3 Multi Layer Perceptron.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classifying Urban sounds using Deep Learning\n",
    "\n",
    "## 3 Model Training and Evaluation "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load Preprocessed data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# retrieve the preprocessed data from previous notebook\n",
    "\n",
    "%store -r x_train \n",
    "%store -r x_test \n",
    "%store -r y_train \n",
    "%store -r y_test \n",
    "%store -r yy \n",
    "%store -r le"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       ...,\n",
       "       [0., 0., 1.],\n",
       "       [0., 0., 1.],\n",
       "       [0., 1., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1., 0., 0.],\n",
       "       [0., 0., 1.],\n",
       "       [0., 0., 1.],\n",
       "       ...,\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 0., 1.]], dtype=float32)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def add_noise(data,x):\n",
    "    noise = np.random.randn(len(data))\n",
    "    data_noise = data + x * noise\n",
    "    return data_noise\n",
    "\n",
    "def shift(data,x):\n",
    "    return np.roll(data, x)\n",
    "\n",
    "def stretch(data, rate):\n",
    "    data = librosa.effects.time_stretch(data, rate)\n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import librosa\n",
    "import pandas as pd\n",
    "import os\n",
    "metadata = pd.read_csv('./metadata_compiled.csv')\n",
    "X_=[]\n",
    "y_=[]\n",
    "#dir_= './clinical/converted_seg/cneg/'\n",
    "for index, row in metadata.iterrows():\n",
    "    if row[\"status\"] == \"healthy\":\n",
    "        file_name = os.path.join(os.path.abspath('./improved_dataset'),str(row[\"uuid\"]+\".wav\"))\n",
    "        # for soundDir in (os.listdir(dir_)):\n",
    "        data_x, sampling_rate = librosa.load(file_name,res_type='kaiser_fast')\n",
    "        mfccs = np.mean(librosa.feature.mfcc(y=data_x, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs)\n",
    "        y_.append('healthy')\n",
    "    else:\n",
    "        p='COVID-19'\n",
    "        \n",
    "        file_name = os.path.join(os.path.abspath('./improved_dataset'),str(row[\"uuid\"]+\".wav\"))\n",
    "        # for soundDir in (os.listdir(dir_)):\n",
    "        data_x, sampling_rate = librosa.load(file_name,res_type='kaiser_fast')\n",
    "        mfccs = np.mean(librosa.feature.mfcc(y=data_x, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs)\n",
    "        y_.append(p)\n",
    "\n",
    "        data_noise = add_noise(data_x,0.005)\n",
    "        mfccs_noise = np.mean(librosa.feature.mfcc(y=data_noise, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs_noise)\n",
    "        y_.append(p)\n",
    "\n",
    "        data_shift = shift(data_x,1600)\n",
    "        mfccs_shift = np.mean(librosa.feature.mfcc(y=data_shift, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs_shift)\n",
    "        y_.append(p)\n",
    "\n",
    "        data_stretch = stretch(data_x,1.2)\n",
    "        mfccs_stretch = np.mean(librosa.feature.mfcc(y=data_stretch, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs_stretch)\n",
    "        y_.append(p)\n",
    "\n",
    "        data_stretch_2 = stretch(data_x,0.8)\n",
    "        mfccs_stretch_2 = np.mean(librosa.feature.mfcc(y=data_stretch_2, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "        X_.append(mfccs_stretch_2)\n",
    "        y_.append(p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import numpy as np\n",
    "# import librosa\n",
    "# import os\n",
    "\n",
    "# dir_= './clinical/converted_seg/cpos/'\n",
    "# p='positive'\n",
    "# for soundDir in (os.listdir(dir_)):\n",
    "#     data_x, sampling_rate = librosa.load(dir_+soundDir,res_type='kaiser_fast')\n",
    "#     mfccs = np.mean(librosa.feature.mfcc(y=data_x, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "#     X_.append(mfccs)\n",
    "#     y_.append('positive')\n",
    "\n",
    "#     data_noise = add_noise(data_x,0.005)\n",
    "#     mfccs_noise = np.mean(librosa.feature.mfcc(y=data_noise, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "#     X_.append(mfccs_noise)\n",
    "#     y_.append(p)\n",
    "\n",
    "#     data_shift = shift(data_x,1600)\n",
    "#     mfccs_shift = np.mean(librosa.feature.mfcc(y=data_shift, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "#     X_.append(mfccs_shift)\n",
    "#     y_.append(p)\n",
    "\n",
    "#     data_stretch = stretch(data_x,1.2)\n",
    "#     mfccs_stretch = np.mean(librosa.feature.mfcc(y=data_stretch, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "#     X_.append(mfccs_stretch)\n",
    "#     y_.append(p)\n",
    "\n",
    "#     data_stretch_2 = stretch(data_x,0.8)\n",
    "#     mfccs_stretch_2 = np.mean(librosa.feature.mfcc(y=data_stretch_2, sr=sampling_rate, n_mfcc=40).T,axis=0) \n",
    "#     X_.append(mfccs_stretch_2)\n",
    "#     y_.append(p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = np.array(X_)\n",
    "y = np.array(y_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-5.63286072e+02  1.82369003e+01 -2.34143949e+00 ...  2.08231524e-01\n",
      "  -3.72024238e-01  1.91061288e-01]\n",
      " [-7.36395081e+02  5.60351982e+01  9.28143120e+00 ... -1.26430750e-01\n",
      "   8.30247402e-01  9.10993099e-01]\n",
      " [-2.79818207e+02  1.02286911e+02 -2.72372284e+01 ... -3.43401885e+00\n",
      "  -2.21963358e+00 -1.87745678e+00]\n",
      " ...\n",
      " [-4.77796448e+02  3.58470688e+01 -5.85082474e+01 ... -2.01891646e-01\n",
      "  -2.20986962e-01  2.13427007e-01]\n",
      " [-4.71484192e+02  3.25430870e+01 -5.61231918e+01 ... -5.69744349e-01\n",
      "  -4.61468399e-01 -6.55625939e-01]\n",
      " [-2.65285828e+02  1.31167923e+02 -4.36617393e+01 ...  1.49560201e+00\n",
      "   9.85208750e-01 -8.23351368e-02]]\n",
      "['healthy' 'healthy' 'healthy' ... 'COVID-19' 'COVID-19' 'healthy']\n"
     ]
    }
   ],
   "source": [
    "print(X)\n",
    "print(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0. 1.]\n",
      " [0. 1.]\n",
      " [0. 1.]\n",
      " ...\n",
      " [1. 0.]\n",
      " [1. 0.]\n",
      " [0. 1.]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.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",
    "print(yy)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split the dataset \n",
    "from sklearn.model_selection import train_test_split \n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(X, yy, test_size=0.3, random_state = 42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stored 'x_train' (ndarray)\n",
      "Stored 'x_test' (ndarray)\n",
      "Stored 'y_train' (ndarray)\n",
      "Stored 'y_test' (ndarray)\n",
      "Stored 'yy' (ndarray)\n",
      "Stored 'le' (LabelEncoder)\n"
     ]
    }
   ],
   "source": [
    "%store x_train \n",
    "%store x_test \n",
    "%store y_train \n",
    "%store y_test \n",
    "%store yy \n",
    "%store le"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "%store -r x_train \n",
    "%store -r x_test \n",
    "%store -r y_train \n",
    "%store -r y_test \n",
    "%store -r yy \n",
    "%store -r le"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Initial model architecture - MLP\n",
    "\n",
    "We will start with constructing a Multilayer Perceptron (MLP) Neural Network using Keras and a Tensorflow backend. \n",
    "\n",
    "Starting with a `sequential` model so we can build the model layer by layer. \n",
    "\n",
    "We will begin with a simple model architecture, consisting of three layers, an input layer, a hidden layer and an output layer. All three layers will be of the `dense` layer type which is a standard layer type that is used in many cases for neural networks. \n",
    "\n",
    "The first layer will receive the input shape. As each sample contains 40 MFCCs (or columns) we have a shape of (1x40) this means we will start with an input shape of 40. \n",
    "\n",
    "The first two layers will have 256 nodes. The activation function we will be using for our first 2 layers is the `ReLU`, or `Rectified Linear Activation`. This activation function has been proven to work well in neural networks.\n",
    "\n",
    "We will also apply a `Dropout` value of 50% on our first two layers. This will randomly exclude nodes from each update cycle which in turn results in a network that is capable of better generalisation and is less likely to overfit the training data.\n",
    "\n",
    "Our output layer will have 2 nodes (num_labels) which matches the number of possible classifications. The activation is for our output layer is `softmax`. Softmax makes the output sum up to 1 so the output can be interpreted as probabilities. The model will then make its prediction based on which option has the highest probability."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    }
   ],
   "source": [
    "print(yy.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21842\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential,load_model\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers import Convolution2D, MaxPooling1D\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from keras.utils import np_utils\n",
    "from sklearn import metrics \n",
    "\n",
    "\n",
    "num_labels = yy.shape[1]\n",
    "filter_size = 1\n",
    "\n",
    "\n",
    "# Construct model \n",
    "model = Sequential()\n",
    "\n",
    "model.add(Dense(256, input_shape=(40,)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(128))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(128))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "# model.add(MaxPooling2D(pool_size=1))\n",
    "model.add(Dropout(0.20))\n",
    "\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "# model.add(MaxPooling1D(pool_size=1))\n",
    "model.add(Dropout(0.20))\n",
    "\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "# model.add(MaxPooling2D(pool_size=1))\n",
    "model.add(Dropout(0.20))\n",
    "\n",
    "\n",
    "model.add(Dense(num_labels))\n",
    "model.add(Activation('softmax'))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compiling the model \n",
    "\n",
    "For compiling our model, we will use the following three parameters: \n",
    "\n",
    "* Loss function - we will use `categorical_crossentropy`. This is the most common choice for classification. A lower score indicates that the model is performing better.\n",
    "\n",
    "* Metrics - we will use the `accuracy` metric which will allow us to view the accuracy score on the validation data when we train the model. \n",
    "\n",
    "* Optimizer - here we will use `adam` which is a generally good optimizer for many use cases.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_20 (Dense)             (None, 256)               10496     \n",
      "_________________________________________________________________\n",
      "activation_20 (Activation)   (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dropout_18 (Dropout)         (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_21 (Dense)             (None, 128)               32896     \n",
      "_________________________________________________________________\n",
      "activation_21 (Activation)   (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_19 (Dropout)         (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_22 (Dense)             (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "activation_22 (Activation)   (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_20 (Dropout)         (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_23 (Dense)             (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "activation_23 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_21 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_24 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "activation_24 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_22 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_25 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "activation_25 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_23 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_26 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "activation_26 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_24 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_27 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "activation_27 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_25 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_28 (Dense)             (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "activation_28 (Activation)   (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_26 (Dropout)         (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_29 (Dense)             (None, 2)                 130       \n",
      "_________________________________________________________________\n",
      "activation_29 (Activation)   (None, 2)                 0         \n",
      "=================================================================\n",
      "Total params: 89,090\n",
      "Trainable params: 89,090\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Pre-training accuracy: 59.4531%\n"
     ]
    }
   ],
   "source": [
    "# Display model architecture summary \n",
    "model.summary()\n",
    "\n",
    "# Calculate pre-training accuracy \n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "print(\"Pre-training accuracy: %.4f%%\" % accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training \n",
    "\n",
    "Here we will train the model. \n",
    "\n",
    "We will start with 100 epochs which is the number of times the model will cycle through the data. The model will improve on each cycle until it reaches a certain point. \n",
    "\n",
    "We will also start with a low batch size, as having a large batch size can reduce the generalisation ability of the model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.7258 - accuracy: 0.5793 - val_loss: 0.6753 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.67527, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 2/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6740 - accuracy: 0.5983 - val_loss: 0.6755 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 0.67527\n",
      "Epoch 3/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6731 - accuracy: 0.5998 - val_loss: 0.6734 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.67527 to 0.67337, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 4/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6722 - accuracy: 0.6008 - val_loss: 0.6758 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.67337\n",
      "Epoch 5/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6702 - accuracy: 0.5998 - val_loss: 0.6688 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00005: val_loss improved from 0.67337 to 0.66880, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 6/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6703 - accuracy: 0.6012 - val_loss: 0.6674 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00006: val_loss improved from 0.66880 to 0.66742, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 7/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6704 - accuracy: 0.6020 - val_loss: 0.6642 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.66742 to 0.66416, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 8/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6681 - accuracy: 0.6015 - val_loss: 0.6833 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.66416\n",
      "Epoch 9/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6654 - accuracy: 0.6018 - val_loss: 0.6609 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.66416 to 0.66090, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 10/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6610 - accuracy: 0.6010 - val_loss: 0.6998 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.66090\n",
      "Epoch 11/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6542 - accuracy: 0.6010 - val_loss: 0.6545 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.66090 to 0.65453, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 12/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6486 - accuracy: 0.6013 - val_loss: 0.6345 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00012: val_loss improved from 0.65453 to 0.63452, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 13/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6424 - accuracy: 0.6015 - val_loss: 0.6698 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00013: val_loss did not improve from 0.63452\n",
      "Epoch 14/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6408 - accuracy: 0.6027 - val_loss: 0.6249 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00014: val_loss improved from 0.63452 to 0.62487, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 15/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6378 - accuracy: 0.6018 - val_loss: 0.6246 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.62487 to 0.62461, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 16/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6354 - accuracy: 0.6023 - val_loss: 0.6229 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00016: val_loss improved from 0.62461 to 0.62292, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 17/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6316 - accuracy: 0.6021 - val_loss: 0.6616 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.62292\n",
      "Epoch 18/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6321 - accuracy: 0.6019 - val_loss: 0.6219 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00018: val_loss improved from 0.62292 to 0.62193, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 19/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6323 - accuracy: 0.6020 - val_loss: 0.6902 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00019: val_loss did not improve from 0.62193\n",
      "Epoch 20/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6324 - accuracy: 0.6025 - val_loss: 0.6343 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00020: val_loss did not improve from 0.62193\n",
      "Epoch 21/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6275 - accuracy: 0.6023 - val_loss: 0.6210 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00021: val_loss improved from 0.62193 to 0.62104, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 22/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6280 - accuracy: 0.6025 - val_loss: 0.6198 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00022: val_loss improved from 0.62104 to 0.61977, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 23/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6282 - accuracy: 0.6027 - val_loss: 0.6198 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.61977\n",
      "Epoch 24/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6278 - accuracy: 0.6025 - val_loss: 0.6193 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00024: val_loss improved from 0.61977 to 0.61925, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 25/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6252 - accuracy: 0.6022 - val_loss: 0.6216 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.61925\n",
      "Epoch 26/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6253 - accuracy: 0.6025 - val_loss: 0.6171 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00026: val_loss improved from 0.61925 to 0.61708, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 27/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6245 - accuracy: 0.6024 - val_loss: 0.6257 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00027: val_loss did not improve from 0.61708\n",
      "Epoch 28/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6270 - accuracy: 0.6024 - val_loss: 0.6166 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.61708 to 0.61663, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 29/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6308 - accuracy: 0.6020 - val_loss: 0.6195 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00029: val_loss did not improve from 0.61663\n",
      "Epoch 30/130\n",
      "683/683 [==============================] - 2s 3ms/step - loss: 0.6277 - accuracy: 0.6023 - val_loss: 0.6198 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00030: val_loss did not improve from 0.61663\n",
      "Epoch 31/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6270 - accuracy: 0.6023 - val_loss: 0.6177 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00031: val_loss did not improve from 0.61663\n",
      "Epoch 32/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6278 - accuracy: 0.6025 - val_loss: 0.6186 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00032: val_loss did not improve from 0.61663\n",
      "Epoch 33/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6236 - accuracy: 0.6022 - val_loss: 0.6166 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00033: val_loss improved from 0.61663 to 0.61660, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 34/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6262 - accuracy: 0.6024 - val_loss: 0.6400 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00034: val_loss did not improve from 0.61660\n",
      "Epoch 35/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6272 - accuracy: 0.6020 - val_loss: 0.6245 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00035: val_loss did not improve from 0.61660\n",
      "Epoch 36/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6223 - accuracy: 0.6023 - val_loss: 0.6254 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00036: val_loss did not improve from 0.61660\n",
      "Epoch 37/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6239 - accuracy: 0.6024 - val_loss: 0.6260 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00037: val_loss did not improve from 0.61660\n",
      "Epoch 38/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6199 - accuracy: 0.6023 - val_loss: 0.6156 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00038: val_loss improved from 0.61660 to 0.61563, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 39/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6247 - accuracy: 0.6022 - val_loss: 0.6351 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00039: val_loss did not improve from 0.61563\n",
      "Epoch 40/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6220 - accuracy: 0.6023 - val_loss: 0.6134 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00040: val_loss improved from 0.61563 to 0.61340, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 41/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6229 - accuracy: 0.6023 - val_loss: 0.6348 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00041: val_loss did not improve from 0.61340\n",
      "Epoch 42/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6254 - accuracy: 0.6025 - val_loss: 0.6150 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00042: val_loss did not improve from 0.61340\n",
      "Epoch 43/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6230 - accuracy: 0.6024 - val_loss: 0.6162 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00043: val_loss did not improve from 0.61340\n",
      "Epoch 44/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6227 - accuracy: 0.6021 - val_loss: 0.6214 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00044: val_loss did not improve from 0.61340\n",
      "Epoch 45/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6198 - accuracy: 0.6020 - val_loss: 0.6212 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00045: val_loss did not improve from 0.61340\n",
      "Epoch 46/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6201 - accuracy: 0.6024 - val_loss: 0.6149 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00046: val_loss did not improve from 0.61340\n",
      "Epoch 47/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6219 - accuracy: 0.6023 - val_loss: 0.6138 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00047: val_loss did not improve from 0.61340\n",
      "Epoch 48/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6252 - accuracy: 0.6023 - val_loss: 0.6158 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00048: val_loss did not improve from 0.61340\n",
      "Epoch 49/130\n",
      "683/683 [==============================] - 2s 4ms/step - loss: 0.6227 - accuracy: 0.6025 - val_loss: 0.6246 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00049: val_loss did not improve from 0.61340\n",
      "Epoch 50/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6198 - accuracy: 0.6025 - val_loss: 0.6209 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00050: val_loss did not improve from 0.61340\n",
      "Epoch 51/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6190 - accuracy: 0.6023 - val_loss: 0.6190 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00051: val_loss did not improve from 0.61340\n",
      "Epoch 52/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6236 - accuracy: 0.6024 - val_loss: 0.6626 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00052: val_loss did not improve from 0.61340\n",
      "Epoch 53/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6239 - accuracy: 0.6024 - val_loss: 0.6148 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00053: val_loss did not improve from 0.61340\n",
      "Epoch 54/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6248 - accuracy: 0.6026 - val_loss: 0.6165 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00054: val_loss did not improve from 0.61340\n",
      "Epoch 55/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6189 - accuracy: 0.6024 - val_loss: 0.6127 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00055: val_loss improved from 0.61340 to 0.61274, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 56/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6198 - accuracy: 0.6025 - val_loss: 0.6247 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00056: val_loss did not improve from 0.61274\n",
      "Epoch 57/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6183 - accuracy: 0.6025 - val_loss: 0.6138 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00057: val_loss did not improve from 0.61274\n",
      "Epoch 58/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6190 - accuracy: 0.6022 - val_loss: 0.6188 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00058: val_loss did not improve from 0.61274\n",
      "Epoch 59/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6175 - accuracy: 0.6025 - val_loss: 0.6132 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00059: val_loss did not improve from 0.61274\n",
      "Epoch 60/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6199 - accuracy: 0.6025 - val_loss: 0.6152 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00060: val_loss did not improve from 0.61274\n",
      "Epoch 61/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6193 - accuracy: 0.6025 - val_loss: 0.6137 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00061: val_loss did not improve from 0.61274\n",
      "Epoch 62/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6206 - accuracy: 0.6025 - val_loss: 0.6176 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00062: val_loss did not improve from 0.61274\n",
      "Epoch 63/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6182 - accuracy: 0.6024 - val_loss: 0.6419 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00063: val_loss did not improve from 0.61274\n",
      "Epoch 64/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6277 - accuracy: 0.6025 - val_loss: 0.6152 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00064: val_loss did not improve from 0.61274\n",
      "Epoch 65/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6242 - accuracy: 0.6024 - val_loss: 0.6149 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00065: val_loss did not improve from 0.61274\n",
      "Epoch 66/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6193 - accuracy: 0.6024 - val_loss: 0.6148 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00066: val_loss did not improve from 0.61274\n",
      "Epoch 67/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6218 - accuracy: 0.6025 - val_loss: 0.6114 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00067: val_loss improved from 0.61274 to 0.61138, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 68/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6189 - accuracy: 0.6024 - val_loss: 0.6160 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00068: val_loss did not improve from 0.61138\n",
      "Epoch 69/130\n",
      "683/683 [==============================] - 6s 8ms/step - loss: 0.6216 - accuracy: 0.6024 - val_loss: 0.6133 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00069: val_loss did not improve from 0.61138\n",
      "Epoch 70/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6207 - accuracy: 0.6024 - val_loss: 0.6151 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00070: val_loss did not improve from 0.61138\n",
      "Epoch 71/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6197 - accuracy: 0.6023 - val_loss: 0.6428 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00071: val_loss did not improve from 0.61138\n",
      "Epoch 72/130\n",
      "683/683 [==============================] - 6s 9ms/step - loss: 0.6191 - accuracy: 0.6025 - val_loss: 0.6298 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00072: val_loss did not improve from 0.61138\n",
      "Epoch 73/130\n",
      "683/683 [==============================] - 6s 9ms/step - loss: 0.6178 - accuracy: 0.6024 - val_loss: 0.6367 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00073: val_loss did not improve from 0.61138\n",
      "Epoch 74/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6171 - accuracy: 0.6024 - val_loss: 0.6197 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00074: val_loss did not improve from 0.61138\n",
      "Epoch 75/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6209 - accuracy: 0.6025 - val_loss: 0.6277 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00075: val_loss did not improve from 0.61138\n",
      "Epoch 76/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6219 - accuracy: 0.6026 - val_loss: 0.6149 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00076: val_loss did not improve from 0.61138\n",
      "Epoch 77/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6192 - accuracy: 0.6024 - val_loss: 0.6199 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00077: val_loss did not improve from 0.61138\n",
      "Epoch 78/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6172 - accuracy: 0.6024 - val_loss: 0.6134 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00078: val_loss did not improve from 0.61138\n",
      "Epoch 79/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6153 - accuracy: 0.6024 - val_loss: 0.6218 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00079: val_loss did not improve from 0.61138\n",
      "Epoch 80/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6167 - accuracy: 0.6024 - val_loss: 0.6178 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00080: val_loss did not improve from 0.61138\n",
      "Epoch 81/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6194 - accuracy: 0.6026 - val_loss: 0.6125 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00081: val_loss did not improve from 0.61138\n",
      "Epoch 82/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6170 - accuracy: 0.6025 - val_loss: 0.6186 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00082: val_loss did not improve from 0.61138\n",
      "Epoch 83/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6199 - accuracy: 0.6024 - val_loss: 0.6129 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00083: val_loss did not improve from 0.61138\n",
      "Epoch 84/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6176 - accuracy: 0.6025 - val_loss: 0.6289 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00084: val_loss did not improve from 0.61138\n",
      "Epoch 85/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6168 - accuracy: 0.6023 - val_loss: 0.6152 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00085: val_loss did not improve from 0.61138\n",
      "Epoch 86/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6155 - accuracy: 0.6024 - val_loss: 0.6162 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00086: val_loss did not improve from 0.61138\n",
      "Epoch 87/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6193 - accuracy: 0.6022 - val_loss: 0.6271 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00087: val_loss did not improve from 0.61138\n",
      "Epoch 88/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6217 - accuracy: 0.6024 - val_loss: 0.6124 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00088: val_loss did not improve from 0.61138\n",
      "Epoch 89/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6216 - accuracy: 0.6024 - val_loss: 0.6110 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00089: val_loss improved from 0.61138 to 0.61097, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 90/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6172 - accuracy: 0.6024 - val_loss: 0.6110 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00090: val_loss did not improve from 0.61097\n",
      "Epoch 91/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6183 - accuracy: 0.6023 - val_loss: 0.6136 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00091: val_loss did not improve from 0.61097\n",
      "Epoch 92/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6183 - accuracy: 0.6025 - val_loss: 0.6181 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00092: val_loss did not improve from 0.61097\n",
      "Epoch 93/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6150 - accuracy: 0.6024 - val_loss: 0.6096 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00093: val_loss improved from 0.61097 to 0.60964, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 94/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6172 - accuracy: 0.6026 - val_loss: 0.6208 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00094: val_loss did not improve from 0.60964\n",
      "Epoch 95/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6170 - accuracy: 0.6025 - val_loss: 0.6104 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00095: val_loss did not improve from 0.60964\n",
      "Epoch 96/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6201 - accuracy: 0.6024 - val_loss: 0.6188 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00096: val_loss did not improve from 0.60964\n",
      "Epoch 97/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6175 - accuracy: 0.6024 - val_loss: 0.6111 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00097: val_loss did not improve from 0.60964\n",
      "Epoch 98/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6164 - accuracy: 0.6024 - val_loss: 0.6217 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00098: val_loss did not improve from 0.60964\n",
      "Epoch 99/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6162 - accuracy: 0.6025 - val_loss: 0.6092 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00099: val_loss improved from 0.60964 to 0.60921, saving model to saved_models\\weights.best.basic_mlp.h5\n",
      "Epoch 100/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6133 - accuracy: 0.6024 - val_loss: 0.6121 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00100: val_loss did not improve from 0.60921\n",
      "Epoch 101/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6123 - accuracy: 0.6025 - val_loss: 0.6173 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00101: val_loss did not improve from 0.60921\n",
      "Epoch 102/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6226 - accuracy: 0.6025 - val_loss: 0.6110 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00102: val_loss did not improve from 0.60921\n",
      "Epoch 103/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6178 - accuracy: 0.6025 - val_loss: 0.6488 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00103: val_loss did not improve from 0.60921\n",
      "Epoch 104/130\n",
      "683/683 [==============================] - 5s 8ms/step - loss: 0.6230 - accuracy: 0.6024 - val_loss: 0.6250 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00104: val_loss did not improve from 0.60921\n",
      "Epoch 105/130\n",
      "683/683 [==============================] - 5s 8ms/step - loss: 0.6170 - accuracy: 0.6024 - val_loss: 0.6245 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00105: val_loss did not improve from 0.60921\n",
      "Epoch 106/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6146 - accuracy: 0.6025 - val_loss: 0.6294 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00106: val_loss did not improve from 0.60921\n",
      "Epoch 107/130\n",
      "683/683 [==============================] - 5s 7ms/step - loss: 0.6178 - accuracy: 0.6025 - val_loss: 0.6292 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00107: val_loss did not improve from 0.60921\n",
      "Epoch 108/130\n",
      "683/683 [==============================] - 5s 8ms/step - loss: 0.6144 - accuracy: 0.6023 - val_loss: 0.6176 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00108: val_loss did not improve from 0.60921\n",
      "Epoch 109/130\n",
      "683/683 [==============================] - 4s 7ms/step - loss: 0.6155 - accuracy: 0.6024 - val_loss: 0.6173 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00109: val_loss did not improve from 0.60921\n",
      "Epoch 110/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6142 - accuracy: 0.6025 - val_loss: 0.6360 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00110: val_loss did not improve from 0.60921\n",
      "Epoch 111/130\n",
      "683/683 [==============================] - 4s 6ms/step - loss: 0.6165 - accuracy: 0.6025 - val_loss: 0.6157 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00111: val_loss did not improve from 0.60921\n",
      "Epoch 112/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6139 - accuracy: 0.6023 - val_loss: 0.6094 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00112: val_loss did not improve from 0.60921\n",
      "Epoch 113/130\n",
      "683/683 [==============================] - 4s 5ms/step - loss: 0.6154 - accuracy: 0.6026 - val_loss: 0.6232 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00113: val_loss did not improve from 0.60921\n",
      "Epoch 114/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6148 - accuracy: 0.6025 - val_loss: 0.6205 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00114: val_loss did not improve from 0.60921\n",
      "Epoch 115/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6218 - accuracy: 0.6023 - val_loss: 0.6100 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00115: val_loss did not improve from 0.60921\n",
      "Epoch 116/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6140 - accuracy: 0.6024 - val_loss: 0.6979 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00116: val_loss did not improve from 0.60921\n",
      "Epoch 117/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6300 - accuracy: 0.6022 - val_loss: 0.6222 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00117: val_loss did not improve from 0.60921\n",
      "Epoch 118/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6233 - accuracy: 0.6025 - val_loss: 0.6234 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00118: val_loss did not improve from 0.60921\n",
      "Epoch 119/130\n",
      "683/683 [==============================] - 3s 5ms/step - loss: 0.6264 - accuracy: 0.6025 - val_loss: 0.6222 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00119: val_loss did not improve from 0.60921\n",
      "Epoch 120/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6199 - accuracy: 0.6024 - val_loss: 0.6254 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00120: val_loss did not improve from 0.60921\n",
      "Epoch 121/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6183 - accuracy: 0.6024 - val_loss: 0.6126 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00121: val_loss did not improve from 0.60921\n",
      "Epoch 122/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6194 - accuracy: 0.6023 - val_loss: 0.6337 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00122: val_loss did not improve from 0.60921\n",
      "Epoch 123/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6178 - accuracy: 0.6024 - val_loss: 0.6339 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00123: val_loss did not improve from 0.60921\n",
      "Epoch 124/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6172 - accuracy: 0.6025 - val_loss: 0.6374 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00124: val_loss did not improve from 0.60921\n",
      "Epoch 125/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6161 - accuracy: 0.6025 - val_loss: 0.6148 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00125: val_loss did not improve from 0.60921\n",
      "Epoch 126/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6186 - accuracy: 0.6024 - val_loss: 0.6385 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00126: val_loss did not improve from 0.60921\n",
      "Epoch 127/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6260 - accuracy: 0.6024 - val_loss: 0.6179 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00127: val_loss did not improve from 0.60921\n",
      "Epoch 128/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6187 - accuracy: 0.6025 - val_loss: 0.6160 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00128: val_loss did not improve from 0.60921\n",
      "Epoch 129/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6182 - accuracy: 0.6024 - val_loss: 0.6160 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00129: val_loss did not improve from 0.60921\n",
      "Epoch 130/130\n",
      "683/683 [==============================] - 3s 4ms/step - loss: 0.6230 - accuracy: 0.6025 - val_loss: 0.6248 - val_accuracy: 0.5945\n",
      "\n",
      "Epoch 00130: val_loss did not improve from 0.60921\n",
      "Training completed in time:  0:06:54.909422\n"
     ]
    }
   ],
   "source": [
    "from keras.callbacks import ModelCheckpoint \n",
    "from datetime import datetime \n",
    "\n",
    "num_epochs = 130\n",
    "num_batch_size = 32\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.basic_mlp.h5', \n",
    "                               verbose=1, save_best_only=True)\n",
    "start = datetime.now()\n",
    "\n",
    "history = 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",
    "\n",
    "duration = datetime.now() - start\n",
    "print(\"Training completed in time: \", duration)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test the model \n",
    "\n",
    "Here we will review the accuracy of the model on both the training and test data sets. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: saved_models/mlp_ka_sapna\\assets\n"
     ]
    }
   ],
   "source": [
    "model.save('saved_models/mlp_ka_sapna')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy:  0.6024631261825562\n",
      "Testing Accuracy:  0.5945310592651367\n"
     ]
    }
   ],
   "source": [
    "# Evaluating the model on the training and testing set\n",
    "model=load_model('saved_models/mlp_ka_sapna')\n",
    "\n",
    "training_accuracy = model.evaluate(x_train, y_train, verbose=0)\n",
    "print(\"Training Accuracy: \", training_accuracy[1])\n",
    "\n",
    "testing_accuracy = model.evaluate(x_test, y_test, verbose=0)\n",
    "\n",
    "print(\"Testing Accuracy: \", testing_accuracy[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The initial Training and Testing accuracy scores are quite high. As there is not a great difference between the Training and Test scores (~5%) this suggests that the model has not suffered from overfitting. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.5484351  0.45156488]\n",
      " [0.5484351  0.45156488]\n",
      " [0.99102205 0.00897799]\n",
      " ...\n",
      " [0.5484351  0.45156488]\n",
      " [0.5484351  0.45156488]\n",
      " [0.5484351  0.45156488]]\n"
     ]
    }
   ],
   "source": [
    "#predicted_vector = model.predict(x_test)\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "predict_x=model.predict(x_test) \n",
    "predicted_vector=np.argmax(predict_x,axis=1)\n",
    "le = LabelEncoder()\n",
    "yy_pred = to_categorical(le.fit_transform(predicted_vector)) \n",
    "print(predict_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized confusion matrix\n",
      "[[1. 0.]\n",
      " [1. 0.]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import plot_confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    print(cm)\n",
    "    np.set_printoptions(precision=2)\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, cm[i, j],\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "\n",
    "\n",
    "cnf_matrix=confusion_matrix(y_test.argmax(axis=1),yy_pred.argmax(axis=1))\n",
    "np.set_printoptions(precision=2)\n",
    "plt.figure()\n",
    "plot_confusion_matrix(cnf_matrix, ['negative','positive'],\n",
    "                      'Confusion matrix, without normalization')\n",
    "plt.show()\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.59      0.75      9362\n",
      "           1       0.00      0.00      0.00         0\n",
      "\n",
      "    accuracy                           0.59      9362\n",
      "   macro avg       0.50      0.30      0.37      9362\n",
      "weighted avg       1.00      0.59      0.75      9362\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(yy_pred.argmax(axis=1),y_test.argmax(axis=1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x272a8959850>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    " \n",
    "history_dict = history.history\n",
    "loss_values = history_dict['loss']\n",
    "val_loss_values = history_dict['val_loss']\n",
    "accuracy = history_dict['accuracy']\n",
    "val_accuracy = history_dict['val_accuracy']\n",
    " \n",
    "epochs = range(1, len(loss_values) + 1)\n",
    "fig, ax = plt.subplots(1, 2, figsize=(14, 6))\n",
    "#\n",
    "# Plot the model accuracy vs Epochs\n",
    "#\n",
    "ax[0].plot(epochs, accuracy, 'ro', label='Training accuracy')\n",
    "ax[0].plot(epochs, val_accuracy, 'b', label='Validation accuracy')\n",
    "ax[0].set_title('Training & Validation Accuracy', fontsize=16)\n",
    "ax[0].set_xlabel('Epochs', fontsize=16)\n",
    "ax[0].set_ylabel('Accuracy', fontsize=16)\n",
    "ax[0].legend()\n",
    "#\n",
    "# Plot the loss vs Epochs\n",
    "#\n",
    "ax[1].plot(epochs, loss_values, 'ro', label='Training loss')\n",
    "ax[1].plot(epochs, val_loss_values, 'b', label='Validation loss')\n",
    "ax[1].set_title('Training & Validation Loss', fontsize=16)\n",
    "ax[1].set_xlabel('Epochs', fontsize=16)\n",
    "ax[1].set_ylabel('Loss', fontsize=16)\n",
    "ax[1].legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predictions  \n",
    "\n",
    "Here we will build a method which will allow us to test the models predictions on a specified audio .wav file. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "import librosa \n",
    "import numpy as np \n",
    "\n",
    "def extract_feature(file_name):\n",
    "   \n",
    "    try:\n",
    "        audio_data, sample_rate = librosa.load(file_name, res_type='kaiser_fast') \n",
    "        mfccs = librosa.feature.mfcc(y=audio_data, sr=sample_rate, n_mfcc=40)\n",
    "        mfccsscaled = np.mean(mfccs.T,axis=0)\n",
    "        \n",
    "    except Exception as e:\n",
    "        print(\"Error encountered while parsing file: \", file_name)\n",
    "        return None, None\n",
    "\n",
    "    return np.array([mfccsscaled])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_prediction(file_name):\n",
    "    prediction_feature = extract_feature(file_name) \n",
    "\n",
    "    #predicted_vector = model.predict_classes(prediction_feature)\n",
    "    predict_x=model.predict(prediction_feature) \n",
    "    predicted_vector=np.argmax(predict_x,axis=1)\n",
    "    predicted_class = le.inverse_transform(predicted_vector) \n",
    "    print(\"The predicted class is:\", predicted_class[0], '\\n') \n",
    "\n",
    "    predicted_proba_vector = model.predict(prediction_feature) \n",
    "    predicted_proba = predicted_proba_vector[0]\n",
    "    for i in range(len(predicted_proba)): \n",
    "        category = le.inverse_transform(np.array([i]))\n",
    "        print(category[0], \"\\t\\t : \", format(predicted_proba[i], '.32f') )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Validation \n",
    "\n",
    "#### Test with sample data \n",
    "\n",
    "Initial sainity check to verify the predictions using a subsection of the sample audio files we explored in the first notebook. We expect the bulk of these to be classified correctly. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\yajwi\\anaconda3\\lib\\site-packages\\librosa\\core\\audio.py:162: UserWarning: PySoundFile failed. Trying audioread instead.\n",
      "  warnings.warn(\"PySoundFile failed. Trying audioread instead.\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Error encountered while parsing file:  ./clinical/converted/neg/neg-0421-088-cough-f-66.wav\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "Failed to find data adapter that can handle input: (<class 'tuple'> containing values of types {\"<class 'NoneType'>\"}), <class 'NoneType'>",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-27-037e57a34688>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      2\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mfilename\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'./clinical/converted/neg/neg-0421-088-cough-f-66.wav'\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 4\u001b[1;33m \u001b[0mprint_prediction\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfilename\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-26-85ccf61e4314>\u001b[0m in \u001b[0;36mprint_prediction\u001b[1;34m(file_name)\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m     \u001b[1;31m#predicted_vector = model.predict_classes(prediction_feature)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 5\u001b[1;33m     \u001b[0mpredict_x\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpredict\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mprediction_feature\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      6\u001b[0m     \u001b[0mpredicted_vector\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpredict_x\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0maxis\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      7\u001b[0m     \u001b[0mpredicted_class\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mle\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0minverse_transform\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpredicted_vector\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py\u001b[0m in \u001b[0;36mpredict\u001b[1;34m(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)\u001b[0m\n\u001b[0;32m   1596\u001b[0m                         '. Consider setting it to AutoShardPolicy.DATA.')\n\u001b[0;32m   1597\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1598\u001b[1;33m       data_handler = data_adapter.DataHandler(\n\u001b[0m\u001b[0;32m   1599\u001b[0m           \u001b[0mx\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1600\u001b[0m           \u001b[0mbatch_size\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\data_adapter.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, x, y, sample_weight, batch_size, steps_per_epoch, initial_epoch, epochs, shuffle, class_weight, max_queue_size, workers, use_multiprocessing, model, steps_per_execution)\u001b[0m\n\u001b[0;32m   1097\u001b[0m       \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_steps_per_execution_value\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msteps_per_execution\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnumpy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1098\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1099\u001b[1;33m     \u001b[0madapter_cls\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mselect_data_adapter\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1100\u001b[0m     self._adapter = adapter_cls(\n\u001b[0;32m   1101\u001b[0m         \u001b[0mx\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\anaconda3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\data_adapter.py\u001b[0m in \u001b[0;36mselect_data_adapter\u001b[1;34m(x, y)\u001b[0m\n\u001b[0;32m    959\u001b[0m   \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0madapter_cls\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    960\u001b[0m     \u001b[1;31m# TODO(scottzhu): This should be a less implementation-specific error.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 961\u001b[1;33m     raise ValueError(\n\u001b[0m\u001b[0;32m    962\u001b[0m         \u001b[1;34m\"Failed to find data adapter that can handle \"\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    963\u001b[0m         \"input: {}, {}\".format(\n",
      "\u001b[1;31mValueError\u001b[0m: Failed to find data adapter that can handle input: (<class 'tuple'> containing values of types {\"<class 'NoneType'>\"}), <class 'NoneType'>"
     ]
    }
   ],
   "source": [
    "# Class: neg\n",
    "\n",
    "filename = './clinical/converted/neg/neg-0421-088-cough-f-66.wav' \n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Class: positive\n",
    "\n",
    "filename = './clinical/converted/pos/pos-0421-094-cough-m-51.wav'\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # Class: Street music \n",
    "\n",
    "# filename = '../UrbanSound Dataset sample/audio/101848-9-0-0.wav'\n",
    "# print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#  Class: Car Horn \n",
    "\n",
    "# filename = '../UrbanSound Dataset sample/audio/100648-1-0-0.wav'\n",
    "# print_prediction(filename) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Observations \n",
    "\n",
    "From this brief sanity check the model seems to predict well. One errror was observed whereby a car horn was incorrectly classifed as a dog bark. \n",
    "\n",
    "We can see from the per class confidence that this was quite a low score (43%). This allows follows our early observation that a dog bark and car horn are similar in spectral shape. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Other audio\n",
    "\n",
    "Here we will use a sample of various copyright free sounds that we not part of either our test or training data to further validate our model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#filename = '../Evaluation audio/pos/az5zELVN7ObicxGrBLMoX8ki1LF2_shallow_1.wav'\n",
    "#print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Tanya(20-4-2021).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#filename = '../Evaluation audio/pos/Fi2lLmV3LHR5brvXs0z1JSKDvnI2_heavy_1.wav'\n",
    "\n",
    "#print_prediction(filename) \n",
    "\n",
    "# sample data weighted towards gun shot - peak in the dog barking sample is simmilar in shape to the gun shot sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#filename = '../Evaluation audio/neg/0zAgZhOXaGhmL1SnhkzJOIoowKg2_heavy_1.wav'\n",
    "\n",
    "#print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Ghaziabad.wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/coughtest2.wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "#filename = '../Evaluation audio/neg/zBiORuLmvGhOciaUXFeAKRhMyhj2_heavy_5.wav'\n",
    "\n",
    "#print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Divyam.wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/DJ(03-05-21).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/DJ2.wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/DJ3.wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Vardhika(03-05-21).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Tanyas_Grandma(29-4-2021).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Mom(29-4-2021).wav'\n",
    "\n",
    "print_prediction(filename) \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Avi(29-4-2021).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Bhaiya(04-05-21).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Papa(29-4-2021).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Shreya2(08-05-21).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../Evaluation audio/Satyashree3(08-05-21).wav'\n",
    "\n",
    "print_prediction(filename) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Observations \n",
    "\n",
    "The performance of our initial model is satisfactorry and has generalised well, seeming to predict well when tested against new audio data. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### *In the next notebook we will refine our model*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}