added logging in loudness_norm and fixed some minor issues (ecb41a64) · Commits · IVAS Codec Public Collaboration / IVAS Processing Scripts

ivas_processing_scripts/audiotools/convert/init.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -251,7 +251,7 @@ def process_audio(
		logger.debug(
		f"Applying loudness adjustment to {loudness} LKFS for format {loudness_fmt} using ITU STL bs1770demo"
		)
		x.audio = loudness_norm(x, loudness, loudness_fmt)
		x.audio = loudness_norm(x, loudness, loudness_fmt, logger=logger)

		"""low-pass filtering"""
		if fc is not None:

ivas_processing_scripts/audiotools/wrappers/bs1770.py

+38 −16

Original line number	Diff line number	Diff line
		@@ -45,9 +45,6 @@ from ivas_processing_scripts.audiotools.wrappers.filter import resample_itu
		from ivas_processing_scripts.constants import DEFAULT_CONFIG_BINARIES
		from ivas_processing_scripts.utils import find_binary, get_devnull, run

		logger = logging.getLogger("__main__")
		logger.setLevel(logging.DEBUG)


		def bs1770demo(
		input: audio.Audio,
		@@ -131,7 +128,7 @@ def bs1770demo(
		write(tmp_file, tmp_sig, 48000)

		# run command
		result = run(cmd, logger=logger)
		result = run(cmd)

		# parse output
		measured_loudness = float(result.stdout.splitlines()[3].split(":")[1])
		@@ -198,6 +195,8 @@ def loudness_norm(
		input: audio.Audio,
		target_loudness: Optional[float] = -26,
		loudness_format: Optional[str] = None,
		logger: Optional[logging.Logger] = None,
		file_name_logging: Optional[Union[str, Path]] = None,
		) -> np.ndarray:
		"""
		Iterative loudness normalization using ITU-R BS.1770-4
		@@ -212,6 +211,10 @@ def loudness_norm(
		Desired loudness level in LKFS
		loudness_format: Optional[str]
		Loudness format to render to for loudness computation (default input format)
		logger: Optional[logging.Logger]
		Logger to log loudness information
		file_name_logging: Optional[Union[str, Path]]
		Name of processed file for logging information

		Returns
		-------
		@@ -224,22 +227,38 @@ def loudness_norm(
		scale_factor = 1
		num_iter = 1

		while np.abs(measured_loudness - target_loudness) > 0.5 and num_iter < 10:
		measured_loudness, scale_factor_new = get_loudness(
		# save loudness before and after scaling for the logger info
		loudness_before, scale_factor_new = get_loudness(
		input, target_loudness, loudness_format
		)

		while np.abs(measured_loudness - target_loudness) > 0.5 and num_iter < 10:

		# scale input
		input.audio *= scale_factor_new

		# update scale factor
		scale_factor *= scale_factor_new

		# measure loudness and get scaling factor
		measured_loudness, scale_factor_new = get_loudness(
		input, target_loudness, loudness_format
		)

		num_iter += 1

		loudness_after = measured_loudness

		# log loudness before and after adjustment
		if logger:
		if file_name_logging:
		logger.debug(f"File {file_name_logging} loudness; before: {loudness_before}, after: {loudness_after}")
		else:
		logger.debug(f"Loudness; before: {loudness_before}, after: {loudness_after}")

		if num_iter >= 10:
		warn(
		f"Loudness did not converge to desired value, stopping at: {measured_loudness:.2f}"
		f"Loudness did not converge to desired value, stopping at: {loudness_after:.2f}"
		)

		return input.audio
		@@ -249,6 +268,7 @@ def scale_files(
		file_list: list[list[Union[Path, str]]],
		fmt: str,
		loudness: float,
		loudness_format: Optional[str] = None,
		fs: Optional[int] = 48000,
		in_meta: Optional[list] = None,
		logger: Optional[logging.Logger] = None,
		@@ -264,15 +284,22 @@ def scale_files(
		Audio format of files in list
		loudness: float
		Desired loudness level in LKFS/dBov
		loudness_format: Optional[str]
		Format for loudness measurement
		fs: Optional[int]
		Sampling rate
		in_meta: Optional[list]
		Metadata for ISM with same structure as file_list but one layer more
		for the list of metadata for one file
		logger: Optional[logging.Logger]
		Logger to log loudness information
		"""

		if fmt.startswith("ISM") and in_meta:
		if fmt.startswith("ISM"):
		if in_meta:
		meta_bool = True
		else:
		raise ValueError("No metadata available for loudness measurement")
		else:
		in_meta = copy.copy(file_list)
		meta_bool = False
		@@ -285,13 +312,8 @@ def scale_files(
		else:
		audio_obj = audio.fromfile(fmt, file, fs)

		# TODO: just for testing: remove later
		if logger:
		measured_loudness, _ = get_loudness(audio_obj)
		logger.info(f"measured loudness of file {file.name} for condition {file.parent.name}: {measured_loudness}")

		# adjust loudness
		scaled_audio = loudness_norm(audio_obj, loudness)
		scaled_audio = loudness_norm(audio_obj, loudness, loudness_format, logger, file)

		# write into file
		write(file, scaled_audio, audio_obj.fs)

ivas_processing_scripts/processing/processing.py

+1 −0

Original line number	Diff line number	Diff line
		@@ -354,6 +354,7 @@ def reverse_process_2(cfg, logger):
		out_paths_splits,
		cfg.postprocessing["fmt"],
		cfg.postprocessing["loudness"],
		cfg.postprocessing.get("loudness_fmt", None),
		cfg.postprocessing["fs"],
		out_meta_splits,
		logger,