7-cnn-model-augmented.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": [
    "# Training CNN model with augmented data\n",
    "\n",
    "We are now able to train our basic -but yet well performing- CNN architecture with the new augmented data. While the initial dataset had 8,732 samples, the augmented dataset has 69,856, that is 8x more information than previous experiments.<br/>\n",
    "We'll use Log-Mel Spectrograms as features because they showed to perform better than MFCC in our experiments.<br/>\n",
    "<br/>\n",
    "As an important note, we are going to train from augmented data but we are only going to test with samples from the original dataset: we want to know how the augmented data helps to improve predictions of the original data.<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": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define general variables\n",
    "\n",
    "# Set your path to the dataset\n",
    "us8k_path = os.path.abspath('./UrbanSound8K')\n",
    "orig_metadata_path = os.path.join(us8k_path, 'metadata/UrbanSound8K.csv')\n",
    "aug_metadata_path = os.path.abspath('data/augmented-data.csv')\n",
    "models_path = os.path.abspath('./models')\n",
    "data_path = os.path.abspath('./data')\n",
    "audio_path = os.path.join(us8k_path, 'audio')\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": 5,
   "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"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "nbpresent": {
     "id": "87d3ebd1-0d75-4ea4-8f12-9f891a72d3e2"
    }
   },
   "outputs": [],
   "source": [
    "# Pre-processed MEL SPEC coefficients\n",
    "X = np.load(\"data/X-mel_spec-augmented.npy\")\n",
    "y = np.load(\"data/y-mel_spec-augmented.npy\")\n",
    "\n",
    "# Metadata\n",
    "metadata = pd.read_csv(orig_metadata_path)\n",
    "aug_metadata = pd.read_csv(aug_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",
    "\n",
    "We are doing split data with the mentioned criteria in an extremely simple fashion: splitting 20% of the augmented dataset into test and then removing all reference to augmented samples from it. We'll end up with an amount of test data that will always be around 1700+ samples.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test split: 1770 \t\t Train split: 55885\n",
      "X test shape: (12201, 40, 174) \t X train shape: (55885, 40, 174)\n",
      "y test shape: (12201,) \t\t y train shape: (55885,)\n"
     ]
    }
   ],
   "source": [
    "indexes = []\n",
    "total = len(aug_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 metadata\n",
    "test_meta = aug_metadata.iloc[test_split_idx]\n",
    "train_meta = aug_metadata.iloc[train_split_idx]\n",
    "\n",
    "# Remove augmented data from test metadata split\n",
    "test_meta = test_meta[test_meta['augment'] != 'none']\n",
    "aug_on_test_idx = test_meta.index.tolist()\n",
    "\n",
    "# Split the features with the same indexes\n",
    "X_test = np.take(X, aug_on_test_idx, axis=0)\n",
    "y_test = np.take(y, aug_on_test_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",
    "\n",
    "# Print status\n",
    "print(\"Test split: {} \\t\\t Train split: {}\".format(len(test_meta)-len(aug_on_test_idx), 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": 14,
   "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": 15,
   "metadata": {
    "nbpresent": {
     "id": "d742676a-5a23-4fc8-856a-ad5505238d45"
    }
   },
   "outputs": [],
   "source": [
    "# How data should be organized\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": {
    "nbpresent": {
     "id": "23ae43f0-1d59-44ab-818d-54faeecf4f1c"
    }
   },
   "source": [
    "### 2. Basic CNN model\n",
    "\n",
    "We are using the exact same model than the previous notebooks\n",
    "\n",
    "#### 2.1 Model definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "nbpresent": {
     "id": "845659a2-b4b4-4380-8d8f-51144a91ccab"
    },
    "scrolled": false
   },
   "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",
    "    # 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 = 0.001\n",
    "\n",
    "model = create_model(spatial_dropout_rate_1, spatial_dropout_rate_2, l2_rate)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.2 Setup model optimizer and loss function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "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": {},
   "source": [
    "#### 2.3 Compile model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 51227 samples, validate on 4658 samples\n",
      "Epoch 1/48\n",
      "51227/51227 [==============================] - 23s 458us/step - loss: 1.7937 - accuracy: 0.4203 - val_loss: 2.3999 - val_accuracy: 0.2645\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 2.39989, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 2/48\n",
      "51227/51227 [==============================] - 23s 445us/step - loss: 1.4683 - accuracy: 0.5375 - val_loss: 1.3264 - val_accuracy: 0.6069\n",
      "\n",
      "Epoch 00002: val_loss improved from 2.39989 to 1.32639, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 3/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 1.3209 - accuracy: 0.5906 - val_loss: 1.1690 - val_accuracy: 0.6597\n",
      "\n",
      "Epoch 00003: val_loss improved from 1.32639 to 1.16902, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 4/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 1.2217 - accuracy: 0.6303 - val_loss: 1.0841 - val_accuracy: 0.7097\n",
      "\n",
      "Epoch 00004: val_loss improved from 1.16902 to 1.08408, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 5/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 1.1400 - accuracy: 0.6627 - val_loss: 1.0136 - val_accuracy: 0.7359\n",
      "\n",
      "Epoch 00005: val_loss improved from 1.08408 to 1.01362, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 6/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 1.0739 - accuracy: 0.6894 - val_loss: 0.9484 - val_accuracy: 0.7563\n",
      "\n",
      "Epoch 00006: val_loss improved from 1.01362 to 0.94843, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 7/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 1.0206 - accuracy: 0.7087 - val_loss: 0.8849 - val_accuracy: 0.7688\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.94843 to 0.88489, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 8/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.9655 - accuracy: 0.7286 - val_loss: 0.8349 - val_accuracy: 0.7903\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.88489 to 0.83493, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 9/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.9224 - accuracy: 0.7426 - val_loss: 0.7912 - val_accuracy: 0.8031\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.83493 to 0.79115, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 10/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.8846 - accuracy: 0.7578 - val_loss: 0.7574 - val_accuracy: 0.8113\n",
      "\n",
      "Epoch 00010: val_loss improved from 0.79115 to 0.75737, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 11/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.8458 - accuracy: 0.7682 - val_loss: 0.7223 - val_accuracy: 0.8175\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.75737 to 0.72228, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 12/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.8127 - accuracy: 0.7801 - val_loss: 0.6889 - val_accuracy: 0.8356\n",
      "\n",
      "Epoch 00012: val_loss improved from 0.72228 to 0.68887, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 13/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.7847 - accuracy: 0.7882 - val_loss: 0.6597 - val_accuracy: 0.8431\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.68887 to 0.65974, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 14/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.7534 - accuracy: 0.7987 - val_loss: 0.6343 - val_accuracy: 0.8521\n",
      "\n",
      "Epoch 00014: val_loss improved from 0.65974 to 0.63432, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 15/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.7323 - accuracy: 0.8044 - val_loss: 0.6136 - val_accuracy: 0.8497\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.63432 to 0.61361, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 16/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.7075 - accuracy: 0.8123 - val_loss: 0.5870 - val_accuracy: 0.8665\n",
      "\n",
      "Epoch 00016: val_loss improved from 0.61361 to 0.58703, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 17/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.6824 - accuracy: 0.8201 - val_loss: 0.5706 - val_accuracy: 0.8669\n",
      "\n",
      "Epoch 00017: val_loss improved from 0.58703 to 0.57057, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 18/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.6638 - accuracy: 0.8245 - val_loss: 0.5490 - val_accuracy: 0.8708\n",
      "\n",
      "Epoch 00018: val_loss improved from 0.57057 to 0.54902, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 19/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.6439 - accuracy: 0.8297 - val_loss: 0.5235 - val_accuracy: 0.8802\n",
      "\n",
      "Epoch 00019: val_loss improved from 0.54902 to 0.52348, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 20/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.6212 - accuracy: 0.8401 - val_loss: 0.5036 - val_accuracy: 0.8841\n",
      "\n",
      "Epoch 00020: val_loss improved from 0.52348 to 0.50355, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 21/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.6070 - accuracy: 0.8417 - val_loss: 0.4895 - val_accuracy: 0.8897\n",
      "\n",
      "Epoch 00021: val_loss improved from 0.50355 to 0.48947, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 22/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.5904 - accuracy: 0.8458 - val_loss: 0.4841 - val_accuracy: 0.8843\n",
      "\n",
      "Epoch 00022: val_loss improved from 0.48947 to 0.48409, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 23/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.5736 - accuracy: 0.8513 - val_loss: 0.4661 - val_accuracy: 0.8914\n",
      "\n",
      "Epoch 00023: val_loss improved from 0.48409 to 0.46614, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 24/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.5605 - accuracy: 0.8539 - val_loss: 0.4486 - val_accuracy: 0.8939\n",
      "\n",
      "Epoch 00024: val_loss improved from 0.46614 to 0.44857, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 25/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.5443 - accuracy: 0.8589 - val_loss: 0.4364 - val_accuracy: 0.9000\n",
      "\n",
      "Epoch 00025: val_loss improved from 0.44857 to 0.43642, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 26/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.5354 - accuracy: 0.8608 - val_loss: 0.4306 - val_accuracy: 0.9008\n",
      "\n",
      "Epoch 00026: val_loss improved from 0.43642 to 0.43064, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 27/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.5190 - accuracy: 0.8658 - val_loss: 0.4172 - val_accuracy: 0.9036\n",
      "\n",
      "Epoch 00027: val_loss improved from 0.43064 to 0.41718, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 28/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.5066 - accuracy: 0.8698 - val_loss: 0.3942 - val_accuracy: 0.9135\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.41718 to 0.39423, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 29/48\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4953 - accuracy: 0.8732 - val_loss: 0.3969 - val_accuracy: 0.9085\n",
      "\n",
      "Epoch 00029: val_loss did not improve from 0.39423\n",
      "Epoch 30/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4847 - accuracy: 0.8747 - val_loss: 0.3801 - val_accuracy: 0.9105\n",
      "\n",
      "Epoch 00030: val_loss improved from 0.39423 to 0.38006, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 31/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.4740 - accuracy: 0.8792 - val_loss: 0.3732 - val_accuracy: 0.9137\n",
      "\n",
      "Epoch 00031: val_loss improved from 0.38006 to 0.37319, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 32/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.4670 - accuracy: 0.8791 - val_loss: 0.3643 - val_accuracy: 0.9180\n",
      "\n",
      "Epoch 00032: val_loss improved from 0.37319 to 0.36432, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 33/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4524 - accuracy: 0.8833 - val_loss: 0.3568 - val_accuracy: 0.9180\n",
      "\n",
      "Epoch 00033: val_loss improved from 0.36432 to 0.35684, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 34/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4439 - accuracy: 0.8853 - val_loss: 0.3495 - val_accuracy: 0.9163\n",
      "\n",
      "Epoch 00034: val_loss improved from 0.35684 to 0.34946, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 35/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4363 - accuracy: 0.8885 - val_loss: 0.3394 - val_accuracy: 0.9186\n",
      "\n",
      "Epoch 00035: val_loss improved from 0.34946 to 0.33941, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 36/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.4269 - accuracy: 0.8908 - val_loss: 0.3318 - val_accuracy: 0.9257\n",
      "\n",
      "Epoch 00036: val_loss improved from 0.33941 to 0.33180, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 37/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.4184 - accuracy: 0.8930 - val_loss: 0.3237 - val_accuracy: 0.9266\n",
      "\n",
      "Epoch 00037: val_loss improved from 0.33180 to 0.32375, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 38/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.4127 - accuracy: 0.8940 - val_loss: 0.3126 - val_accuracy: 0.9304\n",
      "\n",
      "Epoch 00038: val_loss improved from 0.32375 to 0.31263, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 39/48\n",
      "51227/51227 [==============================] - 23s 445us/step - loss: 0.4006 - accuracy: 0.8975 - val_loss: 0.3133 - val_accuracy: 0.9281\n",
      "\n",
      "Epoch 00039: val_loss did not improve from 0.31263\n",
      "Epoch 40/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3931 - accuracy: 0.8994 - val_loss: 0.3083 - val_accuracy: 0.9287\n",
      "\n",
      "Epoch 00040: val_loss improved from 0.31263 to 0.30833, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 41/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3879 - accuracy: 0.9012 - val_loss: 0.3015 - val_accuracy: 0.9289\n",
      "\n",
      "Epoch 00041: val_loss improved from 0.30833 to 0.30150, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 42/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3802 - accuracy: 0.9033 - val_loss: 0.2890 - val_accuracy: 0.9362\n",
      "\n",
      "Epoch 00042: val_loss improved from 0.30150 to 0.28900, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 43/48\n",
      "51227/51227 [==============================] - 23s 445us/step - loss: 0.3763 - accuracy: 0.9031 - val_loss: 0.2869 - val_accuracy: 0.9334\n",
      "\n",
      "Epoch 00043: val_loss improved from 0.28900 to 0.28690, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 44/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3668 - accuracy: 0.9062 - val_loss: 0.2746 - val_accuracy: 0.9382\n",
      "\n",
      "Epoch 00044: val_loss improved from 0.28690 to 0.27456, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 45/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3581 - accuracy: 0.9089 - val_loss: 0.2795 - val_accuracy: 0.9375\n",
      "\n",
      "Epoch 00045: val_loss did not improve from 0.27456\n",
      "Epoch 46/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.3546 - accuracy: 0.9106 - val_loss: 0.2684 - val_accuracy: 0.9395\n",
      "\n",
      "Epoch 00046: val_loss improved from 0.27456 to 0.26843, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 47/48\n",
      "51227/51227 [==============================] - 23s 446us/step - loss: 0.3491 - accuracy: 0.9117 - val_loss: 0.2575 - val_accuracy: 0.9438\n",
      "\n",
      "Epoch 00047: val_loss improved from 0.26843 to 0.25750, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Epoch 48/48\n",
      "51227/51227 [==============================] - 23s 447us/step - loss: 0.3438 - accuracy: 0.9128 - val_loss: 0.2571 - val_accuracy: 0.9440\n",
      "\n",
      "Epoch 00048: val_loss improved from 0.25750 to 0.25706, saving model to /home/edu/Projects/urban-audio-classifier/models/aug-train-nb3.hdf5\n",
      "Training completed in time:  0:18:25.702795\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 53\n",
    "num_batch_size = 128\n",
    "model_file = 'aug-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"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                    LOSS      ACCURACY\n",
      "--------------------------------------\n",
      "Training:         0.2284       95.3172\n",
      "Test:             0.2373       95.0086\n",
      "Normal diff          3.75 \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": [
    "With a normal error difference of **3.75** between test and train we seem to have found a sweet spot, just before it starts overfitting, with an awesome accuracy improvement compared to all previous experiments, reaching **~95%** in test. The error has been minimized as much as **~0.22**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.5 Train vs Test history plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "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)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training dynamics have changed in comparison to the last two experiments where the validation error reached a maximum value during the first few epochs before starting to minimize and becoming comparable with train loss. In this case the validation loss minimized since the first epoch, as we have more data representing what we want to predict. Generalization gap is noticeable by two smooth loss lines and the error difference looks quite the same than last experiment with Log-Mel Spectrograms.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Model evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### * Register model output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "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 Confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "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": [
    "# 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": [
    "We now have multiple times more data than before in our test set but still we can see how some particular patterns are still present, hopefully, with a slighter effect. Usual misclassification cases still happen between **Street Music** and **Children Playing**, **Drilling** and **Jack Hammer** or **Dog Bark** and **Children Playing**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.2 Accuracy table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CLASS</th>\n",
       "      <th>ACCURACY</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>Air Conditioner</td>\n",
       "      <td>97.528249</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>Jackhammer</td>\n",
       "      <td>97.522368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>Gun Shot</td>\n",
       "      <td>97.495183</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>Car Horn</td>\n",
       "      <td>97.246127</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>Siren</td>\n",
       "      <td>96.753247</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>Engine Idling</td>\n",
       "      <td>95.549174</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>Drilling</td>\n",
       "      <td>94.868035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>Dog bark</td>\n",
       "      <td>93.152639</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>Children Playing</td>\n",
       "      <td>92.553932</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>Street Music</td>\n",
       "      <td>90.455840</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              CLASS   ACCURACY\n",
       "0   Air Conditioner  97.528249\n",
       "7        Jackhammer  97.522368\n",
       "6          Gun Shot  97.495183\n",
       "1          Car Horn  97.246127\n",
       "8             Siren  96.753247\n",
       "5     Engine Idling  95.549174\n",
       "4          Drilling  94.868035\n",
       "3          Dog bark  93.152639\n",
       "2  Children Playing  92.553932\n",
       "9      Street Music  90.455840"
      ]
     },
     "execution_count": 56,
     "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": [
    "Happily, every class accuracy has raised! What's more interesting is that results are more balanced between classes now.<br/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                  precision    recall  f1-score   support\n",
      "\n",
      " Air Conditioner       0.94      0.98      0.96      1416\n",
      "        Car Horn       0.90      0.97      0.94       581\n",
      "Children Playing       0.91      0.93      0.92      1437\n",
      "        Dog bark       0.93      0.93      0.93      1402\n",
      "        Drilling       0.97      0.95      0.96      1364\n",
      "   Engine Idling       0.98      0.96      0.97      1393\n",
      "        Gun Shot       0.95      0.97      0.96       519\n",
      "      Jackhammer       0.97      0.98      0.97      1453\n",
      "           Siren       0.99      0.97      0.98      1232\n",
      "    Street Music       0.95      0.90      0.93      1404\n",
      "\n",
      "        accuracy                           0.95     12201\n",
      "       macro avg       0.95      0.95      0.95     12201\n",
      "    weighted avg       0.95      0.95      0.95     12201\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": [
    "Similar to class accuracies, we see an increase in the average F1, Precision and Recall scores, and also, more balanced results. The common patterns we have been seeing for the last two experiments are still present but with a much slighter effect.<br/>\n",
    "<br/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Conclusion\n",
    "\n",
    "All along this experiments we have learned how a very simple convolutional neural network can learn to recognize audible events, even when working with quite unbalanced datasets as the UrbanSound8K is. I would still like to experiment with different feature combinations; not that this performance scores are useless, we sure can use this model as it is to learn simple events like orders (\"start\", \"right\", \"update\", \"stop\", ...) or numbers for example.<br/>\n",
    "<br/>\n",
    "The amount of data showed to be a very important factor, until data augmentation training results where much more bound to the distribution of train/test splits: where it was easy to end up with a test split that wouldn't represent the train split. This didn't happen with augmented data.<br/>\n",
    "<br/>\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (deep-learning-p36)",
   "language": "python",
   "name": "deep-learning-p36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}