def main(): block_blob_service = BlockBlobService(account_name=azureStorgeAccountName, account_key=azureStorageKeyName) if (block_blob_service.exists( container_name=azureStorageContainer) == False): block_blob_service.create_container( container_name=azureStorageContainer) audio_path_label_pairs = load_path_labels(str(args.data_dir)) print('loaded: ' + str(len(audio_path_label_pairs)) + ' files') classifier = Cifar10AudioClassifier() batch_size = 8 epochs = 50 classifier.fit(audio_path_label_pairs=audio_path_label_pairs, model_dir_path='model', batch_size=batch_size, epochs=epochs) #zipping and uploading the compiled model to blob storage shutil.make_archive(base_name='model', format='zip', base_dir='model') block_blob_service.create_blob_from_path( azureStorageContainer, str(int(time.time() * 1000)) + 'model.zip', 'model.zip') # copying the model contents to the output directory os.system('cp -r ./model outputs')
def predict(audio_path, model_dir_path): classifier = Cifar10AudioClassifier() classifier.load_model(model_dir_path) predicted_label_id = classifier.predict_class(audio_path) print(predicted_label_id)
def main(): classifier = Cifar10AudioClassifier() classifier.load_model(model_dir_path='./models') classifier.export_tensorflow_model( output_fld='./models/tensorflow_models/cifar10')
def main(): audio_path_label_pairs = load_audio_path_label_pairs() print('loaded: ', len(audio_path_label_pairs)) classifier = Cifar10AudioClassifier() batch_size = 8 epochs = 100 history = classifier.fit(audio_path_label_pairs, model_dir_path='./models_music_speech', batch_size=batch_size, epochs=epochs)
def main(): audio_path_label_pairs = load_audio_path_label_pairs() print('loaded: ', len(audio_path_label_pairs)) classifier = Cifar10AudioClassifier() batch_size = 128 epochs = 200 history = classifier.fit( audio_path_label_pairs, model_dir_path='/newvolume/keras-audio-master/demo/models', batch_size=batch_size, epochs=epochs)
def main(): audio_path_label_pairs = load_audio_path_label_pairs() shuffle(audio_path_label_pairs) print('loaded: ', len(audio_path_label_pairs)) classifier = Cifar10AudioClassifier() classifier.load_model(model_dir_path='./models_music_speech') for i in range(0, 20): audio_path, actual_label_id = audio_path_label_pairs[i] predicted_label_id = classifier.predict_class(audio_path) print(audio_path) predicted_label = gtzan_categories[predicted_label_id] actual_label = gtzan_categories[actual_label_id] print('predicted: ', predicted_label, 'actual: ', actual_label)
def processFile(filename): classifier = Cifar10AudioClassifier() classifier.load_model(model_dir_path='./models') predicted_label_id = classifier.predict_class(filename) predicted_label = labels[predicted_label_id] predicted_scores = classifier.predict(filename) predicted_score = predicted_scores[predicted_label_id] data = {} data['label'] = predicted_label data['score'] = float(predicted_score) data['dateTime'] = str(datetime.now().strftime("%Y-%m-%d %H:%M:%S")) data['modelDate'] = str(time.ctime(os.path.getmtime('models/cifar10-weights.h5'))) data['fileProcessed'] = filename os.remove(filename + "-mg.npy") os.remove(filename) K.clear_session() return json.dumps(data)
def main(): audio_path_label_pairs = load_path_labels(str(args.data_dir)) print('loaded: ' + str(len(audio_path_label_pairs)) + ' files') classifier = Cifar10AudioClassifier() batch_size = 8 epochs = 50000 classifier.fit(audio_path_label_pairs=audio_path_label_pairs, model_dir_path='model', batch_size=batch_size, epochs=epochs) #zipping compiled model and copy back to mounted blob storage shutil.make_archive(base_name='model', format='zip', base_dir='model') os.system('cp -r model.zip ' + args.data_dir + '/' + str(int(time.time() * 1000)) + 'model.zip') # copying the model contents to the output directory os.system('cp -r ./model outputs')