CoCalc -- Urban_Sound_Analysis_Sound

GitHub Repository: aswintechguy/Deep-Learning-Projects
Path: blob/main/Urban Sound Analysis - Sound Classification/Urban_Sound_Analysis_Sound_Classification.ipynb
⁵⁶⁹ views

Kernel: Python 3

Dataset Information

This dataset contains 8732 labeled sound excerpts (<=4s) of urban sounds from 10 classes:

air_conditioner

car_horn

children_playing

dog_bark

drilling

engine_idling

gun_shot

jackhammer

siren

street_music

Mounting Drive

In [1]:

from google.colab import drive
drive.mount('/content/drive')

Out[1]:

Mounted at /content/drive

In [2]:

!pwd

Out[2]:

/content

Unzip data

In [1]:

!unzip 'drive/MyDrive/Colab Notebooks/train.zip'

Import modules

In [16]:

import pandas as pd
import numpy as np
import librosa
import librosa.display
import glob
import IPython.display as ipd
import random
%pylab inline

import warnings
warnings.filterwarnings('ignore')

Out[16]:

Populating the interactive namespace from numpy and matplotlib

Loading the dataset

In [6]:

df = pd.read_csv('Urban Sound Dataset.csv')
df.head()

Out[6]:

In [9]:

ipd.Audio('Train/1.wav')

Out[9]:

Exploratory Data Analysis

In [10]:

data, sampling_rate = librosa.load('Train/1.wav')

In [11]:

data

Out[11]:

array([-0.09602016, -0.14303702,  0.05203498, ..., -0.01646687,
       -0.00915894,  0.09742922], dtype=float32)

In [12]:

sampling_rate

Out[12]:

22050

In [15]:

plt.figure(figsize=(12,4))
librosa.display.waveplot(data, sr=sampling_rate)

Out[15]:

<matplotlib.collections.PolyCollection at 0x7f135da27550>

In [28]:

index = random.choice(df.index)

print('Class:', df['Class'][index])
data, sampling_rate = librosa.load('Train/'+str(df['ID'][index]) + '.wav')

plt.figure(figsize=(12,4))
librosa.display.waveplot(data, sr=sampling_rate)

Out[28]:

Class: dog_bark

<matplotlib.collections.PolyCollection at 0x7f1354f76160>

In [29]:

index = random.choice(df.index)

print('Class:', df['Class'][index])
data, sampling_rate = librosa.load('Train/'+str(df['ID'][index]) + '.wav')

plt.figure(figsize=(12,4))
librosa.display.waveplot(data, sr=sampling_rate)

Out[29]:

Class: gun_shot

<matplotlib.collections.PolyCollection at 0x7f1354f570f0>

In [30]:

index = random.choice(df.index)

print('Class:', df['Class'][index])
data, sampling_rate = librosa.load('Train/'+str(df['ID'][index]) + '.wav')

plt.figure(figsize=(12,4))
librosa.display.waveplot(data, sr=sampling_rate)

Out[30]:

Class: car_horn

<matplotlib.collections.PolyCollection at 0x7f1354eb2438>

In [34]:

import seaborn as sns
plt.figure(figsize=(12,7))
sns.countplot(df['Class'])

Out[34]:

<matplotlib.axes._subplots.AxesSubplot at 0x7f135451c278>

Input Split

In [39]:

import os

def parser(row):
  # path of the file
  file_name = os.path.join('Train', str(row.ID) + '.wav')
  # load the audio file
  x, sample_rate = librosa.load(file_name, res_type='kaiser_fast')
  # extract features from the data
  mfccs = np.mean(librosa.feature.mfcc(y=x, sr=sample_rate, n_mfcc=40).T, axis=0)

  feature = mfccs
  label = row.Class

  return [feature, label]

In [40]:

data = df.apply(parser, axis=1)
data.columns = ['feature','label']

In [42]:

data[0]

Out[42]:

[array([-82.12358939, 139.50591598, -42.43086489,  24.82786139,
        -11.62076447,  23.49708426, -12.19458986,  25.89713885,
         -9.40527728,  21.21042898,  -7.36882138,  14.25433903,
         -8.67870015,   7.75023765, -10.1241154 ,   3.2581183 ,
        -11.35261914,   2.80096779,  -7.04601346,   3.91331351,
         -2.3349743 ,   2.01242254,  -2.79394367,   4.12927394,
         -1.62076864,   4.32620082,  -1.03440959,  -1.23297714,
         -3.11085341,   0.32044827,  -1.787786  ,   0.44295495,
         -1.79164752,  -0.76361758,  -1.24246428,  -0.27664012,
          0.65718559,  -0.50237115,  -2.60428533,  -1.05346291]), 'siren']

In [43]:

# input split
X = np.array(list(zip(*data))[0])
y = np.array(list(zip(*data))[1])

Label encoder

In [44]:

from sklearn.preprocessing import LabelEncoder
from keras.utils import np_utils

le = LabelEncoder()
y = np_utils.to_categorical(le.fit_transform(y))

In [45]:

y.shape

Out[45]:

(5435, 10)

In [46]:

y[0]

Out[46]:

array([0., 0., 0., 0., 0., 0., 0., 0., 1., 0.], dtype=float32)

Model Training

In [54]:

from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten

num_classes = 10

# model creation
model = Sequential()

model.add(Dense(256, input_shape=(40,)))
model.add(Activation('relu'))
model.add(Dropout(0.3))

model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.3))

model.add(Dense(256))
model.add(Activation('relu'))
model.add(Dropout(0.3))

model.add(Dense(num_classes))
model.add(Activation('softmax'))

model.compile(loss='categorical_crossentropy', metrics='accuracy', optimizer='adam')

In [55]:

# train the model
model.fit(X, y, batch_size=32, epochs=100, validation_split=0.25)

Out[55]:

Epoch 1/100
128/128 [==============================] - 1s 6ms/step - loss: 10.9345 - accuracy: 0.1495 - val_loss: 2.0777 - val_accuracy: 0.2597
Epoch 2/100
128/128 [==============================] - 1s 5ms/step - loss: 2.1973 - accuracy: 0.2605 - val_loss: 1.8058 - val_accuracy: 0.4091
Epoch 3/100
128/128 [==============================] - 1s 5ms/step - loss: 1.9379 - accuracy: 0.3309 - val_loss: 1.6051 - val_accuracy: 0.4643
Epoch 4/100
128/128 [==============================] - 1s 5ms/step - loss: 1.7228 - accuracy: 0.3807 - val_loss: 1.4608 - val_accuracy: 0.5077
Epoch 5/100
128/128 [==============================] - 1s 5ms/step - loss: 1.5491 - accuracy: 0.4630 - val_loss: 1.3355 - val_accuracy: 0.5453
Epoch 6/100
128/128 [==============================] - 1s 5ms/step - loss: 1.4185 - accuracy: 0.4843 - val_loss: 1.2097 - val_accuracy: 0.5938
Epoch 7/100
128/128 [==============================] - 1s 5ms/step - loss: 1.3123 - accuracy: 0.5384 - val_loss: 1.1305 - val_accuracy: 0.6350
Epoch 8/100
128/128 [==============================] - 1s 5ms/step - loss: 1.2230 - accuracy: 0.5772 - val_loss: 1.0233 - val_accuracy: 0.6843
Epoch 9/100
128/128 [==============================] - 1s 5ms/step - loss: 1.0966 - accuracy: 0.6136 - val_loss: 0.9303 - val_accuracy: 0.7145
Epoch 10/100
128/128 [==============================] - 1s 5ms/step - loss: 1.0108 - accuracy: 0.6537 - val_loss: 0.9144 - val_accuracy: 0.7020
Epoch 11/100
128/128 [==============================] - 1s 6ms/step - loss: 0.9479 - accuracy: 0.6794 - val_loss: 0.8196 - val_accuracy: 0.7513
Epoch 12/100
128/128 [==============================] - 1s 5ms/step - loss: 0.8572 - accuracy: 0.7047 - val_loss: 0.7841 - val_accuracy: 0.7623
Epoch 13/100
128/128 [==============================] - 1s 5ms/step - loss: 0.7855 - accuracy: 0.7370 - val_loss: 0.7249 - val_accuracy: 0.7815
Epoch 14/100
128/128 [==============================] - 1s 5ms/step - loss: 0.7760 - accuracy: 0.7312 - val_loss: 0.6986 - val_accuracy: 0.7881
Epoch 15/100
128/128 [==============================] - 1s 5ms/step - loss: 0.7491 - accuracy: 0.7465 - val_loss: 0.6526 - val_accuracy: 0.7969
Epoch 16/100
128/128 [==============================] - 1s 5ms/step - loss: 0.7028 - accuracy: 0.7673 - val_loss: 0.6324 - val_accuracy: 0.8035
Epoch 17/100
128/128 [==============================] - 1s 5ms/step - loss: 0.6544 - accuracy: 0.7700 - val_loss: 0.5943 - val_accuracy: 0.8197
Epoch 18/100
128/128 [==============================] - 1s 5ms/step - loss: 0.6218 - accuracy: 0.7817 - val_loss: 0.6158 - val_accuracy: 0.8116
Epoch 19/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5938 - accuracy: 0.8014 - val_loss: 0.5872 - val_accuracy: 0.8094
Epoch 20/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5655 - accuracy: 0.8001 - val_loss: 0.5640 - val_accuracy: 0.8197
Epoch 21/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5385 - accuracy: 0.8160 - val_loss: 0.5352 - val_accuracy: 0.8322
Epoch 22/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5351 - accuracy: 0.8151 - val_loss: 0.5039 - val_accuracy: 0.8411
Epoch 23/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5097 - accuracy: 0.8191 - val_loss: 0.4905 - val_accuracy: 0.8506
Epoch 24/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4892 - accuracy: 0.8348 - val_loss: 0.5233 - val_accuracy: 0.8521
Epoch 25/100
128/128 [==============================] - 1s 5ms/step - loss: 0.5039 - accuracy: 0.8224 - val_loss: 0.5153 - val_accuracy: 0.8484
Epoch 26/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4413 - accuracy: 0.8531 - val_loss: 0.4805 - val_accuracy: 0.8514
Epoch 27/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4459 - accuracy: 0.8451 - val_loss: 0.4908 - val_accuracy: 0.8550
Epoch 28/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4137 - accuracy: 0.8596 - val_loss: 0.4724 - val_accuracy: 0.8528
Epoch 29/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4073 - accuracy: 0.8548 - val_loss: 0.4560 - val_accuracy: 0.8793
Epoch 30/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4153 - accuracy: 0.8538 - val_loss: 0.4550 - val_accuracy: 0.8653
Epoch 31/100
128/128 [==============================] - 1s 5ms/step - loss: 0.4276 - accuracy: 0.8559 - val_loss: 0.4321 - val_accuracy: 0.8631
Epoch 32/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3614 - accuracy: 0.8824 - val_loss: 0.4599 - val_accuracy: 0.8779
Epoch 33/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3620 - accuracy: 0.8733 - val_loss: 0.4363 - val_accuracy: 0.8720
Epoch 34/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3326 - accuracy: 0.8882 - val_loss: 0.4184 - val_accuracy: 0.8779
Epoch 35/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3564 - accuracy: 0.8765 - val_loss: 0.4316 - val_accuracy: 0.8823
Epoch 36/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3115 - accuracy: 0.8869 - val_loss: 0.4323 - val_accuracy: 0.8801
Epoch 37/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3554 - accuracy: 0.8836 - val_loss: 0.4180 - val_accuracy: 0.8764
Epoch 38/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3401 - accuracy: 0.8867 - val_loss: 0.3973 - val_accuracy: 0.8904
Epoch 39/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3368 - accuracy: 0.8922 - val_loss: 0.4650 - val_accuracy: 0.8631
Epoch 40/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3444 - accuracy: 0.8800 - val_loss: 0.4376 - val_accuracy: 0.8896
Epoch 41/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2883 - accuracy: 0.9044 - val_loss: 0.3772 - val_accuracy: 0.8889
Epoch 42/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3086 - accuracy: 0.8952 - val_loss: 0.3987 - val_accuracy: 0.8904
Epoch 43/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2890 - accuracy: 0.8943 - val_loss: 0.3914 - val_accuracy: 0.8859
Epoch 44/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2837 - accuracy: 0.9025 - val_loss: 0.4280 - val_accuracy: 0.8911
Epoch 45/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3258 - accuracy: 0.9006 - val_loss: 0.3884 - val_accuracy: 0.8852
Epoch 46/100
128/128 [==============================] - 1s 6ms/step - loss: 0.2781 - accuracy: 0.9124 - val_loss: 0.4229 - val_accuracy: 0.8940
Epoch 47/100
128/128 [==============================] - 1s 5ms/step - loss: 0.3023 - accuracy: 0.8968 - val_loss: 0.3803 - val_accuracy: 0.8926
Epoch 48/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2292 - accuracy: 0.9182 - val_loss: 0.3708 - val_accuracy: 0.8999
Epoch 49/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2638 - accuracy: 0.9092 - val_loss: 0.3868 - val_accuracy: 0.8874
Epoch 50/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2714 - accuracy: 0.9109 - val_loss: 0.4067 - val_accuracy: 0.8874
Epoch 51/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2402 - accuracy: 0.9187 - val_loss: 0.3848 - val_accuracy: 0.8992
Epoch 52/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2717 - accuracy: 0.9139 - val_loss: 0.3934 - val_accuracy: 0.8926
Epoch 53/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2731 - accuracy: 0.9102 - val_loss: 0.3777 - val_accuracy: 0.9036
Epoch 54/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2456 - accuracy: 0.9216 - val_loss: 0.4055 - val_accuracy: 0.8845
Epoch 55/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2561 - accuracy: 0.9116 - val_loss: 0.3938 - val_accuracy: 0.9043
Epoch 56/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2459 - accuracy: 0.9220 - val_loss: 0.3679 - val_accuracy: 0.8999
Epoch 57/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2536 - accuracy: 0.9210 - val_loss: 0.3828 - val_accuracy: 0.8955
Epoch 58/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2276 - accuracy: 0.9189 - val_loss: 0.3920 - val_accuracy: 0.9021
Epoch 59/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2219 - accuracy: 0.9285 - val_loss: 0.3921 - val_accuracy: 0.9051
Epoch 60/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2325 - accuracy: 0.9197 - val_loss: 0.4338 - val_accuracy: 0.9021
Epoch 61/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2177 - accuracy: 0.9242 - val_loss: 0.3954 - val_accuracy: 0.8999
Epoch 62/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2323 - accuracy: 0.9280 - val_loss: 0.4031 - val_accuracy: 0.8933
Epoch 63/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2495 - accuracy: 0.9152 - val_loss: 0.4132 - val_accuracy: 0.9058
Epoch 64/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2013 - accuracy: 0.9313 - val_loss: 0.4326 - val_accuracy: 0.8940
Epoch 65/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2480 - accuracy: 0.9251 - val_loss: 0.3966 - val_accuracy: 0.9088
Epoch 66/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2965 - accuracy: 0.9232 - val_loss: 0.3706 - val_accuracy: 0.9065
Epoch 67/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2011 - accuracy: 0.9338 - val_loss: 0.3669 - val_accuracy: 0.9102
Epoch 68/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1873 - accuracy: 0.9384 - val_loss: 0.4369 - val_accuracy: 0.8962
Epoch 69/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2712 - accuracy: 0.9157 - val_loss: 0.3485 - val_accuracy: 0.9080
Epoch 70/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2294 - accuracy: 0.9224 - val_loss: 0.3572 - val_accuracy: 0.9102
Epoch 71/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1905 - accuracy: 0.9372 - val_loss: 0.3791 - val_accuracy: 0.9029
Epoch 72/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1879 - accuracy: 0.9304 - val_loss: 0.3752 - val_accuracy: 0.9095
Epoch 73/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2025 - accuracy: 0.9359 - val_loss: 0.4290 - val_accuracy: 0.8970
Epoch 74/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1840 - accuracy: 0.9373 - val_loss: 0.3987 - val_accuracy: 0.9014
Epoch 75/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2182 - accuracy: 0.9357 - val_loss: 0.4522 - val_accuracy: 0.8948
Epoch 76/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2166 - accuracy: 0.9372 - val_loss: 0.3313 - val_accuracy: 0.9183
Epoch 77/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2009 - accuracy: 0.9352 - val_loss: 0.4012 - val_accuracy: 0.9051
Epoch 78/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1828 - accuracy: 0.9363 - val_loss: 0.3910 - val_accuracy: 0.9095
Epoch 79/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1728 - accuracy: 0.9475 - val_loss: 0.3741 - val_accuracy: 0.9051
Epoch 80/100
128/128 [==============================] - 1s 6ms/step - loss: 0.1792 - accuracy: 0.9395 - val_loss: 0.3704 - val_accuracy: 0.9102
Epoch 81/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1769 - accuracy: 0.9407 - val_loss: 0.3389 - val_accuracy: 0.9191
Epoch 82/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2019 - accuracy: 0.9305 - val_loss: 0.4008 - val_accuracy: 0.9021
Epoch 83/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1888 - accuracy: 0.9357 - val_loss: 0.3751 - val_accuracy: 0.9146
Epoch 84/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2088 - accuracy: 0.9357 - val_loss: 0.3838 - val_accuracy: 0.9021
Epoch 85/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1858 - accuracy: 0.9421 - val_loss: 0.3945 - val_accuracy: 0.9110
Epoch 86/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2087 - accuracy: 0.9345 - val_loss: 0.4125 - val_accuracy: 0.9058
Epoch 87/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1704 - accuracy: 0.9406 - val_loss: 0.3750 - val_accuracy: 0.9169
Epoch 88/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1565 - accuracy: 0.9485 - val_loss: 0.3929 - val_accuracy: 0.9095
Epoch 89/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1564 - accuracy: 0.9501 - val_loss: 0.3995 - val_accuracy: 0.9058
Epoch 90/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1976 - accuracy: 0.9330 - val_loss: 0.3827 - val_accuracy: 0.9139
Epoch 91/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1643 - accuracy: 0.9475 - val_loss: 0.3867 - val_accuracy: 0.9095
Epoch 92/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2080 - accuracy: 0.9333 - val_loss: 0.3833 - val_accuracy: 0.9051
Epoch 93/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1938 - accuracy: 0.9378 - val_loss: 0.3767 - val_accuracy: 0.9073
Epoch 94/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2045 - accuracy: 0.9322 - val_loss: 0.3608 - val_accuracy: 0.9191
Epoch 95/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1565 - accuracy: 0.9552 - val_loss: 0.3429 - val_accuracy: 0.9117
Epoch 96/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1859 - accuracy: 0.9393 - val_loss: 0.3583 - val_accuracy: 0.9213
Epoch 97/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1961 - accuracy: 0.9441 - val_loss: 0.3974 - val_accuracy: 0.9088
Epoch 98/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1798 - accuracy: 0.9384 - val_loss: 0.4386 - val_accuracy: 0.8992
Epoch 99/100
128/128 [==============================] - 1s 5ms/step - loss: 0.2149 - accuracy: 0.9316 - val_loss: 0.4006 - val_accuracy: 0.9110
Epoch 100/100
128/128 [==============================] - 1s 5ms/step - loss: 0.1530 - accuracy: 0.9526 - val_loss: 0.3809 - val_accuracy: 0.9169

<tensorflow.python.keras.callbacks.History at 0x7f12c2d88cc0>

In [ ]: