftSp04dDhw4wIIFC5gxY4YVsMslLCkYY3LUpk2bGDt2LN26dWPNmjW0bt3a7ZBMKjanYIzxu0OHDvHBBx/wr3/9i5iYGBITEylZsmTmbzQ5znoKxhi/+uSTT4iJiaFr16789ttvAJYQcjFLCsYYv9i7dy+dOnXijjvuoHDhwixevJjKlSu7HZbJhA0fGWOy3ZkzZ2jQoAGbNm3ihRdeoF+/flx22WVuh2V8YEnBGJNtdu/ezbXXXktYWBgjRoygTJkyxMbGuh2WuQg2fGSMuWRnz55l4sSJVKxYkYkTJwJwxx13WEIIQJkmBREpICL9RWSCs19eRFr5PzRjTCBITEykadOm9OjRg5tuuokWLVq4HZK5BL70FN7As2BOA2d/JzDEbxEZYwLGm2++SZUqVVixYgWTJ0/myy+/JDIy0u2wzCXwJSlUUNUhQDKAqh4j/VXVjDEhplSpUrRo0YK1a9fStWtXuys5CPgy0XzKWRFNAUSkLHDKr1EZY3KlkydP8vLLL3P27FkGDRpE06ZNadq0qdthmWzkS0/hReAzIEJEpgNfA0/7NSpjTK6zdOlSatasyQsvvMDWrVutgF2QyjQpqOqnQHugGzAPqKWqX/o7MGNM7nD06FH69OlD3bp1OXToEB9//DHTpk2zoaIg5cvVR5+r6l5Vna+qH6rqHhH5PCeCM8a4b8uWLSQkJNCjRw/WrFnD7bff7nZIxo8ynFMQkcuA/EBxESnI35PLVwGlciA2Y4xLDh48yNy5c+natSvR0dEkJibaSmgh4kI9hYeBNUBl59+UxyJggv9DM8a4Yf78+URHR9OjRw9vATtLCKEjw6SgqqNUtSTwlKqWUtWSziNGVUfnYIzGmBywZ88eOnbsyJ133kmxYsVYsmSJFbALQZlekqqqo0WkMhCNZzgp5flZ/gzMGJNzzpw5Q/369dm6dSuDBw/mySefJG/evG6HZVyQaVIQkWeA5niGkRYBLYDvAEsKxgS4nTt3ct111xEWFsZrr71GmTJliI6Odjss4yJf7lPoADQBdqlqZ+BGrLqqMQHt7NmzjB8/nsqVKzNhgmeK8LbbbrOEYHxKCsdV9Qxw2rkKaTdgxU2MCVAbNmygSZMm9OzZk9q1a9OqldW3NH/zJSn8LCJX4ymMtwz4EVjh16iMMX4xdepUbrzxRlatWsUbb7zB559/TtmyZd0Oy+QiFxwGEs8tiwNV9SAwTkQWAVepqiUFYwJQmTJlaNWqFePGjeP66693OxyTC10wKaiqisjHQE1nPzFHojLGZIuTJ0/y4osvAjB48GArYGcy5cvw0Y8iUsPvkRhjstUPP/xAtWrVeOmll9i1a5cVsDM+8SUpNMCTGNaLyAoR+VlEbPjImFzqyJEj9O7dmwYNGnDs2DE+++wzpk6dagXsjE98ubT0zqweXERaAq8BYcAUVX0lnTZ3AwPxrNfwi6rem9XPM8bA1q1bmThxIg8//DBDhgyhYMGCbodkAogvdzT/npUDi0gYMA64FdgO/CQiC1R1bao2FYD+QH1VPSAi12bls4wJdQcOHOC9996je/fuREdHk5SURIkSJdwOywQgX4aPsqoWkKiqSap6CngHaJumTTdgnKoeAFDVPX6Mx5igNG/ePKKjo+nZsyfr168HsIRgssyfSeEGYFuq/e3Oc6lVBCqKyPcissQZbjqPiHQXkWUismzv3r1+CteYwLJ7927at2/PP/7xD6677jp+/PFHKlWq5HZYJsD5VK5CRCKACqr6tYjkA8JV9Whmb0vnubSXP4QDFYDGQATwPxGJde6L+PtNqpOASQBxcXF2CYUJeWfOnKFhw4Zs27aNIUOG0LdvXytgZ7KFLwXx/g/oBRQCygGlgQSgWSZv3Q6UTLUfAexMp80SVU0GNonIejxJ4iefojcmxGzfvp0SJUoQFhbGmDFjKFu2rJW3NtnKl+GjR4E6wGEAVd0A+DIh/BNQQUTKOqu4dQQWpGnzIZ5ie4hIUTzDSUm+hW5M6Dh79ixjx46lcuXKjB8/HoBWrVpZQjDZzpekcMKZKAa8VxVlesGzqp7G08NYBKwD3lXVNSIySETaOM0WAftFZC3wNfCEqu6/2JMwJpj99ttvNGrUiEcffZQGDRpwxx13uB2SCWK+zCl8LyJPAvlFpAmeZTo/9uXgqroQWJjmuedSbSvQx3kYY9KYMmUKvXr14vLLL2f69Ol07tzZbkIzfuVLT+FJ4C/gN6A38BUwwJ9BGWM8ypUrR+vWrVm3bh1dunSxhGD8zpeewm147kYe7+9gjAl1J06cYNCgQQAMGTKEJk2a0KRJE5ejMqHEl57C3UCiiLwpIi2cOQVjTDb7/vvvqVatGi+//DJ79+61AnbGFZkmBWcJzorAR8D/AUkiMsHfgRkTKv766y8eeeQRGjZsyMmTJ1m0aBGTJ0+2oSLjCp/uaFbVk8B8YBqeS03v9mNMxoSU7du3M2XKFB555BF+/fVXmjdv7nZIJoRlmhREpJmITAF+B+4DZgDX+TswY4LZ/v37vfcbREVFkZSUxGuvvcaVV17pcmQm1PnSU+gBfAZEqWonVV2Q+r4FY4zvVJW5c+cSHR3No48+6i1gZ0tjmtzClzmFf6rqXFU9nhMBGROsdu3axV133UX79u0pWbIky5YtswJ2JtfJ8JJUEfmvqt4sIgc4t5Cd4Lnv7Bq/R2dMkEgpYLdjxw6GDRvGv//9b8LDfapHaUyOutBPZcrF0UVzIhBjgtG2bdu44YYbCAsLY9y4cZQtW5aKFSu6HZYxGcpw+EhVzzqbU1X1TOoHMDVnwjMmMJ05c4YxY8acU8CuRYsWlhBMrudL/7Vq6h3n5rWb/BOOMYFv3bp1xMfHs3jxYlq1akXr1q3dDskYn2XYUxCRp5z5hKoi8qfzOADsJU2RO2OMx6RJk6hWrRobNmxg5syZfPLJJ5QqVcrtsIzx2YWuPhoGFANGOf8WA4qq6jWq+kROBGdMoKlQoQLt2rVj7dq13HfffXZXsgk4Fxo+Kq+qG0VkJhCT8mTKD7mqrvJzbMbkesePH2fgwIGICK+88ooVsDMB70JJoR8QD4xL5zUFGvklImMCxLfffkvXrl3ZuHEjPXr0QFWtZ2ACXoZJQVXjnX8b5lw4xuR+hw8fpl+/fowfP57IyEi++uorbrnlFrfDMiZb+FL76B8iUtDZ7ici74rIjf4PzZjcaefOnUybNo0+ffqwatUqSwgmqPhS+2igqv4lIvWA1sAcYKJ/wzImd9m3bx8JCQkAVK5cmU2bNjFixAiuuOIKlyMzJnv5khTOOP/eASSo6vtAPv+FZEzuoarMmTOH6OhoHnvsMTZs2ABA8eLFXY7MGP/wJSnsEpFxQEdgoYhc5uP7jAloO3fu5M4776Rjx46ULl2a5cuX2x3JJuj5ckfz3XjWaR6rqgdEpASeK5OMCVpnzpyhUaNG7Nixg+HDh9O7d28rYGdCQqY/5ap6RETWAo1FpDHwP1X91O+RGeOCLVu2EBERQVhYGAkJCURGRlK+fHm3wzImx/hy9VEv4F2glPN4V0R6+jswY3LSmTNnGDlyJFFRUd4Cds2bN7eEYEKOL/3h7kAtVT0CICJDgB+ABH8GZkxOWb16NfHx8fz444/ccccd3HnnnW6HZIxrfJkwFiA51X6y85wxAW/ChAnUqFGDpKQkZs2axYIFC4iIiHA7LGNc40tPYSawRETex5MM7gSm+zUqY/wspSRFVFQU7du3Z/To0RQrVsztsIxxnS8TzcNE5GsgpdxFD1X9yb9hGeMfx44d47nnniMsLIyhQ4dy8803c/PNN7sdljG5hq/3G5x0Hsedf40JON988w1Vq1ZlxIgRHDlyBFXN/E3GhBhfrj4aAMwGrgcigFki0t/fgRmTXQ4dOsSDDz7oLWn9n//8h3HjxllFU2PS4cucwn1ATVU9BiAiLwHLgZf9GZgx2WXXrl289dZb9O3blxdeeIHLL7/c7ZCMybV8GT7awrnJIxxI8uXgItJSRNaLSKKIZHgXtIj8U0RUROJ8Oa4xmdm7dy9jx44FPAXsNm/ezKuvvmoJwZhM+JIUjgFrRGSKiEwGfgUOishIERmZ0ZtEJAzPAj2tgGjgHhGJTqddQeBRYGlWTsCY1FSVWbNmERUVxeOPP+4tYGdXFhnjG1+Gjz5xHimW+HjsWkCiqiYBiMg7QFtgbZp2L+JZD7qvj8c1Jl3btm3joYce4pNPPqF27dpMnTrVCtgZc5F8uSR1ahaPfQOwLdX+dqB26gYiUh0oqaofi0iGSUFEuuO5s5pSpUplMRwTzE6fPk3jxo3ZvXs3o0aN4pFHHiEsLMztsIwJOP4s+5jepR3eawBFJA8wCnggswOp6iRgEkBcXJxdR2i8Nm/eTMmSJQkPD2fixIlERkYSGRnpdljGBCx/rouwHSiZaj8C2JlqvyAQC3wjIpuBOsACm2w2vjh9+jTDhw8nKirKuyJas2bNLCEYc4l87imISD5VvZgb134CKohIWWAHnkV67k15UVUPAUVTHf8boK+qLruIzzAhaNWqVcTHx7Ns2TLatm3LXXfd5XZIxgQNX25eqyUivwIbnf0bRWRsZu9T1dNAL2ARsA54V1XXiMggEWlziXGbEJWQkEDNmjXZsmULc+bMYd68eZQoUcLtsIwJGr70FMbgWZ/5QwBV/UVEmvhycFVdCCxM89xzGbRt7MsxTWhKKWAXGxtLx44dGTVqFEWLFs38jcaYi+JLUsijqlvSlAQ446d4jDnH0aNHeeaZZwgPD+fVV1+lUaNGNGrUyO2wjAlavkw0bxORWoCKSJiIPAZs8HNcxvDVV19RpUoVRo8ezcmTJ62AnTE5wJek8BDQB89SnH/guUroIX8GZULbwYMH6dq1K82aNSM8PJxvv/2WMWPGWAE7Y3KALzev7cFz5ZAxOeKPP/7gnXfe4amnnuL555+nQIECbodkTMjINCk49Y7O67erane/RGRCUkoi6N27N5UqVWLz5s02kWyMC3wZPvoS+Mp5fA9ciy20Y7KJqvLWW28RHR3Nk08+ycaNGwEsIRjjEl+Gj+ak3heRmcAXfovIhIytW7fSo0cPPv30U+rWrcvUqVOpUKGC22EZE9KyUvuoLFA6uwMxoSWlgN2ePXsYM2YMPXv2tAJ2xuQCvswpHODvOYU8wJ9AhgvmGHMhSUlJlC5dmvDwcCZPnky5cuUoU6aM22EZYxwXnFMQzzWANwLFnEdhVY1U1XdzIjgTPE6fPs3QoUOJjo5m3LhxADRt2tQSgjG5zAV7CqqqIjJPVWvmVEAm+KxcuZL4+HhWrFhBu3btaN++vdshGWMy4MvVRz+KSA2/R2KC0uuvv85NN93Ejh07mDt3Lh988AHXX3+922EZYzKQYU9BRMKdSqcNgG4i8jtwFM/iOaqqlihMhlIK2FWtWpVOnToxcuRIrrnmGrfDMsZk4kLDRz8CNYA7cygWEwSOHDnCgAEDyJs3L8OHD7cCdsYEmAsNHwmAqv6e3iOH4jMB5PPPPyc2NpaxY8eSnJxsBeyMCUAX6ikUE5E+Gb2oqiP9EI8JQAcOHKBPnz5MmzaNSpUq8e2339KgQQO3wzLGZMGFegphwJV41lJO72EMAHv27GHu3Ln079+flStXWkIwJoBdqKewS1UH5VgkJqDs3r2b2bNn8+9//9tbwK5IkSJuh2WMuUSZzikYk5qqMn36dKKjo+nfv7+3gJ0lBGOCw4WSQtMci8IEhM2bN9OyZUseeOABoqOjWblypRWwMybIZDh8pKp/5mQgJnc7ffo0TZo0Yd++fYwbN44ePXqQJ48v9z4aYwJJVqqkmhCSmJhI2bJlCQ8P54033iAyMpLSpa1IrjHByv7UM+lKTk5myJAhxMTEeAvYNWnSxBKCMUHOegrmPCtWrCA+Pp6VK1fSvn17OnTo4HZIxpgcYj0Fc44xY8ZQq1Ytdu/ezQcffMC7775L8eLF3Q7LGJNDLCkYAG9JiurVq9OlSxfWrl1Lu3btXI7KGJPTbPgoxP3111/079+ffPnyMWLECBo2bEjDhg3dDssY4xLrKYSwzz77jNjYWBISElBVK2BnjLGkEIr279/P/fffT6tWrbjiiiv4/vvvGTlyJJ7VV40xocySQgjav38/8+bN49lnn+Xnn3+mbt26bodkjMkl/JoURKSliKwXkUQR6ZfO631EZK2IrBKRr0TELoL3k127djF8+HBUlYoVK7JlyxYGDRpEvnz53A7NGJOL+C0piEgYMA5oBUQD94hIdJpmPwNxqloVmAsM81c8oUpVeeONN4iKiuLZZ58lMTERgMKFC7scmTEmN/JnT6EWkKiqSap6CngHaJu6gap+rarHnN0lQIQf4wk5mzZtonnz5sTHx3PjjTfyyy+/WAE7Y8wF+fOS1BuAban2twO1L9A+Hvg0vRdEpDvQHaBUqVLZFV9QO336NLfccgv79+9n/PjxdO/e3QrYGWMy5c+kkN6lLOle8ygi9wFxwM3pva6qk4BJAHFxcXbd5AVs3LiRyMhIwsPDefPNNylXrhwlS5Z0OyxjTIDw55+O24HUv40igJ1pG4lIM2AA0EZVT/oxnqCWnJzM4MGDiY2N5fXXXwegcePGlhCMMRfFnz2Fn4AKIlIW2AF0BO5N3UBEqgM3a4prAAAUwUlEQVQTgZaqusePsQS1ZcuWER8fz6pVq+jYsSP33HOP2yEZYwKU33oKqnoa6AUsAtYB76rqGhEZJCJtnGavAlcC74nIShFZ4K94gtVrr71G7dq12bdvH/Pnz2f27Nlce+21bodljAlQfq19pKoLgYVpnnsu1XYzf35+MFNVRIS4uDji4+MZNmwYV199tdthGWMCnBXECzCHDx/mqaeeIn/+/IwaNYr69etTv359t8MyxgQJu0YxgCxcuJCYmBgmTZpEeHi4FbAzxmQ7SwoBYN++fdx3333cfvvtFCpUiB9++IFXX33VCtgZY7KdJYUAcODAAT766COef/55VqxYQe3aF7oH0Bhjss7mFHKpHTt28Pbbb/PEE09QoUIFtmzZYhPJxhi/s55CLqOqTJ48mejoaAYOHMjvv/8OYAnBGJMjLCnkIr///jtNmzale/fu1KhRg1WrVlG+fHm3wzLGhBAbPsolTp8+TdOmTfnzzz+ZOHEiXbt2tQJ2xpgcZ0nBZevXr6dcuXKEh4czffp0ypUrR0SEVRA3xrjD/hR1yalTp3jhhReoUqUK48aNA+Dmm2+2hGCMcZX1FFzw448/Eh8fz+rVq7n33nvp1KmT2yEZYwxgPYUcN3r0aOrWreu99+Dtt9+maNGibodljDGAJYUck1KSolatWnTr1o01a9Zwxx13uByVMcacy4aP/OzQoUM8+eSTFChQgNGjR1OvXj3q1avndljGGJMu6yn40UcffUR0dDRTpkwhX758VsDOGJPrWVLwg71793LvvffSpk0bihQpwpIlSxg6dKgVsDPG5HqWFPzg0KFDLFy4kBdeeIFly5Zx0003uR2SMcb4xOYUssm2bdt466236NevH+XLl2fLli0UKlTI7bCMMeaiWE/hEp09e5YJEyYQExPD4MGDvQXsLCEYYwKRJYVLsHHjRm655RYeeughatWqxa+//moF7IwxAc2Gj7Lo9OnT3HrrrRw8eJCpU6fyr3/9yyaSjTEBz5LCRVq3bh0VKlQgPDycmTNnUq5cOUqUKOF2WCZEJScns337dk6cOOF2KCaXyJ8/PxEREeTNmzdL77ek4KOTJ08yZMgQhgwZwquvvspjjz1Gw4YN3Q7LhLjt27dTsGBBypQpYz1Vg6qyf/9+tm/fTtmyZbN0DEsKPliyZAnx8fGsXbuWzp0707lzZ7dDMgaAEydOWEIwXiJCkSJF2Lt3b5aPYRPNmRgxYgT16tXjr7/+YuHChcyYMYMiRYq4HZYxXpYQTGqX+vNgSSEDZ8+eBaBu3br06NGD1atX06pVK5ejMsYY/7KkkMbBgweJj4+nd+/eANSrV4+EhASuuuoqlyMzJncKCwujWrVqxMbG0rp1aw4ePOh9bc2aNdxyyy1UrFiRChUq8OKLL55TA+zTTz8lLi6OqKgoKleuTN++fdP9jA8//JBBgwb5/VyySlV59NFHKV++PFWrVmXFihXptpszZw5Vq1YlJiaGJ5980vv81q1badKkCdWrV6dq1aosXLgQgC+++IKaNWtSpUoVatasyX/+8x/ve5o1a8aBAwf8czKB9KhZs6Zmxd0TftC7J/xwwTbz5s3T66+/XsPCwrR///569uzZLH2WMTll7dq1boegV1xxhXe7S5cuOnjwYFVVPXbsmEZGRuqiRYtUVfXo0aPasmVLff3111VV9ddff9XIyEhdt26dqqomJyfruHHj0v2MunXr6t69e32OKTk5OUvnklWffPKJtmzZUs+ePauLFy/WWrVqnddm3759WrJkSd2zZ4+qer5WX375paqqduvWTRMSElRVdc2aNVq6dGlVVV2xYoXu2LFDVT1frxIlSniPN23aNO/XOq30fi6AZerD71ibaAb27NlDr169eO+996hWrRoff/wxNWrUcDssYy7KCx+tYe3Ow9l6zOgSV/F86xif29etW5dVq1YBMGvWLOrXr0/z5s0BuPzyy3n99ddp3LgxDz/8MMOGDWPAgAFUrlwZgPDwcHr27HneMTds2EC+fPm8i1F99NFHDB48mFOnTlGkSBHefvttihcvzsCBA9m5cyebN2+maNGizJw5k379+vHNN99w8uRJHn74YR588EGOHDlC27ZtOXDgAMnJyQwePJi2bdte0tdp/vz5dOnSBRGhTp06HDx4kF27dnH99dd72yQlJVGxYkWKFSsGeP7Sf//992natCkiwuHDnu/doUOHvJe5V69e3fv+mJgYTpw4wcmTJ8mXLx9t2rShYcOGDBgw4JJiT8uSAnD48GG++OILXnrpJZ544oksX99rTCg7c+YMX331FfHx8YBn6KhmzZrntClXrhxHjhzh8OHDrF69mscffzzT437//ffn/JHWoEEDlixZgogwZcoUhg0bxogRIwBYvnw53333HQUKFGDSpEkUKlSIn376iZMnT3oTVMmSJZk3bx5XXXUV+/bto06dOrRp0+a8CdoOHTqwfv368+Lp06cPXbp0Oee5HTt2ULJkSe9+REQEO3bsOCcplC9fnt9++43NmzcTERHBhx9+yKlTpwAYOHAgzZs3Z+zYsRw9epQvv/zyvM99//33qV69Ovny5QOgcOHCnDx5kv3792frxS8hmxS2bt3KzJkzefrppylfvjxbt26lYMGCbodlTJZdzF/02en48eNUq1aNzZs3U7NmTW699VbAMzSd0ZUwF3OFzK5du7x/XYPn3owOHTqwa9cuTp06dc71+G3atKFAgQIAfP7556xatYq5c+cCnr/AN27cSEREBE8//TTffvstefLkYceOHfzxxx9cd91153zunDlzfI5R01krJe05Fi5cmPHjx9OhQwfy5MlDvXr1SEpKAmD27Nk88MADPP744yxevJjOnTuzevVq8uTxTPuuWbOGp556is8///ycY1577bXs3LkzW5OCXyeaRaSliKwXkUQR6ZfO6/lEZI7z+lIRKePPeMBzVVFCQgIxMTEMGTLEW8DOEoIxWVOgQAFWrlzJli1bOHXqFOPGjQM8wx3Lli07p21SUhJXXnklBQsWJCYmhuXLl/t0/NR3bD/yyCP06tWLX3/9lYkTJ57z2hVXXOHdVlXGjh3LypUrWblyJZs2baJ58+a8/fbb7N27l+XLl7Ny5UqKFy+e7h3hHTp0oFq1auc9ZsyYcV7biIgItm3b5t3fvn17upUOWrduzdKlS1m8eDGVKlWiQoUKAEydOpW7774b8AzBnThxgn379nmP1a5dO2bMmEG5cuXOOd6JEye8STC7+C0piEgYMA5oBUQD94hIdJpm8cABVS0PjAKG+isegOPHj3nHM+vWrcuaNWusgJ0x2aRQoUKMGTOG4cOHk5ycTKdOnfjuu++8QyHHjx/n0Ucf9V5188QTTzBkyBA2bNgAeP5gGzly5HnHjYqKIjEx0bt/6NAhbrjhBgCmT5+eYTwtWrRg/PjxJCcnA565iaNHj3Lo0CGuvfZa8ubNy9dff82WLVvSff+cOXO8CSX1I+3QEXh6KDNmzEBVWbJkCYUKFTpn6CjFnj17ADhw4AAJCQl07doVgFKlSvHVV18BnlI6J06coFixYhw8eJDbb7+dl19+mfr1659zLFVl9+7dlClTJsOvQVb4s6dQC0hU1SRVPQW8A6SdzWkLpHxX5wJNxU934qgqq1at4tdff+XNN99k0aJF2f7FNCbUVa9enRtvvJF33nmHAgUKMH/+fAYPHkylSpWoUqUKN910E7169QKgatWqjB49mnvuuYeoqChiY2PZtWvXecds1KgRP//8s3eIZuDAgbRv356GDRt6J5/T07VrV6Kjo6lRowaxsbE8+OCDnD59mk6dOrFs2TLi4uJ4++23vRPdl+K2224jMjKS8uXL061bNxISEryvVatWzbvdu3dvoqOjqV+/Pv369aNixYqA5ybZyZMnc+ONN3LPPfcwbdo0RITXX3+dxMREXnzxRW9PJSWxLF++nDp16hAenr2zAJLeWFi2HFjkn0BLVe3q7HcGaqtqr1RtVjtttjv7vztt9qU5VnegO0CpUqVqZpTZL+SFj9awc+dOBraJTTeDGxOI1q1bR1RUlNth+F3v3r1p3bo1zZo1czuUXKN37960adOGpk2bnvdaej8XIrJcVeMyO64/J5rT+4s/bQbypQ2qOgmYBBAXF5elLOaZhHNnIs4Yc2mefvppli5d6nYYuUpsbGy6CeFS+XP4aDtQMtV+BLAzozYiEg4UAv70Y0zGmABUvHhx2rRp43YYuUq3bt38clx/JoWfgAoiUlZELgM6AgvStFkA3O9s/xP4j/prPMuYIGX/ZUxql/rz4LekoKqngV7AImAd8K6qrhGRQSKSkvKnAkVEJBHoA5x32aoxJmP58+dn//79lhgM8Pd6Cvnz58/yMfw20ewvcXFxmvbaZ2NCla28ZtLKaOW13DDRbIzxs7x582Z5hS1j0mOls40xxnhZUjDGGONlScEYY4xXwE00i8he4OJvafYoCuzLtFVwsXMODXbOoeFSzrm0qhbLrFHAJYVLISLLfJl9DyZ2zqHBzjk05MQ52/CRMcYYL0sKxhhjvEItKUxyOwAX2DmHBjvn0OD3cw6pOQVjjDEXFmo9BWOMMRdgScEYY4xXUCYFEWkpIutFJFFEzqu8KiL5RGSO8/pSESmT81FmLx/OuY+IrBWRVSLylYiUdiPO7JTZOadq908RUREJ+MsXfTlnEbnb+V6vEZFZOR1jdvPhZ7uUiHwtIj87P9+3uRFndhGRN0Rkj7MyZXqvi4iMcb4eq0SkRrYGoKpB9QDCgN+BSOAy4BcgOk2bnsAEZ7sjMMftuHPgnJsAlzvbD4XCOTvtCgLfAkuAOLfjzoHvcwXgZ6Cws3+t23HnwDlPAh5ytqOBzW7HfYnn3AioAazO4PXbgE/xrFxZB1ianZ8fjD2FWkCiqiap6ingHaBtmjZtgenO9lygqYiktzRooMj0nFX1a1U95uwuwbMSXiDz5fsM8CIwDAiG2tK+nHM3YJyqHgBQ1T05HGN28+WcFbjK2S7E+Ss8BhRV/ZYLr0DZFpihHkuAq0Uk2xaeD8akcAOwLdX+due5dNuoZzGgQ0CRHInOP3w559Ti8fylEcgyPWcRqQ6UVNWPczIwP/Ll+1wRqCgi34vIEhFpmWPR+Ycv5zwQuE9EtgMLgUdyJjTXXOz/94sSjOsppPcXf9rrbn1pE0h8Ph8RuQ+IA272a0T+d8FzFpE8wCjggZwKKAf48n0OxzOE1BhPb/B/IhKrqgf9HJu/+HLO9wDTVHWEiNQFZjrnfNb/4bnCr7+/grGnsB0omWo/gvO7k942IhKOp8t5oe5abufLOSMizYABQBtVPZlDsflLZudcEIgFvhGRzXjGXhcE+GSzrz/b81U1WVU3AevxJIlA5cs5xwPvAqjqYiA/nsJxwcqn/+9ZFYxJ4SeggoiUFZHL8EwkL0jTZgFwv7P9T+A/6szgBKhMz9kZSpmIJyEE+jgzZHLOqnpIVYuqahlVLYNnHqWNqgbyWq6+/Gx/iOeiAkSkKJ7hpKQcjTJ7+XLOW4GmACIShScp7M3RKHPWAqCLcxVSHeCQqu7KroMH3fCRqp4WkV7AIjxXLryhqmtEZBCwTFUXAFPxdDET8fQQOroX8aXz8ZxfBa4E3nPm1LeqahvXgr5EPp5zUPHxnBcBzUVkLXAGeEJV97sX9aXx8ZwfByaLyL/xDKM8EMh/5InIbDzDf0WdeZLngbwAqjoBz7zJbUAicAz4V7Z+fgB/7YwxxmSzYBw+MsYYk0WWFIwxxnhZUjDGGONlScEYY4yXJQVjjDFelhRMriUiZ0RkZapHmQu0LZNRVcmcJiJxIjLG2W4sIvVSvdZDRLrkYCzVAr1qqMlZQXefggkqx1W1mttBXCznBrmUm+QaA0eAH5zXJmT354lIuFPDKz3V8JQ1WZjdn2uCk/UUTEBxegT/E5EVzqNeOm1iRORHp3exSkQqOM/fl+r5iSISls57N4vIUKfdjyJS3nm+tHjWoUhZj6KU83x7EVktIr+IyLfOc41F5GOnZ9MD+LfzmQ1FZKCI9BWRKBH5Mc15rXK2a4rIf0VkuYgsSq8CpohME5GRIvI1MFREaonID+JZU+AHEank3AE8COjgfH4HEblCPPX6f3LapldZ1oQyt2uH28MeGT3w3JG70nnMc567HMjvbFfAc1crQBmc+vPAWKCTs30ZUACIAj4C8jrPJwBd0vnMzcAAZ7sL8LGz/RFwv7P9f8CHzvavwA3O9tXOv41TvW8g0DfV8b37znlFOttPAc/guXP1B6CY83wHPHfxpo1zGvAxEObsXwWEO9vNgPed7QeA11O9bwhwX0q8wAbgCre/1/bIPQ8bPjK5WXrDR3mB10WkGp6kUTGd9y0GBohIBPCBqm4UkaZATeAnp8xHASCjGlCzU/07ytmuC/zD2Z6JZ40GgO+BaSLyLvDBxZwcniJudwOv4Pnl3wGohKeQ3xdOnGFARnVt3lPVM852IWC60ytSnLII6WgOtBGRvs5+fqAUsO4iYzdBypKCCTT/Bv4AbsQz/Hne4jmqOktElgK3A4tEpCuecsPTVbW/D5+hGWyf10ZVe4hIbeezVjrJyldz8NSi+sBzKN0oIlWANapa14f3H021/SLwtaq2c4atvsngPQLcparrLyJOE0JsTsEEmkLALvXUyu+M5y/pc4hIJJCkqmPwVJSsCnwF/FNErnXaXCMZr1PdIdW/i53tH/i7cGIn4DvnOOVUdamqPgfs49ySxgB/4SnjfR5V/R1Pb+dZPAkCPKWui4lnXQBEJK+IxGQQZ2qFgB3O9gMX+PxFwCPidEPEUz3XGC9LCibQJAD3i8gSPENHR9Np0wFYLSIrgcp4li5ci2fM/nNnQvcLIKMlDPM5PY3eeHomAI8C/3Le29l5DeBVEfnVuRz2WzxrCKf2EdAuZaI5nc+aA9zH3+sBnMJTzn2oiPyCZ97hvMn0dAwDXhaR7zk3UX4NRKdMNOPpUeQFVjkxv+jDsU0IsSqpxqQingV54lR1n9uxGOMG6ykYY4zxsp6CMcYYL+spGGOM8bKkYIwxxsuSgjHGGC9LCsYYY7wsKRhjjPH6fxuTu1cJY5JHAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(1)\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.plot(fpr_keras, tpr_keras, label='ROC (area = {:.3f})'.format(auc_keras))\n",
    "plt.xlabel('False positive rate')\n",
    "plt.ylabel('True positive rate')\n",
    "plt.title('ROC curve')\n",
    "plt.legend(loc='best')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute ROC for various SNRs\n",
    "\n",
    "num_samples = sirenNoiseVal.shape[0]\n",
    "\n",
    "from sklearn.metrics import roc_curve\n",
    "from numpy import ravel\n",
    "from sklearn.metrics import auc\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "num_rows = 40\n",
    "num_columns = 130 #x_train.shape[2]\n",
    "num_channels = 1\n",
    "\n",
    "def getAUC(sirenSNR):\n",
    "\n",
    "    num_samples = sirenNoiseVal.shape[0]\n",
    "    X, yy = generateData(carNoiseVal, backNoiseVal, sirenNoiseVal, sirenSNR)\n",
    "\n",
    "    X = X.reshape(X.shape[0], num_rows, num_columns, num_channels)\n",
    "\n",
    "    y_pred_keras = model.predict_proba(X)\n",
    "    predicted_proba = y_pred_keras[:,1]\n",
    "    desc_score_indices = np.argsort(predicted_proba, kind=\"mergesort\")[::-1]\n",
    "    y_pred = predicted_proba\n",
    "\n",
    "    y_test2 = yy[:,1]\n",
    "\n",
    "    y_pred_keras = model.predict_proba(X)\n",
    "\n",
    "    fpr_keras, tpr_keras, thresholds_keras = roc_curve(y_test2, y_pred)\n",
    "\n",
    "    auc_keras = auc(fpr_keras, tpr_keras)\n",
    "    \n",
    "    print('AUC for SNR of ', sirenSNR, 'is ', auc_keras)\n",
    "\n",
    "    return auc_keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SNR of 0 dB\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.414706\n",
      "0.9759622562632506\n"
     ]
    }
   ],
   "source": [
    "#SNRs = range(-20, 30)\n",
    "#AUCs = np.empty(len(SNRs))\n",
    "#for i in range(0, len(SNRs)):\n",
    "#    AUCs[i] = SNRs[i]\n",
    "    \n",
    "#print(AUCs)\n",
    "AUCs = getAUC(0)\n",
    "print(AUCs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Uniform SNR with SNR =  -20\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.366896\n",
      "AUC for SNR of  -20 is  0.7723431574546866\n",
      "Uniform SNR with SNR =  -19\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.503924\n",
      "AUC for SNR of  -19 is  0.7396458538624623\n",
      "Uniform SNR with SNR =  -18\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:22.376027\n",
      "AUC for SNR of  -18 is  0.7070082713726895\n",
      "Uniform SNR with SNR =  -17\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:22.823460\n",
      "AUC for SNR of  -17 is  0.7955746663083398\n",
      "Uniform SNR with SNR =  -16\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:20.530934\n",
      "AUC for SNR of  -16 is  0.7985905819821434\n",
      "Uniform SNR with SNR =  -15\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:24.074369\n",
      "AUC for SNR of  -15 is  0.787064409208994\n",
      "Uniform SNR with SNR =  -14\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:19.216662\n",
      "AUC for SNR of  -14 is  0.8015467765534952\n",
      "Uniform SNR with SNR =  -13\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.197814\n",
      "AUC for SNR of  -13 is  0.8359759921168144\n",
      "Uniform SNR with SNR =  -12\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.097764\n",
      "AUC for SNR of  -12 is  0.8045029711248469\n",
      "Uniform SNR with SNR =  -11\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.643236\n",
      "AUC for SNR of  -11 is  0.8220908357968287\n",
      "Uniform SNR with SNR =  -10\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.542187\n",
      "AUC for SNR of  -10 is  0.8613873212099494\n",
      "Uniform SNR with SNR =  -9\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.374965\n",
      "AUC for SNR of  -9 is  0.8853653438442473\n",
      "Uniform SNR with SNR =  -8\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.839332\n",
      "AUC for SNR of  -8 is  0.8967720744124936\n",
      "Uniform SNR with SNR =  -7\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.157368\n",
      "AUC for SNR of  -7 is  0.8873958613276001\n",
      "Uniform SNR with SNR =  -6\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.866152\n",
      "AUC for SNR of  -6 is  0.9257965302039476\n",
      "Uniform SNR with SNR =  -5\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:19.159923\n",
      "AUC for SNR of  -5 is  0.940637224163158\n",
      "Uniform SNR with SNR =  -4\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:18.578169\n",
      "AUC for SNR of  -4 is  0.9310818477709097\n",
      "Uniform SNR with SNR =  -3\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.596731\n",
      "AUC for SNR of  -3 is  0.9699304249156439\n",
      "Uniform SNR with SNR =  -2\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.608075\n",
      "AUC for SNR of  -2 is  0.9887127116366569\n",
      "Uniform SNR with SNR =  -1\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.172527\n",
      "AUC for SNR of  -1 is  0.9868613574606587\n",
      "SNR of 0 dB\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.920038\n",
      "AUC for SNR of  0 is  0.974708113111768\n",
      "SNR of 0 dB\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.998042\n",
      "AUC for SNR of  1 is  0.9731852249992534\n",
      "Uniform SNR with SNR =  2\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.015723\n",
      "AUC for SNR of  2 is  0.9436830003881872\n",
      "Uniform SNR with SNR =  3\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.485482\n",
      "AUC for SNR of  3 is  0.9862342858849175\n",
      "Uniform SNR with SNR =  4\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.666026\n",
      "AUC for SNR of  4 is  0.9827406013915018\n",
      "Uniform SNR with SNR =  5\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.004040\n",
      "AUC for SNR of  5 is  0.9875780106900773\n",
      "Uniform SNR with SNR =  6\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.326362\n",
      "AUC for SNR of  6 is  0.9884439666756248\n",
      "Uniform SNR with SNR =  7\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.107210\n",
      "AUC for SNR of  7 is  0.9981785063752276\n",
      "Uniform SNR with SNR =  8\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.270901\n",
      "AUC for SNR of  8 is  0.9933112365254263\n",
      "Uniform SNR with SNR =  9\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.083612\n",
      "AUC for SNR of  9 is  0.9923855594374272\n",
      "Uniform SNR with SNR =  10\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.342959\n",
      "AUC for SNR of  10 is  0.9948042640867151\n",
      "Uniform SNR with SNR =  11\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.930982\n",
      "AUC for SNR of  11 is  0.9875780106900773\n",
      "Uniform SNR with SNR =  12\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.500373\n",
      "AUC for SNR of  12 is  0.9940577503060707\n",
      "Uniform SNR with SNR =  13\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.186214\n",
      "AUC for SNR of  13 is  0.9937591447938129\n",
      "Uniform SNR with SNR =  14\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.683251\n",
      "AUC for SNR of  14 is  0.9958493833796171\n",
      "Uniform SNR with SNR =  15\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.507870\n",
      "AUC for SNR of  15 is  0.990146018095494\n",
      "Uniform SNR with SNR =  16\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.694448\n",
      "AUC for SNR of  16 is  0.9913703006957508\n",
      "Uniform SNR with SNR =  17\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.513474\n",
      "AUC for SNR of  17 is  0.9931022126668457\n",
      "Uniform SNR with SNR =  18\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.477897\n",
      "AUC for SNR of  18 is  0.9927737466033624\n",
      "Uniform SNR with SNR =  19\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.623589\n",
      "AUC for SNR of  19 is  0.9907730896712352\n",
      "Uniform SNR with SNR =  20\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.430782\n",
      "AUC for SNR of  20 is  0.9953118934575532\n",
      "Uniform SNR with SNR =  21\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.601710\n",
      "AUC for SNR of  21 is  0.9935501209352325\n",
      "Uniform SNR with SNR =  22\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:15.763278\n",
      "AUC for SNR of  22 is  0.9932813759742005\n",
      "Uniform SNR with SNR =  23\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.475283\n",
      "AUC for SNR of  23 is  0.991609185105557\n",
      "Uniform SNR with SNR =  24\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.217638\n",
      "AUC for SNR of  24 is  0.9908626713249126\n",
      "Uniform SNR with SNR =  25\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.226048\n",
      "AUC for SNR of  25 is  0.9928633282570396\n",
      "Uniform SNR with SNR =  26\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.140333\n",
      "AUC for SNR of  26 is  0.9922661172325241\n",
      "Uniform SNR with SNR =  27\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generating training and test data took:  0:00:17.210060\n",
      "AUC for SNR of  27 is  0.9945653796769088\n",
      "Uniform SNR with SNR =  28\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:17.197408\n",
      "AUC for SNR of  28 is  0.9934306787303293\n",
      "Uniform SNR with SNR =  29\n",
      "Finished scrambling data set\n",
      "Extracting MFCCs...\n",
      "Generating training and test data took:  0:00:16.977812\n",
      "AUC for SNR of  29 is  0.9918779300665891\n"
     ]
    }
   ],
   "source": [
    "SNRs = range(-20, 30)\n",
    "AUCs = np.empty(len(SNRs))\n",
    "for i in range(0, len(SNRs)):\n",
    "    AUCs[i] = getAUC(SNRs[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "    # plot\n",
    "    plt.figure(1)\n",
    "    #plt.plot([0, 1], [0, 1], 'k--')\n",
    "    plt.plot(SNRs, AUCs)\n",
    "    plt.xlabel('Signal-to-Noise Ratio (dB)')\n",
    "    plt.ylabel('Area Under Curve (AUC)')\n",
    "    plt.title('Area Under the ROC Curve as a Function of Siren SNR')\n",
    "    #plt.legend(loc='best')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[10, 11, 12, 13, 14, 15, 16, 17, 18, 19]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sirenSet.shape\n",
    "#sirenSet[:,:RATE*amountPlay].shapeS\n",
    "list(range(10,20))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Test various sounds\n",
    "\n",
    "import sounddevice as sd\n",
    "import time\n",
    "startPlay = 0 # Where to start index\n",
    "amountPlay = 5 # seconds\n",
    "for i in list(range(startPlay, amountPlay+startPlay)):\n",
    "    sd.play(sirenSet[i,:], RATE)\n",
    "    time.sleep(3) "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}