def test_speaker_embedding():
# load config
config = load_config(encoder_config_path)
config.audio.resample = True
# create a dummy speaker encoder
model = setup_encoder_model(config)
save_checkpoint(model, None, None, get_tests_input_path(), 0)
# load audio processor and speaker encoder
ap = AudioProcessor(**config.audio)
manager = SpeakerManager(encoder_model_path=encoder_model_path,
encoder_config_path=encoder_config_path)
# load a sample audio and compute embedding
waveform = ap.load_wav(sample_wav_path)
mel = ap.melspectrogram(waveform)
d_vector = manager.compute_embeddings(mel)
assert d_vector.shape[1] == 256
# compute d_vector directly from an input file
d_vector = manager.compute_embedding_from_clip(sample_wav_path)
d_vector2 = manager.compute_embedding_from_clip(sample_wav_path)
d_vector = torch.FloatTensor(d_vector)
d_vector2 = torch.FloatTensor(d_vector2)
assert d_vector.shape[0] == 256
assert (d_vector - d_vector2).sum() == 0.0
# compute d_vector from a list of wav files.
d_vector3 = manager.compute_embedding_from_clip(
[sample_wav_path, sample_wav_path2])
d_vector3 = torch.FloatTensor(d_vector3)
assert d_vector3.shape[0] == 256
assert (d_vector - d_vector3).sum() != 0.0
# remove dummy model
os.remove(encoder_model_path)
# compute speaker embeddings
speaker_mapping = {}
for idx, wav_file in enumerate(tqdm(wav_files)):
if isinstance(wav_file, dict):
class_name = wav_file[class_name_key]
wav_file = wav_file["audio_file"]
else:
class_name = None
wav_file_name = os.path.basename(wav_file)
if args.old_file is not None and wav_file_name in encoder_manager.clip_ids:
# get the embedding from the old file
embedd = encoder_manager.get_embedding_by_clip(wav_file_name)
else:
# extract the embedding
embedd = encoder_manager.compute_embedding_from_clip(wav_file)
# create speaker_mapping if target dataset is defined
speaker_mapping[wav_file_name] = {}
speaker_mapping[wav_file_name]["name"] = class_name
speaker_mapping[wav_file_name]["embedding"] = embedd
if speaker_mapping:
# save speaker_mapping if target dataset is defined
if ".json" not in args.output_path:
mapping_file_path = os.path.join(args.output_path, "speakers.json")
else:
mapping_file_path = args.output_path
os.makedirs(os.path.dirname(mapping_file_path), exist_ok=True)