# -*- coding: utf-8 -*- """FinalProject.ipynb Automatically generated by Colaboratory. Original file is located at https://colab.research.google.com/drive/1cKPkWgKQrZ_jNDWEzWDtVgUudXrTUvQP """ # Import our libraries import librosa import librosa.display import numpy as np import matplotlib.pyplot as plt from matplotlib.pyplot import specgram import pandas as pd import os import IPython.display as ipd # To play sound in the notebook # NOTE: BEFORE RUNNING THIS CELL move my data folder to the path specified below in your google drive and then mount from google.colab import drive drive.mount('/content/drive/') # Source - RAVDESS; Gender - Female; Emotion - Angry path = "drive/My Drive/kaggle/rav/Actor_08/03-01-05-02-01-01-08.wav" X, sample_rate = librosa.load(path, res_type='kaiser_fast',duration=2.5,sr=22050*2,offset=0.5) mfcc = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13) # audio wave plt.figure(figsize=(20, 15)) plt.subplot(3,1,1) librosa.display.waveplot(X, sr=sample_rate) plt.title('Audio sampled at 44100 hrz') # MFCC plt.figure(figsize=(20, 15)) plt.subplot(3,1,1) librosa.display.specshow(mfcc, x_axis='time') plt.ylabel('MFCC') plt.colorbar() ipd.Audio(path) RAV = "drive/My Drive/kaggle/rav/audio_speech_actors_01-24/" dir_list = os.listdir(RAV) dir_list.sort() emotion = [] gender = [] path = [] for i in dir_list: fname = os.listdir(RAV + i) for f in fname: if (f == ".ipynb_checkpoints"): continue part = f.split('.')[0].split('-') print(i, part, f) emotion.append(int(part[2])) temp = int(part[6]) if temp%2 == 0: temp = "female" else: temp = "male" gender.append(temp) path.append(RAV + i + '/' + f) RAV_df = pd.DataFrame(emotion) RAV_df = RAV_df.replace({1:'neutral', 2:'neutral', 3:'happy', 4:'sad', 5:'angry', 6:'fear', 7:'disgust', 8:'surprise'}) RAV_df = pd.concat([pd.DataFrame(gender),RAV_df],axis=1) RAV_df.columns = ['gender','emotion'] RAV_df['labels'] =RAV_df.gender + '_' + RAV_df.emotion RAV_df['source'] = 'RAVDESS' RAV_df = pd.concat([RAV_df,pd.DataFrame(path, columns = ['path'])],axis=1) RAV_df = RAV_df.drop(['gender', 'emotion'], axis=1) RAV_df.labels.value_counts() # Source - RAVDESS; Gender - Female; Emotion - Angry path = RAV + "Actor_08/03-01-05-02-01-01-08.wav" X, sample_rate = librosa.load(path, res_type='kaiser_fast',duration=2.5,sr=22050*2,offset=0.5) female = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13) female = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13), axis=0) print(len(female)) # Source - RAVDESS; Gender - Male; Emotion - Angry path = RAV + "Actor_09/03-01-05-01-01-01-09.wav" X, sample_rate = librosa.load(path, res_type='kaiser_fast',duration=2.5,sr=22050*2,offset=0.5) male = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13) male = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13), axis=0) print(len(male)) # audio wave plt.figure(figsize=(20, 15)) plt.subplot(3,1,1) plt.plot(female, label='female') plt.plot(male, label='male') plt.legend() path = RAV + "Actor_12/03-01-03-01-02-01-12.wav" X, sample_rate = librosa.load(path, res_type='kaiser_fast',duration=2.5,sr=22050*2,offset=0.5) female = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13) female = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13), axis=0) print(len(female)) # Source - RAVDESS; Gender - Male; Emotion - happy path = RAV + "Actor_11/03-01-03-01-02-02-11.wav" X, sample_rate = librosa.load(path, res_type='kaiser_fast',duration=2.5,sr=22050*2,offset=0.5) male = librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13) male = np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13), axis=0) print(len(male)) # Plot the two audio waves together plt.figure(figsize=(20, 15)) plt.subplot(3,1,1) plt.plot(female, label='female') plt.plot(male, label='male') plt.legend() import keras from keras import regularizers from keras.preprocessing import sequence from keras.preprocessing.text import Tokenizer from keras.preprocessing.sequence import pad_sequences from keras.models import Sequential, Model, model_from_json from keras.layers import Dense, Embedding, LSTM from keras.layers import Input, Flatten, Dropout, Activation, BatchNormalization from keras.layers import Conv1D, MaxPooling1D, AveragePooling1D, Convolution2D, MaxPool2D from keras.utils import np_utils, to_categorical from keras.callbacks import ModelCheckpoint # sklearn from sklearn.metrics import confusion_matrix, accuracy_score, classification_report from sklearn.model_selection import train_test_split from sklearn.preprocessing import LabelEncoder # Other import librosa import librosa.display import json import numpy as np import matplotlib.pyplot as plt import tensorflow as tf from matplotlib.pyplot import specgram import pandas as pd import seaborn as sns import glob import os import pickle import IPython.display as ipd # To play sound in the notebook # #Starting TESS dataset # from google.colab import drive # drive.mount('/content/tessdrive') path = 'drive/My Drive/kaggle' # from google.colab import files # files.upload() #this will prompt you to upload the kaggle.json !pip install -q kaggle !mkdir -p ~/.kaggle !cp kaggle.json ~/.kaggle/ !ls ~/.kaggle !chmod 600 /root/.kaggle/kaggle.json # set permission !kaggle datasets download -d ejlok1/cremad -p /content/tessdrive/My\ Drive/kaggle/tess CREMA = path + '/cremad/AudioWAV/' dir_list = os.listdir(CREMA) dir_list.sort() print(dir_list[0:10]) gender = [] emotion = [] path = [] female = [1002,1003,1004,1006,1007,1008,1009,1010,1012,1013,1018,1020,1021,1024,1025,1028,1029,1030,1037,1043,1046,1047,1049, 1052,1053,1054,1055,1056,1058,1060,1061,1063,1072,1073,1074,1075,1076,1078,1079,1082,1084,1089,1091] for i in dir_list: part = i.split('_') if int(part[0]) in female: temp = 'female' else: temp = 'male' gender.append(temp) if part[2] == 'SAD' and temp == 'male': emotion.append('male_sad') elif part[2] == 'ANG' and temp == 'male': emotion.append('male_angry') elif part[2] == 'DIS' and temp == 'male': emotion.append('male_disgust') elif part[2] == 'FEA' and temp == 'male': emotion.append('male_fear') elif part[2] == 'HAP' and temp == 'male': emotion.append('male_happy') elif part[2] == 'NEU' and temp == 'male': emotion.append('male_neutral') elif part[2] == 'SAD' and temp == 'female': emotion.append('female_sad') elif part[2] == 'ANG' and temp == 'female': emotion.append('female_angry') elif part[2] == 'DIS' and temp == 'female': emotion.append('female_disgust') elif part[2] == 'FEA' and temp == 'female': emotion.append('female_fear') elif part[2] == 'HAP' and temp == 'female': emotion.append('female_happy') elif part[2] == 'NEU' and temp == 'female': emotion.append('female_neutral') else: emotion.append('Unknown') path.append(CREMA + i) CREMA_df = pd.DataFrame(emotion, columns = ['labels']) CREMA_df['source'] = 'CREMA' CREMA_df = pd.concat([CREMA_df,pd.DataFrame(path, columns = ['path'])],axis=1) CREMA_df.labels.value_counts() # A fearful track fname = CREMA + '1012_IEO_FEA_HI.wav' data, sampling_rate = librosa.load(fname) plt.figure(figsize=(15, 5)) librosa.display.waveplot(data, sr=sampling_rate) # Lets play the audio ipd.Audio(fname) #Tess Data next # print(path) path = 'drive/My Drive/kaggle' TESS = path + "/tess/data/" dir_list = os.listdir(TESS) dir_list.sort() dir_list path = [] emotion = [] for i in dir_list: fname = os.listdir(TESS + i) for f in fname: if i == 'OAF_angry' or i == 'YAF_angry': emotion.append('female_angry') elif i == 'OAF_disgust' or i == 'YAF_disgust': emotion.append('female_disgust') elif i == 'OAF_Fear' or i == 'YAF_fear': emotion.append('female_fear') elif i == 'OAF_happy' or i == 'YAF_happy': emotion.append('female_happy') elif i == 'OAF_neutral' or i == 'YAF_neutral': emotion.append('female_neutral') elif i == 'OAF_Pleasant_surprise' or i == 'YAF_pleasant_surprised': emotion.append('female_surprise') elif i == 'OAF_Sad' or i == 'YAF_sad': emotion.append('female_sad') else: emotion.append('Unknown') path.append(TESS + i + "/" + f) TESS_df = pd.DataFrame(emotion, columns = ['labels']) TESS_df['source'] = 'TESS' TESS_df = pd.concat([TESS_df,pd.DataFrame(path, columns = ['path'])],axis=1) TESS_df.labels.value_counts() # lets play a fearful track fname = TESS + 'YAF_fear/YAF_dog_fear.wav' data, sampling_rate = librosa.load(fname) plt.figure(figsize=(15, 5)) librosa.display.waveplot(data, sr=sampling_rate) # Lets play the audio ipd.Audio(fname) #Surrey Data path = 'drive/My Drive/kaggle' SAVEE = path + '/surrey/ALL/' dir_list = os.listdir(SAVEE) # parse the filename to get the emotions emotion=[] path = [] for i in dir_list: if i[-8:-6]=='_a': emotion.append('male_angry') elif i[-8:-6]=='_d': emotion.append('male_disgust') elif i[-8:-6]=='_f': emotion.append('male_fear') elif i[-8:-6]=='_h': emotion.append('male_happy') elif i[-8:-6]=='_n': emotion.append('male_neutral') elif i[-8:-6]=='sa': emotion.append('male_sad') elif i[-8:-6]=='su': emotion.append('male_surprise') else: emotion.append('male_error') path.append(SAVEE + i) # Now check out the label count distribution SAVEE_df = pd.DataFrame(emotion, columns = ['labels']) SAVEE_df['source'] = 'SAVEE' SAVEE_df = pd.concat([SAVEE_df, pd.DataFrame(path, columns = ['path'])], axis = 1) SAVEE_df.labels.value_counts() # use the well known Librosa library for this task fname = SAVEE + 'DC_f11.wav' data, sampling_rate = librosa.load(fname) plt.figure(figsize=(15, 5)) librosa.display.waveplot(data, sr=sampling_rate) # Lets play the audio ipd.Audio(fname) # Data Creation Stage def prepare_data(df, n, aug, mfcc): X = np.empty(shape=(df.shape[0], n, 216, 1)) input_length = sampling_rate * audio_duration cnt = 0 for fname in tqdm(df.path): file_path = fname data, _ = librosa.load(file_path, sr=sampling_rate ,res_type="kaiser_fast" ,duration=2.5 ,offset=0.5 ) # Random offset / Padding if len(data) > input_length: max_offset = len(data) - input_length offset = np.random.randint(max_offset) data = data[offset:(input_length+offset)] else: if input_length > len(data): max_offset = input_length - len(data) offset = np.random.randint(max_offset) else: offset = 0 data = np.pad(data, (offset, int(input_length) - len(data) - offset), "constant") # which feature? if mfcc == 1: # MFCC extraction MFCC = librosa.feature.mfcc(data, sr=sampling_rate, n_mfcc=n_mfcc) MFCC = np.expand_dims(MFCC, axis=-1) X[cnt,] = MFCC else: # Log-melspectogram melspec = librosa.feature.melspectrogram(data, n_mels = n_melspec) logspec = librosa.amplitude_to_db(melspec) logspec = np.expand_dims(logspec, axis=-1) X[cnt,] = logspec cnt += 1 return X ''' # 4. Create the 2D CNN model ''' def get_2d_conv_model(n): ''' Create a standard deep 2D convolutional neural network''' nclass = 14 inp = Input(shape=(n,216,1)) #2D matrix of 30 MFCC bands by 216 audio length. x = Convolution2D(32, (4,10), padding="same")(inp) x = BatchNormalization()(x) x = Activation("relu")(x) x = MaxPool2D()(x) x = Dropout(rate=0.2)(x) x = Convolution2D(32, (4,10), padding="same")(x) x = BatchNormalization()(x) x = Activation("relu")(x) x = MaxPool2D()(x) x = Dropout(rate=0.2)(x) x = Convolution2D(32, (4,10), padding="same")(x) x = BatchNormalization()(x) x = Activation("relu")(x) x = MaxPool2D()(x) x = Dropout(rate=0.2)(x) x = Convolution2D(32, (4,10), padding="same")(x) x = BatchNormalization()(x) x = Activation("relu")(x) x = MaxPool2D()(x) x = Dropout(rate=0.2)(x) x = Flatten()(x) x = Dense(64)(x) x = Dropout(rate=0.2)(x) x = BatchNormalization()(x) x = Activation("relu")(x) x = Dropout(rate=0.2)(x) out = Dense(nclass, activation=softmax)(x) model = keras.models.Model(inputs=inp, outputs=out) opt = optimizers.Adam(0.001) model.compile(optimizer=opt, loss=losses.categorical_crossentropy, metrics=['acc']) return model TESS_df['path'].iloc[1400] TESS_df = TESS_df[TESS_df.labels != "Unknown"] df = pd.concat([SAVEE_df, RAV_df, TESS_df], axis = 0) from tqdm import tqdm, tqdm_pandas df # dataframe containing all the data df.head() # # df['source'] == 'TESS' # mfcc # import pickle # # example_dict = {1:"6",2:"2",3:"f"} # pickle_out = open("mfcc.pickle","wb") # pickle.dump(mfcc, pickle_out) # pickle_out.close() sampling_rate=44100 audio_duration=2.5 n_mfcc = 30 # mfcc = prepare_data(df, n = n_mfcc, aug = 0, mfcc = 1) mfcc = pickle.load( open( "drive/My Drive/kaggle/mfcc.pickle", "rb" ) ) X_train, X_test, y_train, y_test = train_test_split(mfcc , df.labels , test_size=0.25 , shuffle=True , random_state=42 ) # one hot encode the target lb = LabelEncoder() y_train = np_utils.to_categorical(lb.fit_transform(y_train)) y_test = np_utils.to_categorical(lb.fit_transform(y_test)) X_train.shape # df.tail # mfcc X, sample_rate = librosa.load("drive/My Drive/kaggle/surrey/ALL/DC_a01.wav" , res_type='kaiser_fast' ,duration=2.5 ,sr=44100 ,offset=0.5 ) sample_rate TESS_df = TESS_df.drop([1400]) TESS_df['path'].iloc[1400] TESS_df.iloc[1398:1406] TESS_df.iloc[1400] TESS_df = TESS_df[TESS_df.labels != "Unknown"] TESS_df.labels # ! mv drive/My\ Drive/mfcc.pickle drive/My\ Drive/kaggle y_train.shape model = get_2d_conv_model(n=30) model_history = model.fit(X_train, y_train, validation_data=(X_test, y_test), batch_size=16, verbose = 2, epochs=10) import keras from keras import regularizers, losses from keras.preprocessing import sequence from keras.preprocessing.text import Tokenizer from keras.preprocessing.sequence import pad_sequences from keras.models import Sequential, Model, model_from_json from keras.layers import Dense, Embedding, LSTM from keras.layers import Input, Flatten, Dropout, Activation, BatchNormalization, AveragePooling2D from keras.layers import Conv1D, MaxPooling1D, AveragePooling1D, Convolution2D, MaxPool2D, Conv2D from keras.utils import np_utils, to_categorical from keras.activations import softmax from keras.callbacks import ModelCheckpoint from keras import optimizers # sklearn from sklearn.metrics import confusion_matrix, accuracy_score, classification_report from sklearn.model_selection import train_test_split from sklearn.preprocessing import LabelEncoder # Other import librosa import librosa.display import json import numpy as np import matplotlib.pyplot as plt import tensorflow as tf from matplotlib.pyplot import specgram import pandas as pd import seaborn as sns import glob import os import pickle import IPython.display as ipd # To play sound in the notebook from __future__ import print_function import keras from keras.layers import Dense, Conv2D, BatchNormalization, Activation from keras.layers import AveragePooling2D, Input, Flatten from keras.optimizers import Adam from keras.callbacks import ModelCheckpoint, LearningRateScheduler from keras.callbacks import ReduceLROnPlateau from keras.preprocessing.image import ImageDataGenerator from keras.regularizers import l2 from keras import backend as K from keras.models import Model import numpy as np import os batch_size = 32 # orig paper trained all networks with batch_size=128 epochs = 200 data_augmentation = True num_classes = 10 subtract_pixel_mean = True n = 3 version = 1 if version == 1: depth = n * 6 + 2 elif version == 2: depth = n * 9 + 2 model_type = 'ResNet%dv%d' % (depth, version) # Load the CIFAR10 data. # (x_train, y_train), (x_test, y_test) = cifar10.load_data() # X_train, y_train, # Input image dimensions. input_shape = X_train.shape[1:] x_test = y_train x_train = X_train print('x_train shape:', x_train.shape) print(x_train.shape[0], 'train samples') print(x_test.shape[0], 'test samples') print('y_train shape:', y_train.shape) # Convert class vectors to binary class matrices. # y_train = keras.utils.to_categorical(y_train, num_classes) # y_test = keras.utils.to_categorical(y_test, num_classes) def lr_schedule(epoch): """Learning Rate Schedule Learning rate is scheduled to be reduced after 80, 120, 160, 180 epochs. Called automatically every epoch as part of callbacks during training. # Arguments epoch (int): The number of epochs # Returns lr (float32): learning rate """ lr = 1e-3 if epoch > 180: lr *= 0.5e-3 elif epoch > 160: lr *= 1e-3 elif epoch > 120: lr *= 1e-2 elif epoch > 80: lr *= 1e-1 print('Learning rate: ', lr) return lr def resnet_layer(inputs, num_filters=16, kernel_size=3, strides=1, activation='relu', batch_normalization=True, conv_first=True): """2D Convolution-Batch Normalization-Activation stack builder # Arguments inputs (tensor): input tensor from input image or previous layer num_filters (int): Conv2D number of filters kernel_size (int): Conv2D square kernel dimensions strides (int): Conv2D square stride dimensions activation (string): activation name batch_normalization (bool): whether to include batch normalization conv_first (bool): conv-bn-activation (True) or bn-activation-conv (False) # Returns x (tensor): tensor as input to the next layer """ conv = Conv2D(num_filters, kernel_size=kernel_size, strides=strides, padding='same', kernel_initializer='he_normal', kernel_regularizer=keras.regularizers.l2(1e-4)) x = inputs if conv_first: x = conv(x) if batch_normalization: x = BatchNormalization()(x) if activation is not None: x = Activation(activation)(x) else: if batch_normalization: x = BatchNormalization()(x) if activation is not None: x = Activation(activation)(x) x = conv(x) return x def resnet_v1(input_shape, depth, num_classes=10): """ResNet Version 1 Model builder [a] Stacks of 2 x (3 x 3) Conv2D-BN-ReLU Last ReLU is after the shortcut connection. At the beginning of each stage, the feature map size is halved (downsampled) by a convolutional layer with strides=2, while the number of filters is doubled. Within each stage, the layers have the same number filters and the same number of filters. Features maps sizes: stage 0: 32x32, 16 stage 1: 16x16, 32 stage 2: 8x8, 64 The Number of parameters is approx the same as Table 6 of [a]: ResNet20 0.27M ResNet32 0.46M ResNet44 0.66M ResNet56 0.85M ResNet110 1.7M # Arguments input_shape (tensor): shape of input image tensor depth (int): number of core convolutional layers num_classes (int): number of classes (CIFAR10 has 10) # Returns model (Model): Keras model instance """ if (depth - 2) % 6 != 0: raise ValueError('depth should be 6n+2 (eg 20, 32, 44 in [a])') # Start model definition. num_filters = 16 num_res_blocks = int((depth - 2) / 6) inputs = Input(shape=input_shape) x = resnet_layer(inputs=inputs) # Instantiate the stack of residual units for stack in range(3): for res_block in range(num_res_blocks): strides = 1 if stack > 0 and res_block == 0: # first layer but not first stack strides = 2 # downsample y = resnet_layer(inputs=x, num_filters=num_filters, strides=strides) y = resnet_layer(inputs=y, num_filters=num_filters, activation=None) if stack > 0 and res_block == 0: # first layer but not first stack # linear projection residual shortcut connection to match # changed dims x = resnet_layer(inputs=x, num_filters=num_filters, kernel_size=1, strides=strides, activation=None, batch_normalization=False) x = keras.layers.add([x, y]) x = Activation('relu')(x) num_filters *= 2 # Add classifier on top. # v1 does not use BN after last shortcut connection-ReLU x = AveragePooling2D(pool_size=8)(x) y = Flatten()(x) outputs = Dense(14, activation='softmax', kernel_initializer='he_normal')(y) # Instantiate model. model = Model(inputs=inputs, outputs=outputs) return model if version == 2: model = resnet_v2(input_shape=input_shape, depth=depth) else: model = resnet_v1(input_shape=input_shape, depth=depth) model.compile(loss='categorical_crossentropy', optimizer=keras.optimizers.Adam(lr=lr_schedule(0)), metrics=['accuracy']) model.summary() print(model_type) # Prepare model model saving directory. save_dir = os.path.join(os.getcwd(), 'saved_models') model_name = 'cifar10_%s_model.{epoch:03d}.h5' % model_type if not os.path.isdir(save_dir): os.makedirs(save_dir) filepath = os.path.join(save_dir, model_name) # Prepare callbacks for model saving and for learning rate adjustment. checkpoint = ModelCheckpoint(filepath=filepath, monitor='val_acc', verbose=1, save_best_only=True) lr_scheduler = LearningRateScheduler(lr_schedule) lr_reducer = ReduceLROnPlateau(factor=np.sqrt(0.1), cooldown=0, patience=5, min_lr=0.5e-6) callbacks = [checkpoint, lr_reducer, lr_scheduler] # Run training, with or without data augmentation. if not data_augmentation: print('Not using data augmentation.') model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, validation_data=(x_test, y_test), shuffle=True, callbacks=callbacks) else: print('Using real-time data augmentation.') # This will do preprocessing and realtime data augmentation: datagen = ImageDataGenerator( # set input mean to 0 over the dataset featurewise_center=False, # set each sample mean to 0 samplewise_center=False, # divide inputs by std of dataset featurewise_std_normalization=False, # divide each input by its std samplewise_std_normalization=False, # apply ZCA whitening zca_whitening=False, # epsilon for ZCA whitening zca_epsilon=1e-06, # randomly rotate images in the range (deg 0 to 180) rotation_range=0, # randomly shift images horizontally width_shift_range=0.1, # randomly shift images vertically height_shift_range=0.1, # set range for random shear shear_range=0., # set range for random zoom zoom_range=0., # set range for random channel shifts channel_shift_range=0., # set mode for filling points outside the input boundaries fill_mode='nearest', # value used for fill_mode = "constant" cval=0., # randomly flip images horizontal_flip=True, # randomly flip images vertical_flip=False, # set rescaling factor (applied before any other transformation) rescale=None, # set function that will be applied on each input preprocessing_function=None, # image data format, either "channels_first" or "channels_last" data_format=None, # fraction of images reserved for validation (strictly between 0 and 1) validation_split=0.0) # Compute quantities required for featurewise normalization # (std, mean, and principal components if ZCA whitening is applied). # datagen.fit(x_train) print("Yes") history = model.fit(x_train, y_train, validation_split=0.2, batch_size=16, verbose = 2, epochs=40) # Fit the model on the batches generated by datagen.flow(). # model.fit_generator(datagen.flow(x_train, y_train, batch_size=batch_size), # validation_data=(x_test, y_test), # epochs=epochs, verbose=1, workers=4, # callbacks=callbacks) # Score trained model. # scores = model.evaluate(x_test, y_test, verbose=1) # print('Test loss:', scores[0]) # print('Test accuracy:', scores[1]) import matplotlib.pyplot as plt plt.plot(history.history['loss'], label='loss') plt.plot(history.history['val_loss'], label = 'val_loss') plt.xlabel('Epoch') plt.ylabel('Loss') # plt.ylim([0.5, 1]) plt.legend(loc='lower right') import matplotlib.pyplot as plt plt.plot(history.history['acc'], label='acc') plt.plot(history.history['val_acc'], label = 'val_acc') plt.xlabel('Epoch') plt.ylabel('Accuracy') # plt.ylim([0.5, 1]) plt.legend(loc='lower right') x_test.shape len(df['labels'].values) (df.labels.unique()) def prepare_data(df, n, aug, mfcc): X = np.empty(shape=(df.shape[0], n, 216, 1)) input_length = sampling_rate * audio_duration cnt = 0 for fname in tqdm(df.path): file_path = fname data, _ = librosa.load(file_path, sr=sampling_rate ,res_type="kaiser_fast" ,duration=2.5 ,offset=0.5 ) # Random offset / Padding if len(data) > input_length: max_offset = len(data) - input_length offset = np.random.randint(max_offset) data = data[offset:(input_length+offset)] else: if input_length > len(data): max_offset = input_length - len(data) offset = np.random.randint(max_offset) else: offset = 0 data = np.pad(data, (offset, int(input_length) - len(data) - offset), "constant") # which feature? if mfcc == 1: # MFCC extraction MFCC = librosa.feature.mfcc(data, sr=sampling_rate, n_mfcc=n_mfcc) MFCC = np.expand_dims(MFCC, axis=-1) X[cnt,] = MFCC else: # Log-melspectogram melspec = librosa.feature.melspectrogram(data, n_mels = n_melspec) logspec = librosa.amplitude_to_db(melspec) logspec = np.expand_dims(logspec, axis=-1) X[cnt,] = logspec cnt += 1 return X X, sample_rate = librosa.load("drive/My Drive/kaggle/surrey/ALL/DC_a08.wav" , res_type='kaiser_fast' ,duration=2.5 ,sr=44100 ,offset=0.5 ) len(np.mean(librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=13), axis=0)) # len((librosa.feature.mfcc(y=X, sr=sample_rate, n_mfcc=16))[0]) input_length = sampling_rate * audio_duration # sampling_rate # audio_duration input_length > len(X) X model.history.history