Initial commit

constants.py

+from pandas import datetime
+
+MFCC_FEATURE_START = 1
+MFCC_FEATURE_END = 21
+
+TRAINING_DATA_DIRECTORY_NAME = 'DemoTrainingDataset'
+TESTING_DATA_DIRECTORY_NAME = 'TEST'
+
+
+CAR = 'Car'
+BUS = 'Bus'
+TRUCK = 'Truck'
+MOTORCYCLE = 'Motorcycle'
+TRAM = 'Tram'
+
+LIGHT_WEIGHT = 'Light-Weight'
+MEDIUM_WEIGHT = 'Medium-Weight'
+HEAVY_WEIGHT = 'Heavy-Weight'
+TWO_WHEELED = 'Two-Wheeled'
+RAIL_BOUND = 'Rail-Bound'
+
+FEATURES_CSV_NAME = 'features.csv'
+TEST_CSV_NAME = 'test.csv'
+
+HIDDEN_LAYER_1_DIMENSIONS = 32
+HIDDEN_LAYER_2_DIMENSIONS = 32
+HIDDEN_LAYER_3_DIMENSIONS = 32
+OUTPUT_LAYER_DIMENSIONS = 5
+
+ACTIVATION_RELU = 'relu'
+ACTIVATION_SOFTMAX = 'softmax'
+
+OPTIMIZER_ADAM = 'adam'
+LOSS_FUNCTION_SPARSE = 'sparse_categorical_crossentropy'
+ACCURACY_METRICS = 'accuracy'
+
+LOG_DIR_PATH = "logs/scalars/" + datetime.now().strftime("%Y%m%d-%H%M%S")
+
+TRAINED_MODEL = 'trained_model.h5'

main.py

+import librosa.feature
+import pandas as pd
+import numpy as np
+from pathlib import Path
+from os import chdir
+import os
+import csv
+from tensorflow import keras
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import LabelEncoder, StandardScaler
+from keras import models
+from keras import layers
+import matplotlib.pyplot as plt
+from sklearn.metrics import classification_report
+import constants
+
+
+def create_csv_header():
+    header = 'filename '
+    for i in range(constants.MFCC_FEATURE_START, constants.MFCC_FEATURE_END):
+        header += f' mfcc{i}'
+    header += ' label'
+    header = header.split()
+    file = open(constants.FEATURES_CSV_NAME, 'w', newline='')
+    with file:
+        writer = csv.writer(file)
+        writer.writerow(header)
+
+
+def extract_features(trainingDataDir, trainingDataSubDirs):
+    create_csv_header()
+    # Looping over every file inside the subdirectories for feature extraction
+    for trainingDataSubDir in trainingDataSubDirs:
+        for fileName in os.listdir(trainingDataDir/f'{trainingDataSubDir}'):
+            if fileName.endswith(".wav"):
+                audioFile = trainingDataDir/f'{trainingDataSubDir}/{fileName}'
+                print("Extracting Features from Directory "+trainingDataSubDir+" and file "+audioFile.name)
+                y, sr = librosa.load(audioFile, mono=True)
+                mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=(constants.MFCC_FEATURE_END - constants.MFCC_FEATURE_START))
+                to_append = f'{audioFile.name}'
+                for g in mfcc:
+                    to_append += f' {np.mean(g)}'
+                if trainingDataSubDir == constants.CAR:
+                    to_append += f' {constants.LIGHT_WEIGHT}'
+                elif trainingDataSubDir == constants.BUS:
+                    to_append += f' {constants.MEDIUM_WEIGHT}'
+                elif trainingDataSubDir == constants.TRUCK:
+                    to_append += f' {constants.HEAVY_WEIGHT}'
+                elif trainingDataSubDir == constants.MOTORCYCLE:
+                    to_append += f' {constants.TWO_WHEELED}'
+                elif trainingDataSubDir == constants.TRAM:
+                    to_append += f' {constants.RAIL_BOUND}'
+
+                file = open(constants.FEATURES_CSV_NAME, 'a', newline='')
+                with file:
+                    writer = csv.writer(file)
+                    writer.writerow(to_append.split())
+
+
+def preprocessing_csv_data():
+    print("Reading Features... ")
+    data = pd.read_csv(constants.FEATURES_CSV_NAME)
+    data.head()
+    # Dropping unnecessary columns (Column Filename is dropped)
+    data = data.drop(['filename'], axis=1)
+    data.head()
+    return data
+
+
+def encode_labels(data):
+    # Extracting classes/label column as y from csv and converting string labels to numbers using LabelEncoder
+    audio_list = data.iloc[:, -1]
+    encoder = LabelEncoder()
+    target_labels = encoder.fit_transform(audio_list)
+    return target_labels, encoder
+
+
+def normalize_data(data):
+    # normalizing - Extracting Remaining Columns as X and normalizing them to a common scale
+    scaler = StandardScaler()
+    X = scaler.fit_transform(np.array(data.iloc[:, :-1], dtype=float))
+    return X
+
+
+def train_test_data_split(X, y):
+    # splitting of dataset into train and test dataset
+    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20)
+    return X_train, X_test, y_train, y_test
+
+
+def create_and_compile_model():
+    print("Creating a Model")
+    # creating a model
+    model = models.Sequential()
+    model.add(layers.Dense(constants.HIDDEN_LAYER_1_DIMENSIONS, activation=constants.ACTIVATION_RELU, input_shape=(X.shape[1],)))
+    model.add(layers.Dense(constants.HIDDEN_LAYER_2_DIMENSIONS, activation=constants.ACTIVATION_RELU))
+    model.add(layers.Dense(constants.HIDDEN_LAYER_3_DIMENSIONS, activation=constants.ACTIVATION_RELU))
+    model.add(layers.Dense(constants.OUTPUT_LAYER_DIMENSIONS, activation=constants.ACTIVATION_SOFTMAX))
+
+    print("Compiling a Model")
+    model.compile(optimizer= constants.OPTIMIZER_ADAM, loss= constants.LOSS_FUNCTION_SPARSE, metrics=[constants.ACCURACY_METRICS])
+    return model
+
+
+def train_and_save_model(model, X_train, y_train, X_test, y_test):
+    logdir = constants.LOG_DIR_PATH
+    tensorboard_callback = keras.callbacks.TensorBoard(log_dir=logdir)
+    print("Start Training...")
+    history = model.fit(X_train, y_train, epochs=35, validation_data=(X_test, y_test), callbacks=[tensorboard_callback])
+    # Saving the trained model to avoid re-training
+    model.save(constants.TRAINED_MODEL)
+    return history
+
+
+def predict(X_test, y_test):
+    print("Predictions.....")
+    predictions = np.argmax(model.predict(X_test), axis=-1)
+    target_names = [constants.LIGHT_WEIGHT, constants.MEDIUM_WEIGHT, constants.HEAVY_WEIGHT,constants.TWO_WHEELED, constants.RAIL_BOUND]
+    print(classification_report(y_test, predictions, target_names=target_names))
+
+
+def plot_model_accuracy(history):
+    # Plot graph Model Accuracy
+    plt.plot(history.history['accuracy'])
+    plt.plot(history.history['val_accuracy'])
+    plt.title('Model Accuracy')
+    plt.ylabel('Accuracy')
+    plt.xlabel('Epoch')
+    plt.legend(['Train', 'Test'], loc='upper left')
+    plt.show()
+
+
+def plot_model_loss(history):
+    # Plot graph Model Loss
+    plt.plot(history.history['loss'])
+    plt.plot(history.history['val_loss'])
+    plt.title('Model loss')
+    plt.ylabel('Loss')
+    plt.xlabel('Epoch')
+    plt.legend(['Train', 'Test'], loc='upper right')
+    plt.show()
+
+
+# Changing Directory to Training Dataset Folder
+chdir(constants.TRAINING_DATA_DIRECTORY_NAME)
+trainingDataDir = Path.cwd()
+trainingDataSubDirs = os.listdir(trainingDataDir)
+
+extract_features(trainingDataDir, trainingDataSubDirs)
+data = preprocessing_csv_data()
+target_labels, encoder = encode_labels(data)
+X = normalize_data(data)
+X_train, X_test, y_train, y_test = train_test_data_split(X, target_labels)
+model = create_and_compile_model()
+history = train_and_save_model(model, X_train, y_train, X_test, y_test)
+predict(X_test, y_test)
+plot_model_accuracy(history)
+plot_model_loss(history)
+
+
+

requirements.txt

+absl-py==0.11.0
+appdirs==1.4.4
+astunparse==1.6.3
+audioread==2.1.9
+cached-property==1.5.2
+cachetools==4.1.1
+certifi==2020.6.20
+cffi==1.14.3
+chardet==3.0.4
+cycler==0.10.0
+decorator==4.4.2
+gast==0.3.3
+google-auth==1.23.0
+google-auth-oauthlib==0.4.2
+google-pasta==0.2.0
+grpcio==1.33.2
+h5py==2.10.0
+idna==2.10
+importlib-metadata==2.0.0
+inline==0.0.1
+joblib==0.17.0
+Keras==2.4.3
+Keras-Preprocessing==1.1.2
+kiwisolver==1.2.0
+librosa==0.8.0
+llvmlite==0.34.0
+Markdown==3.3.3
+matplotlib==3.3.2
+numba==0.51.2
+numpy==1.18.5
+oauthlib==3.1.0
+opt-einsum==3.3.0
+packaging==20.4
+pandas==1.1.4
+pathlib==1.0.1
+Pillow==8.0.1
+pooch==1.2.0
+protobuf==3.13.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pycparser==2.20
+pyparsing==2.4.7
+python-dateutil==2.8.1
+pytz==2020.4
+PyYAML==5.3.1
+requests==2.24.0
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rsa==4.6
+scikit-learn==0.23.2
+scipy==1.5.3
+six==1.15.0
+SoundFile==0.10.3.post1
+tensorboard==2.3.0
+tensorboard-plugin-wit==1.7.0
+tensorflow==2.3.1
+tensorflow-estimator==2.3.0
+termcolor==1.1.0
+threadpoolctl==2.1.0
+urllib3==1.25.11
+Werkzeug==1.0.1
+wrapt==1.12.1
+zipp==3.4.0

test.py

+import librosa.feature
+import pandas as pd
+import numpy as np
+import os
+from pathlib import Path
+import csv
+from tensorflow import keras
+from sklearn.preprocessing import LabelEncoder, StandardScaler
+import constants
+
+
+def create_csv_header():
+    header=''
+    for i in range(constants.MFCC_FEATURE_START, constants.MFCC_FEATURE_END):
+        header += f' mfcc{i}'
+    header = header.split()
+    file = open(constants.TEST_CSV_NAME, 'w', newline='')
+    with file:
+        writer = csv.writer(file)
+        writer.writerow(header)
+
+
+def extract_features(workingDir, subDirectories):
+    create_csv_header()
+    for subDirectory in subDirectories:
+        if subDirectory == constants.TESTING_DATA_DIRECTORY_NAME:
+            for fileName in os.listdir(workingDir/f'{subDirectory}'):
+                if fileName.endswith(".wav"):
+                    audioFile = workingDir / f'{subDirectory}/{fileName}'
+                    y, sr = librosa.load(audioFile, mono=True)
+                    mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=(constants.MFCC_FEATURE_END - constants.MFCC_FEATURE_START))
+                    to_append = ''
+                    for g in mfcc:
+                        to_append += f' {np.mean(g)}'
+                    file = open(constants.TEST_CSV_NAME, 'a', newline='')
+                    with file:
+                        writer = csv.writer(file)
+                        writer.writerow(to_append.split())
+
+
+def preprocessing_csv_data():
+    # reading dataset from csv
+    print("Reading Features... ")
+    data = pd.read_csv(constants.TEST_CSV_NAME)
+    data.head()
+    return data
+
+
+def normalize_data(data):
+    # # normalizing - Extracting Remaining Columns as X and normalizing them to a common scale
+    scaler = StandardScaler()
+    X = scaler.fit_transform(np.array(data.iloc[:, :], dtype=float))
+    print(X)
+    print(X.shape)
+    return X
+
+
+WorkingDir = Path.cwd()
+subDirectories = os.listdir(WorkingDir)
+extract_features(WorkingDir, subDirectories)
+data = preprocessing_csv_data()
+X = normalize_data(data)
+model = keras.models.load_model('./DemoTrainingDataset/trained_model.h5')
+model.summary()
+predictions = np.argmax(model.predict(X), axis=-1)
+encoder = LabelEncoder()
+labels = ['Light-Weight', 'Medium-Weight', 'Heavy-Weight', 'Two-Wheeled', 'Rail-Bound']
+encoder.fit_transform(labels)
+print(predictions)
+print(encoder.inverse_transform(predictions))