Dalam artikel ini, saya menunjukkan cara menggunakan aliran tensor untuk meramal gaya muzik.
Dalam contoh saya, saya membandingkan muzik tekno dan klasik.
Anda boleh mencari kod pada github saya :
https://github.com/victordalet/sound_to_partition
Untuk langkah pertama, anda perlu mencipta satu penyimpan set data dan di dalamnya tambah satu folder untuk gaya muzik, contohnya saya menambah satu folder tekno dan folder klasik di mana letak wav soung saya.
Saya mencipta fail kereta api, dengan argumen max_epochs untuk diselesaikan.
Ubah suai kelas dalam pembina yang sepadan dengan direktori anda dalam folder set data.
Dalam kaedah pemuatan dan pemprosesan, saya mendapatkan semula fail wav daripada direktori lain dan mendapatkan spektogram.
Untuk tujuan latihan, saya menggunakan konvolusi dan model Keras.
import os import sys from typing import List import librosa import numpy as np from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense from tensorflow.keras.models import Model from tensorflow.keras.optimizers import Adam from sklearn.model_selection import train_test_split from tensorflow.keras.utils import to_categorical from tensorflow.image import resize class Train: def __init__(self): self.X_train = None self.X_test = None self.y_train = None self.y_test = None self.data_dir: str = 'dataset' self.classes: List[str] = ['techno','classic'] self.max_epochs: int = int(sys.argv[1]) @staticmethod def load_and_preprocess_data(data_dir, classes, target_shape=(128, 128)): data = [] labels = [] for i, class_name in enumerate(classes): class_dir = os.path.join(data_dir, class_name) for filename in os.listdir(class_dir): if filename.endswith('.wav'): file_path = os.path.join(class_dir, filename) audio_data, sample_rate = librosa.load(file_path, sr=None) mel_spectrogram = librosa.feature.melspectrogram(y=audio_data, sr=sample_rate) mel_spectrogram = resize(np.expand_dims(mel_spectrogram, axis=-1), target_shape) data.append(mel_spectrogram) labels.append(i) return np.array(data), np.array(labels) def create_model(self): data, labels = self.load_and_preprocess_data(self.data_dir, self.classes) labels = to_categorical(labels, num_classes=len(self.classes)) # Convert labels to one-hot encoding self.X_train, self.X_test, self.y_train, self.y_test = train_test_split(data, labels, test_size=0.2, random_state=42) input_shape = self.X_train[0].shape input_layer = Input(shape=input_shape) x = Conv2D(32, (3, 3), activation='relu')(input_layer) x = MaxPooling2D((2, 2))(x) x = Conv2D(64, (3, 3), activation='relu')(x) x = MaxPooling2D((2, 2))(x) x = Flatten()(x) x = Dense(64, activation='relu')(x) output_layer = Dense(len(self.classes), activation='softmax')(x) self.model = Model(input_layer, output_layer) self.model.compile(optimizer=Adam(learning_rate=0.001), loss='categorical_crossentropy', metrics=['accuracy']) def train_model(self): self.model.fit(self.X_train, self.y_train, epochs=self.max_epochs, batch_size=32, validation_data=(self.X_test, self.y_test)) test_accuracy = self.model.evaluate(self.X_test, self.y_test, verbose=0) print(test_accuracy[1]) def save_model(self): self.model.save('weight.h5') if __name__ == '__main__': train = Train() train.create_model() train.train_model() train.save_model()
Untuk menguji dan menggunakan model, saya telah mencipta kelas ini untuk mendapatkan berat dan meramal gaya muzik.
Jangan lupa untuk menambah kelas yang betul pada pembina.
from typing import List import librosa import numpy as np from tensorflow.keras.models import load_model from tensorflow.image import resize import tensorflow as tf class Test: def __init__(self, audio_file_path: str): self.model = load_model('weight.h5') self.target_shape = (128, 128) self.classes: List[str] = ['techno','classic'] self.audio_file_path: str = audio_file_path def test_audio(self, file_path, model): audio_data, sample_rate = librosa.load(file_path, sr=None) mel_spectrogram = librosa.feature.melspectrogram(y=audio_data, sr=sample_rate) mel_spectrogram = resize(np.expand_dims(mel_spectrogram, axis=-1), self.target_shape) mel_spectrogram = tf.reshape(mel_spectrogram, (1,) + self.target_shape + (1,)) predictions = model.predict(mel_spectrogram) class_probabilities = predictions[0] predicted_class_index = np.argmax(class_probabilities) return class_probabilities, predicted_class_index def test(self): class_probabilities, predicted_class_index = self.test_audio(self.audio_file_path, self.model) for i, class_label in enumerate(self.classes): probability = class_probabilities[i] print(f'Class: {class_label}, Probability: {probability:.4f}') predicted_class = self.classes[predicted_class_index] accuracy = class_probabilities[predicted_class_index] print(f'The audio is classified as: {predicted_class}') print(f'Accuracy: {accuracy:.4f}')
Atas ialah kandungan terperinci Ramalan muzik aliran tensor. Untuk maklumat lanjut, sila ikut artikel berkaitan lain di laman web China PHP!