4-cnn-model-mel_spec.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "66105088-717f-47be-b719-d070a95e56d3"
    }
   },
   "source": [
    "##  Classifying audio data with convolutional neural networks\n",
    "\n",
    "<br/>\n",
    "by Eduardo Garcia Rajo<br/>\n",
    "<br/>\n",
    "\n",
    "This notebook is part of the project \"Urban sounds classification with Covnolutional Neural Networks\" on [my Github](https://github.com/GorillaBus/urban-audio-classifier).<br/>\n",
    "<br/>\n",
    "Licensed under the GNU LESSER GENERAL PUBLIC LICENSE Version 3, 29 June 2007<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "bdb8f444-143c-49a8-948d-4b56e2cb457d"
    }
   },
   "source": [
    "# Basic CNN Model - Log-Mel Spectrograms\n",
    "\n",
    "To compare the results from the previous notebook between MFCC and Log-Mel Spectrograms features I'm going to replicate the same basic CNN model training procedure of that notebook.\n",
    "<br/>\n",
    "\n",
    "INDEX\n",
    "\n",
    "1. Data preparation: features + metadata\n",
    "2. A basic CNN model\n",
    "4. Model evaluation\n",
    "7. Conclusion\n",
    "8. Next\n",
    "\n",
    "\n",
    "<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "bb86170e-2226-491a-8aec-e01925008746"
    }
   },
   "source": [
    "#### * Load required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "nbpresent": {
     "id": "8afd8752-1ca6-4fe1-b14d-b16615d0199b"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "import os\n",
    "import IPython\n",
    "import math\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import random\n",
    "from datetime import datetime\n",
    "from include import helpers\n",
    "\n",
    "from keras import backend as keras_backend\n",
    "from keras.models import Sequential, load_model\n",
    "from keras.layers import Dense, Dropout, Flatten, LeakyReLU, SpatialDropout2D, Activation, Conv2D, MaxPooling2D, BatchNormalization, GlobalAveragePooling2D\n",
    "from keras.optimizers import Adam\n",
    "from keras.utils import np_utils, to_categorical, plot_model\n",
    "from keras.callbacks import ModelCheckpoint \n",
    "from keras.regularizers import l2\n",
    "\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report, precision_recall_fscore_support, confusion_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### * Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "nbpresent": {
     "id": "d0c78fb3-3f82-445b-9e39-7bb9c9f1edd3"
    }
   },
   "outputs": [],
   "source": [
    "# Define general variables\n",
    "\n",
    "# Set your path to the dataset\n",
    "us8k_path = os.path.abspath('./UrbanSound8K')\n",
    "audio_path = os.path.join(us8k_path, 'audio')\n",
    "metadata_path = os.path.join(us8k_path, 'metadata/UrbanSound8K.csv')\n",
    "models_path = os.path.abspath('./models')\n",
    "data_path = os.path.abspath('./data')\n",
    "\n",
    "# Ensure \"channel last\" data format on Keras\n",
    "keras_backend.set_image_data_format('channels_last')\n",
    "\n",
    "# Define a labels array for future use\n",
    "labels = [\n",
    "        'Air Conditioner',\n",
    "        'Car Horn',\n",
    "        'Children Playing',\n",
    "        'Dog bark',\n",
    "        'Drilling',\n",
    "        'Engine Idling',\n",
    "        'Gun Shot',\n",
    "        'Jackhammer',\n",
    "        'Siren',\n",
    "        'Street Music'\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Required for the RTX GPU -- uncomment if using GTX or CPU\n",
    "import keras.backend as K\n",
    "import tensorflow as tf\n",
    "config = tf.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "K.set_session(tf.Session(config=config))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "a7b58aa0-2b3c-4854-a099-31d117ba74e1"
    }
   },
   "source": [
    "#### * Load data\n",
    "\n",
    "Data is an NumPy float32 array of shape (8732, 40, 174), 8732 samples with coefficients from Mel Spectrogram of 174 frames and 40 MEL bands. Each feature was scaled between -1 and 1 and zero-centered during extraction.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "nbpresent": {
     "id": "87d3ebd1-0d75-4ea4-8f12-9f891a72d3e2"
    }
   },
   "outputs": [],
   "source": [
    "# Pre-processed MFCC coefficients\n",
    "X = np.load(\"data/X-mel_spec.npy\")\n",
    "y = np.load(\"data/y-mel_spec.npy\")\n",
    "\n",
    "# Metadata\n",
    "metadata = pd.read_csv(metadata_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "f8043a00-6bfa-41ca-807f-a49d7430b65a"
    }
   },
   "source": [
    "### 1. Data preparation: features + metadata"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1.1 Train / Test split\n",
    "Note that we are using the same index order for both the Mel Spectrogram arrays and the metadata to keep track of the origin of each feature.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test split: 1746 \t\t Train split: 6986\n",
      "X test shape: (1746, 40, 174) \t X train shape: (6986, 40, 174)\n",
      "y test shape: (1746,) \t\t y train shape: (6986,)\n"
     ]
    }
   ],
   "source": [
    "indexes = []\n",
    "total = len(metadata)\n",
    "indexes = list(range(0, total))\n",
    "\n",
    "# Randomize indexes\n",
    "random.shuffle(indexes)\n",
    "\n",
    "# Divide the indexes into Train and Test\n",
    "test_split_pct = 20\n",
    "split_offset = math.floor(test_split_pct * total / 100)\n",
    "\n",
    "# Split the metadata\n",
    "test_split_idx = indexes[0:split_offset]\n",
    "train_split_idx = indexes[split_offset:total]\n",
    "\n",
    "\n",
    "# Split the features with the same indexes\n",
    "X_test = np.take(X, test_split_idx, axis=0)\n",
    "y_test = np.take(y, test_split_idx, axis=0)\n",
    "X_train = np.take(X, train_split_idx, axis=0)\n",
    "y_train = np.take(y, train_split_idx, axis=0)\n",
    "\n",
    "# Also split metadata\n",
    "test_meta = metadata.iloc[test_split_idx]\n",
    "train_meta = metadata.iloc[train_split_idx]\n",
    "\n",
    "# Print status\n",
    "print(\"Test split: {} \\t\\t Train split: {}\".format(len(test_meta), len(train_meta)))\n",
    "print(\"X test shape: {} \\t X train shape: {}\".format(X_test.shape, X_train.shape))\n",
    "print(\"y test shape: {} \\t\\t y train shape: {}\".format(y_test.shape, y_train.shape))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1.2 One hot encode labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "nbpresent": {
     "id": "65b227e7-2fe5-4582-bd79-c9e9b1b85d4e"
    }
   },
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "y_test_encoded = to_categorical(le.fit_transform(y_test))\n",
    "y_train_encoded = to_categorical(le.fit_transform(y_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "720605ff-4689-47dc-b142-023ea206a50c"
    }
   },
   "source": [
    "#### 1.3 Reshape data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "nbpresent": {
     "id": "d742676a-5a23-4fc8-856a-ad5505238d45"
    }
   },
   "outputs": [],
   "source": [
    "# How data should be structured\n",
    "num_rows = 40\n",
    "num_columns = 174 \n",
    "num_channels = 1\n",
    "\n",
    "# Reshape to fit the network input (channel last)\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",
    "# Total number of labels to predict (equal to the network output nodes)\n",
    "num_labels = y_train_encoded.shape[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Basic CNN network\n",
    "\n",
    "We are using the exact same architecture than the previous notebook.\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "23ae43f0-1d59-44ab-818d-54faeecf4f1c"
    }
   },
   "source": [
    "#### 2.1 Model definition\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "nbpresent": {
     "id": "845659a2-b4b4-4380-8d8f-51144a91ccab"
    }
   },
   "outputs": [],
   "source": [
    "def create_model(spatial_dropout_rate_1=0, spatial_dropout_rate_2=0, l2_rate=0):\n",
    "\n",
    "    # Create a secquential object\n",
    "    model = Sequential()\n",
    "\n",
    "\n",
    "    # Conv 1\n",
    "    model.add(Conv2D(filters=32, \n",
    "                     kernel_size=(3, 3), \n",
    "                     kernel_regularizer=l2(l2_rate), \n",
    "                     input_shape=(num_rows, num_columns, num_channels)))\n",
    "    model.add(LeakyReLU(alpha=0.1))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(SpatialDropout2D(spatial_dropout_rate_1))\n",
    "    model.add(Conv2D(filters=32, \n",
    "                     kernel_size=(3, 3), \n",
    "                     kernel_regularizer=l2(l2_rate)))\n",
    "    model.add(LeakyReLU(alpha=0.1))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "\n",
    "    # Max Pooling #1\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "    model.add(SpatialDropout2D(spatial_dropout_rate_1))\n",
    "    model.add(Conv2D(filters=64, \n",
    "                     kernel_size=(3, 3), \n",
    "                     kernel_regularizer=l2(l2_rate)))\n",
    "    model.add(LeakyReLU(alpha=0.1))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(SpatialDropout2D(spatial_dropout_rate_2))\n",
    "    model.add(Conv2D(filters=64, \n",
    "                     kernel_size=(3,3), \n",
    "                     kernel_regularizer=l2(l2_rate)))\n",
    "    model.add(LeakyReLU(alpha=0.1))\n",
    "    model.add(BatchNormalization())\n",
    "    \n",
    "   \n",
    "    # Reduces each h×w feature map to a single number by taking the average of all h,w values.\n",
    "    model.add(GlobalAveragePooling2D())\n",
    "\n",
    "\n",
    "    # Softmax output\n",
    "    model.add(Dense(num_labels, activation='softmax'))\n",
    "    \n",
    "    return model\n",
    "\n",
    "# Regularization rates\n",
    "spatial_dropout_rate_1 = 0.07\n",
    "spatial_dropout_rate_2 = 0.14\n",
    "l2_rate = 0.0005\n",
    "\n",
    "model = create_model(spatial_dropout_rate_1, spatial_dropout_rate_2, l2_rate)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "8131485b-3d88-4cdd-97d9-3f0c5d877be5"
    }
   },
   "source": [
    "#### 2.2 Setup model optimizer and loss function\n",
    "\n",
    "Using [ADAM](https://machinelearningmastery.com/adam-optimization-algorithm-for-deep-learning/) with default parameters to start from somewhere.\n",
    "Loss function is categorical cross-entropy as we are predicting between 10 classes.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "nbpresent": {
     "id": "b56bca3e-3fd0-4033-9d13-20db795d02f0"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_5 (Conv2D)            (None, 38, 172, 32)       320       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_5 (LeakyReLU)    (None, 38, 172, 32)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 38, 172, 32)       128       \n",
      "_________________________________________________________________\n",
      "spatial_dropout2d_4 (Spatial (None, 38, 172, 32)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 36, 170, 32)       9248      \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_6 (LeakyReLU)    (None, 36, 170, 32)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 36, 170, 32)       128       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 18, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "spatial_dropout2d_5 (Spatial (None, 18, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 16, 83, 64)        18496     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_7 (LeakyReLU)    (None, 16, 83, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_7 (Batch (None, 16, 83, 64)        256       \n",
      "_________________________________________________________________\n",
      "spatial_dropout2d_6 (Spatial (None, 16, 83, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 14, 81, 64)        36928     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_8 (LeakyReLU)    (None, 14, 81, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_8 (Batch (None, 14, 81, 64)        256       \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_2 ( (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 66,410\n",
      "Trainable params: 66,026\n",
      "Non-trainable params: 384\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "adam = Adam(lr=1e-4, beta_1=0.99, beta_2=0.999)\n",
    "model.compile(\n",
    "    loss='categorical_crossentropy', \n",
    "    metrics=['accuracy'], \n",
    "    optimizer=adam)\n",
    "\n",
    "# Display model architecture summary \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "94bf431f-27a8-4b01-b7aa-41a1947c9233"
    }
   },
   "source": [
    "#### 2.3 Training the model\n",
    "\n",
    "With Log-Mel Spectrograms it was possible to train for some more epochs before overfitting compared to the previous experiment with MFCC.\n",
    "<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 6403 samples, validate on 583 samples\n",
      "Epoch 1/358\n",
      "6403/6403 [==============================] - 3s 524us/step - loss: 2.0584 - accuracy: 0.2833 - val_loss: 2.4071 - val_accuracy: 0.1269\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 2.40712, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 2/358\n",
      "6403/6403 [==============================] - 3s 442us/step - loss: 1.8319 - accuracy: 0.3762 - val_loss: 2.7125 - val_accuracy: 0.1063\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 2.40712\n",
      "Epoch 3/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.7360 - accuracy: 0.4115 - val_loss: 3.0600 - val_accuracy: 0.1063\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 2.40712\n",
      "Epoch 4/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.6666 - accuracy: 0.4443 - val_loss: 3.3099 - val_accuracy: 0.1063\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 2.40712\n",
      "Epoch 5/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.6257 - accuracy: 0.4549 - val_loss: 3.4758 - val_accuracy: 0.1063\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 2.40712\n",
      "Epoch 6/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.5877 - accuracy: 0.4696 - val_loss: 3.3831 - val_accuracy: 0.1081\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 2.40712\n",
      "Epoch 7/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.5468 - accuracy: 0.4912 - val_loss: 3.0599 - val_accuracy: 0.1132\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 2.40712\n",
      "Epoch 8/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.5289 - accuracy: 0.4923 - val_loss: 2.6312 - val_accuracy: 0.1389\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 2.40712\n",
      "Epoch 9/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4998 - accuracy: 0.5034 - val_loss: 2.2798 - val_accuracy: 0.1887\n",
      "\n",
      "Epoch 00009: val_loss improved from 2.40712 to 2.27984, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 10/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4721 - accuracy: 0.5223 - val_loss: 1.9819 - val_accuracy: 0.2624\n",
      "\n",
      "Epoch 00010: val_loss improved from 2.27984 to 1.98186, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 11/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4610 - accuracy: 0.5177 - val_loss: 1.7022 - val_accuracy: 0.4237\n",
      "\n",
      "Epoch 00011: val_loss improved from 1.98186 to 1.70222, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 12/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4342 - accuracy: 0.5296 - val_loss: 1.5210 - val_accuracy: 0.5232\n",
      "\n",
      "Epoch 00012: val_loss improved from 1.70222 to 1.52098, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 13/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4174 - accuracy: 0.5416 - val_loss: 1.3959 - val_accuracy: 0.5729\n",
      "\n",
      "Epoch 00013: val_loss improved from 1.52098 to 1.39586, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 14/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.4031 - accuracy: 0.5415 - val_loss: 1.3252 - val_accuracy: 0.5832\n",
      "\n",
      "Epoch 00014: val_loss improved from 1.39586 to 1.32515, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 15/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.3900 - accuracy: 0.5476 - val_loss: 1.2840 - val_accuracy: 0.6244\n",
      "\n",
      "Epoch 00015: val_loss improved from 1.32515 to 1.28399, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 16/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.3706 - accuracy: 0.5576 - val_loss: 1.2486 - val_accuracy: 0.6398\n",
      "\n",
      "Epoch 00016: val_loss improved from 1.28399 to 1.24858, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 17/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.3558 - accuracy: 0.5607 - val_loss: 1.2319 - val_accuracy: 0.6364\n",
      "\n",
      "Epoch 00017: val_loss improved from 1.24858 to 1.23185, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 18/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.3366 - accuracy: 0.5665 - val_loss: 1.2209 - val_accuracy: 0.6467\n",
      "\n",
      "Epoch 00018: val_loss improved from 1.23185 to 1.22088, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 19/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.3269 - accuracy: 0.5752 - val_loss: 1.1909 - val_accuracy: 0.6621\n",
      "\n",
      "Epoch 00019: val_loss improved from 1.22088 to 1.19094, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 20/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.3203 - accuracy: 0.5777 - val_loss: 1.1809 - val_accuracy: 0.6655\n",
      "\n",
      "Epoch 00020: val_loss improved from 1.19094 to 1.18091, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 21/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.3030 - accuracy: 0.5789 - val_loss: 1.1807 - val_accuracy: 0.6775\n",
      "\n",
      "Epoch 00021: val_loss improved from 1.18091 to 1.18071, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 22/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2876 - accuracy: 0.5833 - val_loss: 1.1397 - val_accuracy: 0.6758\n",
      "\n",
      "Epoch 00022: val_loss improved from 1.18071 to 1.13975, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 23/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2784 - accuracy: 0.5843 - val_loss: 1.1152 - val_accuracy: 0.6861\n",
      "\n",
      "Epoch 00023: val_loss improved from 1.13975 to 1.11517, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 24/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2629 - accuracy: 0.5939 - val_loss: 1.1216 - val_accuracy: 0.6827\n",
      "\n",
      "Epoch 00024: val_loss did not improve from 1.11517\n",
      "Epoch 25/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2546 - accuracy: 0.5947 - val_loss: 1.0980 - val_accuracy: 0.7101\n",
      "\n",
      "Epoch 00025: val_loss improved from 1.11517 to 1.09799, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 26/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2440 - accuracy: 0.5986 - val_loss: 1.0924 - val_accuracy: 0.7136\n",
      "\n",
      "Epoch 00026: val_loss improved from 1.09799 to 1.09241, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 27/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2304 - accuracy: 0.6136 - val_loss: 1.0924 - val_accuracy: 0.7256\n",
      "\n",
      "Epoch 00027: val_loss improved from 1.09241 to 1.09240, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 28/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.2307 - accuracy: 0.6089 - val_loss: 1.0867 - val_accuracy: 0.7067\n",
      "\n",
      "Epoch 00028: val_loss improved from 1.09240 to 1.08666, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 29/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2205 - accuracy: 0.6103 - val_loss: 1.0750 - val_accuracy: 0.7324\n",
      "\n",
      "Epoch 00029: val_loss improved from 1.08666 to 1.07498, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 30/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.2116 - accuracy: 0.6203 - val_loss: 1.0528 - val_accuracy: 0.7170\n",
      "\n",
      "Epoch 00030: val_loss improved from 1.07498 to 1.05284, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 31/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.1984 - accuracy: 0.6194 - val_loss: 1.0656 - val_accuracy: 0.7273\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00031: val_loss did not improve from 1.05284\n",
      "Epoch 32/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.2060 - accuracy: 0.6222 - val_loss: 1.0511 - val_accuracy: 0.7358\n",
      "\n",
      "Epoch 00032: val_loss improved from 1.05284 to 1.05112, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 33/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.1805 - accuracy: 0.6238 - val_loss: 1.0357 - val_accuracy: 0.7238\n",
      "\n",
      "Epoch 00033: val_loss improved from 1.05112 to 1.03575, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 34/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1862 - accuracy: 0.6295 - val_loss: 1.0502 - val_accuracy: 0.7307\n",
      "\n",
      "Epoch 00034: val_loss did not improve from 1.03575\n",
      "Epoch 35/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.1718 - accuracy: 0.6294 - val_loss: 1.0290 - val_accuracy: 0.7221\n",
      "\n",
      "Epoch 00035: val_loss improved from 1.03575 to 1.02902, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 36/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1611 - accuracy: 0.6330 - val_loss: 1.0134 - val_accuracy: 0.7444\n",
      "\n",
      "Epoch 00036: val_loss improved from 1.02902 to 1.01343, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 37/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1544 - accuracy: 0.6383 - val_loss: 1.0054 - val_accuracy: 0.7410\n",
      "\n",
      "Epoch 00037: val_loss improved from 1.01343 to 1.00544, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 38/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1437 - accuracy: 0.6394 - val_loss: 1.0051 - val_accuracy: 0.7427\n",
      "\n",
      "Epoch 00038: val_loss improved from 1.00544 to 1.00511, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 39/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1408 - accuracy: 0.6495 - val_loss: 0.9980 - val_accuracy: 0.7530\n",
      "\n",
      "Epoch 00039: val_loss improved from 1.00511 to 0.99803, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 40/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1253 - accuracy: 0.6553 - val_loss: 0.9929 - val_accuracy: 0.7444\n",
      "\n",
      "Epoch 00040: val_loss improved from 0.99803 to 0.99287, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 41/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1281 - accuracy: 0.6422 - val_loss: 0.9818 - val_accuracy: 0.7410\n",
      "\n",
      "Epoch 00041: val_loss improved from 0.99287 to 0.98175, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 42/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.1138 - accuracy: 0.6525 - val_loss: 0.9807 - val_accuracy: 0.7393\n",
      "\n",
      "Epoch 00042: val_loss improved from 0.98175 to 0.98073, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 43/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1126 - accuracy: 0.6430 - val_loss: 0.9768 - val_accuracy: 0.7461\n",
      "\n",
      "Epoch 00043: val_loss improved from 0.98073 to 0.97683, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 44/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.1069 - accuracy: 0.6594 - val_loss: 0.9544 - val_accuracy: 0.7461\n",
      "\n",
      "Epoch 00044: val_loss improved from 0.97683 to 0.95440, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 45/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0864 - accuracy: 0.6650 - val_loss: 0.9668 - val_accuracy: 0.7513\n",
      "\n",
      "Epoch 00045: val_loss did not improve from 0.95440\n",
      "Epoch 46/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.0887 - accuracy: 0.6598 - val_loss: 0.9661 - val_accuracy: 0.7530\n",
      "\n",
      "Epoch 00046: val_loss did not improve from 0.95440\n",
      "Epoch 47/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0804 - accuracy: 0.6645 - val_loss: 0.9466 - val_accuracy: 0.7496\n",
      "\n",
      "Epoch 00047: val_loss improved from 0.95440 to 0.94660, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 48/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0796 - accuracy: 0.6658 - val_loss: 0.9455 - val_accuracy: 0.7581\n",
      "\n",
      "Epoch 00048: val_loss improved from 0.94660 to 0.94553, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 49/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0739 - accuracy: 0.6744 - val_loss: 0.9290 - val_accuracy: 0.7530\n",
      "\n",
      "Epoch 00049: val_loss improved from 0.94553 to 0.92902, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 50/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.0753 - accuracy: 0.6678 - val_loss: 0.9389 - val_accuracy: 0.7530\n",
      "\n",
      "Epoch 00050: val_loss did not improve from 0.92902\n",
      "Epoch 51/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.0656 - accuracy: 0.6687 - val_loss: 0.9352 - val_accuracy: 0.7547\n",
      "\n",
      "Epoch 00051: val_loss did not improve from 0.92902\n",
      "Epoch 52/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 1.0565 - accuracy: 0.6780 - val_loss: 0.9123 - val_accuracy: 0.7599\n",
      "\n",
      "Epoch 00052: val_loss improved from 0.92902 to 0.91230, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 53/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0529 - accuracy: 0.6714 - val_loss: 0.9219 - val_accuracy: 0.7650\n",
      "\n",
      "Epoch 00053: val_loss did not improve from 0.91230\n",
      "Epoch 54/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.0428 - accuracy: 0.6825 - val_loss: 0.9276 - val_accuracy: 0.7633\n",
      "\n",
      "Epoch 00054: val_loss did not improve from 0.91230\n",
      "Epoch 55/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.0400 - accuracy: 0.6898 - val_loss: 0.9150 - val_accuracy: 0.7599\n",
      "\n",
      "Epoch 00055: val_loss did not improve from 0.91230\n",
      "Epoch 56/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.0283 - accuracy: 0.6923 - val_loss: 0.9145 - val_accuracy: 0.7667\n",
      "\n",
      "Epoch 00056: val_loss did not improve from 0.91230\n",
      "Epoch 57/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 1.0174 - accuracy: 0.6931 - val_loss: 0.9019 - val_accuracy: 0.7616\n",
      "\n",
      "Epoch 00057: val_loss improved from 0.91230 to 0.90186, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 58/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0213 - accuracy: 0.6862 - val_loss: 0.8932 - val_accuracy: 0.7599\n",
      "\n",
      "Epoch 00058: val_loss improved from 0.90186 to 0.89319, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 59/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 1.0167 - accuracy: 0.6970 - val_loss: 0.8845 - val_accuracy: 0.7702\n",
      "\n",
      "Epoch 00059: val_loss improved from 0.89319 to 0.88451, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 60/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.0140 - accuracy: 0.6980 - val_loss: 0.8985 - val_accuracy: 0.7667\n",
      "\n",
      "Epoch 00060: val_loss did not improve from 0.88451\n",
      "Epoch 61/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9932 - accuracy: 0.7026 - val_loss: 0.8852 - val_accuracy: 0.7702\n",
      "\n",
      "Epoch 00061: val_loss did not improve from 0.88451\n",
      "Epoch 62/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9944 - accuracy: 0.7034 - val_loss: 0.8760 - val_accuracy: 0.7650\n",
      "\n",
      "Epoch 00062: val_loss improved from 0.88451 to 0.87600, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 63/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 1.0051 - accuracy: 0.6989 - val_loss: 0.8829 - val_accuracy: 0.7599\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00063: val_loss did not improve from 0.87600\n",
      "Epoch 64/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9911 - accuracy: 0.7034 - val_loss: 0.8551 - val_accuracy: 0.7599\n",
      "\n",
      "Epoch 00064: val_loss improved from 0.87600 to 0.85508, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 65/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9799 - accuracy: 0.7047 - val_loss: 0.8566 - val_accuracy: 0.7770\n",
      "\n",
      "Epoch 00065: val_loss did not improve from 0.85508\n",
      "Epoch 66/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9760 - accuracy: 0.7098 - val_loss: 0.8457 - val_accuracy: 0.7650\n",
      "\n",
      "Epoch 00066: val_loss improved from 0.85508 to 0.84566, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 67/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 0.9780 - accuracy: 0.7067 - val_loss: 0.8450 - val_accuracy: 0.7667\n",
      "\n",
      "Epoch 00067: val_loss improved from 0.84566 to 0.84500, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 68/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9685 - accuracy: 0.7073 - val_loss: 0.8360 - val_accuracy: 0.7770\n",
      "\n",
      "Epoch 00068: val_loss improved from 0.84500 to 0.83603, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 69/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9615 - accuracy: 0.7194 - val_loss: 0.8477 - val_accuracy: 0.7753\n",
      "\n",
      "Epoch 00069: val_loss did not improve from 0.83603\n",
      "Epoch 70/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9537 - accuracy: 0.7161 - val_loss: 0.8374 - val_accuracy: 0.7719\n",
      "\n",
      "Epoch 00070: val_loss did not improve from 0.83603\n",
      "Epoch 71/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9590 - accuracy: 0.7086 - val_loss: 0.8440 - val_accuracy: 0.7804\n",
      "\n",
      "Epoch 00071: val_loss did not improve from 0.83603\n",
      "Epoch 72/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9467 - accuracy: 0.7172 - val_loss: 0.8435 - val_accuracy: 0.7633\n",
      "\n",
      "Epoch 00072: val_loss did not improve from 0.83603\n",
      "Epoch 73/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9494 - accuracy: 0.7172 - val_loss: 0.8249 - val_accuracy: 0.7719\n",
      "\n",
      "Epoch 00073: val_loss improved from 0.83603 to 0.82491, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 74/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9399 - accuracy: 0.7204 - val_loss: 0.8267 - val_accuracy: 0.7736\n",
      "\n",
      "Epoch 00074: val_loss did not improve from 0.82491\n",
      "Epoch 75/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9379 - accuracy: 0.7133 - val_loss: 0.8120 - val_accuracy: 0.7873\n",
      "\n",
      "Epoch 00075: val_loss improved from 0.82491 to 0.81199, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 76/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9306 - accuracy: 0.7247 - val_loss: 0.8155 - val_accuracy: 0.7753\n",
      "\n",
      "Epoch 00076: val_loss did not improve from 0.81199\n",
      "Epoch 77/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9289 - accuracy: 0.7283 - val_loss: 0.8035 - val_accuracy: 0.7856\n",
      "\n",
      "Epoch 00077: val_loss improved from 0.81199 to 0.80353, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 78/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9174 - accuracy: 0.7336 - val_loss: 0.8069 - val_accuracy: 0.7804\n",
      "\n",
      "Epoch 00078: val_loss did not improve from 0.80353\n",
      "Epoch 79/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9181 - accuracy: 0.7342 - val_loss: 0.8095 - val_accuracy: 0.7770\n",
      "\n",
      "Epoch 00079: val_loss did not improve from 0.80353\n",
      "Epoch 80/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9096 - accuracy: 0.7331 - val_loss: 0.8046 - val_accuracy: 0.7753\n",
      "\n",
      "Epoch 00080: val_loss did not improve from 0.80353\n",
      "Epoch 81/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.9147 - accuracy: 0.7265 - val_loss: 0.7924 - val_accuracy: 0.7856\n",
      "\n",
      "Epoch 00081: val_loss improved from 0.80353 to 0.79237, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 82/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9127 - accuracy: 0.7303 - val_loss: 0.7951 - val_accuracy: 0.7787\n",
      "\n",
      "Epoch 00082: val_loss did not improve from 0.79237\n",
      "Epoch 83/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.9071 - accuracy: 0.7328 - val_loss: 0.7906 - val_accuracy: 0.7822\n",
      "\n",
      "Epoch 00083: val_loss improved from 0.79237 to 0.79057, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 84/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.9130 - accuracy: 0.7298 - val_loss: 0.7797 - val_accuracy: 0.7890\n",
      "\n",
      "Epoch 00084: val_loss improved from 0.79057 to 0.77968, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 85/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8924 - accuracy: 0.7340 - val_loss: 0.7814 - val_accuracy: 0.7839\n",
      "\n",
      "Epoch 00085: val_loss did not improve from 0.77968\n",
      "Epoch 86/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8906 - accuracy: 0.7368 - val_loss: 0.7834 - val_accuracy: 0.7942\n",
      "\n",
      "Epoch 00086: val_loss did not improve from 0.77968\n",
      "Epoch 87/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8920 - accuracy: 0.7428 - val_loss: 0.7692 - val_accuracy: 0.7925\n",
      "\n",
      "Epoch 00087: val_loss improved from 0.77968 to 0.76919, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 88/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8937 - accuracy: 0.7356 - val_loss: 0.7710 - val_accuracy: 0.7839\n",
      "\n",
      "Epoch 00088: val_loss did not improve from 0.76919\n",
      "Epoch 89/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8789 - accuracy: 0.7426 - val_loss: 0.7699 - val_accuracy: 0.7993\n",
      "\n",
      "Epoch 00089: val_loss did not improve from 0.76919\n",
      "Epoch 90/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.8866 - accuracy: 0.7398 - val_loss: 0.7687 - val_accuracy: 0.7993\n",
      "\n",
      "Epoch 00090: val_loss improved from 0.76919 to 0.76866, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 91/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8856 - accuracy: 0.7431 - val_loss: 0.7778 - val_accuracy: 0.7839\n",
      "\n",
      "Epoch 00091: val_loss did not improve from 0.76866\n",
      "Epoch 92/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8692 - accuracy: 0.7481 - val_loss: 0.7571 - val_accuracy: 0.7925\n",
      "\n",
      "Epoch 00092: val_loss improved from 0.76866 to 0.75711, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 93/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8716 - accuracy: 0.7412 - val_loss: 0.7713 - val_accuracy: 0.7907\n",
      "\n",
      "Epoch 00093: val_loss did not improve from 0.75711\n",
      "Epoch 94/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8661 - accuracy: 0.7461 - val_loss: 0.7684 - val_accuracy: 0.7890\n",
      "\n",
      "Epoch 00094: val_loss did not improve from 0.75711\n",
      "Epoch 95/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8609 - accuracy: 0.7498 - val_loss: 0.7510 - val_accuracy: 0.7890\n",
      "\n",
      "Epoch 00095: val_loss improved from 0.75711 to 0.75098, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 96/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8638 - accuracy: 0.7528 - val_loss: 0.7596 - val_accuracy: 0.8010\n",
      "\n",
      "Epoch 00096: val_loss did not improve from 0.75098\n",
      "Epoch 97/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8569 - accuracy: 0.7500 - val_loss: 0.7556 - val_accuracy: 0.7925\n",
      "\n",
      "Epoch 00097: val_loss did not improve from 0.75098\n",
      "Epoch 98/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8565 - accuracy: 0.7468 - val_loss: 0.7472 - val_accuracy: 0.7942\n",
      "\n",
      "Epoch 00098: val_loss improved from 0.75098 to 0.74720, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 99/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8506 - accuracy: 0.7528 - val_loss: 0.7523 - val_accuracy: 0.7959\n",
      "\n",
      "Epoch 00099: val_loss did not improve from 0.74720\n",
      "Epoch 100/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8507 - accuracy: 0.7521 - val_loss: 0.7286 - val_accuracy: 0.8045\n",
      "\n",
      "Epoch 00100: val_loss improved from 0.74720 to 0.72862, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 101/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8431 - accuracy: 0.7564 - val_loss: 0.7353 - val_accuracy: 0.7976\n",
      "\n",
      "Epoch 00101: val_loss did not improve from 0.72862\n",
      "Epoch 102/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8432 - accuracy: 0.7606 - val_loss: 0.7322 - val_accuracy: 0.8010\n",
      "\n",
      "Epoch 00102: val_loss did not improve from 0.72862\n",
      "Epoch 103/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8465 - accuracy: 0.7568 - val_loss: 0.7267 - val_accuracy: 0.8062\n",
      "\n",
      "Epoch 00103: val_loss improved from 0.72862 to 0.72668, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 104/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8417 - accuracy: 0.7567 - val_loss: 0.7225 - val_accuracy: 0.8027\n",
      "\n",
      "Epoch 00104: val_loss improved from 0.72668 to 0.72252, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 105/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8285 - accuracy: 0.7598 - val_loss: 0.7202 - val_accuracy: 0.8010\n",
      "\n",
      "Epoch 00105: val_loss improved from 0.72252 to 0.72019, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 106/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 0.8280 - accuracy: 0.7575 - val_loss: 0.7323 - val_accuracy: 0.7873\n",
      "\n",
      "Epoch 00106: val_loss did not improve from 0.72019\n",
      "Epoch 107/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8317 - accuracy: 0.7628 - val_loss: 0.7167 - val_accuracy: 0.7907\n",
      "\n",
      "Epoch 00107: val_loss improved from 0.72019 to 0.71670, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 108/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.8187 - accuracy: 0.7607 - val_loss: 0.7268 - val_accuracy: 0.7942\n",
      "\n",
      "Epoch 00108: val_loss did not improve from 0.71670\n",
      "Epoch 109/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.8243 - accuracy: 0.7632 - val_loss: 0.7201 - val_accuracy: 0.7976\n",
      "\n",
      "Epoch 00109: val_loss did not improve from 0.71670\n",
      "Epoch 110/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8208 - accuracy: 0.7590 - val_loss: 0.7133 - val_accuracy: 0.8062\n",
      "\n",
      "Epoch 00110: val_loss improved from 0.71670 to 0.71333, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 111/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8132 - accuracy: 0.7654 - val_loss: 0.7190 - val_accuracy: 0.7959\n",
      "\n",
      "Epoch 00111: val_loss did not improve from 0.71333\n",
      "Epoch 112/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8180 - accuracy: 0.7654 - val_loss: 0.6983 - val_accuracy: 0.8027\n",
      "\n",
      "Epoch 00112: val_loss improved from 0.71333 to 0.69832, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 113/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8043 - accuracy: 0.7765 - val_loss: 0.7086 - val_accuracy: 0.7993\n",
      "\n",
      "Epoch 00113: val_loss did not improve from 0.69832\n",
      "Epoch 114/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.8069 - accuracy: 0.7657 - val_loss: 0.6877 - val_accuracy: 0.8165\n",
      "\n",
      "Epoch 00114: val_loss improved from 0.69832 to 0.68775, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 115/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.8094 - accuracy: 0.7710 - val_loss: 0.6942 - val_accuracy: 0.8285\n",
      "\n",
      "Epoch 00115: val_loss did not improve from 0.68775\n",
      "Epoch 116/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7997 - accuracy: 0.7721 - val_loss: 0.6923 - val_accuracy: 0.8062\n",
      "\n",
      "Epoch 00116: val_loss did not improve from 0.68775\n",
      "Epoch 117/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7939 - accuracy: 0.7762 - val_loss: 0.6857 - val_accuracy: 0.8113\n",
      "\n",
      "Epoch 00117: val_loss improved from 0.68775 to 0.68573, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 118/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7878 - accuracy: 0.7751 - val_loss: 0.6937 - val_accuracy: 0.8165\n",
      "\n",
      "Epoch 00118: val_loss did not improve from 0.68573\n",
      "Epoch 119/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7898 - accuracy: 0.7759 - val_loss: 0.6899 - val_accuracy: 0.8216\n",
      "\n",
      "Epoch 00119: val_loss did not improve from 0.68573\n",
      "Epoch 120/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7923 - accuracy: 0.7706 - val_loss: 0.6849 - val_accuracy: 0.8096\n",
      "\n",
      "Epoch 00120: val_loss improved from 0.68573 to 0.68491, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 121/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7735 - accuracy: 0.7826 - val_loss: 0.6777 - val_accuracy: 0.8165\n",
      "\n",
      "Epoch 00121: val_loss improved from 0.68491 to 0.67769, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 122/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7827 - accuracy: 0.7759 - val_loss: 0.6792 - val_accuracy: 0.8130\n",
      "\n",
      "Epoch 00122: val_loss did not improve from 0.67769\n",
      "Epoch 123/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7725 - accuracy: 0.7821 - val_loss: 0.6845 - val_accuracy: 0.8079\n",
      "\n",
      "Epoch 00123: val_loss did not improve from 0.67769\n",
      "Epoch 124/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7670 - accuracy: 0.7809 - val_loss: 0.6746 - val_accuracy: 0.8096\n",
      "\n",
      "Epoch 00124: val_loss improved from 0.67769 to 0.67463, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 125/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7676 - accuracy: 0.7817 - val_loss: 0.6692 - val_accuracy: 0.8130\n",
      "\n",
      "Epoch 00125: val_loss improved from 0.67463 to 0.66922, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 126/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7692 - accuracy: 0.7843 - val_loss: 0.6793 - val_accuracy: 0.8079\n",
      "\n",
      "Epoch 00126: val_loss did not improve from 0.66922\n",
      "Epoch 127/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7646 - accuracy: 0.7843 - val_loss: 0.6713 - val_accuracy: 0.8113\n",
      "\n",
      "Epoch 00127: val_loss did not improve from 0.66922\n",
      "Epoch 128/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.7637 - accuracy: 0.7885 - val_loss: 0.6700 - val_accuracy: 0.8250\n",
      "\n",
      "Epoch 00128: val_loss did not improve from 0.66922\n",
      "Epoch 129/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7650 - accuracy: 0.7796 - val_loss: 0.6655 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00129: val_loss improved from 0.66922 to 0.66545, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 130/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7658 - accuracy: 0.7818 - val_loss: 0.6631 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00130: val_loss improved from 0.66545 to 0.66311, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 131/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7643 - accuracy: 0.7873 - val_loss: 0.6647 - val_accuracy: 0.7993\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00131: val_loss did not improve from 0.66311\n",
      "Epoch 132/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7445 - accuracy: 0.7870 - val_loss: 0.6626 - val_accuracy: 0.8130\n",
      "\n",
      "Epoch 00132: val_loss improved from 0.66311 to 0.66256, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 133/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7568 - accuracy: 0.7823 - val_loss: 0.6469 - val_accuracy: 0.8199\n",
      "\n",
      "Epoch 00133: val_loss improved from 0.66256 to 0.64686, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 134/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7519 - accuracy: 0.7815 - val_loss: 0.6616 - val_accuracy: 0.8010\n",
      "\n",
      "Epoch 00134: val_loss did not improve from 0.64686\n",
      "Epoch 135/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7442 - accuracy: 0.7898 - val_loss: 0.6537 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00135: val_loss did not improve from 0.64686\n",
      "Epoch 136/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7462 - accuracy: 0.7885 - val_loss: 0.6554 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00136: val_loss did not improve from 0.64686\n",
      "Epoch 137/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7468 - accuracy: 0.7887 - val_loss: 0.6545 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00137: val_loss did not improve from 0.64686\n",
      "Epoch 138/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7389 - accuracy: 0.7912 - val_loss: 0.6477 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00138: val_loss did not improve from 0.64686\n",
      "Epoch 139/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7356 - accuracy: 0.7867 - val_loss: 0.6491 - val_accuracy: 0.8199\n",
      "\n",
      "Epoch 00139: val_loss did not improve from 0.64686\n",
      "Epoch 140/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7396 - accuracy: 0.7879 - val_loss: 0.6364 - val_accuracy: 0.8233\n",
      "\n",
      "Epoch 00140: val_loss improved from 0.64686 to 0.63640, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 141/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7398 - accuracy: 0.7890 - val_loss: 0.6440 - val_accuracy: 0.8302\n",
      "\n",
      "Epoch 00141: val_loss did not improve from 0.63640\n",
      "Epoch 142/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7441 - accuracy: 0.7893 - val_loss: 0.6339 - val_accuracy: 0.8165\n",
      "\n",
      "Epoch 00142: val_loss improved from 0.63640 to 0.63389, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 143/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7313 - accuracy: 0.7949 - val_loss: 0.6539 - val_accuracy: 0.8096\n",
      "\n",
      "Epoch 00143: val_loss did not improve from 0.63389\n",
      "Epoch 144/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7301 - accuracy: 0.7890 - val_loss: 0.6415 - val_accuracy: 0.8148\n",
      "\n",
      "Epoch 00144: val_loss did not improve from 0.63389\n",
      "Epoch 145/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7254 - accuracy: 0.7982 - val_loss: 0.6275 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00145: val_loss improved from 0.63389 to 0.62748, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 146/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7199 - accuracy: 0.7990 - val_loss: 0.6286 - val_accuracy: 0.8250\n",
      "\n",
      "Epoch 00146: val_loss did not improve from 0.62748\n",
      "Epoch 147/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7289 - accuracy: 0.7951 - val_loss: 0.6260 - val_accuracy: 0.8268\n",
      "\n",
      "Epoch 00147: val_loss improved from 0.62748 to 0.62601, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 148/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7199 - accuracy: 0.7974 - val_loss: 0.6276 - val_accuracy: 0.8268\n",
      "\n",
      "Epoch 00148: val_loss did not improve from 0.62601\n",
      "Epoch 149/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7085 - accuracy: 0.8001 - val_loss: 0.6292 - val_accuracy: 0.8268\n",
      "\n",
      "Epoch 00149: val_loss did not improve from 0.62601\n",
      "Epoch 150/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7100 - accuracy: 0.7985 - val_loss: 0.6122 - val_accuracy: 0.8388\n",
      "\n",
      "Epoch 00150: val_loss improved from 0.62601 to 0.61217, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 151/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7232 - accuracy: 0.7942 - val_loss: 0.6271 - val_accuracy: 0.8216\n",
      "\n",
      "Epoch 00151: val_loss did not improve from 0.61217\n",
      "Epoch 152/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7059 - accuracy: 0.8049 - val_loss: 0.6166 - val_accuracy: 0.8199\n",
      "\n",
      "Epoch 00152: val_loss did not improve from 0.61217\n",
      "Epoch 153/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7120 - accuracy: 0.7959 - val_loss: 0.6168 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00153: val_loss did not improve from 0.61217\n",
      "Epoch 154/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.7035 - accuracy: 0.8035 - val_loss: 0.6303 - val_accuracy: 0.8353\n",
      "\n",
      "Epoch 00154: val_loss did not improve from 0.61217\n",
      "Epoch 155/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.7020 - accuracy: 0.8056 - val_loss: 0.6067 - val_accuracy: 0.8302\n",
      "\n",
      "Epoch 00155: val_loss improved from 0.61217 to 0.60675, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 156/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.7031 - accuracy: 0.8034 - val_loss: 0.6152 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00156: val_loss did not improve from 0.60675\n",
      "Epoch 157/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6956 - accuracy: 0.8109 - val_loss: 0.6192 - val_accuracy: 0.8336\n",
      "\n",
      "Epoch 00157: val_loss did not improve from 0.60675\n",
      "Epoch 158/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6900 - accuracy: 0.8081 - val_loss: 0.6079 - val_accuracy: 0.8302\n",
      "\n",
      "Epoch 00158: val_loss did not improve from 0.60675\n",
      "Epoch 159/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6949 - accuracy: 0.8084 - val_loss: 0.6010 - val_accuracy: 0.8250\n",
      "\n",
      "Epoch 00159: val_loss improved from 0.60675 to 0.60096, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 160/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6912 - accuracy: 0.8027 - val_loss: 0.5984 - val_accuracy: 0.8336\n",
      "\n",
      "Epoch 00160: val_loss improved from 0.60096 to 0.59842, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 161/358\n",
      "6403/6403 [==============================] - 3s 444us/step - loss: 0.6920 - accuracy: 0.8054 - val_loss: 0.5969 - val_accuracy: 0.8405\n",
      "\n",
      "Epoch 00161: val_loss improved from 0.59842 to 0.59692, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 162/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6957 - accuracy: 0.8065 - val_loss: 0.5922 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00162: val_loss improved from 0.59692 to 0.59219, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 163/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6996 - accuracy: 0.7981 - val_loss: 0.6060 - val_accuracy: 0.8439\n",
      "\n",
      "Epoch 00163: val_loss did not improve from 0.59219\n",
      "Epoch 164/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6840 - accuracy: 0.8068 - val_loss: 0.6003 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00164: val_loss did not improve from 0.59219\n",
      "Epoch 165/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6801 - accuracy: 0.8049 - val_loss: 0.5969 - val_accuracy: 0.8370\n",
      "\n",
      "Epoch 00165: val_loss did not improve from 0.59219\n",
      "Epoch 166/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6816 - accuracy: 0.8070 - val_loss: 0.5983 - val_accuracy: 0.8336\n",
      "\n",
      "Epoch 00166: val_loss did not improve from 0.59219\n",
      "Epoch 167/358\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6721 - accuracy: 0.8187 - val_loss: 0.5897 - val_accuracy: 0.8388\n",
      "\n",
      "Epoch 00167: val_loss improved from 0.59219 to 0.58970, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 168/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6781 - accuracy: 0.8151 - val_loss: 0.5825 - val_accuracy: 0.8439\n",
      "\n",
      "Epoch 00168: val_loss improved from 0.58970 to 0.58246, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 169/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6703 - accuracy: 0.8131 - val_loss: 0.5910 - val_accuracy: 0.8336\n",
      "\n",
      "Epoch 00169: val_loss did not improve from 0.58246\n",
      "Epoch 170/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6767 - accuracy: 0.8092 - val_loss: 0.5972 - val_accuracy: 0.8388\n",
      "\n",
      "Epoch 00170: val_loss did not improve from 0.58246\n",
      "Epoch 171/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6789 - accuracy: 0.8060 - val_loss: 0.5863 - val_accuracy: 0.8422\n",
      "\n",
      "Epoch 00171: val_loss did not improve from 0.58246\n",
      "Epoch 172/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6681 - accuracy: 0.8110 - val_loss: 0.5884 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00172: val_loss did not improve from 0.58246\n",
      "Epoch 173/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6690 - accuracy: 0.8146 - val_loss: 0.5796 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00173: val_loss improved from 0.58246 to 0.57958, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 174/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6639 - accuracy: 0.8137 - val_loss: 0.5780 - val_accuracy: 0.8439\n",
      "\n",
      "Epoch 00174: val_loss improved from 0.57958 to 0.57795, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 175/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6743 - accuracy: 0.8057 - val_loss: 0.5771 - val_accuracy: 0.8422\n",
      "\n",
      "Epoch 00175: val_loss improved from 0.57795 to 0.57712, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 176/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6621 - accuracy: 0.8140 - val_loss: 0.5736 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00176: val_loss improved from 0.57712 to 0.57356, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 177/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6632 - accuracy: 0.8173 - val_loss: 0.5762 - val_accuracy: 0.8508\n",
      "\n",
      "Epoch 00177: val_loss did not improve from 0.57356\n",
      "Epoch 178/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6551 - accuracy: 0.8182 - val_loss: 0.5690 - val_accuracy: 0.8405\n",
      "\n",
      "Epoch 00178: val_loss improved from 0.57356 to 0.56902, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 179/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6484 - accuracy: 0.8176 - val_loss: 0.5665 - val_accuracy: 0.8405\n",
      "\n",
      "Epoch 00179: val_loss improved from 0.56902 to 0.56645, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 180/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6524 - accuracy: 0.8190 - val_loss: 0.5829 - val_accuracy: 0.8370\n",
      "\n",
      "Epoch 00180: val_loss did not improve from 0.56645\n",
      "Epoch 181/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6476 - accuracy: 0.8199 - val_loss: 0.5691 - val_accuracy: 0.8439\n",
      "\n",
      "Epoch 00181: val_loss did not improve from 0.56645\n",
      "Epoch 182/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6538 - accuracy: 0.8173 - val_loss: 0.5821 - val_accuracy: 0.8319\n",
      "\n",
      "Epoch 00182: val_loss did not improve from 0.56645\n",
      "Epoch 183/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.6447 - accuracy: 0.8207 - val_loss: 0.5718 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00183: val_loss did not improve from 0.56645\n",
      "Epoch 184/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6476 - accuracy: 0.8191 - val_loss: 0.5729 - val_accuracy: 0.8422\n",
      "\n",
      "Epoch 00184: val_loss did not improve from 0.56645\n",
      "Epoch 185/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6447 - accuracy: 0.8190 - val_loss: 0.5600 - val_accuracy: 0.8508\n",
      "\n",
      "Epoch 00185: val_loss improved from 0.56645 to 0.55997, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 186/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6357 - accuracy: 0.8218 - val_loss: 0.5673 - val_accuracy: 0.8491\n",
      "\n",
      "Epoch 00186: val_loss did not improve from 0.55997\n",
      "Epoch 187/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6529 - accuracy: 0.8160 - val_loss: 0.5581 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00187: val_loss improved from 0.55997 to 0.55808, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 188/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6398 - accuracy: 0.8270 - val_loss: 0.5645 - val_accuracy: 0.8491\n",
      "\n",
      "Epoch 00188: val_loss did not improve from 0.55808\n",
      "Epoch 189/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6409 - accuracy: 0.8212 - val_loss: 0.5563 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00189: val_loss improved from 0.55808 to 0.55630, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 190/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6483 - accuracy: 0.8145 - val_loss: 0.5531 - val_accuracy: 0.8388\n",
      "\n",
      "Epoch 00190: val_loss improved from 0.55630 to 0.55314, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 191/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6431 - accuracy: 0.8202 - val_loss: 0.5604 - val_accuracy: 0.8302\n",
      "\n",
      "Epoch 00191: val_loss did not improve from 0.55314\n",
      "Epoch 192/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6440 - accuracy: 0.8170 - val_loss: 0.5524 - val_accuracy: 0.8576\n",
      "\n",
      "Epoch 00192: val_loss improved from 0.55314 to 0.55244, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 193/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6354 - accuracy: 0.8220 - val_loss: 0.5490 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00193: val_loss improved from 0.55244 to 0.54899, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 194/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6234 - accuracy: 0.8277 - val_loss: 0.5453 - val_accuracy: 0.8491\n",
      "\n",
      "Epoch 00194: val_loss improved from 0.54899 to 0.54531, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 195/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6200 - accuracy: 0.8290 - val_loss: 0.5522 - val_accuracy: 0.8593\n",
      "\n",
      "Epoch 00195: val_loss did not improve from 0.54531\n",
      "Epoch 196/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6215 - accuracy: 0.8288 - val_loss: 0.5495 - val_accuracy: 0.8525\n",
      "\n",
      "Epoch 00196: val_loss did not improve from 0.54531\n",
      "Epoch 197/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6190 - accuracy: 0.8290 - val_loss: 0.5396 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00197: val_loss improved from 0.54531 to 0.53959, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 198/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6287 - accuracy: 0.8265 - val_loss: 0.5367 - val_accuracy: 0.8628\n",
      "\n",
      "Epoch 00198: val_loss improved from 0.53959 to 0.53667, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 199/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6195 - accuracy: 0.8312 - val_loss: 0.5404 - val_accuracy: 0.8508\n",
      "\n",
      "Epoch 00199: val_loss did not improve from 0.53667\n",
      "Epoch 200/358\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.6243 - accuracy: 0.8296 - val_loss: 0.5430 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00200: val_loss did not improve from 0.53667\n",
      "Epoch 201/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6204 - accuracy: 0.8284 - val_loss: 0.5401 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00201: val_loss did not improve from 0.53667\n",
      "Epoch 202/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6075 - accuracy: 0.8334 - val_loss: 0.5376 - val_accuracy: 0.8508\n",
      "\n",
      "Epoch 00202: val_loss did not improve from 0.53667\n",
      "Epoch 203/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6136 - accuracy: 0.8301 - val_loss: 0.5345 - val_accuracy: 0.8559\n",
      "\n",
      "Epoch 00203: val_loss improved from 0.53667 to 0.53453, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 204/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6030 - accuracy: 0.8370 - val_loss: 0.5394 - val_accuracy: 0.8628\n",
      "\n",
      "Epoch 00204: val_loss did not improve from 0.53453\n",
      "Epoch 205/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6032 - accuracy: 0.8330 - val_loss: 0.5396 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00205: val_loss did not improve from 0.53453\n",
      "Epoch 206/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6018 - accuracy: 0.8295 - val_loss: 0.5443 - val_accuracy: 0.8439\n",
      "\n",
      "Epoch 00206: val_loss did not improve from 0.53453\n",
      "Epoch 207/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6095 - accuracy: 0.8296 - val_loss: 0.5335 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00207: val_loss improved from 0.53453 to 0.53348, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 208/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.6102 - accuracy: 0.8291 - val_loss: 0.5361 - val_accuracy: 0.8491\n",
      "\n",
      "Epoch 00208: val_loss did not improve from 0.53348\n",
      "Epoch 209/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.6139 - accuracy: 0.8243 - val_loss: 0.5354 - val_accuracy: 0.8559\n",
      "\n",
      "Epoch 00209: val_loss did not improve from 0.53348\n",
      "Epoch 210/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6175 - accuracy: 0.8296 - val_loss: 0.5182 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00210: val_loss improved from 0.53348 to 0.51816, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 211/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6017 - accuracy: 0.8324 - val_loss: 0.5173 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00211: val_loss improved from 0.51816 to 0.51730, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 212/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.6103 - accuracy: 0.8301 - val_loss: 0.5180 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00212: val_loss did not improve from 0.51730\n",
      "Epoch 213/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5978 - accuracy: 0.8334 - val_loss: 0.5334 - val_accuracy: 0.8611\n",
      "\n",
      "Epoch 00213: val_loss did not improve from 0.51730\n",
      "Epoch 214/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6136 - accuracy: 0.8285 - val_loss: 0.5373 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00214: val_loss did not improve from 0.51730\n",
      "Epoch 215/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6011 - accuracy: 0.8357 - val_loss: 0.5451 - val_accuracy: 0.8576\n",
      "\n",
      "Epoch 00215: val_loss did not improve from 0.51730\n",
      "Epoch 216/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.6096 - accuracy: 0.8257 - val_loss: 0.5440 - val_accuracy: 0.8525\n",
      "\n",
      "Epoch 00216: val_loss did not improve from 0.51730\n",
      "Epoch 217/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5970 - accuracy: 0.8379 - val_loss: 0.5423 - val_accuracy: 0.8542\n",
      "\n",
      "Epoch 00217: val_loss did not improve from 0.51730\n",
      "Epoch 218/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5957 - accuracy: 0.8345 - val_loss: 0.5337 - val_accuracy: 0.8491\n",
      "\n",
      "Epoch 00218: val_loss did not improve from 0.51730\n",
      "Epoch 219/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5954 - accuracy: 0.8332 - val_loss: 0.5251 - val_accuracy: 0.8525\n",
      "\n",
      "Epoch 00219: val_loss did not improve from 0.51730\n",
      "Epoch 220/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5897 - accuracy: 0.8412 - val_loss: 0.5256 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00220: val_loss did not improve from 0.51730\n",
      "Epoch 221/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5954 - accuracy: 0.8329 - val_loss: 0.5297 - val_accuracy: 0.8593\n",
      "\n",
      "Epoch 00221: val_loss did not improve from 0.51730\n",
      "Epoch 222/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5884 - accuracy: 0.8384 - val_loss: 0.5248 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00222: val_loss did not improve from 0.51730\n",
      "Epoch 223/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.5923 - accuracy: 0.8343 - val_loss: 0.5180 - val_accuracy: 0.8456\n",
      "\n",
      "Epoch 00223: val_loss did not improve from 0.51730\n",
      "Epoch 224/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5812 - accuracy: 0.8385 - val_loss: 0.5033 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00224: val_loss improved from 0.51730 to 0.50331, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 225/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5878 - accuracy: 0.8388 - val_loss: 0.5112 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00225: val_loss did not improve from 0.50331\n",
      "Epoch 226/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5835 - accuracy: 0.8441 - val_loss: 0.5134 - val_accuracy: 0.8611\n",
      "\n",
      "Epoch 00226: val_loss did not improve from 0.50331\n",
      "Epoch 227/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5740 - accuracy: 0.8421 - val_loss: 0.5113 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00227: val_loss did not improve from 0.50331\n",
      "Epoch 228/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5802 - accuracy: 0.8348 - val_loss: 0.5071 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00228: val_loss did not improve from 0.50331\n",
      "Epoch 229/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5764 - accuracy: 0.8462 - val_loss: 0.5106 - val_accuracy: 0.8628\n",
      "\n",
      "Epoch 00229: val_loss did not improve from 0.50331\n",
      "Epoch 230/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5720 - accuracy: 0.8446 - val_loss: 0.5021 - val_accuracy: 0.8611\n",
      "\n",
      "Epoch 00230: val_loss improved from 0.50331 to 0.50211, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 231/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5659 - accuracy: 0.8423 - val_loss: 0.5019 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00231: val_loss improved from 0.50211 to 0.50192, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 232/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5689 - accuracy: 0.8413 - val_loss: 0.5005 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00232: val_loss improved from 0.50192 to 0.50047, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 233/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5669 - accuracy: 0.8440 - val_loss: 0.4878 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00233: val_loss improved from 0.50047 to 0.48775, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 234/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5639 - accuracy: 0.8423 - val_loss: 0.4918 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00234: val_loss did not improve from 0.48775\n",
      "Epoch 235/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5596 - accuracy: 0.8474 - val_loss: 0.4941 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00235: val_loss did not improve from 0.48775\n",
      "Epoch 236/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5687 - accuracy: 0.8451 - val_loss: 0.4885 - val_accuracy: 0.8765\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00236: val_loss did not improve from 0.48775\n",
      "Epoch 237/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5740 - accuracy: 0.8421 - val_loss: 0.4944 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00237: val_loss did not improve from 0.48775\n",
      "Epoch 238/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5610 - accuracy: 0.8484 - val_loss: 0.5049 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00238: val_loss did not improve from 0.48775\n",
      "Epoch 239/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5584 - accuracy: 0.8395 - val_loss: 0.4858 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00239: val_loss improved from 0.48775 to 0.48578, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 240/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5580 - accuracy: 0.8469 - val_loss: 0.4896 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00240: val_loss did not improve from 0.48578\n",
      "Epoch 241/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5574 - accuracy: 0.8460 - val_loss: 0.5021 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00241: val_loss did not improve from 0.48578\n",
      "Epoch 242/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5542 - accuracy: 0.8521 - val_loss: 0.4953 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00242: val_loss did not improve from 0.48578\n",
      "Epoch 243/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5547 - accuracy: 0.8498 - val_loss: 0.4960 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00243: val_loss did not improve from 0.48578\n",
      "Epoch 244/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5612 - accuracy: 0.8405 - val_loss: 0.4969 - val_accuracy: 0.8645\n",
      "\n",
      "Epoch 00244: val_loss did not improve from 0.48578\n",
      "Epoch 245/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5524 - accuracy: 0.8465 - val_loss: 0.4981 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00245: val_loss did not improve from 0.48578\n",
      "Epoch 246/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5458 - accuracy: 0.8474 - val_loss: 0.4966 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00246: val_loss did not improve from 0.48578\n",
      "Epoch 247/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5484 - accuracy: 0.8532 - val_loss: 0.4876 - val_accuracy: 0.8731\n",
      "\n",
      "Epoch 00247: val_loss did not improve from 0.48578\n",
      "Epoch 248/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5475 - accuracy: 0.8455 - val_loss: 0.4947 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00248: val_loss did not improve from 0.48578\n",
      "Epoch 249/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5448 - accuracy: 0.8543 - val_loss: 0.4846 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00249: val_loss improved from 0.48578 to 0.48461, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 250/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5366 - accuracy: 0.8532 - val_loss: 0.4876 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00250: val_loss did not improve from 0.48461\n",
      "Epoch 251/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5389 - accuracy: 0.8512 - val_loss: 0.4870 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00251: val_loss did not improve from 0.48461\n",
      "Epoch 252/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5361 - accuracy: 0.8537 - val_loss: 0.4835 - val_accuracy: 0.8679\n",
      "\n",
      "Epoch 00252: val_loss improved from 0.48461 to 0.48352, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 253/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5394 - accuracy: 0.8469 - val_loss: 0.4778 - val_accuracy: 0.8731\n",
      "\n",
      "Epoch 00253: val_loss improved from 0.48352 to 0.47782, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 254/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5414 - accuracy: 0.8519 - val_loss: 0.4828 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00254: val_loss did not improve from 0.47782\n",
      "Epoch 255/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5421 - accuracy: 0.8501 - val_loss: 0.4876 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00255: val_loss did not improve from 0.47782\n",
      "Epoch 256/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5439 - accuracy: 0.8469 - val_loss: 0.4915 - val_accuracy: 0.8748\n",
      "\n",
      "Epoch 00256: val_loss did not improve from 0.47782\n",
      "Epoch 257/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5352 - accuracy: 0.8543 - val_loss: 0.4757 - val_accuracy: 0.8748\n",
      "\n",
      "Epoch 00257: val_loss improved from 0.47782 to 0.47571, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 258/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5493 - accuracy: 0.8505 - val_loss: 0.4814 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00258: val_loss did not improve from 0.47571\n",
      "Epoch 259/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5483 - accuracy: 0.8504 - val_loss: 0.4889 - val_accuracy: 0.8748\n",
      "\n",
      "Epoch 00259: val_loss did not improve from 0.47571\n",
      "Epoch 260/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5467 - accuracy: 0.8454 - val_loss: 0.4867 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00260: val_loss did not improve from 0.47571\n",
      "Epoch 261/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5425 - accuracy: 0.8533 - val_loss: 0.4786 - val_accuracy: 0.8696\n",
      "\n",
      "Epoch 00261: val_loss did not improve from 0.47571\n",
      "Epoch 262/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5312 - accuracy: 0.8533 - val_loss: 0.4796 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00262: val_loss did not improve from 0.47571\n",
      "Epoch 263/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5333 - accuracy: 0.8552 - val_loss: 0.4742 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00263: val_loss improved from 0.47571 to 0.47423, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 264/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5328 - accuracy: 0.8541 - val_loss: 0.4751 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00264: val_loss did not improve from 0.47423\n",
      "Epoch 265/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5341 - accuracy: 0.8552 - val_loss: 0.4725 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00265: val_loss improved from 0.47423 to 0.47245, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 266/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5220 - accuracy: 0.8618 - val_loss: 0.4768 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00266: val_loss did not improve from 0.47245\n",
      "Epoch 267/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5188 - accuracy: 0.8574 - val_loss: 0.4622 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00267: val_loss improved from 0.47245 to 0.46225, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 268/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5266 - accuracy: 0.8549 - val_loss: 0.4674 - val_accuracy: 0.8731\n",
      "\n",
      "Epoch 00268: val_loss did not improve from 0.46225\n",
      "Epoch 269/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5279 - accuracy: 0.8526 - val_loss: 0.4662 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00269: val_loss did not improve from 0.46225\n",
      "Epoch 270/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5283 - accuracy: 0.8543 - val_loss: 0.4665 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00270: val_loss did not improve from 0.46225\n",
      "Epoch 271/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5323 - accuracy: 0.8588 - val_loss: 0.4641 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00271: val_loss did not improve from 0.46225\n",
      "Epoch 272/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5214 - accuracy: 0.8574 - val_loss: 0.4699 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00272: val_loss did not improve from 0.46225\n",
      "Epoch 273/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5271 - accuracy: 0.8577 - val_loss: 0.4593 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00273: val_loss improved from 0.46225 to 0.45926, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 274/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5208 - accuracy: 0.8565 - val_loss: 0.4562 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00274: val_loss improved from 0.45926 to 0.45619, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 275/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5103 - accuracy: 0.8608 - val_loss: 0.4633 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00275: val_loss did not improve from 0.45619\n",
      "Epoch 276/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5148 - accuracy: 0.8594 - val_loss: 0.4571 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00276: val_loss did not improve from 0.45619\n",
      "Epoch 277/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5173 - accuracy: 0.8573 - val_loss: 0.4533 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00277: val_loss improved from 0.45619 to 0.45333, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 278/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5154 - accuracy: 0.8599 - val_loss: 0.4609 - val_accuracy: 0.8765\n",
      "\n",
      "Epoch 00278: val_loss did not improve from 0.45333\n",
      "Epoch 279/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5176 - accuracy: 0.8644 - val_loss: 0.4585 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00279: val_loss did not improve from 0.45333\n",
      "Epoch 280/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5221 - accuracy: 0.8571 - val_loss: 0.4572 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00280: val_loss did not improve from 0.45333\n",
      "Epoch 281/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5177 - accuracy: 0.8587 - val_loss: 0.4512 - val_accuracy: 0.8679\n",
      "\n",
      "Epoch 00281: val_loss improved from 0.45333 to 0.45121, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 282/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5123 - accuracy: 0.8588 - val_loss: 0.4637 - val_accuracy: 0.8748\n",
      "\n",
      "Epoch 00282: val_loss did not improve from 0.45121\n",
      "Epoch 283/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5211 - accuracy: 0.8582 - val_loss: 0.4506 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00283: val_loss improved from 0.45121 to 0.45056, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 284/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5099 - accuracy: 0.8598 - val_loss: 0.4564 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00284: val_loss did not improve from 0.45056\n",
      "Epoch 285/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5190 - accuracy: 0.8588 - val_loss: 0.4686 - val_accuracy: 0.8662\n",
      "\n",
      "Epoch 00285: val_loss did not improve from 0.45056\n",
      "Epoch 286/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5175 - accuracy: 0.8577 - val_loss: 0.4577 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00286: val_loss did not improve from 0.45056\n",
      "Epoch 287/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4963 - accuracy: 0.8677 - val_loss: 0.4563 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00287: val_loss did not improve from 0.45056\n",
      "Epoch 288/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5085 - accuracy: 0.8599 - val_loss: 0.4627 - val_accuracy: 0.8748\n",
      "\n",
      "Epoch 00288: val_loss did not improve from 0.45056\n",
      "Epoch 289/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5052 - accuracy: 0.8646 - val_loss: 0.4490 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00289: val_loss improved from 0.45056 to 0.44904, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 290/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5035 - accuracy: 0.8651 - val_loss: 0.4419 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00290: val_loss improved from 0.44904 to 0.44190, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 291/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.5027 - accuracy: 0.8640 - val_loss: 0.4557 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00291: val_loss did not improve from 0.44190\n",
      "Epoch 292/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4974 - accuracy: 0.8652 - val_loss: 0.4487 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00292: val_loss did not improve from 0.44190\n",
      "Epoch 293/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4987 - accuracy: 0.8638 - val_loss: 0.4448 - val_accuracy: 0.8714\n",
      "\n",
      "Epoch 00293: val_loss did not improve from 0.44190\n",
      "Epoch 294/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5049 - accuracy: 0.8633 - val_loss: 0.4634 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00294: val_loss did not improve from 0.44190\n",
      "Epoch 295/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.5036 - accuracy: 0.8640 - val_loss: 0.4546 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00295: val_loss did not improve from 0.44190\n",
      "Epoch 296/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5003 - accuracy: 0.8679 - val_loss: 0.4665 - val_accuracy: 0.8611\n",
      "\n",
      "Epoch 00296: val_loss did not improve from 0.44190\n",
      "Epoch 297/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.5030 - accuracy: 0.8587 - val_loss: 0.4525 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00297: val_loss did not improve from 0.44190\n",
      "Epoch 298/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4928 - accuracy: 0.8677 - val_loss: 0.4419 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00298: val_loss improved from 0.44190 to 0.44190, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 299/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4996 - accuracy: 0.8655 - val_loss: 0.4453 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00299: val_loss did not improve from 0.44190\n",
      "Epoch 300/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4955 - accuracy: 0.8644 - val_loss: 0.4391 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00300: val_loss improved from 0.44190 to 0.43905, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 301/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4963 - accuracy: 0.8629 - val_loss: 0.4418 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00301: val_loss did not improve from 0.43905\n",
      "Epoch 302/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4877 - accuracy: 0.8657 - val_loss: 0.4328 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00302: val_loss improved from 0.43905 to 0.43276, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 303/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4993 - accuracy: 0.8635 - val_loss: 0.4598 - val_accuracy: 0.8799\n",
      "\n",
      "Epoch 00303: val_loss did not improve from 0.43276\n",
      "Epoch 304/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4927 - accuracy: 0.8644 - val_loss: 0.4515 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00304: val_loss did not improve from 0.43276\n",
      "Epoch 305/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4890 - accuracy: 0.8649 - val_loss: 0.4407 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00305: val_loss did not improve from 0.43276\n",
      "Epoch 306/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4838 - accuracy: 0.8705 - val_loss: 0.4296 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00306: val_loss improved from 0.43276 to 0.42964, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 307/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4853 - accuracy: 0.8657 - val_loss: 0.4289 - val_accuracy: 0.8937\n",
      "\n",
      "Epoch 00307: val_loss improved from 0.42964 to 0.42887, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 308/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4834 - accuracy: 0.8719 - val_loss: 0.4362 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00308: val_loss did not improve from 0.42887\n",
      "Epoch 309/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.4856 - accuracy: 0.8732 - val_loss: 0.4310 - val_accuracy: 0.8868\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00309: val_loss did not improve from 0.42887\n",
      "Epoch 310/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4770 - accuracy: 0.8721 - val_loss: 0.4305 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00310: val_loss did not improve from 0.42887\n",
      "Epoch 311/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4836 - accuracy: 0.8635 - val_loss: 0.4398 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00311: val_loss did not improve from 0.42887\n",
      "Epoch 312/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4752 - accuracy: 0.8696 - val_loss: 0.4270 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00312: val_loss improved from 0.42887 to 0.42698, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 313/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4745 - accuracy: 0.8719 - val_loss: 0.4233 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00313: val_loss improved from 0.42698 to 0.42326, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 314/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4819 - accuracy: 0.8680 - val_loss: 0.4247 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00314: val_loss did not improve from 0.42326\n",
      "Epoch 315/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.4898 - accuracy: 0.8649 - val_loss: 0.4333 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00315: val_loss did not improve from 0.42326\n",
      "Epoch 316/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4803 - accuracy: 0.8707 - val_loss: 0.4340 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00316: val_loss did not improve from 0.42326\n",
      "Epoch 317/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4792 - accuracy: 0.8708 - val_loss: 0.4243 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00317: val_loss did not improve from 0.42326\n",
      "Epoch 318/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4710 - accuracy: 0.8763 - val_loss: 0.4334 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00318: val_loss did not improve from 0.42326\n",
      "Epoch 319/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4770 - accuracy: 0.8721 - val_loss: 0.4312 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00319: val_loss did not improve from 0.42326\n",
      "Epoch 320/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4749 - accuracy: 0.8744 - val_loss: 0.4236 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00320: val_loss did not improve from 0.42326\n",
      "Epoch 321/358\n",
      "6403/6403 [==============================] - 3s 449us/step - loss: 0.4720 - accuracy: 0.8719 - val_loss: 0.4182 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00321: val_loss improved from 0.42326 to 0.41820, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 322/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4773 - accuracy: 0.8660 - val_loss: 0.4294 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00322: val_loss did not improve from 0.41820\n",
      "Epoch 323/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.4836 - accuracy: 0.8679 - val_loss: 0.4250 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00323: val_loss did not improve from 0.41820\n",
      "Epoch 324/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4774 - accuracy: 0.8713 - val_loss: 0.4221 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00324: val_loss did not improve from 0.41820\n",
      "Epoch 325/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4722 - accuracy: 0.8694 - val_loss: 0.4308 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00325: val_loss did not improve from 0.41820\n",
      "Epoch 326/358\n",
      "6403/6403 [==============================] - 3s 449us/step - loss: 0.4682 - accuracy: 0.8708 - val_loss: 0.4186 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00326: val_loss did not improve from 0.41820\n",
      "Epoch 327/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.4702 - accuracy: 0.8729 - val_loss: 0.4099 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00327: val_loss improved from 0.41820 to 0.40989, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 328/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4696 - accuracy: 0.8680 - val_loss: 0.4183 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00328: val_loss did not improve from 0.40989\n",
      "Epoch 329/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4681 - accuracy: 0.8697 - val_loss: 0.4249 - val_accuracy: 0.8971\n",
      "\n",
      "Epoch 00329: val_loss did not improve from 0.40989\n",
      "Epoch 330/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4622 - accuracy: 0.8754 - val_loss: 0.4271 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00330: val_loss did not improve from 0.40989\n",
      "Epoch 331/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4681 - accuracy: 0.8754 - val_loss: 0.4110 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00331: val_loss did not improve from 0.40989\n",
      "Epoch 332/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4631 - accuracy: 0.8765 - val_loss: 0.4082 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00332: val_loss improved from 0.40989 to 0.40824, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 333/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4720 - accuracy: 0.8696 - val_loss: 0.4308 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00333: val_loss did not improve from 0.40824\n",
      "Epoch 334/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4729 - accuracy: 0.8738 - val_loss: 0.4265 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00334: val_loss did not improve from 0.40824\n",
      "Epoch 335/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4628 - accuracy: 0.8765 - val_loss: 0.4140 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00335: val_loss did not improve from 0.40824\n",
      "Epoch 336/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4689 - accuracy: 0.8737 - val_loss: 0.4117 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00336: val_loss did not improve from 0.40824\n",
      "Epoch 337/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4585 - accuracy: 0.8740 - val_loss: 0.4196 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00337: val_loss did not improve from 0.40824\n",
      "Epoch 338/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4621 - accuracy: 0.8732 - val_loss: 0.4161 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00338: val_loss did not improve from 0.40824\n",
      "Epoch 339/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4598 - accuracy: 0.8772 - val_loss: 0.4356 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00339: val_loss did not improve from 0.40824\n",
      "Epoch 340/358\n",
      "6403/6403 [==============================] - 3s 448us/step - loss: 0.4675 - accuracy: 0.8760 - val_loss: 0.4194 - val_accuracy: 0.8782\n",
      "\n",
      "Epoch 00340: val_loss did not improve from 0.40824\n",
      "Epoch 341/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4637 - accuracy: 0.8746 - val_loss: 0.4148 - val_accuracy: 0.8937\n",
      "\n",
      "Epoch 00341: val_loss did not improve from 0.40824\n",
      "Epoch 342/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4562 - accuracy: 0.8730 - val_loss: 0.4147 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00342: val_loss did not improve from 0.40824\n",
      "Epoch 343/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4547 - accuracy: 0.8780 - val_loss: 0.4034 - val_accuracy: 0.8919\n",
      "\n",
      "Epoch 00343: val_loss improved from 0.40824 to 0.40342, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 344/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4525 - accuracy: 0.8776 - val_loss: 0.4058 - val_accuracy: 0.8988\n",
      "\n",
      "Epoch 00344: val_loss did not improve from 0.40342\n",
      "Epoch 345/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4443 - accuracy: 0.8852 - val_loss: 0.4051 - val_accuracy: 0.9022\n",
      "\n",
      "Epoch 00345: val_loss did not improve from 0.40342\n",
      "Epoch 346/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4559 - accuracy: 0.8790 - val_loss: 0.4245 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00346: val_loss did not improve from 0.40342\n",
      "Epoch 347/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4610 - accuracy: 0.8766 - val_loss: 0.4189 - val_accuracy: 0.8885\n",
      "\n",
      "Epoch 00347: val_loss did not improve from 0.40342\n",
      "Epoch 348/358\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4643 - accuracy: 0.8755 - val_loss: 0.4224 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00348: val_loss did not improve from 0.40342\n",
      "Epoch 349/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4623 - accuracy: 0.8737 - val_loss: 0.3998 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00349: val_loss improved from 0.40342 to 0.39983, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Epoch 350/358\n",
      "6403/6403 [==============================] - 3s 445us/step - loss: 0.4584 - accuracy: 0.8776 - val_loss: 0.4171 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00350: val_loss did not improve from 0.39983\n",
      "Epoch 351/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4454 - accuracy: 0.8787 - val_loss: 0.4207 - val_accuracy: 0.8868\n",
      "\n",
      "Epoch 00351: val_loss did not improve from 0.39983\n",
      "Epoch 352/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4518 - accuracy: 0.8738 - val_loss: 0.4210 - val_accuracy: 0.8902\n",
      "\n",
      "Epoch 00352: val_loss did not improve from 0.39983\n",
      "Epoch 353/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4582 - accuracy: 0.8765 - val_loss: 0.4371 - val_accuracy: 0.8816\n",
      "\n",
      "Epoch 00353: val_loss did not improve from 0.39983\n",
      "Epoch 354/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4431 - accuracy: 0.8779 - val_loss: 0.4369 - val_accuracy: 0.8834\n",
      "\n",
      "Epoch 00354: val_loss did not improve from 0.39983\n",
      "Epoch 355/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4423 - accuracy: 0.8796 - val_loss: 0.4096 - val_accuracy: 0.8851\n",
      "\n",
      "Epoch 00355: val_loss did not improve from 0.39983\n",
      "Epoch 356/358\n",
      "6403/6403 [==============================] - 3s 447us/step - loss: 0.4406 - accuracy: 0.8818 - val_loss: 0.4051 - val_accuracy: 0.9005\n",
      "\n",
      "Epoch 00356: val_loss did not improve from 0.39983\n",
      "Epoch 357/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4503 - accuracy: 0.8801 - val_loss: 0.4044 - val_accuracy: 0.8988\n",
      "\n",
      "Epoch 00357: val_loss did not improve from 0.39983\n",
      "Epoch 358/358\n",
      "6403/6403 [==============================] - 3s 446us/step - loss: 0.4411 - accuracy: 0.8810 - val_loss: 0.3997 - val_accuracy: 0.9005\n",
      "\n",
      "Epoch 00358: val_loss improved from 0.39983 to 0.39970, saving model to /home/edu/Projects/urban-audio-classifier/models/simple-train-nb3.hdf5\n",
      "Training completed in time:  0:17:16.401907\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 360\n",
    "num_batch_size = 128\n",
    "model_file = 'simple-train-nb3.hdf5'\n",
    "model_path = os.path.join(models_path, model_file)\n",
    "\n",
    "\n",
    "# Save checkpoints\n",
    "checkpointer = ModelCheckpoint(filepath=model_path, \n",
    "                               verbose=1, \n",
    "                               save_best_only=True)\n",
    "start = datetime.now()\n",
    "history = model.fit(X_train, \n",
    "                    y_train_encoded, \n",
    "                    batch_size=num_batch_size, \n",
    "                    epochs=num_epochs, \n",
    "                    validation_split=1/12.,\n",
    "                    callbacks=[checkpointer], \n",
    "                    verbose=1)\n",
    "\n",
    "duration = datetime.now() - start\n",
    "print(\"Training completed in time: \", duration)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.4 General model evaluation\n",
    "Note that we first load the checkpoint with the lowest validation loss achieved during training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                    LOSS      ACCURACY\n",
      "--------------------------------------\n",
      "Training:         0.3556       92.0126\n",
      "Test:             0.3686       91.6380\n",
      "Normal diff        3.64               \n"
     ]
    }
   ],
   "source": [
    "# Load best saved model\n",
    "model = load_model(model_path)\n",
    "\n",
    "helpers.model_evaluation_report(model, X_train, y_train_encoded, X_test, y_test_encoded)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Compared with the previous evaluation the model was able to train for more epochs when using Mel Spectrograms, and consequently, a higher accuracy and lower error where achieved.<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.5 Train vs Test history plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "nbpresent": {
     "id": "19f196f8-c2b5-4c56-ba54-4c145b7023ad"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "helpers.plot_train_history(history, x_ticks_vertical=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Although the shape of the curves are similar to the ones of the previous experiment, the generalization gap is thinner. If we look at the error minimums we also find out that there is a smaller difference between validation and train errors.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "nbpresent": {
     "id": "765aa27d-e4da-45a9-a1d2-e13085b275a8"
    }
   },
   "source": [
    "### 3. Model evaluation\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### * Register model output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/edu/anaconda3/envs/tf-gpu/lib/python3.6/site-packages/ipykernel_launcher.py:9: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  if __name__ == '__main__':\n"
     ]
    }
   ],
   "source": [
    "# Predict probabilities for test set\n",
    "y_probs = model.predict(X_test, verbose=0)\n",
    "\n",
    "# Get predicted labels\n",
    "yhat_probs = np.argmax(y_probs, axis=1)\n",
    "y_trues = np.argmax(y_test_encoded, axis=1)\n",
    "\n",
    "# Add \"pred\" column\n",
    "test_meta['pred'] = yhat_probs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.1 Confussion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import importlib\n",
    "importlib.reload(helpers)\n",
    "\n",
    "# Sets decimal precision (for printing output only)\n",
    "np.set_printoptions(precision=2)\n",
    "\n",
    "# Compute confusion matrix data\n",
    "cm = confusion_matrix(y_trues, yhat_probs)\n",
    "\n",
    "helpers.plot_confusion_matrix(cm,\n",
    "                          labels, \n",
    "                          normalized=False, \n",
    "                          title=\"Model Performance\", \n",
    "                          cmap=plt.cm.Blues,\n",
    "                          size=(12,12))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The patterns found on working with MFCC are replicated, but most of the true positives have increased and false negatives have decreased.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.2 Accuracy table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CLASS</th>\n",
       "      <th>ACCURACY</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>Gun Shot</td>\n",
       "      <td>97.619048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>Drilling</td>\n",
       "      <td>95.212766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>Air Conditioner</td>\n",
       "      <td>95.192308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>Jackhammer</td>\n",
       "      <td>94.036697</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>Siren</td>\n",
       "      <td>92.708333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>Engine Idling</td>\n",
       "      <td>91.397849</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>Car Horn</td>\n",
       "      <td>90.361446</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>Dog bark</td>\n",
       "      <td>87.096774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>Street Music</td>\n",
       "      <td>85.863874</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>Children Playing</td>\n",
       "      <td>83.809524</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              CLASS   ACCURACY\n",
       "6          Gun Shot  97.619048\n",
       "4          Drilling  95.212766\n",
       "0   Air Conditioner  95.192308\n",
       "7        Jackhammer  94.036697\n",
       "8             Siren  92.708333\n",
       "5     Engine Idling  91.397849\n",
       "1          Car Horn  90.361446\n",
       "3          Dog bark  87.096774\n",
       "9      Street Music  85.863874\n",
       "2  Children Playing  83.809524"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Find per-class accuracy from the confusion matrix data\n",
    "accuracies = helpers.acc_per_class(cm)\n",
    "\n",
    "pd.DataFrame({\n",
    "    'CLASS': labels,\n",
    "    'ACCURACY': accuracies\n",
    "}).sort_values(by=\"ACCURACY\", ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In a similar fashion, most class-accuracies have increased."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.3 Classification performance: Presicion, Recall and F1 scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                  precision    recall  f1-score   support\n",
      "\n",
      " Air Conditioner       0.87      0.95      0.91       208\n",
      "        Car Horn       0.90      0.90      0.90        83\n",
      "Children Playing       0.88      0.84      0.86       210\n",
      "        Dog bark       0.88      0.87      0.87       186\n",
      "        Drilling       0.93      0.95      0.94       188\n",
      "   Engine Idling       0.97      0.91      0.94       186\n",
      "        Gun Shot       0.92      0.98      0.95        84\n",
      "      Jackhammer       0.96      0.94      0.95       218\n",
      "           Siren       0.98      0.93      0.95       192\n",
      "    Street Music       0.84      0.86      0.85       191\n",
      "\n",
      "        accuracy                           0.91      1746\n",
      "       macro avg       0.91      0.91      0.91      1746\n",
      "    weighted avg       0.91      0.91      0.91      1746\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Build classification report\n",
    "re = classification_report(y_trues, yhat_probs, labels=[0,1,2,3,4,5,6,7,8,9], target_names=labels)\n",
    "\n",
    "print(re)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Scores have increased a bit in average, we still see the same general patterns.\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Conclusions\n",
    "\n",
    "Log scaled Mel Spectrograms did work better than MFCC coefficients in our experiments, as it also happened in some of the papers I referenced. On a different experiment I also used **128** MEL bands (the usual number of bands used on papers that use Mel Spectrograms), in my particular case I didn't notice a difference that could justify processing so many more channels (+88) per sample.<br/>\n",
    "<br/> \n",
    "\n",
    "### 5. Next\n",
    "\n",
    "We already found out that Mel Spectrograms features work better, now we want to see how much this model can improve if we use augmented data.<br/>\n",
    "<br/>\n",
    "<br/>"
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}